Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Exposure to occupational noise and risk of cardiovascular disease : a retrospective cohort study Davies, Hugh William 2002

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_2002-75006X.pdf [ 8.65MB ]
Metadata
JSON: 831-1.0076817.json
JSON-LD: 831-1.0076817-ld.json
RDF/XML (Pretty): 831-1.0076817-rdf.xml
RDF/JSON: 831-1.0076817-rdf.json
Turtle: 831-1.0076817-turtle.txt
N-Triples: 831-1.0076817-rdf-ntriples.txt
Original Record: 831-1.0076817-source.json
Full Text
831-1.0076817-fulltext.txt
Citation
831-1.0076817.ris

Full Text

E X P O S U R E T O O C C U P A T I O N A L N O I S E A N D R I S K OF C A R D I O V A S C U L A R D I S E A S E : A R E T R O S P E C T I V E C O H O R T S T U D Y by H U G H W I L L I A M D A V I E S B . S c , The University of Alberta, 1992 M . S c , The University of British Columbia, 1995 A THESIS S U B M I T T E D I N P A R T I A L F U L F I L M E N T OF T H E R E Q U I R E M E N T S F O R T H E D E G R E E OF D O C T O R OF P H I L O S O P H Y in T H E F A C U L T Y OF G R A D U A T E STUDIES (Interdisciplinary Studies [School of Occupational and Environmental Hygiene/Health Care and Epidemiology/Mechanical Engineering]) We accept this thesis as conforming to the required standard. T H E U N I V E R S I T Y OF B R I T I S H C O L U M B I A September 2002 © Hugh Wil l iam Davies, 2002 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department" of „':••-.-:,.. ~»f - -..--4 -y ^ . v 1 "*• \. «, %.(..,( t The University of British Columbia Vancouver, Canada Date DE-6 (2/88) Abstract Introduction: Noise is a powerful stressor that consistently produces physiological responses typical of a stress reaction in experimental settings. However, results of observational studies of a hypothesized link between noise and cardiovascular disease have been inconsistent. This retrospective study examined associations between occupational noise exposure and cardiovascular mortality in a pre-existing cohort of 27,499 lumber mill workers. Methods: Subjects were males from 14 lumber mills employed at least one year between 1950 and 1995. Historical exposure levels were estimated by a determinants-of-exposure regression model, developed using 1,900 personal dosimetry measurements. An exposure data matrix was created for 3,809 mill/job-title/time period combinations, then combined with work histories to calculate various cumulative exposure metrics. SMR's were calculated using the regional general population as referents. Exposure-response relationships were examined using Poisson regression, adjusted for age, calendar year and ethnicity. Results: There were 2,519 circulatory disease deaths, of which 910 were due to acute MI. SMR's for circulatory diseases were close to 1, although lower risks were anticipated due to the "healthy worker" effect. SMR's for acute MI were elevated 30-40% for workers exposed more than 30 years over 95 dBA, or for 20 years over 100 dBA. Within the cohort, relative risk (RR) of acute MI was increased 30-60% for those exposed more than 30 years over 95 dBA. To reduce exposure misclassification resulting from hearing protector use, a sub-cohort who terminated cohort employment before 1970 was created. Sub-cohort RR's for those exposed more than 20 years above 85, 90, and 95 dBA ranged from 1.3-1.5 (Ptrend's, <0.05). RR's for cumulative exposure reached 1.6 in the highest exposure group (>115 dBA*yr, Ptrend <0.001). RR's were greatest when follow-up was limited to the period of employment, ranging from 2-A (Ptrend's <0.01). RR's for ischemic heart disease were elevated but not as strongly or consistently as for acute MI. RR's for stroke and hypertensive diseases did not show consistent patterns of increased risk. Smoking did not appear to be a confounder. Conclusion: Occupational exposure to noise above 85 dBA is associated with increased risk of acute MI. Risks appear highest for those actively employed. Table of Contents ABSTRACT ii TABLE OF CONTENTS iii LIST OF TABLES vi LIST OF FIGURES x ABBREVIATIONS xi ACKNOWLEDGMENTS xiii DEDICATION xiv CHAPTER 1. GENERAL INTRODUCTION AND SCOPE OF DISSERTATION 1 CHAPTER 2. BACKGROUND: NOISE, STRESS AND DISEASE 3 INTRODUCTION 3 STRESS A N D C A R D I O V A S C U L A R D I S E A S E 4 A General Model of Stress and Disease 5 N O I S E AS A STRESSOR 7 Attributes of the Noise Signal Influencing the Stress Response 9 S U M M A R Y 10 CHAPTER 3. NOISE AND CARDIOVASCULAR DISEASE - A REVIEW OF THE LITERATURE 11 M E T H O D S 11 E A R L Y E P I D E M I O L O G Y STUDIES O F NOISE A N D C V D 11 EPIDEMIOLOGICAL STUDIES O F T H E H E M O D Y N A M I C E F F E C T S O F E X P O S U R E T O N O I S E 12 Noise and Blood-Pressure Change 12 Noise and Hypertension 21 Noise and Other Hemodynamic outcomes 22 Limitations of Studies of Hemodynamic Outcomes 22 EPIDEMIOLOGICAL STUDIES O N T H E E F F E C T S O F N O I S E O N E N D O C R I N E A N D B L O O D LIPIDS 25 EPIDEMIOLOGICAL STUDIES O F NOISE A N D C H R O N I C D I S E A S E ENDPOINTS 25 N O I S E E X P O S U R E F A C T O R S 31 S U M M A R Y 33 CHAPTER 4. CHARACTERIZING NOISE EXPOSURE IN BC LUMBER MILLS 36 C O H O R T S T U D Y S E T T I N G 36 M E T H O D S 3 9 Sampling Strategy 39 Noise Measurement 40 Frequency Spectrum Analysis 40 Hearing Protection Device Usage 40 Determinants of Noise Exposure 40 R E S U L T S 4 1 General Mill Characteristics 41 Noise Exposure Levels 42 Octave Band Analysis 44 Hearing Protector Device Usage 44 Exposure Determinant Modeling f. 46 D I S C U S S I O N 4 6 S U M M A R Y 5 0 CHAPTER 5. RETROSPECTIVE EXPOSURE ASSESSMENT OF NOISE IN BC LUMBER MILLS 52 I N T R O D U C T I O N 5 2 M E T H O D S 5 4 Exposure Data Acquisition 54 Exposure Data Preparation '. 54 Regression Analysis 55 Estimating Historical Exposure Levels 59 R E S U L T S 6 3 Overall BC Lumber Mill Data - Including Non-Cohort Data 63 Regressio'n Modeling 68 Non-Modeled Adjustment Factors 72 Building the Exposure File 74 S U M M A R Y 7 8 CHAPTER 6. CARDIOVASCULAR DISEASE IN BC LUMBER MILL WORKERS: EPIDEMIOLOGICAL ANALYSES 80 I N T R O D U C T I O N 8 0 M E T H O D S 8 0 Study Population 80 Follow-up 81 Exposure assessment 82 Health Outcomes 82 Potential Confounders 83 Statistical Analyses 84 R E S U L T S 8 5 Analyses of the Full Cohort (n=27,499) 86 Analyses of the Pre-1970 Employment Sub-cohort (n-8,700) 90 Cumulative Exposure to Noise 93 Analyses Of Outcomes Occurring During Employment, or Within One Month of Termination 96 Disease Latency 97 Hearing Protection Use - Adjusted Analyses 105 C-Weighted Noise Exposure 105 Assessment of the Impact of Smoking 113 D I S C U S S I O N 1 1 6 Acute Myocardial Infarction 116 Ischemic Heart Disease 118 Hypertensive Diseases 118 Stroke 118 Strengths and Limitations 118 S U M M A R Y 1 2 1 CHAPTER 7. CONCLUSION AND RECOMMENDATIONS FOR FURTHER WORK 123 BIBLIOGRAPHY 128 v List of Tables T A B L E 1: S U M M A R Y O F S T U D I E S O F N O I S E A N D H E M O D Y N A M I C A N D B L O O D C H E M I S T R Y O U T C O M E S ; O C C U P A T I O N A L E X P O S U R E S ; E N G L I S H L A N G U A G E P U B L I C A T I O N S , 1 9 8 0 - 2 0 0 2 1 3 T A B L E 2 : S U M M A R Y O F S T U D I E S O F N O I S E A N D H E M O D Y N A M I C A N D B L O O D C H E M I S T R Y O U T C O M E S , E N V I R O N M E N T A L E X P O S U R E S ; E N G L I S H L A N G U A G E P U B L I C A T I O N S , 1 9 8 0 - 2 0 0 0 1 9 T A B L E 3 : E P I D E M I O L O G I C A L S T U D I E S O F N O I S E A N D C H R O N I C H E A R T D I S E A S E O U T C O M E S , O C C U P A T I O N A L E X P O S U R E S , E N G L I S H L A N G U A G E P U B L I C A T I O N S , A L L Y E A R S 2 6 T A B L E 4 : S U M M A R Y O F E P I D E M I O L O G I C A L S T U D I E S O F N O I S E A N D C H R O N I C H E A R T D I S E A S E O U T C O M E S , E N V I R O N M E N T A L E X P O S U R E S , E N G L I S H L A N G U A G E P U B L I C A T I O N S , A L L Y E A R S 2 7 T A B L E 5 : C O H O R T M I L L C H A R A C T E R I S T I C S : L O C A T I O N A N D S E L E C T E D D E S C R I P T I V E M I L L P A R A M E T E R S 3 7 T A B L E 6 : S U M M A R Y S T A T I S T I C S : N O I S E D O S I M E T R Y S U R V E Y , 1 9 9 6 - 1 9 9 7 4 2 T A B L E 7: S U M M A R Y O F N O I S E E X P O S U R E B Y M E A N S O U N D P R E S S U R E L E V E L ( L P ) 4 4 T A B L E 8: S E L F - R E P O R T E D H E A R I N G P R O T E C T I O N U S E A M O N G 2 8 3 * B R I T I S H C O L U M B I A L U M B E R M I L L W O R K E R S , 4 5 T A B L E 9 : D E T E R M I N A N T S O F N O I S E E X P O S U R E R E G R E S S I O N M O D E L . F U L L - S H I F T L E Q A N D S E L E C T E D D E T E R M I N A N T S O F E X P O S U R E 4 7 T A B L E 1 0 : L U M B E R M I L L N O I S E E X P O S U R E L E V E L S - P R E V I O U S L Y P U B L I S H E D S T U D I E S 4 8 T A B L E 11 : P R O C E S S G R O U P S D E S C R I P T I O N S , E X A M P L E S O F T Y P I C A L J O B S F O U N D W I T H I N P R O C E S S G R O U P S , A N D T Y P I C A L A D M I N I S T R A T I V E D E P A R T M E N T 5 5 T A B L E 1 2 : S U M M A R Y O F P R E D I C T O R V A R I A B L E S P R E S E N T E D I N D E T E R M I N A N T S O F E X P O S U R E M O D E L I N G 5 7 T A B L E 1 3 : E X P O S U R E D A T A S U M M A R Y . B Y S O U R C E , S A M P L E T Y P E , M I L L T Y P E 6 3 T A B L E 14: N O I S E E X P O S U R E M E A N , ( S D ) A N D F R E Q U E N C Y , B Y S O U R C E , S A M P L I N G M E T H O D A N D M I L L 6 6 T A B L E 1 5 : D E S C R I P T I V E S T A T I S T I C S F O R N O I S E M E A S U R E M E N T D A T A S E T A N D S U B - S E T S 6 6 T A B L E 1 6 : M E A N N O I S E L E V E L S T A N D A R D D E V I A T I O N A N D N U M B E R O F S A M P L E S , B Y Y E A R A N D M I L L 6 7 T A B L E 17: M E A N L E Q , S D A N D N U M B E R O F O B S E R V A T I O N S , M O D E L I N G D A T A S E T ( N = 1 , 9 0 1 ) B Y P R O C E S S G R O U P A N D M I L L 6 9 T A B L E 18: M U L T I P L E L I N E A R R E G R E S S I O N M O D E L O F N O I S E E X P O S U R E I N L U M B E R M I L L W O R K E R S ( N = 1 , 9 0 1 ) 7 0 T A B L E 1 9 : H E A R I N G P R O T E C T I O N F A C T O R S ( D B A ) F O R H E A R I N G P R O T E C T O R U S E I N B C L U M B E R M I L L W O R K E R S 7 3 T A B L E 2 0 : L O W - F R E Q U E N C Y W E I G H T I N G F A C T O R S , A N D P R E D I C T E D A - W E I G H T E D A N D E S T I M A T E D C -W E I G H T E D P R O C E S S G R O U P M E A N S ( B A S E D O N F U L L M O D E L I N G D A T A S E T , N = 1 , 9 0 1 ) 7 4 T A B L E 2 1 : E X P O S U R E D A T A M A T R I X : D E S C R I P T I V E S T A T I S T I C S , A N D D I S T R I B U T I O N O F C U M U L A T I V E E X P O S U R E B Y P R O C E S S G R O U P 7 5 T A B L E 2 2 : C U M U L A T I V E E X P O S U R E D I S T R I B U T I O N B Y D E C A D E 7 6 vi T A B L E 2 3 : P R O V I N C E - W I D E S M O K I N G R A T E S ( M E N , A G E 1 2 + ) , C O M P A R E D T O 1 9 9 7 Q U E S T I O N N A I R E S U B -C O H O R T ( M E N , A G E 3 5 + ) 8 4 T A B L E 2 4 : I N D E X O F S T A T I S T I C A L A N A L Y S E S , A N D T A B L E S O F R E S U L T S I N C H A P T E R 6 8 5 T A B L E 2 5 : D E M O G R A P H I C C H A R A C T E R I S T I C S F O R ( A ) T H E B C L U M B E R M I L L W O R K E R S C O H O R T ( N = 2 7 , 4 9 9 M A L E S ) ; A N D ( B ) A S U B - C O H O R T O F W O R K E R S W H O C O M P L E T E D E M P L O Y M E N T P R I O R T O 1 9 7 0 ( N = 8 , 7 0 0 M A L E S ) 8 6 T A B L E 2 6 : S T A N D A R D I Z E D M O R T A L I T Y R A T I O S ( S M R ' S ) A N D 9 5 % C O N F I D E N C E I N T E R V A L S ( C I ' S ) F O R A L L -C A U S E A N D C A U S E - S P E C I F I C M O R T A L I T Y A M O N G B C L U M B E R M I L L W O R K E R S ( A ) F U L L C O H O R T ( 2 7 , 4 9 9 M A L E S ) , A N D ( B ) P R E - 1 9 7 0 E M P L O Y M E N T S U B - C O H O R T ( N = 8 , 7 0 0 ) . R E F E R E N C E R A T E S F R O M T H E G E N E R A L B C P O P U L A T I O N . . 8 7 T A B L E 2 7 : S T A N D A R D I Z E D M O R T A L I T Y R A T I O S ( S M R ' S ) A N D 9 5 % C O N F I D E N C E I N T E R V A L S ( C I ) F O R A L L C A U S E S , A N D S E L E C T E D C A R D I O V A S C U L A R D I S E A S E S B Y D U R A T I O N O F E M P L O Y M E N T . B C L U M B E R M I L L W O R K E R S , ( A ) F U L L C O H O R T , ( 2 7 , 4 9 9 M A L E S ) ; ( B ) P R E - 1 9 7 0 E M P L O Y M E N T S U B - C O H O R T ( N = 8 , 7 0 0 ) . R E F E R E N C E R A T E S F R O M G E N E R A L B C P O P U L A T I O N 8 8 T A B L E 2 8 : S T A N D A R D I Z E D M O R T A L I T Y R A T I O S ( S M R ' S ) A N D 9 5 % C O N F I D E N C E I N T E R V A L S ( C I ) F O R A L L C A U S E S A N D S E L E C T E D C A R D I O V A S C U L A R D I S E A S E S B Y D U R A T I O N O F E X P O S U R E A B O V E T H R E S H O L D S O F 8 5 , 9 0 , 9 5 A N D 1 0 0 D B A . B C L U M B E R M I L L W O R K E R S C O H O R T ( 2 7 , 4 9 9 M A L E S ) . R E F E R E N C E R A T E S F R O M T H E G E N E R A L B C P O P U L A T I O N 8 9 T A B L E 2 9 : P O I S S O N R E G R E S S I O N : R E L A T I V E R I S K S A N D 9 5 % C O N F I D E N C E I N T E R V A L S ( C I ) F O R D U R A T I O N O F E X P O S U R E T O N O I S E A B O V E T H R E S H O L D S O F 8 5 , 9 0 , 9 5 A N D 1 0 0 D B A I N B C L U M B E R M I L L W O R K E R S C O H O R T F U L L C O H O R T ( 2 7 , 4 9 9 M A L E S ) 9 1 T A B L E 3 0 : S T A N D A R D I Z E D M O R T A L I T Y R A T I O S ( S M R ' S ) A N D 9 5 % C O N F I D E N C E I N T E R V A L S F O R A L L C A U S E S A N D S E L E C T E D C A R D I O V A S C U L A R D I S E A S E S B Y D U R A T I O N O F E X P O S U R E A B O V E T H R E S H O L D O F 8 5 , 9 0 , 9 5 A N D I O O D B A , I N A S U B - C O H O R T O F M A L E W O R K E R S W H O S E D A T E O F L A S T E M P L O Y M E N T W A S P R I O R T O 1 9 7 0 ( N = 8 7 0 0 M A L E S ) . R E F E R E N C E R A T E S F R O M G E N E R A L B C P O P U L A T I O N 9 2 T A B L E 3 1 : P O I S S O N R E G R E S S I O N : R E L A T I V E R I S K S A N D 9 5 % C O N F I D E N C E I N T E R V A L S ( C I ) F O R D U R A T I O N O F E X P O S U R E T O N O I S E A B O V E T H R E S H O L D S O F 8 5 , 9 0 , 9 5 A N D 1 0 0 D B A I N B C L U M B E R M I L L W O R K E R S C O H O R T . P R E - 1 9 7 0 S U B - C O H O R T ( 8 , 7 0 0 M A L E S ) 9 4 T A B L E 3 2 : S T A N D A R D I Z E D M O R T A L I T Y R A T I O S ( S M R ' S ) A N D 9 5 % C O N F I D E N C E I N T E R V A L S ( 9 5 % C I ) F O R A L L C A U S E S A N D S E L E C T E D C A R D I O V A S C U L A R D I S E A S E S B Y C U M U L A T I V E E X P O S U R E A B O V E 8 5 D B A I N ( A ) B C L U M B E R M I L L W O R K E R S C O H O R T ( 2 7 , 4 9 9 M A L E S ) , A N D ( B ) P R E - 1 9 7 0 E M P L O Y M E N T S U B - C O H O R T ( 8 , 7 0 0 M A L E S ) . R E F E R E N C E R A T E S F R O M G E N E R A L B C P O P U L A T I O N 9 5 T A B L E 3 3 : P O I S S O N R E G R E S S I O N : R E L A T I V E R I S K A N D 9 5 % C O N F I D E N C E I N T E R V A L S ( C I ) B Y C U M U L A T I V E E X P O S U R E ( D B A * Y R ) A B O V E 8 5 D B A . ( A ) F U L L C O H O R T ( 2 7 , 4 9 9 M A L E S ) ; ( B ) P R E - 1 9 7 0 E M P L O Y M E N T S U B - C O H O R T ( N = 8 , 7 0 0 M A L E S ) 9 6 vii T A B L E 34: P O I S S O N R E G R E S S I O N : R E L A T I V E R I S K S A N D 95% C O N F I D E N C E I N T E R V A L S (CI) F O R S E L E C T E D CVD's W H E N F O L L O W - U P R E S T R I C T E D T O P E R I O D O F E M P L O Y M E N T P L U S O N E M O N T H . R E L A T I V E R I S K F O R S E L E C T E D C A R D I O V A S C U L A R D I S E A S E S § I N T H E F U L L C O H O R T 98 T A B L E 35: C U M U L A T I V E E X P O S U R E A N D R E S T R I C T E D F O L L O W - U P . R E L A T I V E R I S K A N D 95% C O N F I D E N C E I N T E R V A L S F O R S E L E C T E D C A R D I O V A S C U L A R D I S E A S E S B Y C U M U L A T I V E E X P O S U R E A B O V E 85 DBA. B C L U M B E R M I L L W O R K E R S C O H O R T , (N=27,499 M A L E S ) 99 T A B L E 36: L A T E N C Y A N A L Y S I S . P O I S S O N R E G R E S S I O N , A C U T E M Y O C A R D I A L I N F A R C T I O N A N D D U R A T I O N O F E X P O S U R E T O N O I S E I N B C L U M B E R M I L L W O R K E R S F U L L C O H O R T (N=27,499 M A L E S ) 100 T A B L E 37: L A T E N C Y A N A L Y S I S . P O I S S O N R E G R E S S I O N . H Y P E R T E N S I V E D I S E A S E A N D D U R A T I O N O F E X P O S U R E T O N O I S E I N B C L U M B E R M I L L W O R K E R S F U L L C O H O R T (N=27,499 M A L E S ) 101 T A B L E 38: L A T E N C Y A N A L Y S I S : P O I S S O N R E G R E S S I O N I S C H E M I C H E A R T D I S E A S E A N D D U R A T I O N O F E X P O S U R E T O N O I S E I N B C L U M B E R M I L L W O R K E R S F U L L C O H O R T (N=27,499 M A L E S ) 102 T A B L E 39: L A T E N C Y A N A L Y S I S . P O I S S O N R E G R E S S I O N . S T R O K E ( C E R E B R O V A S C U L A R A C C I D E N T ) A N D D U R A T I O N O F E X P O S U R E T O N O I S E I N B C L U M B E R M I L L W O R K E R S F U L L C O H O R T (N=27,499 M A L E S ) . . . . 103 T A B L E 40: L A T E N C Y A N A L Y S I S . P O I S S O N R E G R E S S I O N . A C U T E M Y O C A R D I A L I N F A R C T I O N A N D D U R A T I O N O F E X P O S U R E T O N O I S E I N A S U B - C O H O R T O F L U M B E R M I L L W O R K E R S E M P L O Y E D P R I O R T O 1970 (N=8700 M A L E S ) 104 T A B L E 41: L A T E N C Y A N A L Y S I S . P O I S S O N R E G R E S S S I O N . H Y P E R T E N S I O N A N D D U R A T I O N O F E X P O S U R E T O N O I S E I N A S U B - C O H O R T O F L U M B E R M I L L W O R K E R S E M P L O Y E D P R I O R T O 1970 (N=8700 M A L E S ) 106 T A B L E 42: L A T E N C Y A N A L Y S I S . P O I S S O N R E G R E S S I O N . I S C H E M I C H E A R T D I S E A S E A N D D U R A T I O N O F E X P O S U R E T O N O I S E I N A S U B - C O H O R T O F L U M B E R M I L L W O R K E R S E M P L O Y E D P R I O R T O 1970 (N=8700 M A L E S ) 107 T A B L E 43: L A T E N C Y A N A L Y S I S . P O I S S O N R E G R E S S I O N . S T R O K E ( C E R E B R O V A S C U L A R A C C I D E N T ) A N D D U R A T I O N O F E X P O S U R E T O N O I S E I N A S U B - C O H O R T O F L U M B E R M I L L W O R K E R S E M P L O Y E D P R I O R T O 1970 (N=8700 M A L E S ) 108 T A B L E 44: L A T E N C Y A N A L Y S I S . P O I S S O N R E G R E S S I O N . R E L A T I V E R I S K A N D 95% C O N F I D E N C E I N T E R V A L S F O R S E L E C T E D C A R D I O V A S C U L A R D I S E A S E S B Y C U M U L A T I V E E X P O S U R E A B O V E 85 (A). B C L U M B E R M I L L W O R K E R S C O H O R T (N=27,499 M A L E S ) 109 T A B L E 45: P O I S S O N R E G R E S S I O N : R E L A T I V E R I S K S A N D 95% CPs F O R S E L E C T E D C A R D I O V A S C U L A R D I S E A S E R E L A T I N G T O E X P O S U R E S O V E R T H R E S H O L D S O F 85, 90, 95 A N D 100 DBA A F T E R A D J U S T I N G F O R H E A R I N G P R O T E C T O R U S E . B C L U M B E R M I L L W O R K E R S F U L L C O H O R T (N=27,499 M A L E S ) 110 T A B L E 46: P O I S S O N R E G R E S S I O N : C U M U L A T I V E E X P O S U R E , H E A R I N G P R O T E C T O R - A D J U S T E D . R E L A T I V E R I S K A N D 95% C O N F I D E N C E I N T E R V A L S F O R S E L E C T E D C A R D I O V A S C U L A R D I S E A S E S B Y C U M U L A T I V E E X P O S U R E A B O V E 85 (A). B C L U M B E R M I L L W O R K E R S F U L L C O H O R T (N=27,499 M A L E S ) I l l T A B L E 47: P O I S S O N R E G R E S S I O N : R E L A T I V E R I S K S A N D 95% CPs F O R S E L E C T E D C A R D I O V A S C U L A R D I S E A S E R E L A T I N G T O E X P O S U R E S O V E R T H R E S H O L D S O F 85,90,95 A N D 100 DBA A F T E R A D J U S T I N G F O R C-W E I G H T I N G S C A L E . B C L U M B E R M I L L W O R K E R S F U L L C O H O R T (N=27,499 M A L E S ) 112 viii T A B L E 4 8 : M E A N D U R A T I O N O F E X P O S U R E B Y 1 9 9 7 S M O K I N G S T A T U S A N D A G E . ( A ) S U B J E C T S O F E U R O P E A N A N D O T H E R A N C E S T R Y ; ( B ) S U B J E C T S O F S O U T H A S I A N A N C E S T R Y 1 1 3 T A B L E 4 9 : M E A N C U M U L A T I V E E X P O S U R E ( D B A R E F 2 0 U P A . Y R ) B Y 1 9 9 7 S M O K I N G S T A T U S A N D E T H N I C I T Y 1 1 3 T A B L E 5 0 : M E A N C U M U L A T I V E E X P O S U R E ( D B A * Y R ) ( N ) B Y 1 9 9 7 S M O K I N G S T A T U S , E T H N I C I T Y A N D A G E .... 114 T A B L E 51: M E A N P A C K Y E A R S ( N ) B Y C U M U L A T I V E E X P O S U R E C A T E G O R Y , E T H N I C I T Y A N D A G E 1 1 4 T A B L E 5 2 : M E A N E X P O S U R E D U R A T I O N ( I N Y E A R S ) > 9 5 D B ( Y E A R S ) B Y 1 9 9 7 S M O K I N G S T A T U S , E T H N I C I T Y A N D A G E 1 1 5 T A B L E 5 3 : M E A N E X P O S U R E D U R A T I O N ( I N Y E A R S ) > 9 5 D B A ( N ) B Y 1 9 9 7 S M O K I N G S T A T U S , E T H N I C I T Y A N D A G E 1 1 5 T A B L E 5 4 : M E A N P A C K Y E A R S B Y E X P O S U R E D U R A T I O N > 9 5 D B , ( N ) , B Y A G E A N D E T H N I C I T Y 1 1 5 T A B L E 5 5 : L U N G C A N C E R R E L A T I V E R I S K S ( O B S E R V E D D E A T H S ) F R O M S E L E C T E D P A R A L L E L I N T E R N A L ( P O I S S O N R E G R E S S I O N ) A N A L Y S E S . A D J U S T E D F O R A G E , C A L E N D A R Y E A R A N D E T H N I C I T Y 1 1 6 ix List of Figures F I G U R E 1: N O I S E A N D C A R D I O V A S C U L A R D I S E A S E R I S K M O D E L ( A F T E R B A B I S C H , 1 9 9 8 ) 6 F I G U R E 2: L U M B E R M I L L D E P A R T M E N T S , P R O C E S S F L O W A N D R E P O R T E D N O I S E S O U R C E S 3 8 F I G U R E 3 : C U M U L A T I V E D I S T R I B U T I O N F U N C T I O N O F 8 - H O U R T W A N O I S E E X P O S U R E ( L ^ ) 4 3 F I G U R E 4 : M E A N N O I S E E X P O S U R E L E V E L S B Y L U M B E R M I L L D E P A R T M E N T , 4 3 F I G U R E 5 : N O I S E F R E Q U E N C Y A S S E S S M E N T . O C T A V E B A N D A N A L Y S I S B Y L U M B E R M I L L D E P A R T M E N T 4 5 F I G U R E 6: H I S T O R I C A L T R E N D S I N S E L F - R E P O R T E D H E A R I N G P R O T E C T I O N D E V I C E U S E F R O M 1 9 7 9 T O 1 9 9 6 , F R O M D A T A C O L L E C T E D A T T I M E O F A N N U A L H E A R I N G T E S T ( N = 3 0 , 4 5 9 ) 4 6 F I G U R E 7: R E T R O S P E C T I V E E X P O S U R E A S S E S S M E N T : P R O C E S S D I A G R A M , A N D E X P O S U R E D A T A F L O W 5 6 F I G U R E 8: L O G I C D I A G R A M : A S S I G N I N G V A L U E S T O W O R K H I S T O R Y R E C O R D S W I T H M I S S I N G J O B - T I T L E O R M I S S I N G J O B T I T L E A N D D E P A R T M E N T 6 0 F I G U R E 9 : R E T R O S P E C T I V E E X P O S U R E A S S E S S M E N T : E X P O S U R E D A T A M A T R I X C O N S T R U C T I O N , W I T H S A M P L E D A T A S H O W I N G R E L A T I O N S B E T W E E N K E Y F I L E S , A N D U S E O F M O D E L A N D A D J U S T M E N T F A C T O R S T O A S S I G N E X P L E V E L S 6 1 F I G U R E 1 0 : N O I S E E X P O S U R E D A T A , C O H O R T M I L L S : S A M P L E F R E Q U E N C Y B Y Y E A R , M E A S U R E M E N T T Y P E , A N D D A T A S O U R C E 6 4 F I G U R E 11: N O I S E E X P O S U R E D A T A , N O N - C O H O R T M I L L S : S A M P L E F R E Q U E N C Y B Y Y E A R , M E A S U R E M E N T T Y P E , A N D D A T A S O U R C E 6 5 F I G U R E 1 2 : F R E Q U E N C Y D I S T R I B U T I O N O F P R E D I C T E D V A L U E S I N C O M P L E T E M O D E L I N G D A T A S E T , W I T H P R O J E C T E D N O R M A L C U R V E ( N = 1 , 9 0 1 ) 7 1 F I G U R E 13 : A C O M P A R I S O N O F O B S E R V E D ( L E F T B A R ) V S . P R E D I C T E D V A L U E S ( R I G H T B A R ) F O R J O B S W I T H > 1 0 O B S E R V A T I O N S I N T H E M O D E L V A L I D A T I O N D A T A S E T 7 3 F I G U R E 14: Box P L O T S H O W I N G D I S T R I B U T I O N O F P R E D I C T E D E X P O S U R E L E V E L S F O R 3 , 8 0 9 E X P O S U R E K E Y S B Y M I L L 7 5 F I G U R E 1 5 : Box P L O T S S H O W I N G D I S T R I B U T I O N O F P R E D I C T E D E X P O S U R E L E V E L S F O R 3 , 8 0 9 E X P O S U R E K E Y S B Y P R O C E S S G R O U P 7 6 F I G U R E 1 6 : E X A M P L E S O F P R E D I C T E D E X P O S U R E L E V E L S F O R 5 S A M P L E J O B S A T O N E M I L L , E S T I M A T E D F O R T H E Y E A R S 1 9 5 0 , 1 9 6 0 , 1 9 7 0 , 1 9 8 0 A N D 1 9 9 0 7 7 F I G U R E 17: E X A M P L E S O F P R E D I C T E D E X P O S U R E L E V E L S F O R 5 S A M P L E J O B S A T O N E M I L L , E S T I M A T E D F O R T H E Y E A R S 1 9 5 0 , 1 9 6 0 , 1 9 7 0 , 1 9 8 0 A N D 1 9 9 0 , F O L L O W I N G A D J U S T M E N T F O R H E A R I N G P R O T E C T O R U S E . 7 7 F I G U R E 18: F R E Q U E N C Y D I S T R I B U T I O N O F I N D I V I D U A L C U M U L A T I V E E X P O S U R E S F O R M A L E C O H O R T S U B J E C T S W I T H C U M U L A T I V E E X P O S U R E > 8 5 D B A * Y R ( N = 2 7 , 2 4 7 ) 7 8 F I G U R E 1 9 : R E L A T I V E R I S K O F A C U T E M Y O C A R D I A L I N F A R C T I O N I N P R E - 1 9 7 0 S U B - C O H O R T ( N = 8 , 7 0 0 ) , B Y N O I S E T H R E S H O L D A N D D U R A T I O N O F E X P O S U R E 1 1 7 x Acute myocardial infarction British Columbia Blood pressure Beats per minute Confidence interval Cardiovascular disease 2 Decibel. Units of sound pressure level. Equal to 101og10 —where Pref Prms is root-mean-square sound pressure and P r e f = 20 uPa. Unit of cumulative noise exposure. Equal to 101og10 K 10 1 0 V where k is equal to the number of jobs held by and individual, 7} is time spent in that job (years) and LeqJ is TWA noise exposure for that job. Diastolic blood pressure Generalized estimating equation High density lipoprotein Hypothalamopituitary-adrenal system Hearing protective device Heart rate Hypertension, or hypertensive Hertz. Unit of sound frequency International Classifications of Diseases Ischemic Heart Disease k H z Kilohertz. Unit of sound frequency L e q A measured equivalent sound pressure over a period of time (usually an 8 or 12-hour shift in this study) L P Sound pressure level; in this study an instantaneous reading or averaged over very short period. M I Myocardial infarction N I H L Noise induced hearing loss ns Not significant O R Odds ratio P O R Prevalence odds ratio R R Relative risk S A M Sympathoadrenomedullary system S B P Systolic blood pressure SD Standard deviation S M R Standardized mortality ratio U B C University of British Columbia W C B Workers' Compensation Board of British Columbia xi i Acknowledgments First and foremost, I would like to acknowledge the support and help o f the workers o f the participating lumber mills, the supervisors and managers of the participating mills, and the I W A , P P W C and C E P labour unions, without whom this study would not have been possible. At U B C , my sincere thanks go to my supervisory committee: Susan Kennedy, Murray Hodgson, Kay Teschke, and in particular my research supervisor, Paul Demers. They provided expert guidance, generous support and patient encouragement. Moreover, they provided a wonderfully collegial and inclusive academic environment that made my PhD student experience especially rich. This study was built upon the prior work of many others, those instrumental in the original development of the B C lumber mil l cohort. M y thanks go to Clyde Hertzman, Aleck Ostry, Shona Kel ly , and Ruth Hershler. Also thanks to Bob Hirtle, Martine Dennekamp and Todd Y i p , for their assistance in collecting field data. Thanks too, to the all of the staff, faculty and students of the U B C School o f Occupational and Environmental Hygiene for making this a great place to work, and to learn. Finally, I would like to acknowledge the funding support for this work provided by the B C Medical Services Foundation, and the Canadian Institutes for Health Research. Tor a(C those who have taught me, and for the person 1 have (earned most from, Lit \ xiv Chapter 1. General Introduction and Scope of Dissertation Noise is a perhaps the most ubiquitous of occupational hazards. Its association with sensorineural hearing loss has been well understood for some time (Atherly and Noble, 1985). More recently, however, links between noise and so-called "non-auditory" effects (those affecting physiological systems other than hearing) have been hypothesized and investigated. Effects on the reproductive, cardiovascular and immunological systems have been posited, as well as performance and psychiatric effects (Passchier-Vermeer and Passchier, 2000). Cardiovascular health outcomes have received the bulk of the research attention. This research has demonstrated high blood pressure (e.g. Zhao et al, 1991), increased heart rate (e.g. Green et al, 1991), elevated levels of catecholamines and Cortisol - the so called "stress hormones" (e.g. Cavatorta et al, 1987, Sudo et al, 1996) - and increased risk for coronary heart disease (e.g. Ising et al, 1997). However, overall results have been contradictory, and this inconsistency among studies has been attributed in large part to methodological problems in study design and execution. Reviewers of the noise-cardiovascular disease (CVD) literature have called for specific methodological improvements including use of large follow-up studies of analytical (cohort and case-control) design, more studies of chronic disease mortality, and improved control of confounding (Babisch, 1998; Lercher et al, 1998; van Dijk, 1990; Thompson, 1993). In addition they noted the need for improved exposure assessment, including use of personal dosimetry, analyses of exposure parameters other than only intensity (such as noise frequency or "pitch") and accounting for the use of hearing protectors when assessing more recent exposures. Researchers in the Department of Health Care and Epidemiology, and the School of Occupational and Environmental Hygiene at UBC had previously enumerated a cohort of more than 27,000 lumber mill workers in British Columbia (BC), Canada, for studies of occupational exposures to anti-fungal chemicals (Hertzman et al, 1997), job strain (Ostry, 1998), and wood dust (Friesen et al, 2002). This occupational cohort was known to be highly exposed to noise, and Ostry's work had suggested an association between subjectively-estimated noise exposure, and both self-reported cardiovascular health and coronary heart disease mortality (Ostry, 1998; Ostry etal, 2001). It was decided that this cohort would be an excellent candidate to examine a link between noise and heart disease, while addressing several of the previously mentioned methodological concerns. First, it was an analytical (vs. descriptive) study design of incident, rather than prevalent disease. Its large size gave excellent analytical power, with an over 90 percent probability of detecting an increased risk of ischemic heart disease of 1.2 (a = 0.05). Good case ascertainment was achieved through probability linkage to the Canadian national vital statistics registry. Detailed work histories had been abstracted for all subjects, and earlier studies of the cohort had provided an excellent understanding of industry 1 processes, jobs and tasks. In addition, access was available to all participating sites. Occupational noise exposure at the mills was very high on average, so misclassification of exposure due to non-occupational exposures would be minimized. Quantitative noise dosimetry data for the participating mills (including noise frequency spectra) could be obtained from noise surveys of the mills, and additional personal dosimetry data covering a 20 year period was obtained from the Workers' Compensation Board of British Columbia (WCB), and from the management of some participating mills. Furthermore, detailed information on hearing protector use was available from the WCB Hearing Conservation Branch, and from questionnaire data from cohort mills. Access to mills and their staff provided an opportunity to obtain historical data regarding determinants of exposure through site-visits, walkthroughs and interviews with knowledgeable employees. The motivation to determine an association between noise and CVD, if one exists, is powerful. Although relative risks for an association are likely quite small, it would have considerable public health significance because of the severity of the disease, the ubiquity of the exposure, and its amenability for control through noise-control measures. For my PhD dissertation, I undertook an investigation of the cardiovascular effects of noise exposure in the lumber mill cohort using a retrospective cohort study design. The primary hypothesis was that exposure to high levels of noise in the workplace was associated with increased risk of cardiovascular disease. This dissertation is presented in 7 chapters, as follows. Following this introductory chapter, Chapter 2 presents a biological model that describes the hypothesized pathophysiology linking noise and cardiovascular disease through a generalized stress reaction. Chapter 3 is a review of the epidemiologic literature with respect to noise and cardiovascular disease. Chapter 4 describes the work done to characterize current noise exposure in lumber mills, and its determinants. Data on the use of hearing protectors in the participating mills is also given. Chapter 5 describes the retrospective exposure assessment phase of the study. Chapter 6 describes the epidemiologic analyses and health-outcome results. Finally, Chapter 7 gives concluding remarks and recommendations for future research. I will briefly add here an introduction to the noise measurement nomenclature that is used widely throughout this text. Simple sound pressure level measurements, usually obtained with a sound level meter as a "grab sample", and that is averaged over a very short period is referred to by the standard abbreviation "L p ". Time-weighted average exposures, or "dosimetry", are usually obtained over the period of a full shift, or a representative portion thereof, and are given the standard abbreviation "Leq". 2 Chapter 2. Background: Noise, Stress and Disease Introduction Noise is commonly defined as sound that is unwanted by the listener because it is unpleasant, annoying, interferes with tasks and communication, or is believed to be harmful (Cohen and Weinstein, 1986). Noise is pervasive in work environments throughout the industrialized world, and though data is sparse, studies have shown that very large numbers of workers are likely exposed. It was estimated that thirty million workers were exposed to hazardous levels of noise in the US in 1992 (Franks, 2000). The US Occupational Safety and Health Administration estimates that one-quarter of workers in construction, mining, manufacturing and utilities, transportation and the military are exposed in the 90-100 dBA range (OSHA, 1981). The US National Institute for Occupational Safety and Health reported that one-quarter of workers in textile mills, petroleum and coal production, lumber and wood production, and food production were exposed to greater than 90 dBA (NIOSH, 2000). The problem of occupational noise does not appear to be declining. There have been no major efforts to reduce ambient noise levels as the primary noise-related disease, noise-induced hearing loss, is considered controlled by the use of hearing protectors (Franks, 1998). As well, as richer nations move to a post-industrial society, the burden of this exposure moves to the developing world (Alberti, 1998). Complaints about noise, and the fear that it is detrimental to health and well being, appear to be constant concerns in industrialized countries (Schwarze and Thompson, 1993). And, while the effects of noise on hearing are well established, suspected associations with disease end-points other than those of the auditory system - so-called "non-auditory" effects - have proven much harder to elucidate. It is believed that there may be a variety of such effects. A recent Dutch review of the literature concluded that the strongest evidence exists for associations of noise with hypertension, ischemic heart disease, annoyance, and performance and sleep disturbances. Weaker evidence was cited for hormonal and metabolic perturbations, immune system effects, low birth-weights, and psychiatric disorders (Passchier-Vermeer and Passchier, 2000). Of these outcomes, cardiovascular disease has received the bulk of the attention with respect to noise-related health research. This is presumably because of its high prevalence, and high associated mortality. Heart disease is also the leading cause of death in most developed countries (Marmot and Mustard, 1994) and accounted for 36 percent of all deaths in Canada in 1997 (and a similar proportion in the United States), and it is a growing epidemic in industrializing countries (Wielgosz and Nolan, 2000). Heart disease is an important contributor to mortality in working populations; it was the cause of 50 percent of deaths among 35-45 year old males in a recent Pennsylvania study (Traven et al, 1995). 3 Chronic hypertension is a risk factor for ischemic heart disease, but as well for congestive heart failure, stroke, and renal failure. Approximately 20 percent of the North American population has high blood pressure (HSF, 1999; AHA, 2001), and it is considered "one of the most important underlying risk factors for morbidity and mortality in the world today" (Elliot, 2000). It is hypothesized that the cardiovascular disease consequences of exposure to noise are primarily mediated through a stress response of the body to noise stimuli that either overload sensory capacity, or are perceived in some way as threatening, annoying or challenging1. Because the stress response is identical in virtually all species (Folkow, 1989), it is assumed that it has not evolved in Homo sapiens much beyond what was needed for our existence as hunter-gatherers, tens of thousands of years ago. However, stressors that humans face today are different to those more primitive ones. Modern stressors are more probably psychological than physical; they are subtler, more complex, and perhaps more chronically felt. Some stressors, like noise, are a relatively recent advent, being with us on a wide scale for fewer than 10 generations. It is thought that stress-response patterns that once conferred a survival advantage to humans faced with real life-and-death stressors may today be physiologically detrimental. The responses prepare the body for physical demands, either for immediate action, i.e. "fight or flight", or for the sustained demands of survival in extreme conditions. However, these responses may be inappropriate for psychological stressors, as they do not (or cannot) physiologically resolve themselves as intended, or as required. Consequently, biochemical and physiological adaptations may now harm the organism they evolved to protect (Folkow, 1989). Stress and Cardiovascular Disease The central tenets of the early hypotheses of the role of stress in disease (Cannon, 1935, and Selye, 1979) have been borne out by decades of research. Hinkle, reviewing the field in 1987 said: "[...] the course and manifestation of any disease can be influenced to some extent by the nervous and endocrine systems that are initiated by the actions of the central nervous system in response to information from the environment". The early idea of a single form of stress response has, however, been superseded by a more refined model. It is now believed that different stressors evoke different stress responses. However, the number of these that are "pre-programmed" is still limited, and as the potency of a stressor increases, the specificity of the response decreases (Hinkle, 1987; Folkow, 1989; Chrousos, 1998; Chrousos and Gold, 1992). A generalized model is adequate then to describe the putative psychophysiological effects of noise exposure. 'Stress tenninology: Chrousos (1998) defines the key terms thus: "Life exists by maintaining a complex dynamic equilibrium, or homeostasis, that is being constantly challenged by intrinsic or extrinsic adverse forces or stressors. Stress is thus defined as a physiological and behavioural adaptive response of the organism" 4 A General Model of Stress and Disease The "stress syndrome", or "general adaption syndrome", is understood to be an attempt by the body to protect homeostasis. Behaviourally, it includes increased arousal, alertness and cognition. Concomitant physical adaptations redirect energy, increasing cardiovascular tone (heart rate, blood pressure), respiration, and intermediate metabolism (gluconeogenesis, lipolysis). Digestive function, reproduction and immune system functioning are all inhibited (Chrousos, 1998). Neurologically, the "stress system" processes and distributes signals though a wide variety of both neural (higher cortical, limbic, visual, auditory, olfactory, gustatory, somatosensory, nociceptive, visceral) and humoral (blood composition, hormones, cytokines) signals. The mechanism by which disease results from the stress response is not fully understood. It is presumed to involve a dysregulation of the sympathetic nervous system and neuroendocrine systems resulting from prolonged, or excessive, activation o f the stress response. Henry proposed a model with two major stress reaction patterns that he called the "defense" reaction, an active response dealing with challenge to the organism and equivalent to the "fight or flight" state; and the "defeat" reaction, which is passive, and akin to Selye's chronic "distressed" state (lower part of Figure 1; Henry, 1993). He suggested that both reactions may have pathological consequences i f dysregulated. The "defense" reaction, Henry proposed, is elicited when an organism's control is threatened, and is characterized by activation of the sympathoadrenomedullary ( S A M ) system. The sympathetic nervous system, through innervations of smooth and cardiac muscles, increases sympathetic activity o f the heart, venous vessels, kidney, gut and skin. Vagal activity is suppressed and cardiac output increases, causing a significant increase in arterial blood pressure. Sympathetic activation of the adrenal medulla causes the release o f epinephrine and norepinephrine that bind adrenoreceptors in target tissues further strengthening the sympathetic response. Norepinephrine released by the locus ceruleus to various brain loci results in a heightened state o f attention and vigilance. Salt-water excretion is suppressed, while salt-water absorption in the kidney is enhanced. Sympathetic activity of the kidney may result in activation o f the renin-angiotensin-aldesterone system, providing further positive feedback on the sympathetic nervous system (Folkow, 1989). The defense reaction is typically a transient response, but i f repeatedly activated is believed to lead to increased blood pressure, caused by structural changes to the heart and blood vessels and resetting o f baroreceptors. Structural autoregulation, as this is called, is thought to largely dominate the hypertension process, to the extent that the original underlying genetic or environmental factors "may be difficult to trace" (Folkow, 1989). Henry (1993) proposed that the 'defeat' reaction becomes dominant when the organism feels a loss of control. The reaction is characterized by passivity and depression. It is dominated by the hypothalamopituitary-adrenal neuroendocrine system (HPA) , and by an increased release o f 5 Noise Perception (O D 5 S Sound Parameters Intensity Frequency Dynamics Duration 4-1 0) SI 11 Sound Perception > Noise Perception Situational parameters (external) Individual Parameters (internal) Examples - communications - concentration - recreation Examples -coping potential - vegetative lability - noise sensibility Stress Reactions Interactions and Feedback o CO CU 1— o (A c cu «*-• cu Q Increased: norepinephrine epinephrine testosterone oxytocin free fatty acids renin Increased: ACTH Cortisol Decreased: gonadotropins Peripheral gondal hormones V O CO cu co O Increased Blood Pressure Increased Heart Rate Insulin resistance Hyperinsulinemia Visceral fat accum. Health Outcomes Diseases: Hypertension, Atherosclerosis Stroke, Myocardial Infarction, Cardiac arrhythmias, Diabetes Figure 1: Noise and Cardiovascular Disease Risk Model (after Babisch, 1998) 6 adrenocorticotropin hormone, and elevated levels of Cortisol. The pathogenetic consequences of HP A chronic activation have been much less studied than those of the S A M system, but the central abnormality in HPA is an elevation in Cortisol levels. This in turn is thought to be the most likely cause of visceral fat adiposity, a powerful risk factor for disease (Bjorntorp, 1997). Experiments have shown that increased Cortisol levels are also associated with diabetes, stroke, dyslipidemia and with insulin resistance, itself considered a major risk factor for cardiovascular disease. In addition, the HPA system has been linked directly to increased blood pressure via interactions with sympathetic and renal mechanisms (al'Absi and Arnett, 2000). Noise as a Stressor It has been known for many decades that loud noise (an "audiogenic" stressor) consistently evokes a typical stress reaction similar to that seen when an organism is challenged by a physical threat (e.g. Medoff, 1945). The biological models of stress pathophysiology (Bjorntorp, 1997; Chrousos, 1998; Henry, 1993) have been extended to incorporate noise exposure (Westman and Walters, 1981; Maschke et al, 2000). These models provide numerous testable hypotheses, but are complex, and the actual pathogenic mechanisms have still to be elucidated. It has been proposed that noise elicits a stress response by way of two separate pathways (top part of Figure 1), each distinguished by the nerve paths and neural centers involved (Westman and Walters, 1981). In the 'direct' pathway, noise signals branch off at synaptic junctions along the auditory nerve (the 8 t h cranial) to motor cell nuclei sub-serving reflexes within the brainstem; these are responsible for the startle response (Koch, 1999). Additionally, signals travel along branch pathways to the reticular activating system and connect to the limbic system, and the autonomic nervous system. Thereby the signals integrate into the neuroendocrine system - and hence directly to the parts of the brain controlling the vegetative, autonomic and endocrine functions. This "early-warning" or "alarm" function is the original purpose of the auditory system, and was the earliest aspect to evolve; it is therefore deeply "hardwired", and cannot be consciously extinguished. The second, "indirect", pathway is engaged i f stress results from cognitive processing of noise by the auditory cortex. In this pathway, the level of stress attributed to an incoming noise signal can be influenced by both external, situational parameters and internal, endogenous characteristics of the individual (Ising, 1997; Westman and Walters, 1981). Situational parameters might include communication tasks, or the need to concentrate. Internal characteristics include perceived control over 7 the noise source, attitude toward the noise source, the lability of individuals' autonomic nervous system and their sensitivity to noise2. The relative degree to which each of the two pathways - direct and indirect - feature in the auditory-stress reaction is not understood. It has been proposed that the direct pathway (that bypassing cortical processing) is probably selected more frequently when the noise signal is of high intensity, is novel, or when the organism is not conscious. In other situations, such as when noise intensity is low but has a sufficient source, content, or other parameters, the indirect pathway may predominate. Although the direct auditory pathway presumably evokes the defense stress reaction more frequently than the defeat reaction, either of the pathways might be associated with either of the two stress reactions. Both systems also have feedback components, and they also probably interact in complex ways. Experimental data in humans support the proposed models. For example, Andren et al. (1983) showed elevations of both systolic and diastolic blood pressure in response to noise. It has been shown that this is likely due to vasoconstriction in subjects exposed to noise (Svensson et al, 1987; Singh et al, 1982). Experiments have shown vasoconstriction even in the presence of a - and a-/(3-adrenoreceptor blockades. This suggests that blood pressure increases are critical in the stress response (Eggertson et al, 1987). Subjects exposed to noise have shown increased levels of epinephrine, norepinephrine, dopamine and Cortisol. The levels subsequently dropped when the same subjects reduced their noise exposure by donning hearing protectors (Sudo et al, 1996; Melamed and Bruhis, 1996). In animals, noise-exposure nearly always yields the hypothesized results - elevated blood pressure, hypertension and increased levels of stress hormones (Schwarze and Thompson, 1993); it is so predictable that noise is considered a prototypical stressor in experiments. Rats exposed to a noise representing occupational exposure for 10 weeks showed increased systolic and diastolic blood pressure after one week that remained elevated until the termination of the experiment (Fisher and Tucker, 1991). Rhesus monkeys exposed over a 9-month period to a pattern of noise typical of an occupational environment had increased blood pressure that then remained elevated for at least one month after the cessation of exposure (Peterson et al, 1981). Similar results have been found in dogs (Engeland et al, 1990). In a study of chronic disease outcomes, increased atherosclerosis following exposure to noise was demonstrated in rabbits fed cholesterol-enriched diets (Freidman etal, 1967). More-generally, Cynomolgous monkeys fed atherosclerotic diets were also more likely to develop atherosclerosis if they had been demonstrated to show high cardiac reactivity to an external stressor (in this case, threat of capture, Krantz and Raisen, 1988). 2 Noise sensitivity measures an attitude to noise in general and so is different from annoyance (Stansfeld, 1992). 8 Attributes of the Noise Signal Influencing the Stress Response The magnitude of a noise signal required to elicit a stress response varies greatly. At lower noise levels, a response may depend more on environmental and individual factors. For example, reaction may be limited for workers exposed to continuous 85 dBA noise from a machine that they are operating, while a response may be elicited at levels lower than 60 dBA if the individual is sleeping, or doing tasks requiring mental effort or concentration. The likelihood of a stress response increases with the noise level, however, and there is likely a level at which a response is certain. Experienced artillery gunners, for example, still react (i.e. with a blink reflex) upon the firing of a weapon (Carter, 1988). In the laboratory setting, stress response to noise often appears transient and quickly habituating (i.e. the response decreases with time, eventually returning to baseline). Kryter (1985) concluded that any simple repeated or familiar signal habituates quickly and that therefore only the indirect pathway operates; that is, noise cannot directly cause stress, only the annoyance caused by noise is stressful. However, there are several lines of evidence suggesting habituation is not complete (van Dijk, 1987b; Westman and Waters, 1981; McLean, in Westman and Waters, 1981). Rossi (1959) showed that after habituation occurred for a background noise, a new tone superimposed over top elicits a new stress response. In community studies, it has been shown that long-term neighbourhood residents react to noise at least as strongly as new residents (Cohen and Weinstein, 1981). The probability of provoking a strong response that does not habituate increases when noise is of a high or changing level, or is unpredictable (Carter and Beh, 1989; Carter, 1988; Germano et al., 1988). Low frequency noise is thought to have effect characteristics different to other noises of a similar level. In studies of noise and annoyance, a lower frequency signal elicits stronger reaction, but this may be due to superior propagation of low frequencies, or rattling of objects (Job 1996). Exposure to very low frequencies (10 -20 Hz) has been shown to cause "unusual feelings" (Kryter, 1985). Salivary Cortisol has been shown to be elevated in subjects exposed to low-frequency noise compared to other noise, when both are presented at 40 dBA (Persson-Waye, 2002). The commonly used A-weighting system severely underestimates the perception of noise that contains strong low-frequency components (Kjellberg, 1990). Other parameters of noise appear to be important in determining the scale of the response. Unpredictable noises are more likely to cause response than predictable noise (Glass and Singer, 1972; Schwarze and Thompson, 1993), and sudden, impulsive noises create substantially more reaction (specifically annoyance) than do non-impulsive noises (Job, 1996). An individual's control over the noise source can modify a response; for example the observed annoyance reaction associated with traffic noise is reduced when subjects are told they may close their windows, even if they do not (Job, 1996). One would predict that a worker would react less to noise that they generate themselves, compared to that generated by others. Attitude toward a source can also modify the response. For example, aircraft-noise 9 annoyance levels decreased at an air force base after the local residents were presented with pro-air force 'propaganda' (Westman and Walters, 1981). Summary Henry's model thus suggests a number of mechanisms and a basis by which stress causes disease through perturbations of the neuroendocrine system and resulting physiological changes, such as elevated blood pressure and metabolic dysregulation. Additional mechanisms might exist. Stress might, for example, precipitate acute events in those with pre-existing disease or who have a predisposition to disease. Such a mechanism has been proposed for both arrhythmic disease of the heart and acute myocardial infarction, specifically related to anger (Meerson, 1994; Mittleman et al, 1995) and mental stress (Muller, 1999). It is also possible that noise may increase levels of known CVD risk factors, by causing modifications in behaviour, such as changes in dietary, exercise or smoking habits (Maschke et al, 2000). Finally, of course, noise may elevate as yet unknown risk factors for ischemic heart disease. Analogies to a noise-CVD link can be drawn from other studies of stress-related CVD. Research has shown links between psychological, psychosocial stressors, such as job strain and life-change events, and CVD (Bosma et al, 1997; Kristensen, 1995; Rozanski et al, 1999). Yet despite a large number of observational studies of noise and cardiovascular effects, the evidence for a causal association between noise and heart disease has been judged to be equivocal (Babisch, 1998). 10 Chapter 3. Noise and Cardiovascular Disease - A Review of the Literature Methods A literature search was conducted for the years 1966-2000 using Medline, which abstracts most of the world's biomedical journals. The search used keywords associated with occupational and environmental exposure to noise (i.e. noise, sound, audiogenic, acoustics, acoustic stimulation, auditory, auditory stimulation), and logically combined these with results of cardiovascular disease and related keyword searches (i.e. blood pressure, hypertension, coronary disease, sudden death, cardiovascular disease, cardiovascular system, myocardial infarction, arrhythmia, heart arrest, heart disease, cerebrovascular disease, heart rate, pulse, hemodynamic, epinephrine, norepinephrine, Cortisol, catecholamines). The literature gathered was then searched for references not found by the Medline search. Only English-language, occupational and community-based observational studies of human adults were reviewed. All studies relating to chronic heart disease were reviewed; however because of the very large volume of studies, review of those regarding noise and hemodynamic outcomes such as hypertension were limited to the period 1980 to 2000. This was in part justified by the availability of several reviews of earlier studies and their consensus that the early research in this field was of low methodological quality. Some of the problems of the early work are described in the next section. Further, studies using noise-induced hearing loss as a surrogate for noise exposure were excluded. For ease of comparison, prevalence odds ratios (POR) were calculated if not given by the author (Checkoway etal, 1989). Early Epidemiology Studies of Noise and CVD There have been well over one hundred published epidemiological studies investigating the relationship between noise and cardiovascular disease. The majority of this research has been directed toward change in the level of arterial blood pressure (BP), and hypertension (i.e. BP greater than a clinically defined level, such as those recommended by WHO, 1978). A much smaller fraction has included analyses of other hemodynamic and biochemical effects, and studies investigating heart disease outcomes have, so far, been rare (Babisch, 1998). The consensus among reviewers of early studies in this field was that the methodological quality of work was of low quality, contradictory, and showed slow evolution in sophistication, or in "progress of enquiry" (DeJoy, 1984). Many of these studies were poorly communicated, failed to include detailed 11 methodological descriptions, or contained no quantitative data (Thompson, 1981). Noise exposure levels were poorly specified in many instances, and there was inadequate control for confounding factors. The majority of studies were cross-sectional, limiting their ability to determine causality. Thompson nevertheless concluded that there was sufficient evidence to support further research, and that the strongest evidence was between high noise levels and elevated blood pressure. DeJoy (1984), noted a preponderance o f "exploratory" studies, and concluded that despite numerous positive findings, the existing literature "did not permit inferences of causality, or derivations of dose-response relationships". Three years later, Kristensen concluded that the "epidemiology in this area [was] still o f low quality", and commented on a lack o f longitudinal studies, and limited investigation of outcomes other than blood pressure and hypertension (Kristensen, 1989). However he concluded that the research yielded "reasonable support" for a link between noise and cardiovascular disease. A common problem among early studies was the use of noise-induced hearing loss as a surrogate indicator for exposure to noise (Jonsson and Hansson, 1977; Lees et al, 1979; Hedstrand et al, 1977; Delin, 1984; Kent et al, 1986). This has been widely criticized because o f its low sensitivity, the difficulty in differentiating noise-induced hearing loss from other forms of hearing loss, and because atherosclerosis may be a risk factor for hearing loss (Wu et al. 1987; Kristensen, 1989; van Dijk, 1990; Thompson, 1993). Epidemiological Studies of the Hemodynamic Effects of Exposure to Noise I reviewed thirty-one post-1980 studies that examined arterial blood pressure levels, heart rate, heart rhythm, and hypertension prevalence. Twenty-six of the studies were occupational and five community-based studies. They are described in detail in Tables 1 and 2, respectively. The range o f noise exposure levels studied varied from 66 d B A in community studies, to levels well in excess of 100 d B A in occupational studies. A noise level greater than 80-85 d B A was used in most cases by the authors as a criterion for "exposed" status. Study design was mostly cross-sectional, the exceptions being a case-control study by W u et al. (1987) and the repeated-survey studies o f Hessel and Sluis-Cremer (1994), and Neus et al, 1983. Some studies, though not al l , considered duration of exposure. Some studies looked only at acute effects, such as change in the outcome parameter over a shift, or ambulatory blood pressure changes and concurrent noise exposure. Because many of these studies shared similar methodology and had similar limitations, issues of study design, bias and confounding are discussed at the end of the section. Noise and Blood-Pressure Change Twenty-four studies of blood pressure differences were reviewed (studies o f clinically defined hypertension are described in the following section). Noise exposure levels in community studies were generally lower than in occupational studies, with subjects considered exposed at levels as low as 66 12 JS_BJ O S u a a K £ ffi o ^ i l l L f i l > N g g 8; 2 1 1 * • £ « 8 » I T . 1 I I s s £ c = >• © SM — U 3 jj § § J=S..S_«. E a . to O CB eg tt-§ 3 ° 1 s 2 | b t " s L a. CQ OT ^ i l l f i i E E " E 5! § 0. « & CO OT M Q S I L | 1 J o E i ^ $ J.I.I <N "3 "3 "1 S I i ' O — "1 • • ~ O — o 8 - . 3 a w o\ M "3 N _^  rn "Si a ^ •• ** I I l o £ 53 M O • r fi ' C -E E S 2 E E 111 I o b i y A E s es E S I J n -s s I?.? I i 00 ,r gf E DG o i s B » S ° 1 ° £ -H 2 2 t a 2 « I a 2- 5/3 ™ . . ^ J ^ . . . Q f , , i £ L . •8 •s c u B S a *T3 O 1/1 26 s 5 a a •§ I 01s c i l l •a >, E l i s •a .0. e s. B 73 § B § >, 53 a 8 , 1 1 S S * S 8 -s -a «• B S o S 111-1! UJ u fl ° i I11 .I a !* g 0.8 *- o "8 5 •x S a » 8 •« I l l - . S Q . . . S 2?.. c oo a a J 1 rf « u = g E § I -3 = 0 .2 JS o ao 2 3 i f a 3 1 < 2 x 8 1,9 111 Al g 5 s 1 ! •o a fc ft« 1 2 I t § 1 p -. E 0 -g i l l •a S 'J3 .a....e_» k s a B H 3 a 1 HI I - 5 2 . S I r. 2 H I § | 5 s « II"* * 1.13.1. 73 o • S «> TI » a ij M u 2 s A1 S I a - i l l H E 5 c © U U U •3 B t o g s i 81 S 0. ca M E 3 < CQ » 3 * « — a 5 < > < co 'Eb _ o I ' l l o w 0 o CM £ ;S v i 1 1 a E — 2 <C &n CQ CQ . i l l § 2 V h 1 1 -s "2 P <ao<iSSo § a 1 1 g 8 1 a 1 1 5 1 la 1 H I 19 Is 1 v 13 CQ "3 O «i tn <d 8,0 S. CQ x op x -0 w A- w 0 8 K 0 A •fi 5 X u O C O * ? is '0 I I I 00 a _ o I I 8 «c P E° "T C t3 — °C 4> TJ ^3 00 s i l i i-3 1 ^ 5 E g 5 ° 8 1 * fe e B g •2 -a § 2 1 i s 8 1 1 E -5 w s a » T J a » t t w .B g. a a f T3 1 o i f . i 111: 1 1 1 ! E » 1 -r * ft a o — s i s a s S "9 00.a « E „• 6 • & : 8 ^ i l l l 1 1 1 1 1 3LJB, - 1 SLSL. I 1 I [Jill "a. "0. a, 111 I S 2 8 fN « .2 0000 52 £ E 2 £ 8 OT w Q OT — 00 CQ C J 13 i l 00 f l Jl 3 " — r-i 60 OO > (fl a a E s •o . . • • § & & ° a D 6 0 CB O I | 8 S-S-O T3 S i l l £.1 1 S i l J^ LEL • « -3 J5 S _2 O p i £ | 2 > JS g 3 s ? * 1 i | 1 § | | 8..|.| 5 111 s e o g ~. 3 > ,u£ 53 S < — x a — r- < s ^ 3 - V •S .1 _ s 8 D..S-00 8 8 » a 1 1 •a i § B I J o © g o o "3 2 E O H 03 kg SI 9 £ 1! •o o 5 8 c & •5 ~ a o i l .IS i; S l l w 2 • S E E I I I E if § 1 s 1: s | S3 o. .9 M s 1 1 SC-o . 3 ! 8 § 1 1 1 1 c o IS 1 1 ' I 5 -B 8 2 e 0 0 -2 r-~ a-S 5 6 8 | ; & u -8 £ '5 — 3J3 g M .5 Z 91 wT a .1 < 9 ^ 1 1 15 2 I » •2 SP„ J S 2 1 1 J-ffi.JL o J X 1 1 2 $ •a c o S H 0 ~S U S « I o< .9 - 8 •—- efl It 1 u r . l w S V C . 00 g .2 il Vi s t 5 p i f •c , ,—. ^ . . Oil [U UJ c un <S> £ 1 rN — _ 1 > o ItQ CO . • « l i e O C X S f i l l a .2 o S3 1 S | g e « » 1 S £, .2 £ S BS H ig 1 3 gi 2 « • I 5^ a 8. & 2 < v> _ '3 ca *. ja u j Q , , ' v i W U < £ g IS, U i i ? I i i 831 "111 JJJ_ IS • a g | t i l •8 « M eg .» .S 3 Jg 8 if IS 1 1 1 1 1 1 . s s g i. • 111 • i l l S ao -E « ii Si as ?«< 1 1 s^S JS s ^ ^ .9^ .....^ ?„..!^ ...?°....9!?.„ S2 ^ § a> — « c •M 3^ « c 1 SS-8 a oN 6o 5 so w, w § ° 1 l l « g i § 2 « ON 2 2 g ~ i . a g 1 r j a z VJ "3 i f | 1 s | I I -i I • R "3 5 3 (M I -a 5 ° o a" ffi ja •5 ° J E J a f . • I l i a s S3 § E 8 ¥ 1 12 E 2 3 « E S O L , 3 3 S s S3 1> 6fl S a A I g 5 9 on a ea ° i-a (N -3 ev ill n fl M 2 5 I" 11 §9 -3 ? o *„ 3 u as 0 3.^_._Qj-_.^__9-.._ 60 • I t E „ S I vf2 1 1 1 in JS J S a . a i l ,ltO,, m„J.& I i i s | l i j & *3 -o s« » • 3 S3 oo E 3 a Jlf I 8 § a v .a a g 1 7 6fl — 1 s ux F :ar oo D 2 £>J3_ 2 S 6 i 14 3 ..X.. ..a pg' Of op oo 33 a a • • E S S S 6 E E Q "1 « r-? O O vo C: 9? oo £5 O OD ^ 2 C OJ E a I 3 ^ , -3 2 > 3 § 1 S o u o v .8 <N j - O ;ja 1 5 •B ? s "9 a i l l D . ra E x E B g (0 G 111 a i i 8 I asa •o s -S i 8 x. •a « JJ i l l a * 1 3 Sll * * I OJ v> o i l i l l S 9 IM o> « j = • O S S H a. o a S z i "a J 8 •u O 95 j = 11 ^ a ^ i l • s 2 -a oo | | | -a | H « o 5 -| * 1 .1 i l i i _v r-> 3 •a £ a r i 9 •i..ari.iiiffiiii t ^ Oi CN a j K S L i 'fib 3 I !3 *t9 * § a ni i i l l ! I § 2 -fl •3 3. J i l l " a x E E S oo — a <= II 2 1 a°P s s _ MS 00 aoc fro S8JSL. .^ S C2' = — 00 "0 aa ea ca II E £ i l . i t s 6 oo.u 3 3 * o ^> Al K l l P « <*-T3 M ° — tfl 11 i figs p o 11 a a 2 I 1 ! & U TI % II I I I I i OJ VJ E I a i H I »e 9 •§ O B IE 3 a* 1 0 oo£ a E S II * 2 ? T! 1 9 . i l l a a 3 E o a o ° •» o S 53 w ~ a JS 12 2 51 &. l i i i i a' g g m g B S f e , •a S S3 J2 •Sep • ° » 3 00 . a !Q § p ffl a Q £ X >. oo E v J= lag 3 o ft> < t 00 V S A l fe, J5 TS 00 3 w -TH - B S 8§ ? a H o i l i Z u 1 fi a t _ & J 2 _ If a s | < a > T3 fi selec level: dBA 1 selec level: dBA 1 1 al ne - int alyses l a . . J g % I ? o & III .§>! s vi "3 — 8 -5 fi I i a 00 V) v; aa< x x CQ 3 O 5 * S § 1: ! S : * ITS * &B a vi B i s B a CQ J a ) mi (SBP-rN (SBP-ly, after 15 f ft. CQ 3 nurse, once on rest II 1 CQ fl ^ 8 nurse, once on rest £ CQ < > O v i U -8.frI} 2 I i < g x> « « i «S g. cq E E 3 .2f u S =3 * 8 ox«« E fi vj O vi — e 3 « 00 IM a x ,s e a. AI S vf • 1 u 8 r 8 5 g | .&o a i l S 5 H R- a s i CQ v -•1 a s « s CN ^ UJ o g o l l 8 Ji~ al. vj O U 3 0H CQ i i B § f l E >> IS a S cu — oo 00 o S 2 S3 E ' I s S3 OJ « V) L . II § I a cd u o s 1 ft. S ,.1, § J O OJ " O <N jj. | > ^ oo *- "6 9 W .S 8 1 ?'fe oo « 3 S. S o o Ij OJ m s. I.I.J SA'o SI fr > ss •B * o CJ bo * ^ a t bo ' < s co t* a a; L L h i ? 2 5 o .s d .fi ffl CN "S «/ I CN M O X i O .™~si._S^ ...S3_rr;__ I Cfl o 15 •5 Si £ SS o <5 z -2 a £ a « a & 3 2 B l | c 8 1 .S o. i CQ | Oft. is S •a o. I 8 £ 6 I at, O o Bi t oi S O l l c a. 3" a © vo vo vo tr: w ~ cK g r- 3 litis s i gr.| 5 * 1 (3 (rt 3 O a .2 o i z 2 S , ° 1 ! I l l 'C J3 T3 O M « 311 Al a '5 I o £ <fc" A ,-, $ 1 8 2 . | ™ g § 4& g | l i l l CU 1 | M i l 8 oi 2 | § ST5 -<2 I H CL. -S u 8 E .a a § § § § CA bo 3 .9 S S a. & S % § * 3 ' I S L l f t ° | i ca <g £ u g Q 8 U < < cq 1=9 | s a § s s I -S T 3 tu ^ -<N > ° W O 00 • O 3 j i r g 'a 1 1 I! 8 * l a J t S S , Q . -7 1 1 1 a J S J J E i <2 vo a a, * 2 1 1 l?f 3 >- 2 I 'a S .= ON S a § § j« o u 2 3 still « 5 S S g> .9°....J1„'< _«....a_&_. ia a ov S 5 j_s s_e_j ^ (N O M 1^  OO 0\ A -s 1 CA CA _ § 8 '& a * *• § S O N «t _ ON , S S3 o o « i aj_,.,.o — — rt f i a 2 g £ S K & p. g u ^ ^ t o « r-' ft, g a o\ r i >o CO *S C 9 ? 2 ? p ^ r- r- r- tN ? £! © © v-i o S ^  2 2 2 ^ c« — — — ; o u B >> >> § J ^ i o i uajSLrjjg q q X e e a. CO i (N 5 A * •s •5 8 • o m _ ? 00' 3 •—' » a. ° i _fa.fi. I 8 I 9 x 2 8 I a o -•1 J n . S K S . s a 3 2 a § & a 1 I I y CA "o 0 5ft a a 3 3_a_ 1! s l a s a s .tt CA cn i s a- 3*5 «-5 -a * CMJ §>=?! < £ j 2 > g « « 8 I 2 S c £_ 3 o rl O 00 Qj v 1 >• (4-1 Q . * O ^ ] crt O h, 8 B fc. a .s u C 00 8 ^ 8 n 0 u i i 5 . 3 Z ., X. r ^ CU 1—1 w U CA i js .3 <*i S £ 11 TS g i si a I u 2 ! » 8 s i ...11 = u 3 3> 5 6 t CO < -o C* IB - a 1 *c i 00 s « o • 9 M P , 6 1-Q ^ J a, OT i is p > 3 e IJ « 6 • 1 1 cfl > s J 1 ; 1 A l I 0 " I I U 8 m 5 1 1 1 3 X X to B 8 t.i... S 8 • CO 8 CA OT >, •3 BI. M I K d- c "9 P 8 I < U P i ' ob 4; < 3?.... f J . — O o u 00 § X -= 1 § CQ «" CO >^ ® 1 . £ 00 33 I E E e E -t t- s- Q 8 B J ca § I ! _ ! L i _ _ _ o « 1 ? 8T3 81 E 3 — o co .E IT-11 I = i 1 1 S E a. 0 , JSLS-1 § (A — ra D 1 § E » S 0 OT . _ Cb _ ^ •a 1 • a * a • s e a . I f i g | * s ....<s....&..fi,..-S..fi 17 S 9 •a t a! o iu ,—i = I 55 tu 11 l i I D d: is n | Q - 1 . . S 2 E S a B. g •a * o -2 S .o Cu o CD a e OH O o -a 1 1 m is <U cn «<=> 0-CQ x 2 o o (M i o oo S o •a « o a •§ OH u M 3 I J !• c u vf V 3 IB O nn a s 2 a •s o § a a a « ft •s s • I | a OH VI B -5 ca s-2 1 a. oj S -3 •a — s g a "O — 8 2 W (n O (" «« . i f 1 1 1 -85 J2 3 5j . V) -3 a, oo SI 53 A 3 3n*o A s 2 1 i l l i> o n * _ T3 ^3 = 8 S , S 5 1 1 1 1 1 1 3 OO A S . T 3 i t oj 3 Pg OO Si ji o s IS S g a l i i hit E ft. 1 J * • V I -£ 2 < * tS — oo* °»w I > £ E — < 6 0 U ft, 0 0 < S "1 (rt So" •a £ ca -* .2 S 8 Q J5 a ...o...«2..B....Q.1 ooS I i a i a *e l l OJ < x JJ x. g ft" "3 C O 1 3 w _ * i - a i S3 <? E l i 3 (rt Jj £ « * •8.3 » 2 • ! • SPO (rt opw 4 > £ 8 a IS (O > CO a a. •*• CO CN - CO oo a ft. 8 S3 o a « •a I * Si 3 o 2 ? oo i iz 2 § § s. a ma arini J elt no CL, o isure: sure i a a X X aj ^ a s •d us chotomo noise, oi lf-report< chotomo 'rt> >> H i to 5 g 8 111 3 I 1 - a | a <N a B a a S g -a = B CJ i/i c/3 a> 8 -a " S fell « a - I , J S M o <C SJ 3 1 1 « E '£ 5? < O CO GO CO ft. Si R g S T H 1 £2 Ills K i l l : •» S s 13 5 2 o <§ a 5 s + ft." CO a I x B E ft. CO 09 S 3 (rt 2- E o S C o . " ( » " t3 OJ -S a § a g s . X (g o 1 r2 x c « £» — 60 CO < i ft. C O O , 1 2 1 5 » = .E & § I n ! 2 Ml I i l l n R.s j? ° 2 JS a (N P. OJ (rt ® 81 VJ . o 3"S S 8 X C . U oo 19 dBA, while exposed levels in occupational studies ranged from 80 to greater than 90 dBA. Subjective measures of exposure were utilized by two community studies and one occupational study, and two studies measured exposure on a continuous scale (one of these cumulatively). Control groups in occupational studies were typically other occupational groups, simply reported as "not exposed to noise"; although in two studies "unexposed" groups were exposed to levels as high as 80 dBA (Lang et al, 1992; Fogari etal, 1994). Sixteen of the 24 studies found positive associations in at least one of the noise-exposed subgroups they examined. In studies examining simple differences between BP levels in exposed and un-exposed groups, differences in SBP ranged from 1 to 10 mmHg, and differences in DBP from 1 to 7 mmHg (Singh et al, 1982;, Fouriaud et al, 1984; Idzior-Walus, 1987; Wu et al, 1987; Babisch et al, 1988; Lercher et al, 1993; Fogari et al, 1994; Ledesert et al, 1994; and Xu et al, 1999). Exposure-response relationships were examined in several studies where duration of employment or length of residence (used as surrogates of duration of exposure) was available. In their community study, Neus et al (1983) found significant DBP increases of approximately 5.6 mmHg after 27 months of follow up, but the study cohort small and not fixed. Verbeek et al. (1987) demonstrated an exposure response in workers exposed to > 80 dBA, with an increase of 3.8 mmHg systolic and 1.6 mmHg diastolic in those exposed 10-19 years, and 16.5 mmHg systolic and 7.4 mmHg diastolic in those exposed over 20 years (both compared to those exposed 0-9 years, and adjusted for age). Lang et al. (1992) found little effect in noise-exposed workers with less than 20 - 25 years of employment, but an increase of 10.3 mmHg (SBP) and 8.3 mmHg (DBP) was apparent for those exposed > 25 years. Tarter and Robins (1990) found duration of exposure was a significant term in regression analysis of mean blood pressure (0.35 mmHg/year), but only among black workers. Aro (1984) and Talbott et al. (1999) demonstrated exposure-response relationships with increasing noise level and increasing cumulative noise exposure, respectively. Aro assigned area mean exposure levels to subjects; regression analysis gave coefficients of 0.32 - 0.43 mmHg systolic per dB increase, and coefficients of 0.073 - 0.176 mmHg diastolic per dB increase. Talbott found similar increases of 0.16 mmHg systolic per dBA, and 0.15 mmHg diastolic per dBA 3. Finally Green etal. (1991) showed a small acute response by monitoring ambulatory BP over a work-shift and the following evening period, and correlating periodic BP measurements to concurrent noise dosimetry data. dBA cumulative exposure — L e q + 16.61 (Logio (T/To)) dBA, where T=time employed, and To = 1 year 20 In an extension of their 1988 study, Babisch et al, (1990) found that BP increases in a sub-cohort of 255 males were more pronounced in those exposed to traffic noise > 60 dBA if they were also exposed at work to levels above 90 dBA. Eight BP studies reviewed did not find any association with exposure to noise (Talbott et al., 1985; Cavatorta etal, 1987; van Dijk etal, 1987a, 1987b; Hirai etal, 1991; Garcia and Garcia, 1992; Babisch et al, 1993b; Hessel and Sluis Cremer, 1994; and Kristal-Boneh et al, 1995). Noise and Hypertension Thirteen studies (all occupational) assessed hypertension (HT) as a health outcome. Most used the World Health Organization definition for HT of SBP/DBP > 160 mmHg/> 95 mmHg (WHO, 1978). Exceptions to this were Belli et al. (1984), Tarter and Robins (1990), Fogari et al. (1994), and van Dijk et al. (1987a). Nine HT studies found statistically significant positive associations in at least one exposure sub-group (Brini et 67/., 1981; Belli et al; 1984, Idzior-Walus, 1987; Verbeek et al, 1987; Wu et al, 1987; Tomei et al, 1991; Zhao et al, 1991; and Lang et al, 1992; Fogari et al, 1994). The range of prevalence odd ratios (POR) for HT was 1.2 to 6.4, with the majority being between 2 to 3. With respect to exposure-response relationships, Brini et al. (1981) found that HT prevalence increased with duration of exposure to continuous noise, but their analysis was not age-adjusted. Verbeek et al. (1987) showed an age-adjusted POR of 3.6 in subjects exposed > 20 years compared with those exposed less than 9 years. The baseline HT prevalence was 9 percent. Lang et al. (1992) also found prevalence increased with duration of exposure, culminating in a POR of 2.6 after 25 years of exposure (though the authors did not specify if this was age-adjusted). Tomei et al. (1991) reported POR's of 1.4 in those exposed fewer than 3,000 hours, and 2.4 in those exposed more than 3,000 hours. These two groups had baseline HT prevalence of 5 and 13 percent, respectively. Zhao et al. (1991) found no association with employment duration, but did demonstrate an increasing risk of HT with noise level; a POR of 2.5 associated with an increase in exposure of 30 dBA. In this study, the baseline HT prevalence was 5 percent. Brini et al. (1981) also demonstrated an increased HT risk with increasing noise intensity, but only in those exposed to impulse noise. Studies by van Dijk et al (1987a), Tarter and Robins (1990), Hirai et al. (T991) and Ledesert et al. (1994) showed no association between HT prevalence and noise in any analysis sub-group, van Dijk et al. (1987a) estimated cumulative exposure, and Ledesert etal. (1994) "subjective" noise exposure by questionnaire. Duncan et al. (1993) performed a meta-analysis utilizing data from 10 cross-sectional HT studies. Only 4 of these studies met the criteria for inclusion in this review, four others being pre-1980, one using 21 noise-induced hearing loss as a surrogate for noise exposure, and one a 1985 Polish study that could not be obtained. The pooled data showed an exposure-response relationship in males with statistically significant relative risks for hypertension ranging from 1.08 at 64.8 d B A to 1.49 at 80.2 d B A . The range in females was 1.12 to 1.80. Noise and Other Hemodynamic outcomes Singh et al. (1982) found heart rates of noise-exposed workers were on average 3.3 beats per minute ( B P M ) faster than unexposed controls (PO.05) . Green et al. (1991) showed increases o f similar magnitude in an ambulatory study (increases of 2.6 B P M and 2.7 B P M in exposed men < 45 and > 45 years of age respectively, P<0.002). In contrast, Babisch et al. (1988) and Fogari et al. (1994) found no increase in heart rate in noise-exposed groups. Kristal-Boneh et al. (1995) showed no effect in males, but found an exposure-response relationship in females, with B P M increasing by 0.13 per d B A . Singh et al. (1982) also reported an elevated risk of irregular heart rhythm (in either amplitude and duration), with a prevalence odds ratio of 3.9 in exposed subjects (not statistically significant). Limitations of Studies of Hemodynamic Outcomes Numerous potential biases can be identified among the hemodynamic studies, which largely shared a similar study approach. O f greatest concern are biases that result in a positive association being made erroneously. Inadequate adjustment for potential confounding as found, for example, in Singh et al, 1982; W u et al, 1987; and Fogari et al, 1994, can result in a such a positive bias. Smaller studies (e.g. Neus et al, 1984; Lercher et al, 1993; and Singh et al, 1983) are especially prone to bias by uncontrolled confounding, where chance may weight an exposure group with a particular confounder. Studies that used subjective assessment o f exposures (Lercher et al, 1993; Ledesert et al, 1994; and X u et al, 1999) are susceptible to information bias. A common problem is recall bias, in which cases may be more likely to correctly identify an exposure than healthy controls. A n interesting extension o f this is that individuals who are already stressed (i.e. by their disease state) may perceive a given noise as being louder than those who are not stressed (Ising et al., 1997). Differential misclassification bias is a concern where investigators are aware of the subjects' status (either outcome status or exposure status). For example, researchers who were not blinded to subjects' exposure status before measuring blood pressure may be more likely to differentially round sphygmomanometry readings based on status. In fact only four studies specified that medical assessments had been blinded (Neus et al, 1983; W u et al, 1987; Tarter and Robins, 1990; Lercher et al, 1993). Studies that had low participation rates (e.g. Lercher et al, 1993) could be positively biased i f subjects' reasons for participating (or not participating) were associated with exposure status. Unfortunately some studies provided little selection and participation information on 22 which to judge this aspect (e.g. Singh et al, 1983; and Fogari et al, 1994). One study had very high levels of concomitant vibration that makes its results hard to interpret (Idzior-Walus, 1987). Even though most studies that found noise-disease associations could be faulted in some way with respect to potential bias, several stood out as more methodologically rigorous (Zhao et al, 1991; Lang etal, 1992; Fogari etal, 1994; and Talbott etal, 1999). These studies were well documented, and used generally more sophisticated analyses. They had larger sample sizes, conducted objective, quantitative exposure assessment, and adressed several potential confounders. Other types of bias may result in attenuation of the observed magnitude of true relationships, or lead to false negative results. Of these, information bias is likely the greatest threat in the studies reviewed. Inaccuracy in the measurement of exposure, that is not correlated to outcome status, may result in non-differential misclassification of exposure, and attenuation of the observed effect (Rothman and Greenland, 1998). In the case of noise and cardiovascular disease, where associations are likely weak to begin with, this might be a fatal flaw in the study design. Personal noise dosimetry (Talbott et al, 1985; Cavatorta et al, 1987; van Dijk et al, 1987a and 1987b; Tarter and Robins, 1990; and Green etal, 1991) would be expected to give the most valid exposure estimates. Other methods, such as area ("stationary") measurement (Kristel-Boneh et al, 1995), combinations of personal, area and expert assessment (Hessel and Sluis-Cremer, 1994), semi-quantitative classifications based on subjects' job title (Garcia and Garcia, 1992) or estimated noise levels based on traffic volumes (Neus et al, 1983), are potentially much weaker. Simple comparison of exposed and unexposed departments or plants will lead to misclassification if the control group is not truly unexposed (e.g. Talbott et al, 1985 and 1999; and van Dijk, 1987b). Stationary sampling has been shown to underestimate noise exposure levels when compared to dosimetry (Hansen etal, 1989). Methods for aggregating area measurements by weighting according to participants' activity profiles may be adequate in some circumstances (e.g. CSA, 1994). But many studies of non-auditory effects have relied on much less comprehensive area-sampling strategies, such as "on-the-spot" sampling by worksite physicians (Fouriard et al, 1984), use of sampling times as short as 60 seconds (Belli et al, 1984), or assignment of personal exposures based on plant-wide average noise levels (Talbott et al, 1985). Some studies failed to report their exposure assessment methods at all (Brini et al, 1981; Idzior Walus, 1987; and Hirai etal, 1991). The majority of HT studies were cross-sectional in design, studying only prevalence of health outcome. Five were limited to studying only current exposure levels (Belli a/., 1984; Fouraiud etal, 1984; Garcia and Garcia, 1993; Fogari et al, 1994; and Hessel and Sluis-Cremer, 1994), and another six measured current exposure, but restricted the study population to those with a minimum duration of employment (2-15 years) in the noisy environment (Singh et al, 1982; Neus et al, 1983; Aro, 1984; 23 Cavatorta et al, 1987; Idzior-Walus, 1987; and Wu et al, 1987). Such approaches likely result in exposure groups with heterogeneous exposures, mixing those with low risk of adverse cardiovascular effects (because they have little exposure) with those at higher risk (because they have higher exposure; Stewart and Herrick, 1991). Several studies used duration of employment in a noisy environment as a means of assessing chronic exposure to noise (Brini et al, 1983; Talbott et al, 1985; van Dijk et al, 1987a; Verbeek et al, 1987; Tarter and Robins, 1990; Tomei et al, 1991; Zhao et al, 1991; Lang et al, 1992; and Kristel-Boneh et al, 1995). While an improvement, this method is limited because it assumes that exposure levels are unchanged over the duration of employment, and that all members of the study group are exposed at the same level. Again, this can lead to significant misclassification, and it has been shown that situations exist where duration of employment is not, in fact, related to exposure level (Stewart and Herrick, 1991). Only two of the 31 studies reviewed attempted fully quantitative assessments by estimating personal cumulative exposure levels combining measured noise levels and individual work histories (van Dijk et al, 1987b; and Talbott et al, 1999). Because dosimetry measurements are made at the shoulder, use of hearing protection devices can result in substantial overestimation of exposure levels. Their use has been widely prevalent for several years, but few of the studies reported their status in exposed populations. Hearing protector use in noise-exposed populations would likely attenuate expected associations, as actual inter-group exposure differences would be lower than apparent from ambient monitoring. Experimentally, they have been shown to reduce the apparent effect of noise on the cardiovascular system (Sudo et al, 1996). Misclassification of health outcome is of equal concern, because of concerns over accuracy of blood pressure measurement. There is substantial prevalence of intra-individual variability and measurement error can be relatively large (Campbell et al, 1994; Staessen et al, 2000). About half the studies utilized single-reading, casual BP measurement only; those that conducted repeat measurements did so over a fairly short time period. Only two-thirds had subjects rest before measurements were made. One study utilized automatic BP measurement (van Dijk et al, 1987a), and one, ambulatory monitoring (Green et al, 1991). Both of these methods are considered superior to casual BP measurement. Attenuation of the observed effect may also result from over-controlling; i.e. controlling for what appears to be a confounder, but that may in fact be part of the disease pathway (Rothman and Greenland, 1998). Such a factor might be considered an "intermediate" rather than confounder, and should be handled differently in analyses. The potential of this is greater with "internal" rather than "external" risk factors. For example, if noise causes a perturbation of the hypothalamopituitary-adrenal axis, and results in visceral fat adiposity, which results in turn in increased blood pressure, then controlling for "body-mass-index" may weaken any association between noise and HT. However this was commonly done in 24 the majority of studies. Researchers in some studies also controlled for factors that were likely to co-occur with the health outcome. For example, Garcia and Garcia (1992) excluded subjects presenting symptoms of stress, and van Dijk et al, (1987a) controlled for "nervousness". Epidemiological Studies on the Effects of Noise on Endocrine and Blood Lipids Both the neuroendocrine system and blood lipid levels are hypothesized to be affected by exposure to noise, yet observational studies are rare, perhaps because o f the difficulties of obtaining blood samples. Levels of the catecholamines epinephrine and norepinephrine are both predicted to increase with noise exposure, and Cavatorta etal. (1987) found approximately 70 percent increase in both, as well as in their urinary metabolite, vanillylmandelic acid, in noise-exposed subjects. Sudo et al. (1996) showed elevations in epinephrine, norepinephrine, and Cortisol in noise-exposed textile workers. The study group of Cavatorta et al. (1987) had been exposed to noise for 5 - 10 years, and Cortisol increases would have been predicted there, but none were detected. It has been noted, however, that serum Cortisol measurement (as done by Cavatorta) must be carefully planned. There is a natural diurnal variation in Cortisol level, that changes significantly as Cortisol enters the blood system in 'pulses'. Levels of stress-related Cortisol release wi l l differ depending on whether it is superimposed on a natural peak or trough in Cortisol levels. Urinary Cortisol measurements integrate these variations somewhat, but long-period urinary samples are also prone to technical difficulties (Bjorntorp, 1997). Idzior-Walus (1987) found no increase in cholesterol levels, and found triglycerides depressed in his study group, being 1.56 mmol/1 vs. 1.72 mmol/1 in the control group (P < 0.05). These findings of lower levels o f known risk factors are counter to the hypothesis that noise is associated with heart disease. Babisch et al. (1993b), found total triglycerides, platelet count, plasma viscosity, and glucose all to be positively correlated with noise. In an earlier study of similar design, the same research group (Babisch et al., 1988) found increases in oestradiol, total cholesterol, fibrinogen and plasma viscosity, and decreases in testosterone levels in exposed subjects. A l l of these were consistent with the noise/ischemic heart disease hypothesis. However, they also found decreased Cortisol levels and platelet counts with increasing noise, and increased H D L cholesterol, and antithrombin III, that were all counter the hypothesized direction. Epidemiological Studies of Noise and Chronic Disease Endpoints Summaries of studies investigating occupational and environmental noise exposure and chronic heart disease endpoints are given in Tables 3 and 4, respectively. 25 u d o "£ Jo 3 C L . <L> oc 3 00 c —1 00 d 3 CO O CL X tu "5 c o a 3 O o O CO o 03 cu a cu X cj "S 1 u d 03 o Z o -a 3 C/3 ~a o '5b c <L) -o 'a. LU cn _y X) 03 H 81 •2 5? -g 1S I ! « • II?. < C O 8 _ | 8 i I : 35 cj C co o © I S " LU 35 «S J c*-T3 O | | d w H ® 1 l l « J? —I 8 § « i l l i i i i.§_-• f i l l .5 « ° w • a s i © X u, <D r- <u o O CQ o ; o ) ca 1? O i & OT O W g 5 £ * * & • a l 60 o U] C U a 11; 11 cfi a o u a a u 22 .3 §!1 igO - a I E 1 E as 1 1 1 ? * i l l u u u u 2 fi 1 u fir .2 M T 3 -l i e « O O T CH. 3^ O &3 • <u U o y d s (U 41 • O T3 • | | 11 43 .5. S » w I a E 43 11 .S o 1! 9 > 1 CN r-^ —" M v i So u cn oo tu o B P ddd-4> —* t_ © 00 SB ___ £__ _3t 1 II 8 o •s '2 1 | l l E OT OT (J 2 « S ° Oi 1 1 S | 2 a 2 -III i i 41 £ H .5 00 # II u '5 ^ c/3 > < ty_ t/_ cs .g . ° a 5 E i l i 8 i s s CO o If--3 E "> -a s • s i s 1 f i E CL §2 a_ s § § 2 •I S U CTa. <5 a o s o ¥ 2! — g g TS . ' a , i S-J— OT ? s 8 , « O OT 2 'I ' i s " o j a • l l i l l : g -a _ i 5 2 g & I. co a> _ O & u 3 a & S U 3 u co ui. S — CL CD 8 c i a c*3 S T3 1 l l v i l l i §•141' < CO LU CL 6 s Ifg g o . , •§ E C3\ ON 26 . A) — x o •= ^ "g 1g| T3 3 "rt U _o "o u T3 "ft w o « £ E s tn 2 s C*. N » "-*' j "8 - K •a S3 9 •Q 1 / 1 S w y I as a. u u " S i * : 111* l i t ! 1 g 1 c a m I £1 X .2 •c S u I I 5 i J= fr • 5 t i i i — SP 11 — B 5-a a s E .2 B 15 S 1 OJ • = S 'C gf § a I S 1 SI I J j l 8 T O .o tN S ffl E i •3 : Jit ! Si® i fl> M n *h -P £ « i l l § • S i !<& a II113 •c pa 8 s LU " OS * . (A M £ a I i > * X o 18 &s I 8 I I J '5 S e ^ -1 1 * 3 e o t N a ji -9 * -2-S | v | co" os "P o3 - -) a. orj-c-'5 e 5J 5D oo S 2 & 8 12 8 11 u e — CL. ,1 © 5 5 S a l " I s . ,9 O.TJ E « a U o 9 { . P e s §f 11 — crt qj e <2 6 i i i ° s & VJ u O a->. s 6 3 " I I I O cn £ • a l l a a Z B B S * £ O 3 a < -2 ^ a "S = « a- £ 5 e C E ! *0 (« If I S •a •§ JS "r -3 •£ "g " s l i t i s a S i l l « 8 0> « C O W J S _oj 'S -a s 2 § 2 -a' §1 Q H S .2 .9 o a oj T: D .E ' •2 3 9 S X 5 T3 fi 5 S &c3 £ >j & S .a u 53 JS Q • 1 u S g. ua ,a 8 | l 2 *»> * 2 •g .S? Q < on cyj c/a :> O o to ? 00 13 8 8 j 3 E W -I CN S 3 — W) T3 CU tflO op ; 00 « s 8 g {.Sag i D. -o I P 3 Tt J « IF CL a •a - 3 CN BoS «85 8 •S ~ M IS oo « ea C- v 27 »8 g. 19 o Kl S 2 a: O a B a IS. B -3. U c/) I I 3 • a S S 8 U Q O < CO CO < « a g u m p, « & ca v 1 1 ° j 0 11 ft I 111 cu ta OT a. fl3 ON 3 OT , H ^ OT c R T ; 8,-a 2 3 — CQ j§ J £ I I I •11 l l ^ , j= . 3 ; I 3 8 A T» J ? I I SI a ra OT NO -a K °N 0 , 2 a e <i 0 0 ~ £: ta & S NO ra •S -g = a < CO c3 ^ m a j= St CQ O I z u Overall, study quality has improved with time. O f the 13 studies reviewed, the majority o f recent studies were longitudinal, with good classification of health outcomes, attempts to control for potential confounders, and appropriate analytical techniques. Babisch et al. (1994) investigated non-fatal myocardial infarction over a one-year period in Berlin hospitals, comparing residential traffic-noise exposure in cases with controls selected from the general population. Small, non-significant odds ratios of 1.2 - 1.3 were found, with the highest risk in a subset of cases who had not changed residence for more than 15 years. A potential for selection bias was noted, as 30 percent of eligible hospitals did not participate. However, as these were inner-city hospitals in the noisiest part o f the city it was assumed that this would be a conservative bias (i.e. toward the null). Misclassification of noise exposure, which was based on municipal noise maps, would be expected to be non-differential and therefore to dilute any true effect. The authors noted that with such small increases in relative risk, statistical significance would be hard to achieve. In a follow-up study using the same Berl in hospital data, Ising et al. (1997) compared occupational noise exposures in the two groups based on a self-reported qualitative assessment. A n exposure-response relationship was found, with a relative risk for myocardial infarction (MI) of 3.8 in the highest exposure group. The authors expressed concern that subjective measurement o f noise in cases might be biased, by subjects being stressed by their disease status, as previously discussed. A low participation rate among controls (64 percent) might have introduced additional bias; the case and control groups differed with respect to the number of blue-collar workers, 37 percent and 29 percent respectively. Duration of exposure was not considered, but resulting misclassification would likely be random and therefore non-differential, attenuating the effect estimate (as would migration o f sensitive individuals out of exposed settings). The authors estimated a population attributable-risk o f 0.27, suggesting occupational noise might be the most important risk factor for M I after smoking. In a series o f analyses, Babisch etal. (1988, 1993a, 1999) examined residential traffic noise and cardiovascular disease among residents of two U K communities. A prevalence study was conducted in 1988, then the community cohorts were each analyzed after periods o f 4 and 10 years. In addition, subjects from both cohorts were pooled, and this reconstructed cohort group followed from year 4 to year 10. N o increase in risk for any heart disease nor stroke was found in the prevalence study. Most follow-up analyses produced similar results: weak, non-significant elevations in relative risk for ischemic heart disease o f approximately 10 to 30 percent. Noise exposure was estimated for each residence from area noise measurements, but adjusted for individual residence parameters (location o f bedroom, etc.). The highest relative risk found was 1.6 (non-significant), where exposures measures took into account residential layout and window-opening habits. The authors noted that current traffic noise was a crude 29 indictor of past exposure, and that no account of other (e.g. occupational) exposures was available. Further, studies of MI survivors (in their prevalence studies) did not represent all cases of ischemic heart disease, as it has a high mortality (Babisch et al., 1988), and the early follow up studies suffered from a low number of cases, and unstable estimates (Babisch etal, 1988, 1993a). Theriault et al. (1988) examined noise in aluminum smelter workers. A relative risk for ischemic heart disease of 1.7 was found in one specific department, but when noise exposures were investigated, they were not associated with increased risk (RR = 0.9, 95% CI 0.6 - 1.5). This study demonstrated hypothesized exposure-response relationships for smoking and heart disease, suggesting that statistical analyses were appropriate. Company hygienists qualitatively estimated exposure. They reviewed subjects' work histories and calculated lifetime exposures by weighting time spent in each job by an ordinal noise metric, which increased potential for exposure misclassification. No adjustment was made for hearing protector use. The investigations of Ostry (1998, and Ostry et al, 2001) involved the same lumber mill cohort used in this study. Ostry (1998) investigated the association of job-strain with cardiovascular disease, but included noise exposure in his exposure-response models. Noise exposure was assessed subjectively on a 4-point scale by panels of expert job raters. In internal analyses adjusted for age, year of hire, and duration of employment, noise exposure was associated with increased risk of ischemic heart disease mortality (OR=1.29, 95% CI 1.05-1.60). The author considered that risk of other confounding in this study was low. The purpose of the second study was to evaluate predictive validity of various subjective exposure assessment techniques (Ostry et al, 2001). In this cross-sectional study of a sub-cohort of 408 subjects, self-reported noise exposure was significantly associated with self-reported heart disease during the prior six months (OR=1.92, 95% CI 1.28-2.55). Of the earlier chronic disease studies reviewed, the majority was cross-sectional in design. The exception was a small retrospective cohort study that had very low statistical power, and a weak analytical design (Lees et al, 1980). This study found no increased risk of heart disease, but had only 7 cases. Idzior-Walus (1987), found an elevated risk of effort angina in workers exposed to high levels of noise and vibration. However this study had very poor description of exposure measurement and did not control for any potential confounders. Another study was ecological in design. It investigated reasons for visits to general practitioners (GP's) in high and low noise exposed areas around Amsterdam airport during a one-week period (Knipschild 1977b). Visits for cardiovascular diseases in 15-64 year age range were nearly doubled in the exposed area. There is reason however to suspect information bias, as the authors reported strong effects due to physicians' consultation habits, and at least one participating physician was active in an anti-noise 30 campaign group. Two studies used similar methods to examine ischemic heart disease in community groups exposed to aircraft noise (Knipschild, 1977a) and traffic noise (Knipschild and Sajle, 1979). In the first of these, communities were assigned to high or low exposure groups based on environmental aircraft noise data. Prevalence of self-reported medical treatment for heart trouble was significantly elevated in the more exposed communities. Subjects were recruited through letters requesting them to join a heart-disease study, and participation rates were low; 39 percent and 43 percent in high and low exposure communities respectively, and the former was thought to have generally lower socio-economic status. The authors stated that selection bias was unlikely to be correlated to aircraft noise, but as exposure status was based on residence, selection biases operating at the community-level could have resulted in an overestimate of the effect estimate. In the traffic study, classification of'noisy' and 'quiet' streets was made based on traffic volume data at the residence and the study was limited to housewives, who would be expected to spend the majority of the day in or around their residence. Participation rate was high at 86 percent. No difference between the two groups was found for smoking habits, weight or physical activity. Those in the exposed group however were less likely to rely on social health assistance (29 vs. 36%), indicating a difference in socioeconomic status. No significant differences were found in any of the heart disease end-points investigated. It seems possible that the combination of uncontrolled confounding for socioeconomic status, misclassification of noise-exposure, and the general problem of the susceptibility of the cross-sectional study design to migration, could easily obscure small elevation in disease risk. Noise Exposure Factors Overall there was very little data available in the studies reviewed that provided evidence of critical exposure intensities or cumulative exposure levels at which adverse health outcomes would be observed. First, with respect to chronic disease outcomes, only one occupational study provided quantitative exposure data, and this was determined indirectly from a subjective exposure assessment (Ising et al., 1997). These authors found that relative risk for MI was significantly elevated at exposures above 75 dBA. Traffic studies showed weak elevations in chronic heart disease at levels of 66 - 70 dBA (Babisch et al, 1993b, 1998) and 71 - 80 dBA (Babisch et al, 1994). No chronic disease studies provided information on effects of duration of exposure. The picture from the hemodynamic studies was not much clearer. Effects were seen in community studies at very low levels, i.e. 60-70 dBA or lower (Babisch et al, 1988, Neus et al, 1983), but information on duration of exposure was not presented. In occupational studies, associations between noise and CVD were seen as low as 80 dBA (Brini et al, 1981, Verbeek et al, 1987) but there was no apparent effect pattern with respect to increasing exposure intensity. 31 Several authors however demonstrated increasing effects with duration of exposure. Verbeek et al. (1987) and Lang et al. (1992), both showed their largest effect after 20 years of exposure. In contrast, Zhao et al., (1991) found that increase in effects with exposure plateaued after 5 years in duration, and concluded that any increase in blood pressure due to noise must be accomplished within that time frame. While it appears therefore that levels considered hazardous to hearing (i.e. at, or above 80-85 dBA) may elevate risk of HT, it is not clear whether there is a lower limit for these effects in an occupational setting. It is also apparent that hemodynamic changes can be seen as early as within 5 years of the start of exposure. However, longer periods of exposure may be required for reasons that are not understood, but are likely associated with exposure level. Several investigators examined cumulative noise intensity, van Dijk (1987b) calculated a cumulative measure based on the equal energy hypothesis (i.e. that equal amounts of sound energy cause equal hearing loss, regardless of how the energy is distributed across time) in current job, and a second measure that summed exposures at each job using an ordinal 5-point noise scale multiplied by the number of years at that job. Only the latter measure included exposure prior to current job. Neither measure was correlated with blood pressure levels. Using a similar cumulative-ordinal exposure measure, Theriault et al, (1988) found no association with risk of ischemic heart disease. Talbott et al. (1999). calculated cumulative exposure using the equal-energy equation and they found cumulative exposure to be associated with blood pressure, but only in higher exposed groups, which they concluded may be due to a threshold effect. The noise levels in the high and low exposed groups were 93.3 and 82.9 dBA, respectively. Health effects are seen at much lower exposure levels in the community setting, than in occupational settings. This is hypothesized to result from the situational context, i.e. that activities around the home, are disturbed or interrupted at lower levels of noise. Thus, the "indirect" pathway (Figure 1) is assumed to predominate. Consequently, it is suggested that a subjective measure of noise level, rather than an objective one that should be the better predictor of health outcome. To examine this, several studies employed subjective measures of exposure (Garcia and Garcia, 1992; Lercher et al, 1993; Ledesert et al, 1994; and Xu et al, 1999). It is not apparent, however, how much utility findings based on subjective exposure have, unless modifying factors (e.g. coping potential, vegetative lability) and concurrent objective noise levels are measured concurrently to determine the magnitude and direction of their effects. A potential bias of this approach with respect to chronic disease has already been discussed. A subject who is stressed because of their ill health may subsequently be more likely to find noise more annoying or stressful than a comparable healthy subject. 32 Another potentially important parameters of noise exposure is the "impulsive" character of the signal4. Brini et al (1981) found that workers exposed to over 90 dBA of impulse noise had a prevalence odds ratio for hypertension of almost 2, while workers at the same plant exposed to (slightly lower levels of) continuous noise were not affected. They also reported that those exposed to impulse noise showed an exposure-response effect for intensity that was not evident in those exposed to continuous noise. Kristel-Boneh et al. (1993) included a term in their models for impulse exposure, but found no effect. Finally, very little data was available on the effect of noise frequency. Only Belli et al. (1984) reported a frequency spectrum analysis; an average spectrum from 3 sites within the plant showed that the peak noise levels were in the 4, 8 and 16 kHz octave-bands. Tomei et al. (1991) reported some low-frequency harmonic frequencies. Both of these studies found positive associations between noise exposure and hypertension, but provide little evidence that any particular frequency component is more hazardous than another. Summary Various aspects of the noise-cardiovascular disease model have been examined using observational epidemiology, primarily focusing on the effects on blood pressure, and to a lesser degree chronic heart disease. Two-thirds of the hypertension and blood pressure studies I reviewed found positive associations with noise exposure. Overall quality of these studies was mixed however, and so such simple tallies can be misleading (Greenland, 1998). However, a subset of reasonably well-executed studies showed increases in relative risk of hypertension that were between 2 and 3. Some studies also showed exposure-response relationships for HT and duration of exposure (Lang et al, 1992) and intensity (Zhao et al, 1991). Blood pressure studies found average increases of up to 10 mmHg (SPB) and 7mm Hg (DBP) in noise exposed subjects, and exposure-response relationships were found between blood pressure and duration of exposure (Lang et al, 1992) and with cumulative intensity (Talbott et al, 1999). Relative risks for noise and ischemic heart disease were weaker, on the order of 1.1 - 1.3. Generally speaking, existing studies have had insufficient power to achieve statistical significance, but elevations were reasonably consistently shown in community studies. One occupational study found a relative risk of 3.8 for myocardial infarction, but this study used subjective exposure measurement, and 4 Impulse noise consists of short bursts of acoustic energy usually lasting less than 1 second, and is common in industrial environments. 33 information bias was a concern. Another, well conducted study, showed no association (Theriault et al., 1988), but may have been weakened by exposure misclassification. Overall, the epidemiological support for the hypothesized noise-heart disease association is good enough to warrant continued research, but too inconsistent to provide strong evidence for a causal association, especially when the strength of association is also weak. Many reasons for the inconsistency have been suggested. There is wide consensus among reviewers that the overall quality of studies has been weak, and that inadequate control of confounding, low analytical power, misclassification of exposure and disease outcomes, and migration of sensitive subjects all potentially produce bias that could result in over-estimation, or underestimation, of true effects. Because of the complex pathophysiological pathway, over-control of potential confounders might also be a problem. If cardiovascular disease pathology is in general multifactorial, it's reasonable to assume that noise-stress related effects would also be multifactorial (Pickering, 1999). If stress response patterns vary at an individual level (i.e. whether the HPA or SAM pathway is more or less likely to be involved) then more consistent results might be expected by examining the longer-term, chronic outcomes, rather than intermediate states of the disease pathways. It has been suggested that because of the role of psychological and environmental factors in determining whether noise is stressful (particularly at low exposure levels), health effects may be more strongly correlated to subjective measures of noise rather than objective measures. However there are problems with this approach and it does not seem to be compatible with the need for large-scale longitudinal studies required to achieve the needed statistical power. The model suggests that exposure to higher levels of noise might be less influenced by the modifying factors, and so one might expect occupational studies to produce more consistent, and perhaps stronger associations. With respect to chronic heart disease, there is some evidence of higher risk of heart disease compared to community studies, but the studies are still comparatively weak in terms of exposure assessment, and no large-scale occupational study utilizing quantitative exposure assessment has been undertaken. In summary, I agree with reviewers who have strongly recommended that new research in the field should emphasize analytical (i.e. cohort) studies. These would provide the necessary analytical power to examine small relative risks, and give improved control over potential confounders. They also give improved ability over cross-sectional studies to look at temporal issues of cause and effect. Studies of standardized chronic health outcomes, such as CVD mortality, should reduce misclassification of health outcomes. Use of mortality outcomes also overcomes the potential problem of inter-individual 34 differences in preferred stress-response pathway - i.e. that in one person the HPA pathway may dominate, and in another the SAM pathway, giving differing pathophysiologic profiles but both leading to CVD. Exposure assessment also needs to be improved. Future studies should improve validity and reliability in exposure measurement, and require improvement in quantitative, objective measurement of cumulative exposure to noise. The ability to account for the use of hearing protective devices that may strongly impact the ability to measure accurately is also needed. 3 5 Chapter 4. Characterizing Noise Exposure in BC Lumber Mills The first phase of this study was to characterize current noise exposure in the participating lumber mills. A n understanding of their current levels of noise in mills, and determinants o f noise exposure, was important in developing a method for retrospective exposure assessment. Data was also needed for the accurate determination o f hearing protector use, as well, where possible, ancillary data that would allow examination of other important noise parameters. As described in Chapter 3, exposure misclassification is considered a major limitation in many studies on noise and cardiovascular disease. In order to minimize this bias, the use o f exposure assessment methods with the greatest precision, i.e. quantitative, is most desirable (Stewart and Herrick, 1991); they formed the basis of this characterization. Another problem is that measurement o f noise exposure has been complicated by the use o f hearing-protection devices (HPD's ) (van Dijk, 1990; Babisch, 1998). Real-world use o f H P D ' s is expected to provide exposure reduction in the range o f 5 to 10 d B A for ear-plugs, and 10 to 15 d B A for muffs (Berger, 1986); their use has been shown to be associated with reduced cardiovascular effects in noise-exposed populations (Sudo et al., 1996, Melamed et al., 1996). Therefore taking the use o f H P D into account would improve the validity of the overall exposure assessment. Previous observational research related to C V D to date has been done using A-weighting (which substantially discounts low-frequency components o f the noise signal) even though the effect of noise frequency on non-auditory outcomes remains to be elucidated. In this phase we collected spectral data relating to noise exposures in the mil l to permit more flexible analyses with respect to noise frequency. The objectives for the work reported in this chapter were to (i) quantify current noise exposures of B C lumber mi l l workers, using randomized full-shift personal noise dosimetry; (ii) measure representative frequency spectra of noise exposures; (iii) describe current and historical patterns of H P D usage and (iv) investigate the role o f potential determinants o f exposure to noise by multiple linear regression modeling. Cohort Study Setting The fourteen lumber mills included in this study were originally identified and recruited for an investigation of exposure to chorophenol (Hertzman et al, 1997). They were selected based on mi l l size, use of chlorophenol-based fungicides, availability of historical personnel data and their willingness to participate. Table 5 gives basic descriptive data for the fourteen mills. Eleven mills were located on the coast of British Columbia, and three in the interior of the province. Ownership o f the mills was distributed among several forestry companies, and changed for several of the mills during the study period. A l l mills 36 ft •I1 u u Q •a u u - H "u t/3 "O ti ti o •o u o -t—' GO 6 0 O J GO o a. oo H u-O 3 2 & « + • , I E < 61 0 0 o Oi JS JS S-S c 00.2 O 0 0 cu CU a oi c o U S3 cu T3 T3 —1 J M S B ) fi S g PH t-H CA CA CA >- >" O O O O O — O ^ « n UH t-i CD q t-> CD CD i— CU H Wal Wal Wal Wal Wal Wal la Wal Salt Salt 15 GO Salt Salt Salt Salt Salt No Yes Yes Yes Yes No Yes Yes C 3 J3 cu 13 00 u •a GO o o o o o o o o o o o •n-TT H . — o o o Tt-" vo — 0 0 3 o Q PH ^ PH c c c 5 £ 5 u u o o o o o E E E E D. a a GO GO oo u. 5 E 6 oj u. o cu cu K E Q K S T3 ft, E 3 3 o Q oo 3 O a o Q <% *i <3 o o vo" •a u U •a CU a: o o o o o VO — o o o 3 a 0 0 3 o a 0 0 3 O a 0 0 3 o a <% j* cu j£ Jr; S .2 .2 ,2 £ n c e |3 £ E £ .2" C U . U J C U . U H C U . U . J ~ J J u y u 5 J u > 33 E 33 E 33 E > 33 E 33 E 33 E o cn «n Ol <•*-> — O 00 00 T l - O o <N — o m Ov — — — — — r— — — O TT CN r— vo — — 0 0 0 m 00 o m n I N o o CN O O •D-o o o o m as — T l -E E 2 CO JO l l o u 2 JO i o o U u- •« CU CU > > 3 3 8 8 cfl O E cd 43 C C3 43 CS 4b > > 00 U C3 C3 43 i 3 C C g cu cu O O CJ Z 1H U . 1H — " .2 .2 .2 -2 'c "c n c/j cu cu cu o U o o o U o U o U o U 3 o U 3 o U 3 o U m «n cn r- tN O VO »n O Os Os Os CM VO VO r— VO VO O Ov OO 00 OO OO 00 as Ov Ov as ov ov 00 00 as CTv Ov Ov Ov as as O _ m -* CN m •n- in VO 00 Ov Boom Water-based log storage and sorting Boat motors Yards Lumber yard: lumber stored prior to shipment hog loaders, Cranes, Cherry'Picker, Forklift, Lumber dropping, Cut-off saw, Barkers, Trim saws, Logs dropping Boards dropping Radios, Belts and conveyors By-Products: Wood chips usedfor pulp manufacture, sawdust and ' a waste-wood fuel mixture. Log loaders, Splitter, Chip screen, Electric Motors, Hogfuei Belts and Conveyors, compressed air, Ventilation, Chipper Log Processing; Whole logs are debarked and cut to length Cut-off saws, Barkers, Trim saws, Logs Dropping, Beits and conveyors, Chipper 1 Sawmill Primary Breakdown: Logs cut by band-saws into slabs and f ("cants") I Remanufacturing: Slabs and cants sawn into timbers (e.g. > 4 "X4 ") and dimensional lumber by circular "eager" saws and band "resatvs". Lumber trimmed to length by circular "trim " saws. Shipping Lumber is sorted, stacked, packaged Boards, Radios, Headrig saw, Package bander, Package press, Load hoist, Eager saws, Forklift, Bells and conveyors, compressed air, Ventilation Boards dropping Hydraulic system Cable cutter Cut-off saws Trim saws Log Hoist Headrig saw Headrig Carriage Lumber stacker Forklift Splitter ]-Bar sorter Chtp'n'saw Eager saw Re-saw Dropsorter Chip screen Belts and conveyors Ventilation Chipper Planing Lumber is optionally planed to provide a better or specialty surface Boards dropping, Electric Motors, J-Bar sorter, Tilt hoist, Planer in-feed, Planer, Strip shaker, belts and conveyors, compressed air Department Name Department description Reported noise sources . Indicates • • ^ ^ ^ Process Flow Fork lifts Hammering Maintenance Compressors Arc air Hand tools Machine tools Hydraulics Logs Stacker Disk grinding Belts and conveyors Ventilation Traditional Trades: Mill Wrights, welders, machinists etc. Sa willing: specialty "saw-filing" trades responsible for sharpening, truing and replacing saws and knives. Hydraulics Cyclones Barker Compressed air, Ventilation Figure 2: Lumber Mill Departments, Process Flow and Reported Noise Sources had approximately the same production flow (Figure 2), although the number of production lines, and presence or absence of specific facilities (such as planing mills) varied both between mills, and within mills, over time. Mills that stored logs on land made more use of heavy diesel equipment, while those storing logs in water used small tugs and "dozer" boats. There was a fairly large degree of homogeneity in equipment use among the participating mills. In addition, advancement of saw milling technology has been relatively slow. Significant changes have been made in saw-blade technology, automated lumber sorting and computer-aided machine feeds, but much of the technology for handling logs and lumber has been unchanged for decades, and a great deal of older equipment has been retained. Most mills have a mix of old and new machines, and it was not uncommon to see equipment that was several decades old (one mill still operated a reciprocating gang saw built in 1908). The general acoustical environment of a typical mill was complex. It was characterized by very high levels of noise from many sources and apparent combined high frequency sources from saws and hydraulic systems, and lower frequency components from heavy log or lumber movement. There was a considerable impact component to the sound from the movement of logs and lumber. All mills had undertaken to enclose planing machines. Generally speaking, the only other common engineered noise control present were operator booths for certain workers, usually those working with logs (in the so-called "primary" breakdown area). Most other jobs required some degree of manual handling of the product that precluded the use of booths. There were sporadic implementations of machine enclosures, and impact-noise controls such as slides at lumber drop points, noise-dampened bumpers and slack-chains for absorbing board movement. Methods Sampling Strategy Four of the 14 lumber mills participating in the cohort study were randomly selected for a comprehensive personal noise-dosimetry survey. Each of the 4 mills was visited twice during 1996 and 1997, once in the summer (May to October), and once in the winter (November to April). Each site visit lasted approximately six working days. All production and maintenance job categories were eligible for personal dosimetry; we attempted to obtain two samples per job-title per site per visit. All mills worked two shifts, and sampling was scheduled on both. One hundred and eighteen mill-specific job titles were collapsed into 61 standardized job titles. Participants were randomly selected from all possible job category/shift combinations. 39 Noise Measurement Noise dosimetry used four different models of Type-2 dosimeters, the C E L 281 ( C E L Instruments, Mil ford , N H ) ; Quest M - 8 / M - 8 B (Quest Electronics, Oconomowoc, WS) ; B & K 4428 (Brttel & Kjaer, Denmark); and L D L Model 700 (Larson Davis Laboratories, Pleasant Grove, U T ) . A l l were calibrated to manufacturer specifications before and after each sample. Microphones were positioned on participants' shoulders and affixed to their collar or lapel. Dosimeters were mounted at the beginning of the participants' shifts, and were worn throughout the shifts including during breaks spent on site. Participants and equipment were usually observed at the participants' workstations at least twice during a shift. Dosimetry samples over 4 hours duration were assumed to be representative o f the entire shift. The equivalent A-weighted continuous noise level (the average noise exposure integrated over the sample duration, or L e q ) was used as the exposure metric in analyses. For devices that reported only "dose", L e q was calculated as: lOlogio ((dose/100)(criteria-time/sample-time))+criteria-level, where criteria time and level were matched to the relevant dosimeter parameter settings. Jobs were grouped by administrative department. Department means were compared using A N O V A ( S T A T A 7.0, College Station, T X ; command oneway). Frequency Spectrum Analysis Octave-band measurements (32 H z to 8 kHz) were obtained with B & K 2203 S L M and 1613 octave-band analyzer (Briiel & Kjaer, Denmark), and a Rion N A 2 9 E analyzer (Rion Co. Ltd , Tokyo). Instantaneous samples were obtained at participants' positions for most job categories at each mi l l during single observations while participants were doing representative work. Because the total number of measurements per job was small, observations were averaged by department, and inter-department differences were tested by A N O V A ( S T A T A command oneway). Hearing Protection Device Usage Data on the current usage of hearing-protection devices (HPD's ) was obtained by subject interviews during the noise-dosimetry survey. Since 1979 mills have been participating in hearing conservation programs that require annual audiometry for noise-exposed employees. Data on 1979 - 1996 trends in the prevalence o f H P D use was obtained from hearing-test records made available by the Workers' Compensation Board, for 13 of the 14 cohort mills. The missing mi l l ' s data had likely been mixed with other company data, complicating its abstraction. Data on pre-1979 trends in H P D type and use was obtained through interviews with senior workers at all cohort lumber mills. Determinants of Noise Exposure Data regarding determinants o f exposure to noise was collected concurrently with the dosimetry survey, from two primary sources. First, as participants returned sampling equipment, a short 40 questionnaire recorded their activities during that shift. This included the personal use of certain types of equipment (compressed air, chainsaws, diesel equipment), the participant's department, and the amount of time spent in a personal enclosure (booth or vehicle cab). Second, each job-title category was observed in detail at least once at each mill site, in order to collect job-category-specific information. This included job location (indoor vs. outdoor), personal-enclosure construction materials, percentage of enclosure, presence of noise barriers, and identification of major noise sources affecting each job. Common types of noise sources were grouped before analysis into several categories. These are shown in Figure 2 along with lumber mill departments with which they were associated, and process flow. For maintenance personnel, time spent per day inside production areas was also recorded, and the location of their trade-shops. Additional summary determinant variables were generated as follows: "cutting tool type", e.g. circular saw, band saw, etc.; "multiple-bladed saw"; "maintenance worker"; and "booth", indicating that the participant worked in an enclosure. "Time spent in production area" and "trade shop location" were nested within "maintenance worker". Similarly, "booth construction material" was nested in "booth". The relationship between determinants of exposure and dosimetry data was examined using multiple linear regression modeling. The dosimetry data was normally distributed and did not require transformation prior to analysis (underlying sounds pressures are log-normally distributed, but use of the decibel scale is a log-transformation). The modeling phase began with univariate analyses of each determinant and L e q . Because of the large number of potential predictor variables, those with individual P-values greater than 0.25 were not retained for multi-variable analysis. Noise sources with negative univariate coefficients were not offered to the model. Correlation was examined among the remaining variables; within pairs of highly correlated variables (i.e. Pearson r > 0.6), the variable with higher predictive power in the univariate analyses was retained. Because of the large number of independent variables, they were offered to the multiple linear-regression model in two phases; first, all variables except job categories, then, once the model was fitted, job categories were added. L e q was regressed against these determinants by backward-stepwise multiple linear regression (STATA commands sw, regress), retaining independent variables with P < 0.1. Goodness of fit was tested by analysis of residuals and the Cooks-D statistic. The final model was re-estimated using generalized estimating equations to account for correlation among repeated measurements on subjects(GEE; ST ATA command xtgee, i=participant, link=identity, family= gaussian, and correlation=exchangeable). Results General Mill Characteristics Two mills, constructed in 1965 and 1975, were located in the BC interior (which experiences hot dry summers, and cold dry winters) and had an enclosed building envelope. Two were located on the BC 41 coast (warm summer, cool wet winter), were constructed in 1929 and 1967, and were of a more open construction. The basic process flow through a lumber mills is illustrated in Figure 2. Each shaded box represents a mill department (with name given in bold type, e.g. "Boom"), and gives a brief description of typical processes. Typical saw mill design involves four physical levels: basement, machine floor (with some operator stations); operating floor (comprised of catwalks, offices and booths, and operator stations); and a saw-filers' "loft". Planer mills are generally physically separate buildings with a similar construction to the sawmill, but lacking a loft. Only three of the four lumber mills studied had planer mills. Typical construction (with the exception of the basement with a concrete floor) was wood floors (except cat walks, that were typically metal grating) and painted wood walls/ceilings. Noise Exposure Levels Three hundred and ninety four exposure measurements were made during the period October 1996 to August 1997. After eliminating samples of less than 4 hours duration, and those with equipment or handling problems, 343 full-shift dosimetry measurements remained from 286 participants (range, one to three observations per participant) in 61 different job categories. This was short of my goal, but still averaged 5.7 observations per job category (range 1 to 21). Table 6 shows summary noise statistics by mill. Table 6: Summary Statistics: Noise Dosimetry Survey, 1996-1997 M i l l ID # Dosimetry Mean L e q , (SD) Range Observations dBA dBA 1 74 92.9(601) 80.1 - 105.0 3 84 91.6(5.8) 79 .5- 105.0 4 101 90.9 (6.3) 70.4- 109.0 12 84 91.6(5.8) 78 .7- 106.0 A l l 343 91.7 (5.9) 70.4- 109.0 The mean sample time was 450 minutes (range 253 - 935, SD = 75, n= 242) except at mill 3, where 12-hour shifts were the norm. At this site the mean sample time was 523 (range 248 - 726, SD = 135, n=101) minutes. The overall mean L e q was 91.7 dBA, with standard deviation, 5.9 dBA. A cumulative frequency distribution of L e q is shown in Figure 3, and shows that this is a very highly exposed population. Eighty-eight percent of observations among these production and maintenance workers were above 85 dBA, the current BC exposure limit, and 60 percent were above 90 dBA. Of the 52 job-title categories that had at least 2 observations, only 4 had mean L e q ' s below 85 dBA (crane operator, machinist, tallyman and watchman). Twenty-eight jobs had means above 90 dBA, and 4 jobs -planerman, planer feeder, planer tilthoist and chipper operator - were at or above 100 dBA L e q for a full 42 shift. Of 42 participants working 12-hour shifts (identified as those with sample times of at least 600 min), 39 had L e q 's above the 12-hour adjusted exposure limit of 83.2 dBA. Figure 4: Mean Noise Exposure Levels By Lumber Mill Department, With 95% Confidence Intervals (n=343). Excluding hearing protectors, the primary noise control method evident was worker enclosure (i.e. control booths), which was reported used by 73 participants (26%). Figure 4 gives departmental mean L e q ' s and 95% confidence intervals. Inter-department variability was relatively large, with means 43 ranging from 97.2 d B A (planer mill) to 86.0 d B A (shipping), and these differences were statistically significant. Octave Band Analysis One hundred and thirty seven octave-band measurements were obtained. Table 7 gives the mean unweighted sound pressure levels (L p ' s ) and the mean A - and C-weighted total levels. C-weighting emphasizes low frequencies more than A-weighting. Two departments with similar mean A-weighted exposure levels but different C-weighted means have different underlying frequency spectra. For the departments By-products, Log processing, Yards, and Maintenance, the A - and C weighting schemes produce means that differ by more than 10 dB. These differences are shown graphically in Figure 5; the Sawmill and By-products departments have the highest exposures at low frequencies, while the Planer mill, and Sawmill and Sawfiling departments are highest in the high-frequency range. The most common spectral pattern was o f falling intensity with increasing frequency, but Sawfiling, Boom and Planer mi l l all peaked in middle to high frequencies. Differences between the departments were statistically significant at all frequencies from 32 H z - 8kHz. Table 7: Summary of Noise Exposure by Mean Sound Pressure Level (Lp) in Grouped Octave Bands. (n=137) Department Number of samples L p dB* (Unweighted) mean (SD) L p dB(C) mean (SD) L p dBA mean (SD) By-Products 4 107.3 (3.6) 104.0 (5.0) 91.8(11.4) Log Processing 17 99.1 (7.0) 91.9(5.9) 77.4 (9.1) Planer mill 18 98.0 (6.1) 96.5 (8.7) 93.4(11.0) Sawmill 50 97.9 (4.7) 96.2 (5.4) 90.4 (8.0) Yards 17 96.9 (7.9) 96.5 (9.9) 82.9(11.1) Boom 6 95.1 (7.2) 96.1 (7.8) 90.7 (3.5) Maintenance 5 94.3 (5.6) 90.8 (9.0) 77.6(12.6) Saw Filing 13 93.8 (8.7) 91.5 (8.2) 90.6 (8.9) Shipping 8 85.1 (7.2) 84.3 (7.6) 80.8 (8.0) A l l 137 96.8 (7.2) 94.6 (7.9) 87.3 (10.6) Departments *By descending order of unweighted L p Hearing Protector Device Usage Table 8 shows the percentage o f self-reported hearing-protector use derived from the post-dosimetry sample questionnaires in the 1996 - 1997 noise survey. Results are grouped by exposure level ( L e q ) on the day o f sampling. The table shows the number of individuals using H P D , the type worn, and the mean percentage o f the shift for which it was worn, among those using H P D . 44 250 500 1000 Octave Band (Hz) Figure 5: Noise Frequency Assessment. Octave Band Analysis by Lumber Mill Department Table 8: Self-Reported Hearing Protection Use Among 283* British Columbia Lumber Mill Workers, 1996-1997 Noise Exposure n Percentage of Mean Type of Hearing Protection Device Level Participants Percentage None Ear Ear Double Wearing HPD of Time n(%) Muffs Plugs Protection^ Worn n(%) n(%) n(%) <85 dBA 35 60 40.1 14 (40) 5(14) 16 (46) 0(0) 8 5 - 9 4 dBA 166 84 73.4 26(16) 41 (25) 97 (58) 2(1) > 95 dBA 82 99 94.5 1 24 (29) 55 (67) 2(2) Data for 3 subjects was missing. *Wears both Plugs and Muffs Figure 6 summarizes historical trends in HPD use for the past three decades for 13 of the 14 cohort mills for which data were available. Data from 30,459 annual hearing tests was used. The middle (85-94 dBA) exposure category was created by collapsing two 5 dBA-wide categories that showed similar trends. For the last year of audiometry data, 1996, the proportion reporting HPD use was very similar to the dosimetry questionnaire results. Interviews with senior workers from all mills indicated that before 1970 cotton batten was widely used, but that plugs and muffs more or less totally replaced the use of cotton from 1970 on. 45 Exposure Determinant Modeling Table 9 describes the final multiple regression model. The model, with job categories entered, had an R 2 of 0.58, indicating that the model explained 58 percent of variability in L e q (these R 2 values refer to the non-GEE model, as R 2 is not estimated by GEE). The model is of simple additive form, i.e. Leq = a + p]X] + P2X2 + + PkXk, where a is the intercept, P's are coefficient-estimates, and X's are predictor variables. There were 343 observations involving 286 participants. The %2 value was 464.7 with 27 degrees of freedom (P < 0.0001). The within-participant correlation was 0.07. The mean predicted value (Ay) was 91.7 dBA; minimum and maximum predicted values were 79.3 dBA and 103.2 dBA, respectively. Discussion The data from this survey of four British Columbia lumber mills demonstrated that the large majority of workers were exposed to high levels of noise in their work environment. Published studies in comparable mills show similar high levels of noise exposure (Table 10). The much smaller range of exposures they show was probably due to the much smaller number of jobs sampled and fewer samples taken overall. 46 Table 9: Determinants of Noise Exposure Regression Model. FuU-Shift Leq and Selected Determinants of Exposure. Class of Variable* Independent Variables (B's) dBA 95% CI Background 87.5 86.3, 88.8 Sawmill Site Site 12 (n=84) ref Site 3 (n=84) 2.4 1.1,3.6 Site 1 (n= 74) 2.4 1.1,3.7 Site 4 (n=101) 1.2 0.1,2.4 Department Log processing (n=27) 3.1 1.2, 5.0 Sawmill (n=106) 4.4 3.0, 5.7 Planer mill (n=42) 4.2 2.4, 6.1 Saw filing (n=42) 2.1 0.6, 3.6 Noise source Electric motors (n=10) 6.0 3.4, 8.7 Multiple blades (n=52) 1.4 0.1,2.8 Personal Enclosure* mean=25.7, SD=39.1 -0.07 -0.1, -0.04 Booth Material Wood booth ref Metal insulated (n=44) 0.3 -4.2, 4.7 Metal non-insulated (n=27) 6.9 2.2, 10.3 Jobs Planerman (n= 6) 9.0 5.5, 12.5 Chipper operator (n= 7) 8.7 5.6, 11.8 Planer feeder (tv=2) 7.8 2.3, 13.4 Tail sawyer (n=6) 6.8 3.5, 10.1 Small edger operator (n= 7) 6.7 1.2, 12.2 Tilt-hoist operator (n= 7) 6.5 3.3, 9.8 Drop-sort operator (n=ll) 5.2 2.7, 7.7 Resaw operator (n=4) 4.8 2.2, 7.3 Spotter (n=3) 4.6 -0.1,9.3 Grader (n= 11) 3.2 0.7, 5.7 Crane operator (n=8) -2.6 -5.6, 0.3 Tally man (n=10) -4.2 -6.8,-1.6 Machinist (n=6) -5.0 -8.2,-1.8 Watchman (n=2) -5.8 -11.2,-0.4 A l l variables are dichotomous except personal enclosure (mean and standard deviation are given). * Personal enclosure = % enclosure X % time spent in booth (has range of 0-100) CI: confidence interval; ref: reference category, n: number of observations where variable = 1 47 The problem of noise in lumber mills has long been recognized, and BC lumber mills were targeted in a 1980 initiative of the BC Workers' Compensation Board. They created a 9-person "noise-control" section; it laid out a progressive multi-step cooperative plan with industry that they hoped would reduce over-exposure of workers (at the then 90 dBA limit) from the 1981 level of 60 percent to 10 percent by the year 1986 (Tupper, 1981). Unfortunately, the noise-control section and plan were terminated by 1986 (S. Eaton, WCB Engineering Branch, personal communication). This noise survey shows that the percentage of job categories exposed over 90 dBA was still approximately 60 percent in 1997, but that concurrently 90 percent of job categories were exposed above the BC regulatory 8-hour limit of 85 dBA. Some benefits of the noise-control plan and earlier efforts remain, however; planing machines at all sites visited were entirely enclosed, and 26 percent of workers in the four mills surveyed worked in booths, many of these being acoustically treated. Unfortunately, the effectiveness of the control measures in place was often compromised, for example by planer-enclosure doors being left open to allow observation of the machine or to permit "easy" access. Planer-mill jobs (Planer feeder, Planerman and Tilt Hoist Operator) remain among the highest exposed jobs in the lumber mill. Table 10: Lumber mill Noise Exposure Levels - Previously Published Studies Author (year) M i l l Type Noise levels, dBA Range of T W A ' s Comments Dost (1974) Sawmill 87 -104 144 measurements in 8 sawmill jobs, 80 Planermill 93 - 104 measurements in 4 planer mills jobs; West coast US Sawmills comparable to B C Fairfax (1989) Sawmill <85 - 102 Measurements of 6 sawmill and 8 planer mill jobs, Planer mill 91 - 107+ representing levels before and after noise controls implemented. Tubbs(1991) Sawmill 83 - 100 Measurements of 15 sawmill and 7 planer mill jobs, Planer mill 88 -102 representing levels before and after noise controls ininlcmmbid. 'Time Weighted Average ^Combination of dosimetry and sound level measurements Examination of the frequency spectra showed that there was substantial variation between different lumber-mill departments. Including frequency in exposure assessments for analyses of cardiovascular effects is desirable. To investigate the effect of frequency on cardiovascular outcomes, noise levels can be re-weighted using the frequency data gathered, and exposure-response relationships tested with exposure values weighted to emphasize either high or low components in separate analyses. Data from the post-sampling interviews shows that hearing protectors were worn by virtually all participants at all sampling times when exposures were at or above 95 dBA for a shift. The prevalence of use dropped to 84 percent in those exposed between 85 and 94 dBA, and was only 53 percent in those exposed under 85 dBA. Earplugs were more likely to be used than earmuffs, by a ratio of about 2 to 1. Only a very small percentage used "double protection" (plugs and muffs simultaneously). 48 W C B HPD-usage data was available for 13 of the 14 mills, and 67 percent o f the mill/year combinations between 1979 and 1996. Evidence from hearing-test records suggested that H P D use had been consistently high (above 90 percent) in the highest exposure group since at least 1979. Use in the mid-exposure group had increased steadily from about 75 percent to its 1997 level. H P D use in the under 85 d B A exposure category had been more or less steady at 55 - 65 percent. Concern that there may have been a positive reporting bias in the hearing test questionnaire (that was compliance-driven) was mitigated by the similarity in proportions reporting H P D use in the dosimetry study (that was research-driven). The relatively low percentage of time worn reported in the under-85 d B A exposure category may reflect the wearing o f H P D ' s "when required" during the shift and not necessarily inadequate protection. With respect to historical H P D use, senior mil l workers reported that prior to 1970 cotton batten was used widely, but it is known that cotton alone has little protective effect (Berger, 1986). These findings are supported by U S data that showed an increase in H P D use within the lumber industry (SIC code =24) from 7.97 percent in 1972-74 to 67 percent in 1989 (Davis and Sieber, 1998). The reliance on personal protective equipment over noise-control engineering is consistent with the observations of the more general situation made by Franks (1998). A predictive model was constructed that explained 58 percent of variability in L e q . This was comparable with other published models of exposure determinants (Burstyn and Teschke, 1999), and indicates reasonable predictive power. The model indicated that the coastal mills (mills 12 and 3) were on average 1 to 2 d B A quieter than the interior mills. There are many potential explanations for this difference, such as production levels or construction design. Only two o f more than 50 potential noise sources were retained in the final model: "electric motors" and "multi-bladed saws". The ability o f the model to predict nearly 60 percent of the variability in L e q primarily from department and job category, for which I had good historical cohort records, was encouraging in terms o f predictive modeling of historical exposure levels. The lumber-mill environment was acoustically complex, with numerous noise sources, and mill-specific room geometry and surface treatments. I had thought that worker exposures might be better predicted by nearby noise sources that may not be obviously correlated with the job title. For example an "edger" operator may be at some distance from the edger-saw that he or she operates, but directly adjacent to another machine, such as the head-rig. These noise sources would be expected to vary within job category because of differences in physical parameters (size, placement) and technological characteristics (blade type, drive type). These noise-source differences might be expected to be even greater between mills. Because job and department categories are better predictors despite this heterogeneity, confidence was increased in using them as the basis for assigning exposure levels. Within subject correlation was very low (0.07), perhaps in part because 30 percent o f subjects with repeated measures had changed job category between observations. 49 Personal enclosures were found to provide an approximate 7 d B A reduction in exposure, i f they were 100 percent enclosed, and i f the operator reported use for the whole shift. Levels o f predicted protection varied with booth construction type; non-insulated metal enclosures (typical of a vehicle cab) offered virtually no protection, as the coefficient for "metal, non-insulated" construction type (+6.9 d B A ) effectively canceled out the reduction due to being in an enclosure (-7.0 d B A for 100 percent protection). Summary The large majority o f workers at four British Columbia mills tested were highly exposed to noise. This was despite focused government and industry efforts to reduce exposure levels, and with one-quarter o f workers surveyed being in personal enclosures. These four mills were considered representative of current and typical work conditions within the other 10 participating lumber mills, and could be reasonably generalized to comparable mills elsewhere in the province. Noise frequency data suggests that there was sufficient contrast between spectra o f various mi l l departments to enable an analysis of the effects of frequency on health outcome. Patterns of hearing protector use between 1979 and 1996, for individuals exposed at various levels of noise, were obtained from a survey of hearing protector use by dosimetry subjects, and from W C B audiometry records for all mills. Interviews of knowledgeable workers indicated that widespread use of hearing protectors did not begin until around 1970. This knowledge o f hearing protector utilization can be used to refine future exposure estimates of noise exposure and hopefully reduce misclassification of exposure. A determinants-of-exposure model had good predictive power, suggesting that modeling is a good basis for estimating exposure levels of cohort subjects, for combinations o f job title, time period and mil l for which dosimetry measurements were not made. However, despite the large number o f samples obtained, there were still job categories that were not adequately represented in sampling. The current model did not contain a time-parameter, as all the observations were made within a short time period (approximately one year). Because the model was based on data from only four mills, it could not be used to predict externally - which is required i f the model is to predict exposures for all cohort mills - without adding data from all cohort mills, or replacing the mi l l variable with other predictors. Even with improvements to the model, some significant events - such as the enclosure of planing machines, and the effects of the use o f hearing protection - could still not be adequately modeled due to lack of available measurement data. Arithmetic modifiers to exposure predictions developed based on other data sources wi l l be required. 50 However, the model suggested that mill, department, job title, and engineering controls (specifically personal enclosures) were important determinants (predictors) of noise level, while the identity of specific noise sources were less so. This makes adding observations from different time periods and additional mill-sites a simpler proposition, as the relatively hard-to-obtain noise-source determinant data does not appear to be necessary in developing a model of reasonable utility. Chapter 5 (next) describes the 2 n d phase of the study that involved obtaining noise exposure data and determinants data for additional cohort mills and for earlier time periods, and the development of a more comprehensive model suitable for retrospective exposure assessment. 51 Chapter 5. Retrospective Exposure Assessment of Noise in BC Lumber Mills Introduction This 2 n d phase of the study involved the assessment of historical exposure to noise in the lumber mill cohort. An entry criterion for the cohort was one year of employment between 1950 and 1995, and consequently some work histories began early in the 20th century. Although complete personal work-history data was available for most subjects, no exposure data was known to exist for other than the most recent three decades, and even then was incomplete in terms of time period and mills covered. This absence of exposure data is a central difficulty in virtually all retrospective cohort studies, and "poses the greatest challenge" in such studies (Checkoway and Eisen, 1998). Several strategies exist for retrospective exposure assessment (Stewart and Herrick, 1991). Traditionally these have included qualitative approaches such as ever/never exposed, or duration of employment (Merlo et al, 1991), and semi-quantitative approaches, such as ordinal ranking of exposed jobs (Checkoway et al, 1993). More recently, a greater emphasis has been placed on quantitative exposure assessment, because of the benefits it offers. Primarily, it is likely to provide better exposure estimates, thus reducing the likelihood of misclassification. This in turn reduces attenuation of exposure-response relationships, which is important when associations are expected to be weak such as with noise and cardiovascular disease. Quantitative exposure assessment also permits the development of more complex exposure metrics, and allows more flexibility in the creation of exposure groups and exposure cut-points. This may help when underlying mechanisms and effect-levels are poorly understood (Stewart et al, 1996). The approach allows the identification of exposure levels associated with disease, which greatly assists in risk assessment and standard setting. Further benefit is gained from working in a familiar scale, and because the approach forces a more careful assessment of relative exposure levels. Unfortunately, there is no standardized approach to quantitative retrospective exposure assessment, and each study requires a customized approach dependent on the type of information available (Dosemeci etal, 1993; Stewart, 1999). Seixas and Checkoway (1995) however, recommended a 6-step framework for industry-specific retrospective exposure assessment5, which was adapted for this study. For the key step of assessing exposure, a predictive statistical model was developed; Seixas and Checkoway suggest this is the most comprehensive approach, and one that makes the best use of available 5 The six steps were: (1) Gather and characterize all relevant exposure-related information; (2) Evaluate data for errors including bias and precision; (3) Define exposure data matrix for linkage to study subjects; (4) Estimate exposure levels; (5) summarize exposure metrics for each subject; (6) Estimate exposure-response relations. 52 data. It is also suited to the unbalanced nature of the observational data (Burstyn and Teschke, 1999). Similar determinants of exposure models have been successfully developed for the retrospective studies of ethylene oxide (Hornung et al., 1994), granite dust (Eisen et al, 1984) and other occupational toxins (Burstyn and Teschke, 1999). Determinants of exposure models have been developed for noise (Nieuwenhuijsen et al, 1996 and Greenspan et al, 1995) but these were exploratory studies that only served to characterize exposure determinants. The current study is to my knowledge the first to use predictive modeling to assess historical exposure to noise. A modeling approach requires sufficient measurement data and corresponding determinants data, which were available for this study. The preliminary modeling of data collected for 4 of the participating lumber mills in this study (Chapter 4) suggested that mill, department, job title, and engineering controls were important predictors of noise level, while the identity of specific noise sources was less so. As these predictors can be ascertained for historical data more readily than can noise-source data, we supplemented the recent exposure data collected specifically for this study, with historical exposure data obtained from the local regulatory agency (Workers' Compensation Board or WCB) and from data owned by the mills. The determinants of exposure model from Chapter 4 was then rebuilt based on data from as many of the participating mills as possible, and from as broad a time span as possible. The validity of the resulting model was tested with a subset of data held back at the outset for this purpose. Other exposure-related data collected from participating mills, the WCB's Hearing Conservation Branch and from the scientific literature, was used to develop adjustment factors for determinants of exposure that could not be easily dealt with in the model. This included use of hearing protection, time-trends prior to 1980, the effect of the addition of enclosures around planing machines, and the effect of low-frequency noise. The model and the adjustment factors were then used to generate exposure estimates for all mill/job/department/time period combinations represented in cohort work-histories, based on historical predictor values. The objectives of this phase were to: (a) use noise exposure data from research (the UBC samples), mills, and government regulatory sources (WCB) to develop a predictive determinants of exposure model; (b) validate the model; (c) use historical determinant data to generate exposure estimates for all unique mill/job/department/time combinations; (d) use secondary exposure data (hearing protector usage, frequency spectra, engineering control installation data) to adjust model estimates; and (e) calculate cumulative exposure metrics for all cohort subjects. 53 Methods Exposure Data Acquisition Noise exposure data for BC lumber mills was acquired from three sources: (i) dosimetry measurements made by the author during 1996 and 1997 (described in Chapter 3, referred to as "UBC" data); (ii) an exposure database maintained by the Workers' Compensation Board (WCB) (containing measurements made by WCB noise-control and hygiene officers, but also data supplied by the mills and reviewed by WCB officers); (iii) measurement data volunteered by management of 5 of the cohort mills during site-visits (measurements made by professional noise consultants, or by company personnel, referred to as "mill" data). Exposure determinant data was gathered for two purposes: (1) to combine with noise dosimetry data in developing the predictive model (where exposure measurements existed) and (2) for predicting exposure levels where measurements were missing. Data regarding determinants of noise exposure was collected concurrently with noise measurements for those samples obtained by the author. To obtain determinant data for WCB and mill exposure measurements, and for all historical determinant data, visits were made to all participating mills to collect data by direct observation of jobs, site walk-throughs, from mill records, and through interviews with knowledgeable workers. Interviewees were selected on the basis of seniority at the mill and their broad knowledge of the mill's operations (typically foremen and senior maintenance workers). Knowledgeable retirees were sought out where they could be located. Supplemental determinant data was obtained from secondary sources, such as forest products directories. Novel determinants such as degree of mill mechanization and density of machinery were created from primary determinants. Exposure Data Preparation Job titles from noise exposure data were categorized into 81 standardized job codes (incorporating those used in the 1st phase exposure assessment, See Chapter 4). These codes were originally developed in consultation with an expert panel of 6 senior workers, who grouped job-titles based on similarity of performed task (Ostry, 1999). The original job coding was specific to a study of job-strain in the cohort; for this study they were re-categorized where necessary to make them appropriate to noise exposure. Noise measurements for which a standardized job-code could not be assigned were excluded. Because data obtained directly from lumber mills might have previously been reported to the WCB, mill and WCB observations were compared and duplicate entries (based on date and job-title) were excluded. Job-codes were further grouped by task and process into 11 "process groups" (Table 11). The full exposure dataset was only used for long-term time trend analyses. For exposure 54 modeling purposes, only dosimetry observations from cohort-mills were used. Further, this modeling dataset was randomly divided into two subsets (using a random key generated by the ST ATA uniform function); one part was used for model building (the "estimating" dataset) and the other for validation (the "validation" dataset). Figure 7 illustrates the flow of data during the project. Table 11: Process Groups Descriptions, Examples of Typical Jobs Found Within Process Groups, and Typical Administrative Department. Process Group Description Example Jobs Administrative Dep'ts Sawmill - primary processing Front-end of sawmill, primary breakdown of logs Cut-off Saw Operator Head Sawyer Sawmill M i l l - non-sawing Directs lumber flow, manual Lumber Straightener Sawmill sorting Unscrambler Planer mill Sawing Operates secondary band and circular saws Resaw Operator Trimsaw Operator Sawmill Planer mill Lumber Sorting Operates mechanical lumber sorters Stacker Operator J-Bar Sorter Operator Sawmill Planer mill Packaging Operates mechanical packaging machines Package Press Operator Stenciller Shipping Outdoor Working in yards or boom, no vehicles Offbearer Slipman Boom Yard Utility Cleaning, labouring, by-products processing Labourer Cleanup By-Products Sawmill/planer mill Maintenance Skilled tradesmen Millwright Mechanic Maintenance Sawfiling Specialty saw-maintenance trade Sawfiler Grinderman Sawfiling Office Administrative and clerical Clerk Manager/Superintendent Administration Vehicle Operator Operating land or water vehicles (mostly large diesels). Boat Operator Forklift Operator Boom Yard Regression Analysis Descriptive Statistics Mean, median, standard deviation, range and frequency distributions were obtained for all variables. For variables identified for possible inclusion in the predictive model, frequency distributions were characterized, and bi-variate relationships between independent variables were examined. Among variables that were strongly correlated (Pearson r for continuous variables, or Cramer's V for categorical, > 0.6), one of the pair (the one with the lowest univariate R 2, or least informative with respect to prediction) was excluded from further analysis. In this way, five variables were dropped (region, multiple bladed saw, vehicle operator, vehicle cab, number of employees). Variables retained for analysis are shown in Table 12. Because of the large number of observations, all independent variables were initially offered to the model, even if their individual explanatory power was low. Finally, for the purposes of modeling, job titles with only one or two measurements (truck driver, bucker, first aid, clerk, pipe fitter, timber deckman) were combined with similar jobs. 55 Mill Visits Site walk-throughs, job observation, frequency analyses worker interviews, Collection of mill records, schematics. Current determinants of exposure data Exposure Data Sources UBC Data, n=360 Mill Data, n=1,078 WCB Data, n=12,612 Develop independent adjustment factors: - Hearing Protection Devices - Planer Enclosure - Noise Frequency Exposure Measurement Data, includes cohort and non-cohort, dosimetry and grab samples (n=14, 050) Document values of Historical Determinants of Exposure for each "Exposure Key" 12 Historical determinants data coded for all unique mill/job/time combinations in exposure matrix Modeling Process Cohort dosimetry data subset (N=1901) Modeling Data subset (n=1521): Exposure model developed with manual forward step multiple regression analysis Validation Data subset (n=380): Model predictions compared to observed values of sample of original data "held-back" Temporal trend analyses on full exposure dataset (n= 14,050) Adjustment Factors Predictive Model See also Figure 9 Figure 7: Retrospective Exposure Assessment: Process diagram, and Exposure Data Flow 56 Table 12: Summary of Predictor Variables Presented in Determinants of Exposure Modeling Variable Type # of Categories (or range, units) Description Comments Job Title cat 70 Subjects were assigned to one of 70 standardized job titles^ . Jobs with fewer than 3 observations were re-categorized to a similar job title. Process Group cat 11 Standardized job titles were grouped based on task, process or location similarity.' See Table 11 Department cat 14 Subjects' administrative department. Nested in job title: only 5 jobs (offbearer, grader crane operator, cleanup and foreman) and 2 process groups (sorting, and packaging) were distributed across more than one department Job location di 2 Job location - indoor or outdoor Mill cat 13 unique mill identifier Mill area di 2 Work area > 1000 M 2 Work area size were estimated for sawmill, planermill, and sorters. Area was nested in the "indoor" job location Mill productivity cont 35 - 365 MMBM Mill's annual production level (millions of board-feet measure) Intensity cont 0.003 - 3.3 -MMBM/M 2 Indicator of the concentration of equipment in mill Calculated as the annual production level per M 2 sawmill area. Mechanization cont 0.09- 1.7 MMBM/employee MMBM / number of employees Calculated as the annual production level per employee Booth di 2 Subject spent majority of shift in control booth. "Booth" was classified as any protective structure regardless of construction materials, or evidence of soundproofing. Mill age cont 1920-1980 Year of construction of mill of current mill Year di 2 Sample taken after 1983 Sample source cat 3 Sample data obtained by (1) WCB, (2) UBC or (3) Management cat^ categorical, di=dichotomous, cont=continuous ^he modeling file has 11 fewer job-titles represented than the full file, but 95% of cohort jobs are still represented. Model Building A fixed-effect model was estimated by manual forward stepwise regression, considering P-values (using a retention value of P < 0.1), changes in adjusted-R2, partial F-tests, and the effect of variable addition on other coefficients (STATA 7.0, command regress). Variables that were dropped because of low statistical significance were re-entered after a final model was achieved to test if they had "regained" significance. Categorical variables were entered as sets of indicator variables, and were entered or removed as a group (with the exception of job title, which was nested in process group). Reference categories were selected on the basis of largest number of observations. Interactions that were considered, a priori, to be plausible and of predictive value, were examined in the model. These included job location X booth6 and process group X booth. Mill area was nested 6 where X infers a multiplicative interaction between the two variables 57 within "indoor" job location, and specific departments of a priori interest were nested within specific job titles. Model fit and ordinary least squares modeling assumptions were tested with residual diagnostics, Cook's D statistic, and an analysis of outliers. Following model validation (see below), the model was estimated again using the full dataset, combining the "estimating" and "prediction" datasets, and this model used for historical predictions. Model Validation The model's predictive ability was evaluated by applying the model to the subset of data held-off prior to model building (the "validation" dataset). Individual differences between the observed noise level iyi) and predicted value ("y,) were estimated. Bias, precision and accuracy were calculated as follows using the method of Hornung et al, (1994): eqn. 1) Bias = j ^ ' ^ eqn. 2) Precision = v,=i n0-\ eqn. 3) Accuracy = ^ (Precision)2 +(Bias)2 where n 0 = number of observations. Adjustment Factors Not Included in the Determinants Model Effects of several determinants of exposure could not be included in the determinants model, due to lack of data: the effect of the enclosure of planer machines; use of hearing protection devices; noise frequency spectra; and temporal effects prior to 1980. These were estimated independently, using a variety of methods, as follows. Planer Enclosures: Limited exposure data from cohort mills was available to examine the effect of the installation of planer enclosures, as most had installed enclosures prior to the collection of samples used in this exposure assessment. Therefore, data from two previously published noise surveys at sites similar to the cohort mills were used to estimate the effect of enclosures on planer jobs. Temporal Effects, 1970 to 1980: While year-of-sample from 1980 onward was included in the determinants model, temporal effects between 1970 and 1980 were investigated using the entire exposure dataset (n=14,050) in separate multiple regression analyses on a reduced set of independent variables that was available for all observations. This included mill, data source, sample type (i.e. dosimetry vs. sound 58 pressure level, or Leq vs. Lp), and process group. A second regression analysis using only cohort data (n= 3,776) was run, adjusted for the same factors plus the presence of booth. Hearing Protector Devices (HPD): The use of hearing protection devices (HPD) is not reflected in noise measurements, which are made at the shoulder. Because personal data on HPD use was not available, an adjustment factor was estimated as the product of (i) prevalence of HPD use, and (ii) a mean protection factor of HPD's (in dBA). Prevalence of HPD use in cohort mills was known for the years 1979 - 1995 (see Chapter 3, and Figure 6) and linear growth in prevalence from zero use in 1970 to 1979 levels was assumed. A weighted-mean protection factor for HPD use was estimated from a survey of real-world attenuation data (Berger et al., 1996), taking into account the difference in protection between circumaural (muff) and insert (plug) forms of protection, and weighting muff vs. plug use based on 1996/97 usage patterns. Adjustment factors were then calculated for five approximately-equal time periods between 1970 and 1995, and for three exposure-level groups, <85.0 dBA, 85.0-94.9 dBA and >95.0 dBA. Frequency Spectra: In order to examine low-frequency effects, C-weighted exposure levels were estimated. Mean frequency spectra were obtained previously (see Chapter 3, and Figure 5). By assuming that an average spectral "profile" of a process group represented all jobs in that group, an adjustment factor for each process group was calculated as the difference between the A-weighted and C-weighted level for that group (each derived by the standard method7). A-weighted job-title predictions were then converted to C-weighted predicted levels using the adjustment constant for their respective process group. No frequency spectra data was available for the process group "Office", so this group was given an adjustment constant equal to the mean of all other process groups. Estimating Historical Exposure Levels Preparing the Work History Data Job title data from work history data abstracted from mill personnel records was coded to the 81 standardized job titles matching those used in exposure modeling. Records with missing job title information was handled according to the logic shown in Figure 8. This logic assumed that the average exposure for the department gave the best estimate if department was known. If it was not known, then For the process group PG, A-weighted Leq were derived using the standard method (also C-weighted with ^•32/fe./>G~39-4 L 6 3 H Z PQ-26.2 LIKHZ PC-1.1 10 1 0 +10 1 0 +....+ 10 1 0 appropriate scale): APG =101ogX 59 the missing job-title was assumed to be unskilled labour unless it was the first job held by a skilled tradesman (it was then assumed to be that person's trade). It was assumed to be a continuation of a previously held job (i.e. a coding omission on the part of the mill), if that previous job was held within the prior 12 months. Where individuals held multiple job titles simultaneously, and for more than 30 days, the work period was divided equally by the number of jobs held and those job titles assigned to the work period in random order. Where multiple jobs were held for periods of less than 30 days, one job was randomly retained and the rest ignored. After correcting missing job titles and concurrent multiple jobs, all remaining non-lumber mill work-history records were excluded. Is the department specified for this work history record1? Yes Assign department average No Is this the l a record or only work history record for the subject? Yes 1 No If first of 2 or more records and if next process group is utility, maintenance, or sawfiling., assign to next job title to this work history record. Else, assign labouring job title Was the previous job held within the past 12 months? Yes Assign exposures based on previous job and department. No Assign job mill/time period average (weighted by person-years) Figure 8: Logic Diagram: Assigning Values to Work History Records with Missing Job-Title or Missing Job Title and Department Building the Exposure Matrix Figure 9 illustrates the cohort database structure. A file containing vital status and demographic data for each subject (the demographic file), and a file containing work-history records had been previously abstracted for earlier studies in the same cohort. The demographic file contained one record per subject, and was linked in a one-to-many relationship with the work history file. The work history file in turn was linked in a many-to-one relationship to the exposure data matrix8. The exposure matrix thus 8 This file structure was adopted in preparation for input to the analytical software, PC Life Tab/e Analysis System, to be used in the epidemiological analyses, see following chapter. From Table 8 Document values of Historical Determinants of Exposure for each "Exposure Key" Adjustment Factors Predictive Model Exposure data matrix is populated from determinant data collected from mills Determinant data is fed into the predictive model Exposure levels are predicted for each exposure key Exposure Key M i l l P G Job determinant 1 .determinant 2 determinant 70 start date end date Exposure Level record 1 10200 1. 02 0 0 01/01/1951 30/06/1978 94.5 record 2 10020 1 W-.. 02 0 1 01/07/1978 30/06/1993 91.3 record 3 10020 1 02 0 0 01/07/1983 12/31/1997 87.5 record 4 10120 1 10 21 0 1 01/01/1957 30/06/1978 100.3 record 5 10120 10 0 1 01/07/1978 30/06/1993 98.5 0 record 3808 161121 16 1-1'"-, 1 1 01/07/1978 30/06/1993 86.3 record 3809 161121 16 11 0 0 01/07/1983 12/31/1997 83.9 Exposure Data Matrix (n=3,809) Work History File (N=244,875) Subject start date end date Mi l l , . Exposure Key record 1 100011 | .03/22/1951 12/31/1965 1 10200 record 2 100011 i i01/01/1966 08/16/1985 1 10120 record 3 100012 1 05/05/1975 12/31/1975 1 10200 record 4 100012 01/01/1976 06/08/1976 1 11121 record 5 100012 18/19/1977 16/12/1980 1 10930 record 244,875 165959 05/05/1975 12/31/1975 16 10200 A Demographic File (N=27,499) Cohort Mill Personnel Records O H M ) subject sex vital status birthdate person years begin date last observed date of death Cause of Death record 1 100009 1 1 10291950 01131977 12311995 record 2 100010 1 1 08031932 09221959 12311995 record 3 100011 1 2 06041929 03221951 08161985 08161985 410 record 27,499 165959 I 1 06041929 03221951 08161985 08161985 410 Figure 9: Retrospective Exposure Assessment: Exposure Data Matrix Construction, With Sample Data Showing Relations Between Key Files, and Use of Model and Adjustment Factors to Assign Exp Levels. 61 was conceptually a traditional Job XTime-period construct (Checkoway and Eisen, 1998), but operationally each Job-Time combination occupied a row in the file, and in addition to an exposure level, contained all the necessary determinant variable values for that combination of job and time period. The exposure data matrix was constructed as follows. The work history file (n=244,875) contained 2,631 unique mill/job-title/department combinations. After pooling combinations that shared identical exposure-determinant profiles (e.g. identical values for all the determinants identified by the modeling process), this was reduced to 982 unique "exposure-keys". Exposure keys were used to link work history records to the exposure data matrix. Because the values of exposure determinants also varied with time, a total of 3,809 exposure-key/time period combinations were required to describe the cohort's entire work history "experience". The exposure data matrix thus consists of 3,809 exposure-key/time-period records (rows), and approximately 70 variables (columns) representing each of the required determinants of exposure, plus fields containing start and end dates of each time-period and a calculated exposure value. Matrix determinant values were populated with the historical determinant data gathered during mill visits. Estimating Exposure Metrics The regression equation output from the modeling phase was then used to estimate exposure values for the exposure data matrix. Three exposure values were calculated for every exposure key/time-period combination: • Leq,A - An estimate of a time-weighted average full shift exposure, A-weighted. • Leq,A, HPD - An estimate of a time-weighted average full shift exposure, adjusted for hearing protector use, A-weighted. • Leq,c - An estimate of a time-weighted average full shift exposure, C-weighted. Department and Mill Average Levels Department and mill average levels were assigned to work-history records that were missing job or department data. Average noise levels for departments and for mills were calculated for each mill/department/time-period combination and mill/time-period combination. Averages were weighted by the proportionate cumulative person days of each job within the mill/department and mill, respectively. Mill/department/time-period averages were then assigned to work history records for which department, but not job title, was known. Mill/time-period averages were assigned to work history records where neither department nor job title was known (Figure 8). 62 Mills and Jobs for Which no Exposure Measurements Were Obtained No measurement data was obtained for Mill 8 and so no regression coefficient was estimated for this mill in the model. Predictions for Mill 8 were therefore run with the mill variable set to mill 10, which was judged similar with respect to physical size, age of mill, production levels, geographic location, technology and timing of technology change. No measurement data was obtained for the jobs Bucker, Dogger, Setter and Locomotive Operator. These jobs were assigned codes of similar jobs: Jobs Dogger and Setter were set to Head Sawyer, and Locomotive Operator was allowed to default to the "Vehicle operator" process group mean. Because of its unique job characteristics, Bucker, an outdoor job whose primary noise source is chain saw operation, was allowed to default to its process group (sawing) mean, then adjusted by + 6dBA, based on the high reported noise levels expected for this job (Schmidek and Carpenter, 1974). Results Overall BC Lumber Mill Data - Including Non-Cohort Data Sixteen thousand, seven hundred and three exposure observations were initially acquired. Observations with: (1) measurement errors; (2) missing job title or non-lumber industry job - or that were: (3) duplicate entries; (4) area (stationary) measurements; or (5) "experimental' observations (e.g. where the observation was evaluating a control intervention), were excluded from analyses. The majority of observations were from the WCB database (90 percent) and from non-cohort mills (73 percent) (Table 13). Figures 10 and 11 show the number of samples for cohort and non-cohort mills by year, source and sample type. Table 13: Exposure Data Summary. By Source, Sample Type, Mill Type Data Time Period Cohort Non-Cohort Total source Covered Samples L p Lea L p L e a WCB 1970- 1991 1014 1324 6078 4196 12,612 Mill 1970 - 1997 859 219 0 0 1078 UBC* 1996 - 1997 0 360 0 0 360 All 1970 - 1997 1873 1903 6078 4196 14,050 "Data collected by author, See Chapter 3 63 With respect to data for cohort mills, there were significant contributions for the years 1975, 1980, 1981, 1984, and 1996, 1997. Data collected prior to 1980 were primarily L p measurements (only 16 L e q measurements were made in cohort mills prior to 1980, and only 53 in non-cohort mills). Table 14 shows mean noise exposure levels by data type, data source, sampling method and mill. All cohort mills contributed data except Mill 8. There was an average of 290 samples from each of the other 13 cohort mills. Although there was little overall difference in mean levels between L p and L e q measurements, larger differences appeared within mills. The differences may reflect distributions of jobs sampled or differences in sampling strategy. Sample source was entered in the regression model to adjust for any possible effect. There was no obvious trend in the direction of difference between L p and L e q measurements. C o h o r t M i l l s 700 j 1 j "1 : _ 600 C "2 Figure 10: Noise Exposure Data, Cohort Mills: Sample Frequency by Year, Measurement Type, and Data Source. L e q : noise dosimetry, L p : grab sample 64 Figure 11: Noise Exposure Data, Non-Cohort Mills: Sample Frequency by Year, Measurement Type, and Data Source. Le q: noise dosimetry, Lp: grab sample Dosimetric data was considered more likely to be representative of true exposure levels than sound pressure readings, and less likely to be susceptible to errors of bias and precision in WCB and mill-volunteered data, therefore only dosimetry samples were used in model building. Because of lack of determinant data for non-cohort mills, and concerns of generalizability of non-cohort data to cohort mills, exposure modeling was further limited to data from only cohort mills. There were 1903 dosimetric measurements available for cohort mills; after 2 samples were excluded because of missing predictor data, 1,901 observations were used for modeling (Figure 7). This modeling dataset was randomly divided into two subsets: 1521 (80 percent) observations were used for model building (the "estimating" dataset) and 380 observations (20 percent) observations for validation (the "validation" dataset). Table 15 compares descriptive statistics for all data, cohort-dosimetry data, and the 2 modeling fractions ("estimating" and "validation"). Although the modeling dataset has a slightly lower mean and standard deviation than the whole dataset (and has lost the extreme values), it and the two sub-files are reasonably comparable. 65 Table 14: Noise Exposure Mean, (SD) and Frequency, by Source, Sampling Method and Mill. Sample Source Mill* WCB Management UBC Total All Lea L„ Lea L n Lea L„ Lea L„ (Ln+Lea) 1 95.7(3.3) 29 78.2 (7.3) 8 92.9 (5.9) 78 93.6 (5.5) 107 78.2 (7.3) 8 92.6 (6.8) 115 2 89.9 (8.4) 174 89.7(9.3) 119 90.0 (5.0) 50 87.7 (8.6) 50 - 89.9 (7.7) 224 89.1(9.1) 169 89.6 (8.4) 393 3 - 97.0 (2.3) 6 90.1 (4.9) 44 89.9 (8.2) 16 92.0 (5.3) 88 91.4 (5.2) 132 91.9 (7.7) 22 91.4 (5.6) 154 4 91.4 (6.5) 128 89.6 (7.0) 19 - - 90.9 (6.2) 104 91.2(6.4) 232 89.6 (7.0) 19 91.1 (6.4) 251 5 6 95.3 (7.5) 86 89.2 (9.4) 81 88.2(11.0) 236 - - -95.3 (7.5) 86 89.2 (9.4) 81 88.2(11.0) 236 95.3 (7.5) 86 88.4(10.6) 317 7 93.1 (5.9) 118 94.5 (7.8) 58 - - - 93.1(5.9) 118 94.5 (7.8) 58 93.5 (6.6) 176 9 91.8(7.1) 104 94.0 (7.1) 222 - - - 91.8(7.1) 104 94.0 (7.1) 222 93.3 (7.2) 326 10 89.5 (9.7) 71 96.4 (8.1) 99 - - - 89.5 (9.7) 71 96.4(8.1) 99 93.5 (9.4) 170 11 94.6 (6.0) 155 94.3 (6.0) 207 94.2 (7.0) 85 96.0 (7.2) 217 - 94.5 (6.4) 240 95.2 (6.7) 424 94.9 (6.6) 664 12 90.3 (9.4) 296 86.5(10.2) 26 - 90.7 (9.8) 472 91.5(5.8) 90 90.5 (8.7) 386 90.5 (9.9) 498 90.5 (9.4) 884 13 90.3 (7.0) 32 91.4 (7.1) 14 93.6 (6.3) 40 88.6(8.9) 104 - 92.2 (6.8) 72 88.9(8.7) 118 90.2 (8.2) 190 14 95.2 7.5) 50 - - - - 95.2 (7.5) 50 - 95.2 (7.5) 50 Cohort 91.7(7.6) 1324 92.0 (9.0) 1014 92.3 (6.4) 219 91.6 (9.4) 859 91.7(5.9) 360 91.8(7.6) 1903 91.8(9.2) 1873 91.8(8.4) 3776 Non-Cohort 93.2 (8.1) 4196 91.9(9.4) 6078 93.2 (8.1) 4196 91.9(9.4) 6078 92.4 (9.0) 10274 Total 92.9 (8.2) 5520 91.9(9.4) 7092 92.3 (6.4) 219 91.6 (9.4) 859 91.7 (5.9) 360 92.8 (8.0) 6099 91.9 (9.4) 7951 92.3 (8.8) 14050 No exposure data was available for M i l l 8. Sampling method: Leq: dosimetry, Lp, sound pressure level measurement Table 15: Descriptive Statistics for noise measurement dataset and sub-sets A l l Modeling dataset Estimating dataset Validation dataset n 14,050 1901 1521 380 Mean (dBA) 92.3 91.8 91.7 92.4 SD (dBA) 8.82 7.5 7.6 7.2 Min - Max (dBA) 50-140 60-115 60-115 60-109 66 Table 16: Mean Noise Level Standard Deviation and Number of Samples, by Year and Mill Year of Sample Mill* 1972 1980 1981 1982 1983 1984 1985 1996 1997 Total 1 95.7 (3.3) 29 91.7 (5.7) 128 92.5 (5.6) 157 2 89.2 88.1 (7.2) (9.3) 54 84 95.2 (5.0) 36 89.9 (8.4) 174 3 91.1 (5.0) 93 92.0 (5.9) 39 91.4 (5.2) 132 4 91.7 91.0 90.9 91.0 91.2 (5.0) 70 (8.1) (5.7) 58 67 (7.3) 37 (6.4) 232 5 93.4 (8.8) 45 97.4 (5.2) 41 95.3 (7.6) 86 6 89.2 (9.4) 81 89.2 (9.4) 81 7 93.1 (5.9) 118 93.1 (5.9) 118 9 91.8 (7.1) 104 91.8 (7.1) 104 10 89.5 (9.7) 71 89.5 (9.7) 71 11 90.9 94.5 91.6 88.0 96.4 104 94.5 (3.3) 16 (6.1) (8.9) (7.2) 177 9 1 35 (0.0) 2 (6.4) 240 12 89.0 85.8 92.6 91.5 90.5 (9.5) (10.9) 155 15 (8.5) 124 (5.8) 90 (8.7) 384 13 91.6 88.3 96.2 (6.3) (7.5) (4.4) 52 3 16 70.0 1 92.2 (6.8) 72 14 98.2 93.7 (4.4) (8.3) 16 34 95.2 (7.6) 50 Total 90.9 92.1 90.5 95.4 91.1 93.4 92.7 91.0 91.6 91.8 (3.3) 16 (7.3) (9.0) (3.5) (9.3) 676 223 30 240 (7.8) (19.6) (5.3) 259 3 160 (6.0) 294 (7.5) 1,901 * No data obtained for Mill 8; blank cells mean no data was obtained for that year/mill combination Table 16 gives a breakdown of the cohort dosimetry data by mill and sample year. Only 3 mills have substantial amounts of data (i.e. 30 or more observations) for 3 or more years. There were 7 years with substantial data, by the same qualification: 1980, 1981, 1982, 1983, 1984, 1996 and 1997. Mill averages range from 89.2 dBA (Mill 6) to 95.3 dBA (Mill 5). Yearly averages range from 90.5 dBA (1981) to 95.4 dBA (1982), but with no pattern in time trend evident. Table 17 gives mean noise levels for cohort dosimetry by mill and process group. Eighty-one percent of mill/process group combinations contain at least 2 measurements, and 85 percent have at least one measurement. Process group means range from 83.6 dBA for office to 97.3 for the sawing group. Other low-exposed groups are the primary breakdown jobs (a large percentage of which work in booths) maintenance workers, outdoor jobs and packaging. Highly exposed groups include utility jobs, sorting, and mill - non-sawing jobs. Regression Modeling The dependent variable, Leq, was a log-transformed sound pressure level in dBA. The noise level variable was normally distributed and did not require transformation. Nineteen variables were initially identified and considered for inclusion in the regression analyses. Variables "Number of employees", "Planer mill on site", "vehicle-based job" and "coastal location of lumber milF were all excluded because of a high degree of correlation with other independent variables. Variables "Cab on vehicle" and "multi-bladed saw" were excluded because historical information for these factors was difficult to obtain, limiting their predictive value. Table 12 describes the variables offered to the model. The final model was based on 1521 observations and, after the generation of indicator variables from categorical variables, contained 61 independent variables. R2 and adjusted-R2 were 0.53 and 0.51, respectively. Model coefficients, standard errors and P-values are shown in Table 18. The linear regression model is of the simple additive form: Y = j80 + A X , + P2X2 + ftX3 + . . . + RnXn Po, or the" intercept", represents the baseline exposure level when all indicator variables are equal to zero, i.e. Mill 12, process group = sawing, no job-modifier, no booth, year prior to 1983, outdoor job location, work area < 1000m3, and WCB sample. Coefficients (/3's) indicate the change in exposure level (in dBA) associated with one unit change in the corresponding independent variable (X). For example, the effect of mill (with all other variables held constant) varies from 1.0 dBA (Mill 10) lower to 3.5 dBA higher (Mill 5) than the exposure level at the reference mill (Mill 12) Similarly, process group varies from 17.6 dBA lower to 2.6 dBA lower than the reference process group, "sawing". All mills and process groups were forced into the model. 68 c « 3 2 O u o o II cn « « •O T3 O a c o •a « u cn XI O V a s e •a a « Q JS e « s | Total 86.3 6.2 107 O N oo O N vo oo in co ©' 0 0 i —i 86.7 6.7 201 r o O N r-vo v i ^ oo 85.3 8.3 238 m © r-O Tf" ^ O N 97.3 ; 6.4 436 CS O N r-CN C O 2 O N " - 92.1 5.6 193 91.0 4.3 171 91.8 7.5 1,901 T f O N C O \ cn Km'"" O N © '—1 CS O N oo © in oo cn rS oo O N © cs oo v i ^ O N r- N O r o C O © O N O O T f cs cs © O N CS V) © v i S ^ O N co CN f- O N CN O N O O -t O O in t> oo O CN CS — ' Tf" O O m cs N O T f oo oo V) O N C N N O O N O O N O O O O N CS 0 0 © oo oo l > cs O N CS O N — C N — O N cS oo cN CS vb ^ O N CS tN rn h T f v i oo O N r- N O in" in ^ O N oo o vo V O r-- — m T f ~ * oo r- oo CN V> O 0 r- in O N cs oo ^ oo r— O O C N O N C O O O r» r- © vb v i ^ O N vo © — T f ^> O N in N O N O cs" K ^ O N © O O CN - T f ^ O N N O V O T f cn .—• in oo N O VI O O O N - H N O O N ' ^ oo N O © r» O N T f oo v> t -» oo —'< v i O N O N — O N O O O N <N T f od vb ' O N O N t - r-» —' cs O N O N — r-cs v i ^ O N O N © co © T f " * O N 94.5 6.4 240 o O N ~ O N n O N (N vo O N O O O vq T f V ) — • oo © O N oo cn —' o — C O O O N O co r~-S T f ^ O N V) O N V O CS —' O N C O C O C O ©" co" O N ~ -C O O O T f K CN oo >o — O N O N O O O N © oo — m" in O N — oo m in vii ^ oo «o r- r s T f © 0 0 p —;; O N O O O O O O C O © C N vo h n O N N O N O O N © cs" O N in © oo —; Tf" O N C O © oo —' v i O N 91.8 7.1 104 Mill O O O N C S oo C N o in « O N o — vo" O N C N v j r-r-" cn oo r- oo C N r-" v i ~™ oo CN — © in © © N O v i ^ O N © T f C S O N 0 O T f O O —> v i O N O O ~ * O N T f vb O N 93.1 5.9 118 V O V O O N T f vb O N O O oo n oo T f r o O N oo cn r-» CN O O O O r- in cn ON" r-' cs — cn w O N ^ O O © T f r o T f oo 95.9 11.6 16 T f —• O vb co" ' oo r o © cS od N O oo © T f V) CN C O O N (S t -0 < O N ° ° oo in f"; CS O N od fS O N in T J - ri-cn CN O O © T f fS T f — ' O N C O V i T f O N O N * ~ I O O N O 1 — 1 VI O N N O ^ O N T f © 0 O vb c s O N CS O N O N O N N O O O CS VJ VI t-^  vb O N m «n N O •o l - ^ ° ° O N T f O O CN O O N VO" O O cn oo N O rt' c s <N O N © - H cn oo O 0 O N v i v i ^ oo cn in cn v i CN 0 0 0 O O N O O N O V i m 0 0 vo in co © " C O O N oc T f r-. V i vb O N — i r*» oo CS T f ^ O N V) T f r~-© ' C O * " 1 O N 91.2 6.4 232 r o " 3 - - 3 - in od v i ^ oo T T I N r-cn O N cs r-oo oo in cn o v i ^ O N CN v j m T f V) O N CN T f N O ob co oo oo © V J C N " * ™ O N © in oo O N co O N C O oo cs co" —• O N r- © O N T f C O ~ O N «n — O N co" 1 — 1 oo N M — u-i 2 O N C N CN N O O N in ~ O N cn cn vo" (N ^ O N r--. O C O N O N N O ~ O O p — | C N v i cs" O O O O © - H t * i O N n oo C O C O N O v i CN oo O O T f N O v i O O ^ O N CN O N r-ro' CO* ^ O N CS co ~-o v i <^ O N O N co K CN oo 89.9 8.4 174 - H co ~-0 0 o oo in O N C N oo cn CN N O O N N O O N CS © CS O N C N in N O O N O " O N oo •—' CO1 ^ O N co © r-od v i O N C O N O T f T f CN ^ O N O N CN T f © v i ^ O N N O T f O O CN oo T I N O <N - n ^ O N ~ * Process Group Maintenance Mill -non-sawing CU o B o Outdoors Packaging Sawmill -primary Sawfilers 0 0 e 1 00 c o C/3 5 Vehicle operators a O 69 s J T t w-i — CN — — o o O O O O o o o o V o d V — CT\ o o o o o o V fj « h vS <N —' O — m » q T t r-. irj r~ in in m v-j — — ' in f s i «n (N r*S (N o <N vq —• (N —; CN rt H T t ro <N T t o o l (6 t I S * " n q m n ^ oS 00 VO (N O O N >n V O < N < N "1 — I-; T f (N <N — ^ in r*N O O 2 VO i n rn o o o e ON 5 o w — o _ D -A O H & 0 O z § 2 » S O j 5 E « « h a if a CO on .5 t-O cw •— CO 5 2 03 2 o o o g •§ o « 031 > S D o l l vj o o o o CO — VN — —< VN T f — — . - H VO CN T t O O O O O — O O O m O — c> p p p p p p p p o p o o o o ' o o o o ' o o o o o o o o V V V CU u 9 an o c X w tf] ' © z \o " n >o N O I s h vo © o o o o o o u r n d r n H r f j O H Vfi OO V> OO C\ o o o o o : ( N lO TT rn n T j - r ^ o o o r ^ r - ^ o o o c s ^ o - ^ o — ^ o o o ' o ' o — ' o \C Ov \0 O oo v i —• <N T t in o \ o o r N | T f o o ' t o o 9 ^ o < > 9 ^ 1 —• — r-- r~-H K TT T t 't ON Tf T J - v q «/"> I rf r n r n r n r n o i c i C S l O OX k. a v e — — t N m T t m v o r ^ II 1 1 1 1 1 1 1 1 1 1 i l l o •a •a a • ii i t oo „ S 'ob" o o o - j a s rt o _ 5 fees 60 t i e ; " l i s . eg » u -- • - o 0 0 c ° 3 Si Z « ° rs 4) "O — — rt o 0 0 0 S ' 1 * O <z> ¥ § rt ^ E ' " E c » >• * . * 5 " ^ r? C 5 ? , 2 9.1 fe o5 O 3 iu 03 rt rt O C L B i U a i _ c a; a § s | s a. v X f-o H l l S a. 70 Nineteen job-titles remained in the model, adjusting the noise level estimate for the process group in which they were nested. Location and work-area size suggest that working indoors was on average 3.7 dBA noisier than working outdoors, and that L e q was reduced by approximately 0.5 dBA in larger work areas (> 1000m2); presumably the result of larger volume and distance between sources and receivers, and increased air absorption. The effect of booth varied by process group, with an average reductions of 4 - 11 dBA for workers involved in primary breakdown in the sawmill, sawing or working outside, but no substantial reduction for maintenance workers, or sawmill workers involved in non-sawing tasks. Sample source was included in the model to adjust for differences in potential sampling strategies and methods associated with the sources (i.e. compliance vs. research). Research-based sampling (UBC samples) was on average 1.5 dBA higher than samples collected by the WCB. Sample year suggested a slight increase of noise level with time: a 0.6 dBA increase for the period after 1983 compared to period 1980 to 1983 inclusive. The mean predicted value was 91.7 dBA, with a standard deviation of 5.7 dBA and range 69.5 to 103.6 dBA. A frequency distribution of predicted values for the full dataset model (n=l,901) is shown in Figure 12. .125 H 70 80 90 100 110 Predicted Leq Figure 12: Frequency Distribution of Predicted Values in Complete Modeling Dataset, with Projected Normal Curve (n=l,901). 71 Goodness of Fit Residuals plots did not reveal patterns in the data that would indicate violations of the normality assumption. There were 44 observations (2.9 percent) with standardized residuals greater than ±2.5 SD's, an indication of problems with the model's fit (the standard normal distribution suggest 1.25% would be expected). A review of these observations shows that 31 (70 percent) were over-estimates of the measured value. In these cases, the measured values were well below the expected values for these jobs, suggesting that necessary predictors were missing. Other outliers may have been due to misclassification of determinant values such as booth, or indoor location, especially for older observations where exposure determinants were not directly observed. Cook's-D residuals were all below 0.05, indicating that no observation was particularly influential. Evaluation of Model Coefficient magnitude and direction of sign were for the most part as anticipated, i.e. with the logical signs and plausible levels. The ranking of both process groups and job titles by predicted noise level was reasonable. Job titles that were found at more than one location in the mills (trim-saw operators, graders, and foremen) were examined by nesting department within job and process group, and these too ranked logically. Noise exposures for 380 observations in the "validation" dataset were estimated using the modeled regression equation. The correlation between individual observed noise levels and predicted noise levels was 0.67. The bias (average difference) of the prediction was +0.61 dBA suggesting a slight over-estimation of exposure levels, consistent with the observations of outliers. Model precision (standard deviation of the observed - predicted differences) was 5.4 dBA, and overall accuracy was 5.4 dBA. The bias for job means was 0.74. Figure 13 shows a comparison of observed and predicted levels, for job titles for which there were more than 10 observations in the validation dataset. In each pairing, the observed values are on the left9. Non-Modeled Adjustment Factors Lumber mills began enclosing planing machines in the early 1970's and almost all planers in cohort mills were enclosed before any of the available noise measurement was taken. Therefore an adjustment factor to improve estimates of pre-enclosure levels in the planer mills was obtained from the literature (Fairfax, 1991, and Tubbs, 1989). This data suggested that enclosure resulted in only moderate reductions in noise levels, and only in those jobs closest to the planer. Based on the published data, the 9 The boxes represent the median value and interquartile range (IQR), and the whiskers, 1.5(IQR) above and below IQR, respectively; outlier values are plotted separately. job-titles Planer Feeder and Planerman, both jobs in the immediate vicinity of the planer, were assigned an adjustment factor of +6 dBA for exposures prior to the implementation of the enclosure; jobs Grader and Trimsaw Operator were assigned an adjustment factor of +3 dBA. 9 Observed 9 Predicted 110 — 1 Barker Carrier Cleanup Dropsort Foreman Grader Resawyer Trimmer Boat Op Chipper Crane Op EdgerOp Forklift Offbear Tailsaw Job Title Figure 13: A comparison of Observed (left bar) vs. Predicted Values (right bar) for jobs with > 10 observations in the model validation dataset. Calculated hearing protection factors are shown in Table 19. These were based on the mean published noise reduction ratings (for muffs, 23 dBA, for plugs, 15 dBA) and the proportion of individuals reporting use within an exposure group at a particular time. The hearing protection factors represent the mean protection afforded the exposure group as a whole. Interviews with workers indicated widespread use of HPD began around 1970. L e q A , H P D (HPD-adjusted cumulative metric) was estimated by subtracting the appropriate hearing protection factor for a given exposure-level/time-period combination from the predicted value in the exposure data matrix. Table 19: Hearing Protection Factors (dBA) for Hearing Protector Use in BC Lumber mill Workers Unprotected Time Period Exposure Level <1970 1970-1974 1974-1978 1978-1983 1983-1990 >1990 <85 dBA 0 2.6 7.0 9.6 10.5 10.5 85 - 94 dBA 0 3.5 8.8 13.1 14.9 15.8 > 95 dBA 0 4.4 10.5 16.6 16.6 17.5 73 Analysis of the full dataset (including all measurements, n=14,050), showed noise levels after 1978 were approximately 3 dBA lower than for the period 1970-1978. A similar decrease was seen in cohort data (n=3,776), which also included an adjustment for worker enclosures. An adjustment factor of 3 dBA was therefore added to the regression equation for pre-1978 predictions. Mean A-weighted and C-weighted levels for each process group, and the adjustment factor based on the difference between the two, are shown in Table 20. Because the A-weighting scheme discounts low-frequency sound more than the C-weighting scheme, process groups with larger weighting factors were exposed to higher levels of low-frequency noise. The results reflect the anticipated differences between process groups based on their differing job tasks and environment (e.g. between the primary breakdown area which handles whole logs and sawfiling, where grinding and hammering tasks predominate). The process-group mean weighting factors were added to the individual A-weighted predicted exposure levels to estimated C-weighted exposure levels. Table 20: Low-Frequency Weighting Factors, and Predicted A-weighted and Estimated C-weighted Process Group Means (Based on Full Modeling Dataset, n=l,901). Sawmill Utility Maint Vehicle Office Outdoor mill non- Sorting Packaging Sawing Sawfilers Primary Ops. sawing A-weighted 85.3 92.1 86.3 91.0 83.6 86.7 95.9 82.2 86.3 97.3 90.5 mean C-weighted 94.0 99.4 93.0 97.0 88.3 91.3 98.2 84.3 88.4 99.0 91.6 mean Calculated weighting 8.7 7.3 6.7 6.0 4.7 4.6 2.3 2.1 2.1 1.7 1.1 Factor Building the Exposure File Exposure levels were estimated for a total of 3,809 exposure keys. The average number of exposure keys per mill was 272, and each mill had on average 6 time periods (range 4 - 8) of presumed constant exposure, i.e. periods when all identified exposure determinants were constant. Table 21 gives the number of exposure data matrix records per process group, plus mean estimated L e q and standard deviations. Also shown are the total number of years of exposure in the cohort attributed to each process group. Note that the percent of total exposure duration that was allocated to department and mill average levels was very low, less than 6 percent of the total. Distribution of estimated values by mill and process group are shown in Figures 14 and 15, respectively. Mills are grouped by geographic region, and no association between region and noise levels was apparent. Figure 16 gives examples of how estimated exposure levels changed with time, for a sample of jobs from one mill. Figure 17 shows the same exposure levels for the same jobs, but adjusted for hearing protective devices. Because most of the determinant data collected reflected more recent changes, all of the time-dependent changes occur in the last 4 decades of the study. Prior to this period, constant 74 exposure levels were assumed. Table 22 shows how the exposure time was distributed. Approximately half the exposure 'experience' of the cohort occurred prior to 1970, but less than 30 percent prior to 1960. Table 21: Exposure Data Matrix: Descriptive Statistics, and Distribution of Cumulative Exposure by Process Group Process group # E D M records Mean Leq SD Total Cumulative exposure (years) Percent Total Cumulative Years Sawmill Primary Processing 314 90.4 6.9 12,293 4.3 M i l l - Nonsawing 489 97.1 3.1 27,311 9.5 Sawing 543 99.9 4.3 30,304 10.6 Sorting 345 94.3 2.7 3,924 1.4 Packaging 157 87.7 3.1 3,497 1.2 Utility 821 93.2 3.4 63,149 22.0 Vehicles 96 94.0 3.4 17,033 5.9 Administration 52 84.9 2.4 12,275 4.3 Outdoor 404 87.2 3.5 46,283 16.2 Sawfiling 52 91.5 2.4 7,787 2.7 Maintenance 127 86.4 2.9 46,783 16.3 Department Average 364 90.8 4.1 15,171 5.3 M i l l Average 45 91.5 2.5 670 0.2 Total 3,809 93.1 5.7 286,480 100 115-110-105 -c i >L c i c i < >_ c >c >< > c c >c >c >c c >>m C T a> _l 100 -95 -90 -85 -80 -1 2 3 4 5 6 7 8 Mill 9 1 0 11 1 2 1.3 14 Figure 14: Box plot showing distribution of predicted exposure levels for 3,809 exposure keys by mill (Mill 1,2 Interior /North; Mill 3 Interior/Okanagan; Mill 4, 5 Coastal/Garabaldi; Mill 6, 7 Coastal/Vancouver; Mill 8 Coastal/Simon Fraser; Mill 9-13 Coastal/C. Island; Mill 14 Coastal/N Island) 75 admin maint non-sawing packaging sawfiling sorting vehicles! dept avg mill avg outdoor primary sawing utility Process Group Figure 15: Box Plots Showing Distribution of Predicted Exposure Levels for 3,809 Exposure Keys by Process Group. Figure 18 shows the distribution of individuals' cumulative exposure to noise based on the predicted exposure estimates10. The data shown is as used for epidemiological analyses (see next chapter) and represents 27,247 male subjects with a minimum cumulative exposure of 85 dBA*yr. The mean cumulative exposure was 101.3 dBA*yr, standard deviation 7.0 dBA*yr and range 85.1 - 138.7 dBA*yr. The dBA*yr unit increases by 3 for a doubling of either noise level or time, so 101.3 dBA*yr is equivalent to one year of exposure at 101.3 dBA or 2 years at 98.3, 4 years at 95.3, etc. Table 22: Cumulative Exposure Distribution by Decade Time Period Total Cumulative exposure (years) Percent Cumulative Percent Before 1950 26,965 9.4 9.4 1950-1959 52,575 18.3 27.7 1960-1969 57,417 20.0 47.7 1970-1979 72,008 25.1 72.8 1980-1989 54,936 19.2 92.0 1990 and later 22,509 7.9 99.9 Total 286,410 100.0 1 Cumulative exposure was calculated as 10 log ^7^(10 1 0 ) .7=1 ; see next chapter. 76 115 1950 1960 1970 1980 1990 Year Figure 17: Examples of Predicted Exposure Levels for 5 Sample Jobs at One Mill, Estimated for the Yeats 1950,1960,1970,1980 and 1990, Following adjustment for Hearing Protector use. 77 1802 H O H 90 100 110 120 130 140 cumulative exposure dBA*yr Figure 18: Frequency Distribution of Individual Cumulative Exposures for Male Cohort Subjects with Cumulative Exposure > 85 dBA*yr (N=27,247) Summary Exposure data from 3 sources was combined and used to build a predictive statistical model. Adjustment factors for determinants that could not be modeled were independently estimated. Following validation, the model, along with adjustment factors, were applied to predict historical exposure levels for over 3,800 mill/job-title/department/time-period combinations utilizing an exposure data matrix. Combining data from different sources (i.e. government agencies) has been suggested as a means to increase the amount of available data (Stewart, 1999). However, care must be taken, as the reason for sampling (i.e. compliance vs. research) may result in biased exposure measurements. To account for this, sample source was entered in the regression model. It was found that the samples provided by mill management, and data from the WCB, were on average actually 1.5 dB lower than those obtained specifically for this study, after adjusting for many other factors. The reason for this bias is not known, though one might speculate that there was some benefit to mills under-reporting their noise exposures. In predicting noise levels, the 1.5 dB adjustment was included for all estimates. The resulting model had very good predictive power in comparison to other models reported in the literature (Burstyn and Teschke, 1999), and although it had a slightly lower R 2 than that of the preliminary model reported in Chapter 3, this model had broader predictive capability as it included 13 of 14 participating mills and data from a 16-year time span. There were a greater number of outliers in the regression model than expected (3% > ±2.5 SD's) suggesting a problem with model fit. On inspection, the large majority of outliers (> 80%) are over-estimates suggesting that the model was perhaps missing predictors for work factors that remove the worker from the noise source (i.e shutdowns, meetings etc.). Regardless, the model performed well in the evaluation exercise with modest bias and reasonable precision. The most significant limitation of the model was the lack of data for the time period prior to 1980. The main determinants model showed a one-time increase of approximately 0.6 dBA for the years 1983-1997 compared to the years 1980-1983. In a separate analysis of all available data (including non-cohort measurements) a decrease of 3 dB was noted for measurements after 1978. Interestingly, this coincides with the introduction of WCB regulation limiting noise exposure, and a peak in exposure measurement volume (Figures 10, 11)11. However, the earliest measurement data entered in the model was for the year 1970, and so the only predicted changes earlier than 1970 were a result of the implementation of planer enclosures in some mills in the late 1960's. Other than that, exposure estimates for earlier time periods were based on 1970-1978 levels. Mitigating this limitation was the fact that several of the mills were built or substantially rebuilt during the late 1960's or early 1970's and so only limited technological change would have been expected during the period immediately following construction. For older mills there was relatively slow change in saw milling technology until the 1970-1980 period. In his review of labour flexibility at two of the cohort lumber mills, Hayter note that "antiquated" machinery was still in use into the 1970's and that the mills did not undergo significant technology upgrades until the early 1980's. Apparently the timing of this technology shift could be generalized to other mills in the province (Hayter et al., (1994). Slow adoption of noise control measures prior to the advent of widespread hearing protector use in the 1970's also would mean more stable exposures to noise in these earlier periods. Finally, less than 30 percent of the total accumulated exposure time was prior to 1960, further reducing the impact of any potential misclassification. 1 1 Similar peaks in measurement volume correspond to the introduction of compensation for hearing loss (1975) and focused WCB efforts to introduce noise control measures (1980-1984). Chapter 6. Cardiovascular Disease in BC Lumber Mill Workers: Epidemiological Analyses Introduction This chapter reports the findings of phase 3, the epidemiological analyses. The objectives of this phase of the study were: (1) to examine the mortality experience of BC lumber mill workers in comparison with the general population of BC; (2) to examine exposure-response relationships within the cohort using Poisson regression; (3) to examine the effects of noise-frequency (specifically low-frequency noise) on health outcomes; and (4) to examine the temporal relationship between exposure and health effects. Methods Study Population Subjects had been previously enumerated from the personnel records of 14 large British Columbia lumber mills, previously described in Chapter 3. To be eligible for inclusion, subjects had to have been employed for at least one year (or 260 working days, which may have been non-contiguous) between January 1st, 1950 and December 31st, 1995. Subjects were required to be hourly workers, although subjects who later transferred to salaried positions were retained in the study. Nineteen-fifty was selected as the start of follow-up, as the earliest date that Provincial death records were available in electronic form. The cohort, totaling 26,487 workers, was first enumerated in 1988 (Hertzman et al, 1997). It was then re-enumerated in 1997 with personnel records from the same 14 mills, adding workers hired after 1988, and updating work histories for all subjects (Teschke et al, 1998). The updated cohort contained 21 fill subjects who met all eligibility criteria, and who had complete identification information. Ethnicity McKeigue et al, (1989) reported that mortality from chronic heart disease is higher in South Asian immigrants in several parts of the world; they postulated that it may result from a greater prevalence of insulin resistance in this population. For the lumber mill cohort, a research assistant who was of Punjabi origin determined identity of South Asian cohort subjects, based on name recognition. The majority of subjects were of European ancestry, but six percent were of South Asian descent, 80 predominantly I s and 2" generation immigrants from the Punjab region of India . Risk of acute MI and hypertensive diseases were subsequently shown to be elevated in subjects of South Asian descent in this study (Demers, personal communication), and all internal analyses were adjusted for South Asian ethnicity. This ethnic group was also known to have very low smoking rates (BC Ministry of Health, 1997). Gender The cohort was more than 99 percent male. Because of this, and because no deaths occurred among the 174 female workers during the follow-up period, analyses were restricted to males. Accounting for the Use of Hearing Protection The use of hearing protectors can significantly reduce noise exposure, but the effect is not captured by personal noise dosimetry, measured at the shoulder. In Chapter 4,1 described how a numerical correction factor for HPD use in the lumber mill cohort was developed. In addition, a sub-cohort of 8,700 workers was created, comprising workers who had a date-last-employed prior to June 30th, 1970, approximately when widespread use of hearing protectors began in the cohort mills. It was hoped that noise-estimates for this period would be more accurate, reducing risk of misclassification and subsequent attenuation of exposure-response effects. Follow-up Cohort deaths were identified by probabilistic linkage to the Canadian Mortality Data Base (Statistics Canada, Ottawa). Vital status ascertainment through this database was considered to be 97.6 percent for deaths in Canada, and 93.1 percent for deaths overall (Schnatter et al, 1990). Various methods were employed to determine the vital status and date last observed for subjects classified as alive, but not still actively employed. For the initial cohort enumeration, (i.e. to 1988) this included: linkage to pension records; motor vehicle records; and personal inquiries at union halls. For the updated cohort (i.e. to 1997) subjects were linked to provincial health administration data through the British Columbia Linked Health Data Project (Chamberlayne, 1998). This database includes all post-1986 registrations in a provincial medical insurance plan that provides coverage for virtually all BC residents. After this linkage, 15 percent of subjects remained lost to follow-up. Of these, individuals for whom a social insurance number was known, or who had ever linked successfully to an external data source 1 2 This approach was validated by comparing the name-based ethnicity assignments with self-reported ethnicity among a sub-cohort (N=l,953) interviewed in 1997, and it showed a sensitivity and specificity of 99 and 100 percent, respectively (Ostry et al, 2000; Paul Demers, personal communication). following enumeration (i.e., for whom good quality personal identifier data was available, and therefore would have been expected to be linked to mortality database if deceased) were assumed to be alive at end of follow-up. For the 978 subjects (3.5 percent) for whom no social insurance number was known, or who had never been linked to an external data source following enumeration, follow-up was ended at their date of last employment, as this required no unverifiable assumptions about mortality status after being lost to follow-up (Checkoway etal., 1989). Exposure assessment The three exposure measures (LEQ>A, L e q >c, and L E Q >A, HPD) estimated by retrospective noise assessment (Chapter 5) were incorporated in two exposure metrics for epidemiological analyses. First, duration of exposure to noise above a threshold level was calculated, where threshold levels were set at 85, 90, 95 and 100 dB (both A and C-weighted): Eqn. 1) Duration of exposure above threshold = * L-7=1 Where k = number of work history records for a subject, 7}=duration of work history period j, and Lj = 0, unless the noise level corresponding to the job held in work period j was greater than the threshold, then Lj• = 1. Secondly, a cumulative exposure to noise was calculated: Eqn. 2) Cumulative exposure above 85 dB = 101oa Xr,(io >°) .;'=i Where k = number of work history records, 7}= duration of work history period j, and Leqj = noise level above 85 dB(A or C) corresponding to job held in work period j. The units of the cumulative metric are dB(A or C), reference 20 uPa»Year (after Talbott et al., 1999). As with other time-dependent dB units, a doubling of either duration of exposure or sound pressure results in an increase of 3 dB. I use units of "dB*yr" to remind the reader when a cumulative noise exposure is referred to. Although it is believed that cardiovascular effects of noise exposure may be associated with exposures at levels below 85 dB, this was set as the lowest threshold level to address the issue of non-occupational noise exposure. Although no attempt was made to measure non-occupational exposure, by setting a threshold that was higher than most non-occupational exposure levels, misclassification of exposure from this source should be reduced. Health Outcomes Four disease categories were selected for investigation, based on the hypothesized disease 82 mechanism (Chapter 1), and the coding divisions available in the external reference rate data. These were: hypertensive diseases (ICD9 1 3 codes 401-405.9); ischemic heart disease (ICD9 codes 411-414.9, 429.2); acute myocardial infarction (ICD codes 410-410.9); and stroke (cerebrovascular disease, ICD9 430-438.9). In addition summary data is provided for all circulatory diseases (ICD9 codes 390-450.9), and all causes. Potential Confounders Smoking Smoking was a potential confounder for all of the health outcomes under study. A common problem for this and other cohort studies was a lack of personal smoking information. Nevertheless, a number of indirect methods for accounting for the effects of smoking are available (Steenland et al, 1984). Its potential impact on noise-heart disease relationships was examined in three ways, as follows. Although not available for all cohort subjects, smoking data had been obtained for a sub-cohort of 1,960 workers who participated in a 1997 survey for another study. This sample was a random sample of cohort subjects who were employed in a cohort mill in 1979, and still alive in 1997 (Ostry et al, 2001). Smoking rates in the survey sub-cohort were compared with those in the general population of BC 1 4 . Smoking status was similar for all male cohort subjects, though former smokers were more common in the lumber mill groups and never-smokers in the general population (Table 23, P. Demers, personal communication). The difference in former smoking status likely reflects an underlying difference in age distribution between the two groups, as the general population included males aged 12 years and up, while the minimum age in the lumber mill sub-cohort was 35 years. The age-stratified results for 45-64 years and 65 + years are much closer in value. The table also shows that the average number of cigarettes smoked per day by current smokers was very similar. Two additional analyses of smoking were conducted and results are reported in the results section of this chapter. First, mean values for duration of employment, cumulative exposure, and duration of exposure above 95 dBA were compared across the 1997 smoking groups (current smoker, past smoker, never smoked). Further, cigarette consumption was examined by comparing mean pack-years15 across 1 3 International Classification of Diseases, Revision 9 1 4 Smoking rates for the general population were obtained from a Heart and Stroke Foundation of B C & Yukon survey conducted for the B C Ministry o f Health by the Angus Reid Group. It was based on 18,030 randomly selected respondents (BC M O H , 1997). 1 5 Pack years = ((number of cigarettes per day)/20) X (years smoked) 83 exposure categories for one of the duration of exposure thresholds (above 95 dBA), as well as for cumulative exposure. Table 23: Province-Wide Smoking Rates (Men, Age 12+), Compared to 1997 Questionnaire Sub-Cohort (Men, Age 35+) B C Lumber mill % % n A l l Men* Current Smoker 23 20 400 Former Smoker 32 47 911 Never Smoker 45 33 648 1959 S. Asian Men* Current Smoker 11 10 26 Former Smoker 12 14 37 Never Smoker 77 76 197 260 Age 25-44 Current Smoker 29 25 137 Former Smoker 24 35 194 Never Smoker 47 40 217 548 Age 45-64 Current Smoker 21 23 207 Former Smoker 48 42 379 Never Smoker 31 35 319 905 Age 65+ Current Smoker 13 11 56 Former Smoker 63 67 338 Never Smoker 24 22 112 Cigs/day, current smokers* 506 A l l Men 18.2 19.6 373 Age 25-44 16.8 18.2 126 Age 45+ 18.7 20.4 247 * Includes ages 12 and up for general population, but only 35 and up for cohort subjects Second, lung cancer mortality was examined in parallel to cardiovascular disease mortality on the assumption that if excess mortality in cardiovascular diseases was caused by smoking, a concomitant increase in lung cancer mortality rates should be evident. Statistical Analyses All person-years-at-risk calculations began one year after first exposure. The PC Life Table Analysis System (Cassinelli et al., 2001) was used to calculate standardized mortality ratios (SMR's). Mortality referent rates were based on the general British Columbia population for the years 1950 to 1995, compiled by Statistics Canada, Ottawa. Ninety-five percent confidence intervals for SMR's were calculated based upon the assumption that the observed effect followed a Poisson distribution (Breslow and Day, 1987). Exposure response relationships within the lumber mill 84 cohort were examined using Poisson regression (STATA 7, command poisson), and adjusted for age, calendar year, and ethnicity (South Asian descent vs. all other). Age and calendar year were entered in 10-year categories, and were pooled if cells contained fewer than 5 deaths. Tests for linear trend were conducted by entering the exposure category as a continuous variable in the Poisson regression analyses. Analyses examining the effects of noise exposure after latency periods of 5 years and 10 years were performed by "lagging" exposure by the respective amounts. The temporal relationship between exposure and outcome was also examined by restricting follow-up to subjects' working years, to investigate whether noise exposure was associated with increased risk while still employed and exposed to noise. Follow-up in this analysis was ended upon death or one month after termination of employment (to capture subjects who terminated employment due to ill health and died shortly thereafter), whichever occurred first. Thirty-five subjects were excluded from this analysis because they failed to meet eligibility requirements after restricting follow-up, reducing the number of subjects in these sub-analyses to 27,464. Results Because of the number of different exposure metrics and analysis groupings, a summary of major analyses and an index to the results tables are presented in Table 24. Table 24: Index of Statistical Analyses, and Tables of Results in Chapter 6 Cohort Group Follow-up Period Exposure Metric* Adjustment* Latency Period Analysis Table # SMR Internal Full (n=27,499) Complete Dur'n None 0 28 29 Complete Dur'n None 5 36-39 Complete Dur'n None 10 36-39 Complete Dur'n HPD-adj'd 0 45 Complete Cumul HPD-adj'd 0 46 Complete Dur'n C-weighted 0 47 Complete Cumul None 0 32a 33a Complete Cumul None 5 44 Complete Cumul None 10 44 Restricted1 Dur'n None 0 34 Restricted Cumul None 0 35 Pre-1970 (n=8,700) Complete Dur'n None 0 30 31 Complete Dur'n None 5 40-43 Complete Dur'n None 10 40-43 Complete Cumul None 0 32b 33b *dur'n: duration of exposure above threshold; cumul: cumulative exposure above 85 dBA ^Noted i f adjustment made for either hearing protector use or low-frequency (C-weighted) *Follow-up restricted to period of employment Demographic information for the full cohort and the "pre-1970" sub-cohort is provided in Table 25a and 25b. Although the number of subjects in the sub-cohort was only 30 percent of the full cohort, it retained 37 percent of the person years, and 60 percent (3,499) of the deaths. The sub-cohort was demographically 85 similar to the full cohort, although the mean age at entry was 3.3 years higher, reflecting the higher proportion of workers entering at the study start. Similarly, length of follow-up in the sub-cohort was on average 4.2 years longer. Table 25: Demographic Characteristics for (a) the BC Lumber Mill Workers Cohort (n=27,499 males); and (b) a Sub-Cohort of Workers Who Completed Employment Prior to 1970 (n=8,700 males) (a) Full Cohort (b) Sub-cohort, Last Employed prior to 1970 South Asian European A l l South European A l l Descent and Asian Descent and other other Number of 1,628 25,871 27,499 179 8521 8700 Subjects Number of 131 5,741 5,872 40 3459 3499 deaths Age at entry mean 28.8 29.7 29.7 32.5 32.9 32.9 (years) median 26.4 26.1 26.1 27.9 29.3 29.3 range 15.5-69.3 10.6-76.3 10.6-76.3 16.0-65.0 12.6-76.3 12.6-76.3 Age at death mean 55.7 65.9 65.7 59.2 67.8 67.7 (years) median 54.6 68.0 67.9 57.6 69.1 69.0 range 22.1 -99.0 18.1 - 101.7 18.1-101.7 26.2-99.0 19.0-101.7 19.0-101.7 Follow up mean 19.9 24.6 24.3 18.0 28.7 28.4 duration median 20.5 24.0 23.6 13.8 31.0 30.8 (years) range 1.0-46.0 1.0-46.0 1.0-46.0 1.0-46.0 1.0-46.0 1.0-46.0 Duration of mean 12.0 10.3 10.4 6.3 8.3 8.2 employment median 9.4 5.3 5.5 4.2 4.8 4.8 (years) range 1.0-42.3 1.0-51.5 1.0-51.5 1.0-41.5 1.0-51.5 1.0-51.5 Year First mean 1971 1964 1964 1954 1952 1952 Employed median 1993 1966 1967 1953 1952 1952 range 1915-1994 1909 - 1994 1909-1994 1915-1969 1909-1969 1909-1969 Analyses of the Full Cohort (n=27,499) A total of 27,499 subjects provided 668,744 person-years of follow-up between 1950 and 1995. Table 26(a) gives standardized mortality ratios (SMR's) and their 95% confidence intervals (95% CI's) for all causes of death, specific circulatory diseases, as well as lung cancer, all cancers, and other major disease categories. Overall the mortality experience of the lumber mills workers was similar to the general population of British Columbia. The SMR for all causes of death was 0.95. This, and the other slight deficits in mortality for respiratory, digestive, circulatory and infective disease are consistent with the "healthy worker effect" (Checkoway et al, 1989), although not as large as might have been anticipated. For the diseases of interest, SMR's were between 0.96 and 1.03. No SMR's that were greater than one were significant at P=0.05. 86 N? O — c o v o 1^  — — o m —i r t — C N — ; CNI V"> I I I I I I I I O O N C N O N C N V O O O V O p v q v o o \ o o o r ~ « n —< o o o " — o o o g ° p o o — — •n oo o oo — in CN ON ON T t CN CN —< o o — CN —; I I I I I • oo v o t— v o r n — in r~ in oo in p O O O O O — O O O — — — NO ON — o T f ro CN VO OO VO T t CN T t r~ o o CN 1^  ON ON r-- oo CN CN in in T t O oo T t r - VO co ON •n OO o T f r n VO — 00 >n ro v d o 00 CN •n T t ro NO NO T t ON 00 ro co r - T t T t ro ro ro T t CN — r~ CN CN ro n-NOvOO-Osh «/-> ON} 00 CN ON O V~l V I ON V~l ( N ^ r N i n r -CN — m U 0 s -V~l ON a. U XI O r o O v o ON — O — O O r o I I I I I I I I T t O O T f O N t ^ v O O O — O N V O v o o o O N O O r ^ v o O O O O O O O O O O O O — O O O I 00 O — O T t ON O O O — — 00 CN T t ON — O t— ON ON — V O O O O O — — — I I I I I I T t in CN co o v o T t in oo o o o o o o o o o o o o r o T t T t oo OO 00 r~ in in OO O o v o in ON r-~ ON in CN T t T t in 00 o CN 00 in OO r-; 00 CN •n T t t - ; O O N T t v o O —* o oo in v d CN CN CN r-^  in r o oo in CN m oo ON T t r o ON ON v o 00 v o V O r- — — — in oo oo r o — — T t T t T t <N oo O N O V O V O O C O I ^ O N — i n c N — — r o CN m OO ON r o — oo — r o O 00 in T t ON v o O — r o O T t OO CN r-~ — V O r o CN T3 O C3 T3 C JJ O SP T3 •a ca ej •O <4-< o s ca • <u |Q c— O Si > u K Q Q o b * S B ca 3 8 .2 "K xi g .a € £ >-, u BC .2 o u M S3 a 3 o o 5 P 8-.S § i 8 S •2 1 € 'S < > o o S c3-trt on S S a 5 • 0) o H 00 & c 15 T3 O ca 1/1 a. — •a Q § b u S > ca *3 u S3 tL) « 5 | .SP g Q a .S3 23 Q -S — e s s I'll 5 5 o ^ 2 5 s < c o > cu I-Q y c o T 3 u c^  CQ 87 SMR's by duration of employment and their 95% CPs are shown in Table 27(a). SMR's were close to one, except hypertensive (HT) disease for those employed over 20-30 years (SMR=1.5). However, the number of HT disease deaths was small, and the estimates somewhat unstable. SMR's for acute MI were slightly elevated for all duration categories except one. Table 27: Standardized Mortality Ratios (SMR's) and 95% Confidence Intervals (CI) for All Causes, and Selected Cardiovascular Diseases by Duration of Employment. BC Lumber Mill Workers, (a) Full Cohort, (27,499 Males); (b) Pre-1970 Employment Sub-cohort (n=8,700). Reference Rates from General BC Population (a) Full Cohort (n=27,499) (b) Pre-1970 Employment Cohort (n= =8700) Duration of Observed Expected SMR 95% CI Observed Expected SMR 95% CI Employment All causes <10 years 863 750 0.99 0.92-1.06 1921 1779 1.08 1.04-1.13 10-20 years 489 464 0.93 0.85-1.02 821 783 1.05 0.98-1.12 20 - 30 years 792 795 0.92 0.86-0.99 509 506 1.01 0.92-1.10 > 30 years 812 895 0.91 0.85-0.98 248 287 0.86 0.75-0.98 Circulatory Disease <10 years 935 938 1.00 0.94-1.07 785 729 1.08 1.01-1.16 10-20 years 605 628 0.96 0.86-1.04 414 399 1.04 0.94-1.15 20 - 30 years 593 594 1.00 0.92-1.08 290 267 1.09 0.97-1.22 > 30 years 386 419 0.82 0.74-0.91 143 154 0.93 0.78-1.10 HT Disease <10 years 10 9.4 1.07 0.51-1.96 9 7.2 1.26 0.58-2.39 10-20 years 2 5.8 0.34 0.04-1.23 2 3.5 0.58 0.07-2.09 20 - 30 years 10 6.5 1.54 0.74-2.83 * 5 4.2 1.20 0.39-2.80 > 30 years 4 4.9 0.81 0.22-2.07 Ischemic Heart Disease <10 years 322 306 1.05 0.94-1.21 282 246 1.15 1.02-1.29 10-20 years 220 235 0.94 0.83-1.07 167 161 1.04 0.89-1.21 20 - 30 years 168 188 0.90 0.77-1.05 106 99 1.07 0.88-1.31 > 30 years 106 123 0.86 0.70-1.05 51 57 0.9 0.67-1.19 Acute MI <10 years 333 328 1.02 0.92-1.13 273 248 1.10 0.98-1.24 10-20 years 200 188 1.06 0.92-1.22 113 103 1.10 0.91-1.33 20 - 30 years 225 210 1.07 0.94-1.22 90 74 1.21 0.98-1.50 > 30 years 152 155 0.98 0.83-1.15 44 43 1.03 0.75-1.34 Stroke <10 years 130 133 0.97 0.81-1.16 107 105 1.02 0.84-1.23 10-20 years 93 96 0.97 0.78-1.20 69 66 1.04 0.81-1.32 20 - 30 years 95 94 1.01 0.82-1.24 42 47 0.89 0.64-1.21 > 30 years 55 66 0.83 0.63-1.09 23 28 0.82 0.52-1.23 Adjusted for age and calendar year *> 20 years Table 28 shows SMR's and 95% CPs by duration of exposure above the 4 exposure thresholds: 85 dBA, 90 dBA, 95 dBA and 100 dBA. In subjects exposed above 100 dBA for more than 20 years, the SMR for acute MI was 1.4 and statistically significant; for more than 30 years above 95 dBA it was 1.3. Fifty percent increased mortality was found for HT disease in those exposed for more than 20 years above 90 dBA, but again numbers of observed deaths were small, and prevented any analysis above the 90 dBA threshold. SMR's for ischemic heart disease and stroke were close to one at all threshold levels. 88 < m o oo ON P*1 NO Csl CN| ON o Tf NO ON OO Tf O p ON r*N O ON O p Tf O ON p in rvi d « —* - H ~- ~- —i —i —i (N NO OO •4 NO Tf Tf v\ Tf O NO CN ON oo VO 00 NO ON ON NO oo ON Tf Tf ON ON O 00 Tf Tf O d d d d d d d d d d d d d d d d d d d d NO TI- 00 00 00 NO 00 o O0 CN! NO VO NO NO o ON ON ON ON ON O ON ON o Tf ON — ON O d d d d d —' d —• O d d d — ' d d — d d ro cs ON —* r*-n oo - - H rs o t IN « h f S h N ^ NO v O ^ W ( N oo in vo v o o v> m NO • CN cs ON £ T" O ro oo NO oo VJ O ON ON O ON ^ O O —' O Tf NO ro v j oo ON oo oo oo vo O O O O O O O O O O O NO ro NO t Tf t • Tf ro oo oo NO ro — oo Tf ro p*- O o ro v i p- p-CN^  — O ^ CN O ^ O " — NO" NO" VJ" O O Tf -H V ) — O O O O O O O O —4 o oo oo oo O O I v i NO ro P- I V) ro ro — — (N — ro O O O O — o o — ON NO p- NO Tf oo r*- oo NO ro o o o o o t-^ CN V) w ON ON O ON NO O O — ' O O — Vi ON NO Vi CN — — Tf V) — — O O O O O Tf Tf Tf p- NO — Tf — i V ) oo oo —> cN CS <N oo l— NO r~~ O O O O O O NO V I O O P"-O ON 0O O C O —' o o —' o 0O P- CN Vi 0O rO co CN P** CO v> — < Cu ^ — o o — o ON v> — ON 0O 0 O 0O O O O O O O O — O O O O O O O ro C O O0 NO NO co co — C O O O O Tf CN — — NO ON r - v i 00 p*- Tf oo p- VJ »n r-* tn v i cs co" o" P-" o" NO" CS O O O Tf - H co CS — o v i p- oo ON ON O O ON P-O O O O O o o o — o C O P- N O Tf O N O V > O Tf o 0O V ) V ) Tf CN <-i r- r- r n n o o o o o o o o O O r o Tf t Tf N O — N O O N O N CN CS O — ON O O O O O o — o — o O ON NO O — NO ON P- co v j O O O O O v i v j v i ON NO ON 0O P- NO O O ON C O ON Tf 00 P- 00 NO 00 Tf O O O O O V ) Tf O O P- C O ON O O O O P-o —' o — o il 3 w CN| O O0 ON ON oo ON N O O Tf O 1^  O p 0.28-2.64 V ) CN N O CN V ) O O o CN Tf CN 00 oo O Tf o V I O ro ON O Tf CN cs N O o cs ON CS o cs cs cs O —« —• d d d 0.28-2.64 CS —^ cs CS OO ON ON OO OO O0 O0 <A ON o\ 00 ON 00 ON Tf O0 0.28-2.64 O O C O Tf O Tf p-C O CN O N 00 ON 00 ro oo N O p-C O px Tf O N O O Tf ON ro ON ro 00 oo N O Tf oo oo px ON r-cs N O d d d d d d d d d d 0.28-2.64 O O o O o O o o O o o O O o o o o o o Tf p ON r<N ON O N ON O ON N O ON ON ON O N ro p C O p v> ro V J V ) N O O O p-p ro O V I ON ON O O N O O O z Tf ON O O p N O O oo O N ON Tf p p-ON O O ON C O oo d d d d d d d d O o 1—1 — i O o o —1 o " - o o o O o ON ON o Tf o Tf N O CN o Tf ro ON P- V J cs ro O ^_ V I oo o C O Tf ON ON ON VJ 00 Tf Tf ON P- ON — O0 oo r- NO^  o\ jo <n CN P-" •< oo" CO ro — CN cs cs —• v-> at u X o >^  >-> ?o ^ ro O O O O -*S y — CN ro ro ^ I O O A H co — <N H U <u <L> <u <u se e e e ^ CO cd CQ il> u 0 0 0 0 « V O O O O — cs ro ro -52 — CN ro ro I O G> A Q I O O A ro — ' <N w r/i t/i in c3 c3 cd v?3 u u u u « ^. ^. ^. - . — ^. ^ Q r o O O O O * S r O O O O O C V S^ ro ro y •—1 CN ro ro id ^ c3 U V U U I o o A -s co — CN g I O O A m — CS I V) VI (A Ul i L k« i - C-•N, cd cd cd « <U U 4> o o o o •—' CS ro ro I O O A — — rN Duration of exposure at lower levels (85 dBA and 90 dBA) was quite highly correlated with duration of employment at end of follow-up (Pearson r = 0.99 and 0.80, respectively), while durations at higher thresholds (95 dBA and 100 dBA) were not (Pearson r = 0.59 and 0.30 respectively). Table 29 shows relative risks (RR's) from internal analyses of duration of exposure metrics, with lowest exposure category (< 3 years) used as the reference category. All internal analyses were adjusted for age, calendar period and ethnicity. Increasing trends in acute MI mortality were evident with duration of exposure over a threshold of 95 dBA, reaching 1.3 (PTrend=0.059) for those exposed for more than 30 years, and reaching RR=1.6 for those exposed for more than 30 years above 100 dBA (Pirend=0.35). HT disease also showed increasing mortality with increasing duration of exposure for those exposed above 90 dBA (PTrend==0.15). Relative risks reached 1.9 for those exposed between 20 and 30 years, though this dropped off slightly in those exposed greater than 30 years. There was insufficient data to examine HT disease for exposure durations over lOOdBA, and numbers were low for the 95 dBA-threshold analyses. No exposure-response gradients were evident for ischemic heart disease, nor for stroke; relative risks for these outcomes were close to one. Analyses of the Pre-1970 Employment Sub-cohort (n=8,700) A total of 8,700 subjects provided 247,347 person-years of follow-up between 1950 and 1995 in the sub-cohort of workers who completed all their employment prior to 1970. Table 26(b) gives summary SMR's and 95% CI's. Overall, SMR's were higher than for the full cohort. This would be expected as all subjects terminated employment at least 25 years before the end of follow-up and the healthy worker effect diminishes as a population ages. Comparing to the general population of BC, the SMR for "all causes of death" was slightly higher, as were mortality from circulatory diseases, acute MI, all cancers, lung cancer and accidental death. SMR's by duration of employment in the sub-cohort (Table 27b) are higher than those in the full cohort for all disease categories, except stroke. SMR's were significantly above one for all causes of death, circulatory disease and ischemic heart disease, but decreasing risk with increasing duration of employment, illustrative of a healthy worker survivor effect, was still evident. SMR's for all causes of death and selected cardiovascular diseases by duration of exposure above selected noise thresholds in the sub-cohort are shown in Table 30. Some excess mortality for acute MI was evident for all duration-of-exposure metrics, and was highest for exposures above thresholds of 95 and 100 dBA, where for workers exposed for more than 20 years, it reached 1.7. HT disease mortality excesses were still evident, and were greatest in the longest duration category above 90 dBA, but the number of deaths was much smaller in the sub-cohort (n=16). 90 o o c « cT o o 2 73 CA U | > o o z o u 3 o ti .0 *+* a u 3 0 .8 y CA & a V u 8 U •o (fi ti o U KM < < < oa T3 < m •O ti « CO 3 •5 "3 oi ti o 73 Ov ON •<t r-^  c o j= o y < CQ -° < 1) < ca o JO < o 0-ON CN 2 a H O O O N Os m — t~-fN fN fN O O — T f o T vo — O O O — m fN fN fN rn *n vo O O O O — o o — — oo m r t r - rj-m fN vo I I 1 o r - » so m o v i fN O O O O © p T t O N O c3 u o> B3 O £> © fN rn — I i V rn — fN A O N O O N O o O N r -— O O O oo N O r - oo N O O O O o v i r» o — O O N O O N vo — O — © © — vs O N (S) — — vo *0 V I — — T t — m o N O — m fN fN o o o o o — — o — o O O N vo O VO O N t-» m — fN — — — V " ) O O fN O N V I fN fN O p — — — — fN NO O O JL, O O r - NO vO O o o © rn oo NO o p O N oo oo © © © Os V N fN O NO o fN fN —1 1 § Is vi vi e a s (3 V <D 4> >% I I fN «n vo — T t p p p — — fN rn O O O O oo p- T t fN fN fN <n oo O O O O oo r - fN m m fN *n oo > o — — r - m o o o o o — r» oo v i O O N O N O O N — O O — O v> vo in O N Os oo r- vo vo fN — — — O O VO oo p IN <N CN —i P1- 00 VO 00 o\ t » o o o o — O O O O m O — — v"> *n Os O fN T t — — fN fN — Vi Vi e a a 3 u u « E S » >> > • a -* a ^ o o » , ^ u o CN C l >,j 3 •*> - ' | | O I o © " > V CO — <N A VO NO o m in — — fN oo VO T f NO T f © © © o v o r - o o — O N — — o o — — — o fN — O T f h VO h -O O O O — o o — o vo T f i n Os TT fN T f p TJ- os © O O NO O O T f © ' © © o O N O fN t> — o o os o — — o — o — T f — NO oo r~- t— >n o o o o o Ti >n — m o — o o oo oo o — — m T t os os os in Vi Vi e s sa cd u u „ a ^ > o o ^ W O CN CO i ^ 1 o o C<l V CO —i (N I y CO •c u CZ, cj •S O ca -i> o •a § 55 < w -c -t-» 3 O on ca a! T3 ca i f ° i-T P a >> CJ S >-i a CJ CB .>. y c o U ci= t; ik % a l <f ° A l l CJ CJ O H CJ CJ 0- >> 2 _ fe « a. >-I I I Ix o — o o o — O O — os >n O Os v> \0 —' o o o — — o — O T f cs cn cs — — CN — - H o — — O O O O O in C-— — in vo — — o o O N T f — - H IT) T f — V"> O M M Tj-ON ON NO O O O O O O oo o o fS - (N O N — O O O — — — o T f o C N o xn -o rn vo cs T f m * — i O O O O r o oo O N — cs r - T f — CN — O N oo O O O — T f i n CS O O O O — — — o fN r - C N r» T f oo r - cs o o o o T f N O m O O O O vo O O O o — o o >n O N r -rn vo in m vo • * vo in 202,1 30,91 o " 3,965 m T f m m O O NO — ON O (N oo ON T f O O T f oo CO NO o — — — O — fN — — r - O N vo O N — T f in fN fN T f T f — T f m — — — — — fN — fN r n — »-i —* — — —- ^ — — — —" C N o in v£> T f fN T f O oV r-^  Ov fN oo O N O NO in NO N O O N oo O p ON O O O N O O N ON in o O T f oo p oo r - ON O N O N O N O O m o" O O O d o o O O o o o o O O O o O o O — — — O VO ON CN — T f m m vo — ON VO T f vo o — cs oo in oo C N m — O N r -oo r - — o CN m N O N O T f m cs — O o — O O (N CN T f — — — O — m oo — r - cs T f fN — — O N O — — O NO O O rn O O f N — ' ON — — o o — — — o in O N T f cs N O O O N T f oo — r-» t-» T f ™ , o in o r -— O N m — O O O O — o m r n O O fN CO T f T f T f CN fN I I * - » O N O m o © — — o — T f i n cs in o — o o O O O O O O O O O — — O ;3 cn ^ V . o O cn — A G Q 1 § I § 1 ^ 7 o cn 1 o cs V m — A * 2 o o O — A cj I o cn •3 v ?3 Wi " S « != u > N I 2 0 >> «? o o cs ' —' A E " § E ^ CS — 3 " i 6 8 ^ V co — A P3 CO ^ c^  ^ 7 o i CO I o cs v co r; A i o o S •a ca I •a 5 cj ^ tfj Cd 3 9. < S o .2 —< w A n . . C Internal analyses of the sub-cohort (Table 31) showed statistically significant increasing mortality gradients for acute MI, for exposures over thresholds of 85, 90 and 95 dBA. Within duration categories (i.e. within the > 20 years category) relative risks also increased across threshold levels (i.e. from duration above 85 dB to duration above 100 dBA) in a generally monotonic fashion. The sub-cohort also demonstrated statistically significant elevations of risk for ischemic heart disease, reaching 1.4 in those exposed for more than 20 years above 85 dBA (PTrendO .O l ) , and 1.2 for more than 20 years above 90 dBA (P T r end =0.05). Interestingly, relative risks for ischemic heart disease dropped off at higher thresholds, i.e. duration of exposure over 95 and 100 dBA. The relative risk for stroke remained generally close to one. Relative risks for HT disease mortality in the pre-1970 cohort did not differ substantially from the full cohort. The weak exposure-response pattern that was evident for exposure over 90 dBA was weakened further. Cumulative Exposure to Noise SMR's for cumulative exposure above 85 dBA in the full and pre-1970 cohorts are shown in Table 32a and 32b, respectively. The results of analyses of cumulative exposure generally supported the findings of the duration of exposure analyses. In the full cohort, there was little evidence of increased mortality; risk for HT disease peaked in a middle exposure category at approximately 1.5, and no other SMR's exceeded 1.1. In the pre-1970 sub-cohort, SMR's show the same overall pattern as the full cohort, except that statistically significant excesses of 20-30 percent for acute MI became apparent in the higher exposure categories. 93 < © o •a c m ON <=T ON uT 00 < 03 — — ro NO d o rs < N O o —. — o — —. o ii 0 •s O ' 0 ft w a o •a a 03 — — — o vo v> m ro r- — — ' C N — oV P^ -O O O N O O O O C N O N — — O N O O N N O — C O C N CQ p* oo — - o i NO N O O N O O — o o — — O — - H O O N C N N O O — —< O O — ' — — O —i co oo — r- C N Tj-C N — — O N N O O O co C N — £ . 1 73 ( N O O O N N O O0 O0 O 1 < O 00 —' d v> vo d — — o — — •o O w-i C N v i 3 3 5 3 a Q o 8! s 1 £. •a ro Q v trt s 2 ? 94 o c s VO o r o 1 r -SO 1 Tt Ov I o ON 1 00 r-i w o ov d O d d d l j wo — o ov cs oo 2 ov r-. ro o r o v o _ oo t -;_J wo ro — ro vo oo wo Tt ro WO C S cs 00 00 Tt o cs Tt ro VO C S . cs C N . ov C O oo cs r-ro o cs ro cs 00 Tt C O vq Ov O VO C S Ov C S Tt C S vo Tt Tt OV cs cs cn cs' ' -~ ~ *-< ~- cs" d >A Ov r~-VO VO wo vo C S Ti-ro o wo Ov Ov ro o Ov ro Tt Ov oo oo oo cs o o oo OV o oo o ro ro o O o d d d d d d d d d d d d d d d d OV o ro Ov o Tt Tt OV cs cs Tt OV © wo © o o oo C O o Tt o C O cs oo cs •Ti-ro o © cs o Tf OV wo o VO cs o o d d d d OV o cs ro Tt Tt Tt OV C O oo f» Tt cs C O Tt c--VO C O o cs OV vo cs wo Ti-ro cs VO 00 no Ov 00 o o OV 00 00 cs cs o cs UO OV ro vo ro CS 00 cs o so o o o Ov OV Ov © © © © o 00 cs © o CS cs ro cs cs CS —' d d d cs cs' cs' —' 00 CO 00 ro vo T3- oo oo ro cs CO Tt CO oo VO 00 Ov «A OV CS cs oo 00 00 cs OV OV oo 00 Ov 00 OV oo Tt Tt o Ov 00 wo 00 00 OV 00 00 00 SO Tt d d d d d d d d d d d d d d d d d d d d d d d d d d d d d vo ro OV oo o CO cs cs vo ro VO OV o cs Ov cs oo Tt VO r o vo ov Tt o OV OV oo oo o Ov o Ov OS o © wo ro o oo o Ov 00 OV © © © © © OV OV 00 d d d d —' d d d ' —' d d d d d —* d d d d T}- C O oo Ov wo vo wo ro cs ro so so so TJ- — . o wo ov ro — O CS £ cs cs cs « m so oo — wo C S C S C O so cs cs cs —< Ov ro O vo OV 2 Ov SO — oo r - m cs ro O — 00 ro OO ov ov r-~ ov Tt = 3 = st f <r. o o — W-) O o — v 6 V O W1 o '5 2 O o. as ac 1 < C/5 95 Table 33 shows relative risks for all outcomes from internal analyses, using cumulative exposure < 100dBA*yr as the reference category. Only small excesses of risk were evident in the full-cohort (Table 33a), and there were no apparent trends. Acute MI in the pre-1970 sub-cohort (Table 33b), however, showed a clear increasing trend reaching a relative risk of 1.6 in those exposed above 115 dBA*yr (PTre„d=0.001). Table 33: Poisson Regression: Relative Risk and 95% Confidence Intervals (CI) by Cumulative Exposure (dBA*yr ) above 85 dBA. (a) Full Cohort (27,499 Males); (b) Pre-1970 Employment Sub-cohort (n=8,700 males) Outcome (a) Full Cohort (n= =27,499) (b) Pre-1970 (n= 8,700) Health Outcome Cumulative dBA*yr Obs. RR 95%CI Obs RR 95%CI HT Disease < 100 7 1.00 7 1.00 100-105 7 0.73 0.23-2.34 7 0.92 0.26-3.24 >105 12 1.00 0.39-2.36 12 1.02 0.31-3.40 P value for trend 0.904 0.901 Ischemic Heart Disease < 100 230 1.00 191 1.00 100-105 217 0.93 0.77-1.13 183 1.16 0.95-1.43 105-110 205 1.02 0.84-1.24 147 1.31 1.05-1.64 110-115 129 0.92 0.74-1.14 72 1.17 0.89-1.55 >115 35 0.78 0.56-1.08 13 0.92 0.52-1.61 P value for trend 0.276 0.157 Acute Ml < 100 226 1.00 174 1.00 100-105 228 1.09 0.90-1.33 136 1.13 0.89-1.42 105-110 231 1.17 0.96-1.41 120 1.36 1.07-1.72 110-115 165 1.13 0.93-1.39 71 1.48 1.11-1.97 >115 60 1.08 0.82-1.42 19 1.59 0.98-2.57 P value for trend 0.273 0.001 Stroke < 100 99 1.00 76 1.00 100-105 105 1.04 0.78-1.38 77 1.07 0.77-1.49 105-110 93 0.86 0.63-1.16 52 0.98 0.68-1.41 110-115 60 1.00 0.73-1.37 36 0.94* 0.62-1.41 >115 16 0.70 0.43-1.14 P value for trend 0.245 0.695 HT - hypertensive, MI - myocardial infarction, Obs - observed deaths, RR - relative risk, CI - confidence interval Adjusted for Calendar Year, Age, and Race (South East Asian or Other) *>110dBA*yr Analyses Of Outcomes Occurring During Employment, or Within One Month of Termination When follow-up was restricted to subjects' period of employment, relative risks increased markedly (Table 34). SMR's for this group were low, as expected for those currently employed. For example SMR's for those exposed above the 85 dBA threshold ranged from 0.34 - 0.63 (not shown). However, relative risk of acute MI reached 4.0 (95% CI 1.8-9.3) in those exposed greater than 20 years above 85 dBA, and 2.0 (95% CI 1.0-3.7) and 2.7 (95% CI 1.4-4.9) in those exposed greater than 20 years 96 above 90 and 95 dBA, respectively. Tests for trend for duration of exposure over 85, 90 and 95 dBA were all highly significant (P-rrem^O.Ol). Similarly, acute MI showed increased risk for cumulative exposure in the restricted follow-up analysis (Table 35), reaching 2.4 (95% CI 1.0-5.7) in those exposed over 115 dBA*yr (PTrend<0.05). Restricting follow-up also increased risk of ischemic heart disease, but only in those in the 85 dBA threshold group. Relative risk peaked at 2.1 (95% CI 0.87-4.8) but this was in a middle exposure category, and test for trend was not significant. Relative risk of stroke remained close to or below one. There were insufficient deaths to include HT disease in these analyses. Disease Latency Full Cohort Because of uncertainty with respect to the underlying mechanism of disease, latency intervals of 5 years and 10 years were investigated for all outcomes. For acute MI in the full cohort, lagging exposure for 5 and 10 years led to very small increases in relative risk, and marginally strengthened the exposure-response relationship for duration of exposure above 95 dB, so that the test for trend achieved statistical significance (Table 36). For HT disease, ischemic heart disease, and stroke in the full cohort (Tables 37, 38, and 39), there was little change in relative risk for duration of exposure metrics. Pre-1970 Sub-cohort Exposure lagging for both 5-years and 10-years had little effect on relative risks for acute MI and duration of exposure in the pre-1970 sub-cohort (Table 40). Again, however, the tests for trend for all metrics were mildly improved, and for 10-year lagged exposure the relative risk for acute MI in those exposed for more than 20 years at > 100 dB reached 2. 97 2 fl <u a >> _2 a S W T3 O •c u PH 11 U CO OH D I o o e 11 A V •o <u +H u u t/i i H <2 tt J I a « a -3 <u .S fl ' „ cu cu 13 CO <S SI fl « CJ Ql •O fl i> I fl o o OH U ov < CQ -a o o > 1-3 •a > < o < 03 T3 o OV u > , o < U -S OV o •a u < CQ -a > , o < CO _v fl O ll •% s l> •5 A o s u fl O O T t NO CN — ' r t O N O O O N O CN iO NO —I —I O ~ O co co *n v-j r- N O T t O N ~H P- oo r -<o oo o o O ~ - CN o O N r o —' o —" co co T J -ON V") CO ON O T t O — " — ' — —• —« CN CN v i v i T t r - - . r o co T t O T> h i n o o o O P- T t CN O » - O O o ~ o O N V~t r o O N oo r-; O CN co I I I CO T t CO — f- o o o — O N O N O N O O r o T t O N ~ ©' —' —' CN T t r -co r - CN I N O O O V I O co V I N O 1 O V") O CN I - - <N -CN O O O O CN O O O N O NO T t i ^ CA CA CA s s s <D 0> <D co O O O j I I co O (A CA CA CA t ^ ^ U <u u a> a>, >.j CO O O O w — CN CN I I A O ^ O T t —; © CN co w O O O O N CN O N O T t • —' o d ol CO CN CN CN f —' T t V ) CN CN CN o v-j O O O N O —' © © O CO O N CN CN i i O N V") CN CN o" © O O N —I o O i 5o So ?o 5o . co O O O w •—' CN CN I I A -2 — •5 « & s u CM u T3 u ,3 II — Tt- oo co o\ vo vo o o o o C N vq T t v-i »n j o — — o — o o — — C N C N CN <*> CS cn vo co o o O C N O O O C N T f —; — o '"8 HI o •B I w - 1 •° 9 o § 3 99 o U C N N O i - i V ) C N r o V") C N C O O T t p o O C O i—i O C O C O o a >— 1 C N C O ^H C N »—I C N OH =? -1 C O >A r-Ll N O O O _!< V1) O T t V ) O N *n O N O O O O p oo O N CO O N o o O o o o i o o O ii < ca Wot POO vo Wot O T t «/~> o N O o oo r o N O O O • o P 1—1 r- p p oo T t o p O N r o < N > 2 — o " t-J ci - H o - H 1—1 i—i o —t *H CU £ 3 CA o N O •1— o J >n O N i n T t o C O ^o C N o r- oo C O C N O N oo C O C O oo T t C N BC BC BC o o f - T J - C N C N »n C O C N o V I T t in 0 C N C N V") O O C O C N N O oo T t T t T t in i—' i—< <—< — ' 1—< r—i i—I C N <u O O V J <n •4 Tj" o o O O o CO V"> oo O O O N O N O N oo p O N O N O N oo p '3 O N o © © O O o w o O O o 1-H z 0 ffl •M •o ii •r, Ov Ov 00 < N fH o N O O • o o O N C N O O oo o N O O N N O r- < N CQ a > pc p p p C N C O <s p O O C N C N C S p —J p T t <=. o DH —H —H —H c> 1-H i »-H i <—i — — 1—<' i—i c; idx Ab W o CA O O C N T t —i O O N O O r o c o C N O O T t T t T t C O T t r o r o o o '+J o «/~> r—< C O v~> r~- C O i—< «n r o « Dui O N —< C N C O T t C N T t r o oo Dui CI p C O C N _' C N C O C N C N r o r o T t T J 2? N O © O N «A O N N O O N T t N O N O O N fl V I oo O O r- O O O O r- o o O N O N ta O N o © © o o o o o o o O o a < op m •o u IN o Ov oo > O o o U O ?~> O *n C O oo N O o r o V ) C N < N £CH p O N O N p O N r\ p O N p p O N V O p p T t p < N o Cs o O o O O <=j 1-H i—I 1—i —I c> s < •3 Q cn v> • n O N o r-i O O C O o N O o O N o o N O N O o O N oo T t O o r o o S C N 1—1 t— "~™1 oo C O ^ r o C N C N i—i N O 1) N O «n _ C N oo oo T t N O O N o o 3 HH O O P C N C N C N C N C N C N T t T t T t o o —' — ~* 1-H i—I i—I i—i - H < j O O o o O N N O -4 C O T t O N o oo a" <r; N O oo oo r- O O O N O N oo Ov o o o o o o o o o O O o ,0 < *co m CO T3 Regre </~) O O • O v » Regre u O vi r- oo O O vo o C O C O r o oo O C O C N C O T t V O Regre > O oo O N O N oo p O N p O O N p p C N Regre o 2 1™H o © o " o C ! O - H O o CO' a < 0 CO CO 0 CA r o o O N O o o _ H o C N V I N O PH 6 v~> O N o C N Tj- o C O oo o C O oo o CO C N C N C N C N C N C N C N 'co >, >, terval | CA ial enc terval | ars ars ear terval | <u < fr les). cd s Oy ?o >o o fl les). Late C v CA Trend cn CA Trend m cn Trend V O C O O v T t S3 'ears year year cn & for 'ears year year S2 cd * cn ears year year cn a for: J U Cvi II o o o tu 32 o o u <TJ c3 o o o> % ? u o C N C O >, -5 o C N r o >v u o C N r o >> "s >> | 1 o a | | O So | | o H Dui <3 I C O O o C N >3 < 3 1 C O o O C N >3 <3 1 C O o O C N >3' a. ? 8 1 0 0 ON O N CN C o A 0 U 3 OJ . f i £ y CQ o *-3 co O w o e o s p a ' c» 8 u V V B c o SJD e o o Pu 8 cu CO •8 1 s o m — t-^  CN CN CN • — • O O O — r n fN fN fN r n <n vo O O O O — o o — — oo m r t p- r t m fN \0 T t o o o o O O O — T t T f © © T f O N © e3 *™ 3 w u « 12 « ^ c 3 o o u U © fN r n >i ^ —' | | © ^ I O O ^ V m — fN A vO t " l T f O T f fN fN fN ° . r - r - m — ' © © ' © — m m — (N Tj- i o p^ » — © 00 — T f VO V") O O O O — © — — fN oo m v i vo T t T f T f r - iA r n m fN »n o o ' o T f vO T f O O T f 22 S | m un r - vo OO fN vo —; fN v i vo OO O O O O —' O •—' CN •—' oo m vo vo m fN T f r n i n oo T t fN <n CN O O O © T f © fN © O O V O O N — ©' — ©' T f p~ T t oo m B « w « 3 >> > N a W O CN c n >NJ >> — , , V 1 ° ° V c n — CN , BO < % 1 s 3 g S 101 oo CU CO 1 1 .3 S O SO O vo O Os f- r -— O O O O O O O o in o — O Os O Os vO — o — o o — VI OS CN — — VO vj VI — — Tj- — ro O VO — ro CN CN O o o. o o — — o — o O Os VO O N O Os r- ro — C N — — — V> CN VO OO — OO t** vo vo O O O O O ro oo vo O O Os OO —' — o o ro Os r~~- f~) CN CN O — VO O — CN CN — — 2 c5 c3 E3 « 4> « ( U >i cd ^ o o <D O CN ro >• — i , V ro — CN A — r t m CN Os T f — o — CN — CN T t «n o o o O Os 1 - H ' O ro VO — Os oo t*- t— ro O O O O O — T f f -O O Os Os — — o o CN O ro O Os oo oo vo O O O O — — o — o oo ro VO OO VO O vo — CN CN — — •n co in os oo r~- t— O O O O O — O Os — — — o co T t r- o in cs os in CN — CN — — Os 3 ^ O O 3> O CN ro >-> — . , V ro — CN A O ro T f r -O Os VO Os — O O O CN CN ro — »n vO O CN OO t*** CO O O O O in r- Os <=> Os Os in O o o O O O O O ro r- T f p — OS o oo — — o — o T f CN ro ro VO — T f O CN CN — — CN CN CN CN os (•"•- r— t— O O CD O — — o o o m o CN «n £ S3 « o N n V m — CN c/5 cj iG ccj cj 0* 15> >• -a £ ca 1 a" # 8 CN A 102 tt o •fl o U a ll u « I 1-1 O N . . O N 2: * ja l l N O O V J T t r -N O CO CN oq — ' ~ - — ' ~ - ~ - ~ -CN N O N O N6 CO O N O O T t T t O O T t T t © © © © © d T t o N O o O O CN T t N O © O N N O p CN O N © © © © d V ) T t V I V J T t C l CO O N r o CO O N v>. CO V J O N O N CO r o v i O O CO CN V j O N •—•' © —* d CN -1, © T t T t JL, N O r - N O r-- N O —-o o d d © © d d O N O T t v-j r - O CO 00 N O O O p O N oo p CO o p p oo p CO O o © © © d CN O O T t CO V l oo CO T t V J CM V ) T t CN vo CN V J T t CN V I N O T t v i O N O O CO N O o T t CN T t O V J CN V ) - H ^ — ' ^ T f O N o J+ O O V J V ) O O N O O O T t O N N O oo T t d o © o o d d d V J O N O CN o o V I P p O N p r o p CN O N .—« © o © d d r-» V I v> CO O N o T t ~* oo CN O O r - r - CN r o T t T t V J r - N O N O N O v i T t CN V I V I CO -—< •—< •-H i — ! ~ - I—« I—* 1^, T t N O oo N O o 00 r - V I oo N O © o o o © d d d o IT) v j r o V J o T t o O N p *""I p p O O cN p CN p O N © © d oo o CO CN CN o O N O T t O N O N O N V J V J O N O N O O V J e3 2 a> c« e » s M O O JJ ?J O CN r o > "*> — | | O " 1 O O m V m - H r l A . 2 a a a o o o <L» O CN r<> > ^ I I o m 1 o o ™ V f-> — CN A O CN C3N O O CN t— 00 — — d o — o o NJ- CN — CN — ' — ' — O O O O O O O — — o — o o o o o >. — "> 1 o V m — CN f^l I I XJ 3 V XI CJ c 1 CN A 3 Q -o a a a o •a o u 3 73 •3 « u o >> s 3 u •< C O W) o 3 J o < < < J8 < < 73 2 o o r~ os II OO T f —i c o V O OO v o o o o o O O N ON CN O — o r--U 0S-— — f N - ON OO O N O O O — 8 2 OO NO ^ ON o o U 0 s O N r- 00 O N O O O <3 v i — — O ON — r o —J o — —I 3s fr2 ti © o •c cu •o fr X i H cd u -3 ^ o J* O O f N ^ ~ I O ro I f N V ro — A f N — NO OO f N —i ro O f N OO T J - r- —. — — CN i i i Tj- \0 OO OO O N O O O — 2? — WO N O r» «n ro o o m o o o f N V"> O f N • § § £ " O N ' I I O V c o — v q o o v q — — ; P O • i i f - O T f c o —• —• — — ts Ov Ov T f o o o T t c o O O c o c o 3 O u © c s I o c o — A T 3 S c f U T3 o ^ T! 8 CJ 4J a >" 3 CO TT V <; * 104 Relative risks for HT disease (Table 41) after 5 years lagging showed large increases for exposure above 85 and 90 dBA, reaching 3.1 to 3.9 in the highest exposure levels. With small numbers of deaths however, the confidence intervals were wide, and the test for trend was only marginally significant. There was insufficient data to examine 10-year lagging for HT diseases. Interestingly, the exposure response relationship for ischemic heart disease was eroded with increasing latency time (Table 42). Stroke showed no change in exposure response patterns magnitude of response with lagging (Table 43). Cumulative Exposure Five and 10 year lagged cumulative exposure increased relative risks for acute MI to statistically significant levels, and produced a highly significant trend in the mortality gradient for 10-year latency (Table 44). Relative risks for HT disease also increased, while relative risks for stroke and ischemic heart disease were unchanged following lagging. Hearing Protection Use - Adjusted Analyses Exposure-response analyses for duration and cumulative exposure metrics were repeated using exposure levels arithmetically corrected for the use of hearing protection (L e q > A,HPD> see Chapter 4). In duration of exposure analyses (Table 45), relative risks remained essentially unchanged over unadjusted estimates, although for two acute MI analyses risk estimates edged into statistical significance. Relative risks for acute MI increased for higher levels of cumulative exposure, however (Table 46). The increases appeared to plateau at 110 dBA*yr, but produced a statistically significant trend. Relative risk for HT disease also increased in the higher exposure categories. Little difference was seen in the re-analyses for ischemic heart disease and stroke, compared to the original analyses. C-WeightedNoise Exposure In general, C-weighted results (emphasizing low-frequency exposure) mirrored A-weighted findings for duration of exposure analyses (Table 47). Use of the C-weighting scale had the overall effect of increasing exposure levels, and so numbers of subjects, and numbers of deaths were increased in the higher exposure levels. Trends for acute MI at the highest thresholds were strengthened, but relative risks did not increase. Increased relative risks were seen for HT diseases for those exposed more than 20 years above 95 dB, with relative risks of approximately 1.4. No substantial differences in relative risks estimated were noted between A-weighted and C-weighted cumulative exposure analyses (data not shown). 105 o •a a T3 fr "S. S 9 II <N -. 7 <N "? h N » O — •* < < pa 1-2 < 8 —< o O OO CN CN T t co ro NO Tj- o © CN — T t d © © © ON oo ro O OO V) \ o — " d o — " Tf- vj CN V) d • ~ S H O vt o> So rt & ^ o S W O CN So — ! 0 CO | CN V co — A CN Vi —i O t— ro OO © T t © © - H " O ro" OO CN CN T t Vl OO ON ro —< CO CN OO o d d O N V) — T t v j CN Vi CA 0> a >< >^ ° ro | V m o S < * 106 •a fr a o CO .3 ^ a « 31 s 5-« o o CJ cd CO CN t— t— m cN o d d o c n oo SO <N o p V i NO OO — ' O o r- r- — O O O N N O o d d C N r-» C N r-r t oo C N T t T ) IT) C N — O N ro O N T t O O N d —^ d — ro O O — I C N T t C N — — O N V i T t T t ro O 00 N O V l C N I 1 V ) O N T t d d d v i ro oo vj — — — O — r - N O ro r » «— C N C N — — O O S3 > > 0 £ CJ o cs > N —- I O O N —; o N O O N O O C N d C N O N ro d d O O V I N O ro O N d © ro C O —i ^ - ~ - O NO" O O N O V I C N — ' O N N O O O N O N — ' d o O V i T t r- o C N — C N — § § e W O N " | O m | CN V m — A 107 < © ON — © fN NO o l< © CN c*^  © o © © «r> ON © CN O N V I — _ ; © ' © < co J o co "n © ' © ' © ' © ON CN NO © —' — OO —' —' —' © ' o © O O < it < 73 2 o o r-oo II 3 £1 o •c P. TJ cj fr a S | I § CJ > N H CO crt CJ >, 3 >>© CJ © CN — | "> I © V O - t IO — — o O O O — — o o 5 o «n 2 r r <N O S ^ r -r» v~> ro OS f- vo o o o o os oo CN O CN O Os —: —: —; o r- vo o oo oo vo vi co CN o 00 Os i n —' —' •4 ro oo oo vo o O o 3 o cn CN CN — r~ >n — OO vo CN j O © © © cn cn O CN © - - ' <=>! 3 NO V, ; £ cn CN oo v o *r> ON ON OO 00 ^ O O v o CN | O CN CS - ' - ' - ' o | I J 3 6 o c .3 m l © V cn — a" | £ I C4 I T ) C 13 u T3 3 |3. 8 T t O T f OO M n - (N O O O O O rn P- O OO O O O O N OO —'< —'< —'< o o T f «o r - o o o r - o o m (N — — m j °0 j O O O O O O m O T t O N o o o O N r » —' — — o o fN N O N O O T i N oo N (N fN fN — — m T t OO p ~- —' t -T t 0 0 T t t -VO o o o o T t < ca O n N (N OO O O N O O N t -—' o — o o o m O N D W _ H | v 6 >A 6 A •a 2 o -r - T f o — T f NO V) OO — — o O j O O O O o ^ t -O O O N O N — — o o O N CN OO O vo Ti fN rn T f . , fN fN fN — Ttl t> O O N O m Ti rn v© o — o o 1 T f Ti — O rn , — — OO »T) — m oo »n NO m C N m v> ool fN fN fN — m T f m T t O O O O o o o o — o o o o N O T) o o oo o — — oo N O m oo m m m 5 r - vo r - T t j O O O — — O — o V O OO — Ti fN fN m N O _ , fN fN fN — N O 1° o c .a -s TJ c . 60 < 13 109 < CQ O ov V) ON < ca s < J o 1X1 8 O N CN O N O T ) ^ d —' — — o o — C N t-» v o o T t ^ i s r - ; \6 r o T t \d r - -oo CN r- — CN T t r- m d o d d O >n r o r~~ r o I O O N CO CN "~ — d — c N —I oo T t v i r - CN I o o o o — — — o o I CN O — CO CN — O VN — — T t — o o o o O O O — —' T t o — o o oo —' — —J — o O NO OO O o CN CN CN — — C N T t ON c o C N — — —« CN O T t ON ON O O O O T t — —* d —* NO OO ON OO P - r o CN V i T t T t T t O O O O O do—' OO T t T t o d d O ON NO NO o — t~~ O O — — do o — v i f - r o — CN CN v i — —< r - - • 1 o o o o O O O N o © o r- r- r o o oo oo oo — o d d i n r - NO o CN T t T t CO o p - SO' —« — oo >o C N O O o T t d d d d O NO I— ON o o O O o —< —< d —i d O O O N m IT) — ON r*- «n CO — — — NO CN CO C N C N ON — OO r - f - — < N O O T t — CN VO* O N O O O N O o — d CN CN V I CO — oo r- r- NO d o d d O — TT CN O O ON ON — — do o r - o r » N O O OO T t h OC h OO © © © © o co oo r-O ON ON ON — o d d — I NO CN r - — t- © ON — — CN — CN — ON <*o v o VI C N CO — — —- —I — C N C N ON Ji* OO N O N O MO d d d d T t oo N O oo ON r -d d d <n r o T t r o O N oo O N T t is 1 Q -5 CN c o >N, -5 I O V r o — C N A l l « u s o S o " ' o o .. u u fa frt 4» SO S o 3 3^  O CN r o cu SO — - I o M " > ' O O Q v r o — CN A 3 1 o o ^ ^ O CN r o X V r o — ( N A | . « ^ o o ? j O O CN r o 1^7 i i i m l o o CO V r o — f N 3 « CJ T3 IU U 4> V eft ii o a V o u a CO 3£ § . XI T3 O v CJ co .2.3 •a P-"? 2 w <U PH 3 OJO S •3 « <u a 33J CO l " M it oo > « o 1*8 cj 3 CO i i i I! = 1 o g CO 3 .a u 3 I XI a R « in o o o o « " o o o o d o CN » n r ~ » o o o o ON o o — — — l o o ON T t O c n ON O O ON O —J — O T t c n o I r-~ N O N O o o o o l — o o o o l T t Vt NO O T t CN ON CN — CN CN — — c n T t o c n c n WN c n CN _ , CN CN CN —• "O P3 '"8 '1 < ca •a o o •a & CM in o\ © " o\ oo o J3 CO <u H U CU o JM *> . -CM gj co CM S S «fl Os 1? in t N CM " g '•3 *H CM CJ •a V •M u ti Ti C/3 « o &» <* « 3 X) CM •a J Cfi " < < < II tl o j= o O < ea (N tN T j " Tj" h - *0 vo CJ\ C N — T f O O O O I — o o — — oo </-> m so T J -c n m rN so o o o o o T t N O r o O N T H OO OO ON r - (N i n r n CN* VO ON f » r t fN I fN O Vj fN O O O O 2 8 a . •i -C J O J CJ 8 O ' 3 oji CM <+; I o vo o o\ o O T f 00 CT\ —" — o —«' o 8 T t T t «*n o — CN •— rn o o o o — O — O O r - TJ- o r - r O c n CN m I T t — — V I fN OO T t P- VO — CN fN O l o o o o o r - fN r o — O O N o O OO ~ o — —' o CN OO fN O S T t oo oo c n vj I in T t oo T t r o r o —i < o o o o O P- t"» r o P-O O O Os oo — — —' o o — o oo m T f fN — — so O — CN CN — — o o o o O — Os so so O — O Os OO I —" —: — o —. — O ON . . , — — C N vo O — (N CN - H — u cu wj £ £ ^ ^ -3 «j ^ O O " 1 S> o cs c n e s s , ed <L> <L> v> 1 1 £ CO I 42 V c o — CN A o o o o v> vi oo Os v> c n c n os T f — — oo m o o o o O en OO NO oo O ON O O ON —" O —<' — O p - T t N O so so O N oo vo p -fN — — — O N O O O O —• o — — o p - r t o C N p -T f Os O C N T f — — fN fN — O O O O o -< ON r - vo o —' O Os oo —" —' — o o fS — O V) fN T f O N O CN V I — — fN fN — o ^ I o o ^ < V c n — C N A o o o o o o o o O SO Os V j p--O — O0 Os p -— — O O O O Os T f O 0O i> P- v D O O O O , _ T f CN NO ON o — — o • H oi o o o o o c n NO r n _ p — — — oo '1 IK vS> m | V f o N n I I Assessment of the Impact of Smoking Current (1997) Smoking Habits in Sub-cohort of1,960 Workers Table 48 shows mean duration of employment by subjects' 1997 smoking status, stratified by age and ethnicity. For non-South Asian subjects (Table 48a) the slight elevation of overall mean duration for past-smokers was likely due to the age of the sub-cohort (mean 54.7 years, standard deviation 12.6 years). Older workers (with correspondingly longer employment durations) are more likely to be past-smokers (75 percent for over 75 years vs. 36 percent for 25-44 years). Nevertheless within 10-year age-categories (as used in internal analyses), duration of employment was very similar across smoking groups. There are larger differences among the South Asian results (Table 48b), though primarily among the youngest subjects. Estimates for the South Asian group were often based on small numbers of observations, resulting in unstable estimates. Table 48: Mean Duration of Exposure by 1997 Smoking Status and Age. (a) Subjects of European and Other Ancestry; (b) Subjects of South Asian Ancestry Smoking Status Age in 1997 Mean duration (n) 35-44 45-54 55-64 65-74 >75 A l l Ages a) European and Other Current Past Never 9.9(131) 8.2(180) 7.3 (187) 14.6(116) 15.4(195) 14.6(100) 22.2 (74) 21.5(163) 22.1 (81) 25.5 (38) 24.2 (206) 26.6(53) 25.8(16) 28.1 (131) 28.3 (29) 16.0 (375) 19.0(875) 15.2 (450) b) South Asian Current Past Never 4.5 (7) 9.3 (14) 12.4 (30) 18.7(7) 20.8 (12) 16.9 (88) 20.1 (10) 24.5 (9) 20.1 (30) 20.9 (8) 24.0(1) 23.4(19) - (0) 31.4(1) 20.4 (10) 15.6 (26) 17.7 (37) 17.8(197) Tables 49 through 51 give smoking habit information for the 1997 sub-cohort by cumulative exposure to noise, stratified by ethnicity, and age in 10-year categories. Overall means (Table 49) are similar across smoking categories in both ethnic groups, though they are relatively low for non-smoking Europeans and others, and relatively high for non-smoking South Asians. Table 49: Mean Cumulative Exposure (dBA ref 20 ii.Pa.yr) by 1997 Smoking Status and Ethnicity Smoking status European and Other South Asian A l l mean (n) mean (n) mean (n) Current 104.0 (375) 104.7(26) 104.1 (401) Past 104.7(875) 105.6 (37) 104.8 (912) Never 102.2 (450) 107.4(197) 103.8 (647) Table 50 shows mean cumulative exposure, by age category and ethnic group. Means for 113 Europeans and others, within age-categories are generally more similar than the overall means. Within age categories, the largest difference was 2.3 dBA*yr (between current and never smokers, 35-44 year age group). Means for South Asians follow a similar pattern, although non-smokers in the youngest age group have a cumulative exposure almost 8 dBA higher than current smokers. The reason for this in not known, but again was based on relatively few observations. Table 51 shows pack-years of smoking by cumulative exposure category (as used in internal analyses), ethnicity and 10-year age groups. The overall correlation between cumulative exposure and pack-years that was evident for Europeans and others (last column) was likely due to the underlying age-distribution, and correlations were weaker within age strata. Table 50: Mean Cumulative Exposure (dBA*yr) (n) by 1997 Smoking Status, Ethnicity and Age Smoking status Age in 1997 mean (n) 35-44 45-54 55-64 65-74 >75 A l l a) European and Other Current 101.2(131) 104.4(116) 105.6(74) 108.2(38) 108.2(16) 104.0 (375) Past 100.0 (180) 103.8(195) 106.1 (163) 106.6 (206) 107.9(131) 104.7 (875) Never 98.9(187) 103.5 (100) 105.9(81) 106.9(53) 107.7 (29) 102.2 (450) b) South Asian Current 96.8 (7) 107.4 (7) 108.0(10) 106.0 (2) -(0) 104.7 (26) Past 99.9(14) 109.2(12) 108.5 (9) 109.7(1) 111.0(1) 105.6(37) Never 104.5 (30) 107.0 (88) 108.9(30) 109.8(10) 107.9(10) 107.4(197) Table 51: Mean Pack Yeats (n) by Cumulative Exposure Category, Ethnicity and Age dBA*yr Age in 1997 mean (n) 35-44 45-54 55-64 65-74 >75 A l l a) European and Other <100 7.4 (253) 15.2(106) 20.6 (57) 25.1 (52) 28.1 (16) 13.2 (484) 100-105 10.5(132) 17.5(115) 20.8 (67) 28.2 (61) 27.1 (40) 18.3 (415) 105-110 11.0(94) 18.5(118) 25.1 (94) 29.8 (73) 22.0 (54) 20.6 (433) 110-115 11.8(18) 17.5 (60) 24.8 (79) 28.4 (70) 27.3 (48) 23.7 (275) > 115 0.0(1) 16.6(12) 17.1 (21) • 22.6(41) 23.3(18) 20.5 (93) b) South Asian <100 3.8(15) 1.4(10) 7.0 (2) 0.0(1) 0.0(1) 2.9 (29) 100-105 2.8(15) 4.0(17) 0.0(14) 0.0 (2) 0.0(3) 2.1(51) 105-110 0.3(16) 1.0 (41) 5.0 (23) 5.0(7) 0.0 (2) 2.8 (89) 110-115 0.0(5) 2.9 (37) 3.8 (26) 3.0(10) 0.2 (4) 2.8(82) > 115 -(0) 9.0 (2) 0.0 (4) 0.0 (2) 0.0(1) 2.0(9) Tables 52 through 54 give smoking habit information for the 1997 sub-cohort by mean duration of exposure above a threshold of 95 dBA. Mean exposure duration varies from 4.6 to 5.7 years above 95 dBA for Europeans and others and 5.6 to 8.6 for South Asians (Table 52). Table 53 shows exposure duration by age category and ethnic group. For Europeans the largest difference within an age strata was an increase of 1.5 years for 65-74 year old past-smokers vs. current smokers. Overall there was no obvious pattern of association between smoking status and duration of exposure. This was also the case for South Asians, though with greater variability within age category; where there was high variability 114 however, the longest exposure durations are generally in the non-smoking group. Table 54 shows mean pack years of smoking by exposure duration, stratified by ethnicity and age category. No positive trends are apparent, giving little evidence of association. Table 52: Mean Exposure Duration (in Years) > 95 dB (years) by 1997 Smoking status, Ethnicity and Age Smoking status European and Other South Asian A l l Current 4.6 (375) 5.6 (26) 4.6(401) Past 5.7 (875) 6.3 (37) 5.7(912 Never 4.1 (450) 8.6(197) 5.4 (647) Table 53: Mean Exposure Duration (in Years) > 95 dBA (n) by 1997 Smoking status, Ethnicity and Age Smoking status Age in 1997 mean (n) 35-44 45-54 55-64 65-74 >75 A l l a) European and Other Current 2.2(131) 4.3(116) 5.9 (74) 9.5 (38) 8.4(16) 4.6(375) Past 1.8(180) 4.0(195) 6.6(163) 8.0 (206) 8.6(131) 5.7 (875) Never 1.5(187) 3.4(100) 6.7(81) 8.5 (53) 7.3 (29) 4.1 (450) b) South Asian Current 1.0(7) 8.4 (7) 7.6(10) 1.7(2) -(0) 5.6 (26) Past 1.4(14) 9.3(12) 8.4 (9) 11.3(1) 15.1(1) 6.3 (37) Never 4.5 (30) 8.1 (88) 10.2 (50) 12.7(19) 9.0(10) 8.6(197) Table 54: Mean Pack Years by Exposure Duration > 95 dB, (n), by Age and Ethnicity Exposure Duration Age in 1997 mean (n) 35-44 45-54 55-64 65-74 >75 A l l a) European and Other < 3 years 8.7 (397) 17.2 (245) 21.5(147) 25.5(141) 24.7 (89) 16.2(1019) 3 - 1 0 years 9.9(83) 16.5 (109) 24.3 (88) 32.9(55) 29.9 (27) 20.4 (362) 10-20 years 12.7(18) 17.7(53) 29.9 (56) 29.5 (56) 24.6 (27) 24.6 (210) > 20 years -(0) 28.8 (4) 10.6 (27) 23.5 (45) 23.7 (33) 20.5 (109) b) South Asian < 3 years 3.0 (32) 1.8(27) 1.2(19) 0.0 (4) 0.0 (4) 2.0 (86) 3 - 1 0 years 0.0(15) 2.8 (38) 4.6 (20) 8.2(6) 0.0 (2) 3.1 (81) 10-20 years 0.0 (4) 2.0 (36) 1.5 (25) 5.3 (7) 8.3 (3) 3.3 (75) > 20 years -(0) 3.0 (6) 0.0(5) 0.0(5) 0.0 (2) 1.0(18) Lung Cancer Parallel Analyses Lung cancer rates in noise-exposed workers were examined in parallel with cardiovascular diseases during internal analyses (Table 55). A few individual analyses showed increased relative risk for lung cancer, but no patterns were evident, and no relative risks above one were statistically significant. The highest relative risk for lung cancer (RR=1.6) was found in a middle category of cumulative exposure. This was in the restricted follow-up cumulative exposure analyses, and numbers of deaths were 115 relatively small (n=12). No positive exposure-response gradients were evident for any analysis, in fact many negative exposure-response trends were observed. Table 55:Lung Cancer Relative Risks (observed deaths) from Selected Parallel Internal (Poisson Regression) Analyses. Adjusted for Age, Calendar year and Ethnicity. Exposure Metric Analysis group Threshold Exposure Categories Test for Trend (P-value) Duration <3 years 3-10 years 10-20 years 20-30 years >30 years Full-cohort Pre-1970 sub-cohort >95 dBA >85 dBA 1.0 (317) 1.0 (70) 1.2 (90) 1.1 (94) 0.7 (47) 1.3 (63) 0.7 (22) 0.9(41)* 0.5(7) 0.003 0.903 Pre-1970 sub-cohort >95 dBA 1.0(197) 1.02 (37) 1.08 (23) 1.00(11)* 0.6(11)* 0.68 (3)* 0.809 Restricted Follow-up >85 dBA 1.0 (9) + 1.2(12) 0.318 Restricted Follow-up >95 dBA 1.0(19) 0.48 (3) 1.11(7) 0.678 Cumulative <100 dBA 100-105 105-110 110-115 >115 Full-cohort >85 dBA 1.0(151) 1.1 (129) 1.0(118) 0.8 (64) 0.7(21) 0.013 Pre-1970 Restricted Follow-up >85 dBA >85 dBA 1.0(108) 1.0(5) 1.1(78) 1.55(12) 0.97 (49) 1.0(8) 1.0(27) 0.8(5) 0.9 (6) 0.8 (2) 0.898 0.325 1 > 20 years * <10 years Discussion Acute Myocardial Infarction The major finding of this phase of the study was an increased risk of acute MI associated with occupational exposure to noise in lumber mill workers. Increased risk was seen both in comparison with the general population of British Columbia, and in workers exposed to higher noise levels when internally compared to those exposed at lower levels. Overall, relative risks were not high, being generally less than two. However a consistent pattern of increasing risk with increasing exposure to noise was evident, and in several cases, statistically significant tests for linear trend were found. Exposure-response relationships were evident for those exposed over thresholds of 95 and 100 dBA in the full cohort, but the strongest relationships were seen in the pre-1970 employment sub-cohort. By limiting this sub-cohort to workers whose entire employment experience was prior to 1970, it was hoped to avoid the problem of misclassification of exposure due to the use of hearing protectors. Without accounting for their use, exposure estimates for years after 1970 would be systematically overestimated, increasing the likelihood of attenuation of a true exposure-response relationship. In the sub-cohort, strong exposure-response relationships were seen in association with cumulative exposure; exposure-response relationships were evident for exposure duration at all thresholds, and risk also increased across thresholds (Figure 19). Also trying to account for hearing protector use, internal exposure-response relationships were reanalyzed using exposure estimates arithmetically corrected for HPD use. These re-analyses resulted in 116 slight increases in risk estimate, though not as great as seen in the pre-1970 sub-cohort. Figure 19: Relative Risk of Acute Myocardial Infarction in Pre-1970 Sub-cohort (n=8,700), by Noise Threshold and Duration of Exposure The highest relative risks for acute MI however were found when the cohort follow-up period was restricted to only the period of employment. Relative risks rose to between 2 and 4 for those exposed for more than 20 years above 85 to 95 dB. The highest relative risk was found for the lowest threshold (85 dBA), though the more stable estimates were in the range of 2 - 2.7, at 90 and 95 dBA thresholds. Estimates, and tests for linear trend were all statistically significant. Lagging exposure by periods of 5 and 10 years had a weak positive effect on exposure estimates, with small increases in risk estimates and strengthening of some tests for linear trend. The analyses did not, however, provide much evidence that the pathophysiology of noise and heart disease involves a latent period. However, this was an exploratory analysis, as no a priori hypothesis existed. It is consistent with ischemic heart disease as a chronically progressive condition, and with the findings of the restricted follow up analyses (i.e., that risk diminishes with time away from the exposure). The effect of low-frequency noise in the cohort was examined by re-analyzing risk estimates with noise exposures recalculated using the "C" weighting scale (see Chapter 4). Some previous studies had suggested that low-frequency noise might be a more powerful stressor than broadband noise, but no major 117 changes to relative risks were observed here. However, negative findings could also result from misclassification of C-weighted exposures, as the dBC levels were not measured directly, but calculated from department-wide average frequency levels. Alternatively, other frequency weightings may be more appropriate, and need to be tested. Ischemic Heart Disease Mortality from ischemic HD other than acute MI was considered separately. Increased relative risks associated with exposure to noise were evident, but overall, they were weaker and less consistent than was seen for acute MI. Statistically significant exposure-response was found in the pre-1970 sub-cohort, where relative risks reached 1.4. Relative risk was increased to 2 in analyses that restricted follow up to the years of employment, but no exposure-response pattern was evident. Exposure lagging did not increase risk estimates, and no increases were seen in either the analyses of HPD-adjusted, nor C-weighted, exposure metrics. It is not clear whether the difference in pattern of deaths from ischemic heart disease (ICD9 411-414) and acute MI (ICD9 410) represents a true difference in underlying pathophysiology related to noise exposure, or whether it was a result of diagnostic or coding practices with respect to death certification. For example, Boyle and Dobson (1995), reviewing hospital discharge diagnoses, suggested that misclassification might occur from physicians incorrectly diagnosing a coronary event as acute, if a previous acute MI appeared on the subject's chart. In contrast, true acute events may be coded as chronic if the subject had a history of coronary heart disease. Hypertensive Diseases There were only 26 deaths due to hypertensive diseases in the full cohort (and only 16 in the pre-1970 sub-cohort) and so most analyses were limited. Confidence intervals were generally wide, and no risk estimates or tests for trend reached statistical significance. Nevertheless, in both the full cohort and the pre-1970 employment sub-cohort, SMR's and relative risks of between 1.5 and 2.0 were frequently observed, often associated with the higher exposure categories. Exposure lagging had little effect on the full-cohort risk estimates, but in the pre-1970 employment sub-cohort, relative risks climbed to between 3 and 4 for those exposed over 10 years with 5 years lagging. Stroke No consistent patterns of elevated risk of stroke were seen in any of the analyses. Strengths and Limitations The study reported here had several key strengths, particularly when compared to previous research on occupational noise and heart disease. It was large and thus had excellent statistical power for 118 examining weak associations. I was able to conduct a thorough, and quantitative, exposure assessment, reducing potential misclassification of exposure and subsequent bias toward the null. I was also able to account for hearing protector use, by two different methods. The study group had high and consistent exposures, but still provided sufficient exposure contrast for analyses. As described below, I was able to control for ethnicity, and addressed several other potential confounders including smoking, psychosocial stressors and socioeconomic status. Nevertheless, several limitations can be identified with the study. Classification of health outcome was based on death certificates, and coding and diagnosis problems with death certificates are well documented (Sehdev and Hutchins, 2001). Such problems lower the sensitivity and specificity of death certificates with respect to coronary heart disease. However because this was most likely non-differential misclassification, it usually would result in a bias toward the null, and underestimate of true risk. Data from the Framingham study demonstrated specificity and sensitivity of approximately 84 percent, which the authors showed could result in the attenuation of a true relative risks of 2 by as much as 40 percent (Lloyd-Jones etal, 1998). Smoking is a strong risk factor for all of the outcomes under study. As with all large historical cohort studies, no personal smoking data was available, and so the impact of smoking on the observed effects could only be made indirectly. No major differences were observed between smoking rates in the cohort population and the general population of BC, although the analyses were somewhat crude because of the limited age-stratification available for provincial smoking data. However, in internal analyses, comparisons of smoking status and pack-years of smoking by exposure level (both duration and cumulative measures) did not reveal any clear association between smoking habit and noise exposure. Also, Poisson regression analyses for lung cancer were run in parallel with cardiovascular disease analyses, but did not show any consistent elevations in lung cancer risk, nor any significant positive trends with increasing noise exposure level. It was assumed that if smoking were a confounder of the relationship between noise and increased heart disease in this cohort, elevations in lung cancer would also be associated with noise exposure. The indirect analyses are somewhat limited however. The sub-cohort only reflects smoking habits at a single point in time, and represents a group limited to those 35 years and older who survived between 1979 and 1997. However, because the cohort was almost entirely blue-collar, it should be relatively homogeneous with respect to socio-economic status, a predictor of smoking. This improves the likelihood that the survey group was representative of the entire cohort, and increases confidence that the internal (Poisson) analyses minimized the impact of smoking on the results. Finally, because noise exposure was assessed objectively and quantitatively, there should be less chance of associations between noise exposure and smoking habits than might occur if both were self-reported. Smoking habits and exposure to 119 noise have not been previously linked, except in one experimental study that found increased cigarette puffing in seven noise-exposed subjects (Cherek, 1985). Other recognized individual risk factors for cardiovascular diseases include family history of heart disease, high serum cholesterol, physical inactivity, and obesity, and are potential confounders. As with smoking, it was not possible to obtain personal data regarding these risk factors. Nevertheless, for any to confound an association between noise and heart disease, they would also have to be correlated with noise level, and there was no reason to suspect such an association. The large size of the study reduces the likelihood of unequal distribution of individuals with these risk factors across exposure categories. Socio-economic status has been associated with heart disease (Brunner, 1997) and a major problem of previous studies has been in selecting appropriate control groups; in industrial settings, non-noise exposed groups tend to be white collar while those exposed to noise, blue-collar. Subjects in this study were almost entirely blue-collar workers, and considered to be quite homogeneous with respect to socioeconomic status; thus, risk of confounding due to socio-economic status was minimized in performing internal analyses. Many occupational agents have been posited as risk factors for coronary heart disease and several of these may be found in the lumber mill, including psychosocial stressors (Schnall et al, 1994), shift work (Knuttson et al, 1986), temperature extremes (Kristensen, 1989), and carbon monoxide exposure (Stern, 1988). Of these, psychosocial stressors, including psychological and physical demand, have been previously investigated in this cohort (Ostry, 1999). Variability in psychosocial stress in the cohort was low, and no associations were found with cardiovascular disease. It was therefore not considered a potential confounder. Work-shift information was not available on an individual basis. However, the cohort was limited to hourly workers (primarily production and maintenance workers), and would be expected to be fairly homogenous with respect to shift work (mainly day and evening on a 1-week rotation). Further, it seems unlikely that there would be an association between shift-work and noise exposure. An exception might be low noise exposure and non-shift work in administrative jobs, but these jobs contributed relatively little person-time to the study, and should not represent a significant threat to the validity of the results. Carbon monoxide exposure may occur in this cohort, particularly among vehicle operators, but most vehicle operations occur outside, and CO exposures would be expected to be relatively low. Temperature extremes can also occur in the mill environment, particularly in the mills located in the BC interior that experience hot summers and very cold winters; however an association between noise and temperature, if it existed, would likely be associated with a job's indoor or outdoor location, and the determinants model (Chapter 4) found that average noise exposures varied very little by location. 120 Coronary health outcomes are known to vary on a regional basis, and some of the variability may be due to regional differences in acute care provision. As the participating mills were located in both isolated rural, and urban settings, region had the potential to confound. However, exposure modeling (see Chapter 4) showed that while noise levels varied by mill, there was no consistent pattern of noise exposure by geographic region. In terms of other limitations, the arithmetic adjustment for hearing protection use did not perform as hypothesized, and did not result in estimate improvements as large as seen by restricting the cohort to pre-1970 employment. It was therefore difficult to interpret the results of the full cohort analyses, and to know whether the lower disease risks observed in the full cohort vs. the pre-1970 sub-cohort are a result of true lower exposure levels in later years, or due to attenuation of the risk estimates due to exposure misclassification. Utility of an arithmetic adjustment factor for HPD use may have been limited by errors in assessing the true protection factor of HPD, the prevalence of their use in the cohort, and by assigning group means to individual subjects (Chapter 4). Similar difficulties have been noted by other authors (Babisch et al, 1990). A lack of exposure and exposure determinant data for decades prior to the 1970's meant that exposure level estimates for these early years were constant, and equal to the levels predicted for the 1970's. This was mitigated in part by evidence regarding slow technological and noise-control advancement in the mills suggesting that noise levels may not have changed substantially over this period. Many cohort subjects spent some portion of their working life in workplaces outside of the cohort. Some presumably worked at workplaces where they may have had additional exposures to high levels of noise. The probability of this occurring however is inversely proportional to their duration of employment in cohort mills. Such incomplete assessment would result in an underestimate of noise exposure, with the potential to: (a) attenuate exposure-response relationships; and (b) cause disease effects to be attributed to lower than actual exposure levels. Finally, exposure assessment techniques in this study were independent of health outcome status, and therefore any misclassification would be expected to be non-differential, again biasing risk estimates toward the null. Summary In summary, epidemiological analyses demonstrated statistically significant excesses of acute MI and other ischemic heart disease mortality in BC lumber mills workers exposed to high levels of noise. 121 Several important potential confounders were either directly, or indirectly, controlled or accounted for. The results for acute MI supports the hypothesis that chronic exposure to noise was causally associated with increased risk of coronary heart disease. While the overall results suggest a chronic, progressive mechanism, with risks increasing gradually with increasing exposure, relative risk for acute MI was highest after 20 years of exposure to noise, and while still employed and exposed to noise. This pattern of results can be explained if we assume the effects of noise exposure are reversible, and therefore overall risks are lower when considering the un-restricted follow up than when examining only those still exposed to noise. Alternatively, this effect would be seen if noise were playing an immediate, "trigger" role in a coronary event in addition to its chronic stressor role. My exposure data did not permit examination of this hypothesis, but a similar role has been proposed for anger and acute MI, with the hypothesized mechanism being hemodynamic disruption of a vulnerable atherosclerotic plaque during stress (Muller, 1999). 122 Chapter 7. Conclusion and Recommendations for Further Work This study was initiated to investigate a hypothesized association between occupational exposure to noise and cardiovascular disease mortality. It was designed to address several of the long-standing limitations of earlier work in this field. It included use of chronic disease outcomes, and standardized outcome definitions (i.e. mortality). It was able to account for several potential confounders of a noise-CVD association, including smoking. Its large size provided good statistical power. A particular emphasis was placed on achieving high validity in exposure assessment, and accounting for the use of hearing protection. For lumber mill workers with high exposure to noise, the study revealed elevations in mortality for acute MI with respect to the general population of BC. It also found significant and consistent exposure-response trends of increasing risk for acute MI, and to a lesser extent, ischemic heart disease mortality, with increasing levels of noise within the members of the cohort. All analyses were controlled for age, ethnicity and calendar period. The impact of other potential confounders, including smoking, socioeconomic status and psychosocial exposures, was reviewed and determined not to be a likely threat to the validity of the findings. These results support the preliminary findings in the same sawmill population by Ostry, who found associations between subjective measures of noise exposure with both ischemic heart disease mortality (combined acute MI and ischemic heart disease as defined in this study) and with self-reported heart disease. Relative risks of 1.3 - 1.9 found in Ostry's studies were comparable to those found in the current study (Ostry, 1999; Ostry et al., 2001). Few other studies have assessed chronic heart disease in occupational cohorts. Ising et al. (1997) found similar increases of relative risk for acute MI in their case-control study of acute MI survivors of working age, who self-reported current exposure to noise on a 5-point scale. Their highest exposure category correlated to noise levels of approximately 100 dBA, comparable to the lumber mill cohort, and the relative risk found of 3.8 was similar to that found in this study when follow-up was restricted to those currently employed. Idzior-Walus' finding of elevated prevalence of angina in those exposed longer than 2-years at "[higher than] permissible levels" levels of noise, was also consistent with this study's findings. Another well-executed study did not find an association (Theriault et al, 1988), although this study used a qualitative method of exposure assessment, and did not take the use of hearing protection into account; resulting misclassification may have been sufficient to obscure a true effect if it existed. Relative risks for ischemic heart disease in community studies have been lower, on the order of 1.1 to 1.5. It was not clear whether this lower level of risk is real, or whether it results from conservative biases that may be present. Community studies have suffered from problems of exposure 123 misclassification, with most relying on indirect or subjective measures of exposure. Few have obtained concomitant occupational exposure data, even though it may represent a significant proportion of subjects' noise exposure. Because environmental levels are typically much lower than occupational exposures it would not be easy to quantitatively combine the two. The interaction of high occupational noise exposure with high traffic noise was investigated in a subset of the Caerphilly/Speedwell cohort (Babisch et al, 1990), and there did appear to be an interaction effect, at least with respect to blood pressure and blood lipids. In the current study, a baseline of 85 dBA (or dBC) was used so that the likelihood of misclassifying due to unmeasured non-occupational exposure was reduced. The great majority of chronic heart disease cases are of multifactorial etiology, with no single cause but a broad range of risk factors that include individual traits (such as age, gender and genetic predisposition), life-style factors (such as smoking, diet, and physical exercise), and occupational and environmental exposures. Ischemic heart disease is responsible for approximately 22% of overall mortality in males in Canada, and acute MI accounts for half of those deaths. The risk of ischemic heart disease varies considerably depending on which risk factors, and to how many, an individual is exposed. Stamler (1991) showed that risk of coronary heart disease death was approximately three times higher in smokers than non-smokers, and approximately four times higher for those with hypertension or high serum cholesterol. However for smokers who were also in the highest quintiles for hypertension and total serum cholesterol exposure, the risk climbed to 20-fold. The results of this study suggest that noise exposure may also be an important risk factor for acute MI. Workers may have up to a 30 percent increased risk of death (relative risk=1.3) from acute MI after 20 years of exposures over 90 dBA (the current 8-hour OSHA exposure limit). Relative risk of acute MI was increased 50 percent after 20 years at 95 dBA (relative risk = 1.5). Relative risks were higher still for those who experienced long durations of exposure at high noise levels and were still employed. Relative risks of acute MI in this group were 2 - 4 times, after as little as 10 years at 85 dBA. Although the evidence was weaker, workers may have up to a 40 percent increased risk of death from ischemic heart disease after 10 years at 85 dBA and higher. These levels of risk associated with noise are similar to those seen for other occupational cardiotoxins that might be present in the sawmill environment, such as carbon monoxide (RR=1.35, Stern et al, 1988) and environmental tobacco smoke (RR=1.2, Steenland, 1999), although both of these would be found at much lower prevalence than noise. During the follow up period, a total of 910 cohort members died of acute MI. Assuming that there is a causal effect between noise and acute MI, one can estimate the attributable fraction, i.e. the proportion of deaths in the exposed group that might have been avoided if the subjects were not exposed to noise. It can be shown that of the 255 acute MI deaths that occurred among cohort subjects exposed for 124 more than 20 years at 90 dBA, 58 deaths might have been averted or delayed if noise levels had been lower16. Similarly, approximately 30 of the 181 other ischemic heart disease deaths may have been averted or delayed. This might be compared to the 40 cohort subjects killed in cohort sawmills in accidents during the similar period (1950-1990). The public health implication of my findings is illustrated by the relatively high prevalence of this exposure in the workplace. It is estimated that 30 millions workers are exposed to hazardous levels of noise in the US (Franks, 2000). While similar data does not exist for Canada or British Columbia, The number of workers exposed in BC can be crudely estimated by examining the number of annual hearing tests reported to the Workers' Compensation Board (annual hearing tests are required where a worker is exposed above 85 dBA). Approximately 150,000 hearing tests per year were reported to the Board for the years 1998-2001, though this likely underestimates true number as it does not include mining and agriculture workers nor those occupations that are federally regulated. True relative risks for heart disease in this cohort may in fact be higher than observed, because of the various non-differential misclassification biases operating in this study. Several have been reported; for example misclassification of disease outcome because of errors in death certificate coding, and misclassification of noise exposure due to missing exposure data for early parts of the study. Both would be expected to bias the observed risk toward the null. Exposure levels this high are quite common in the workplace (see Chapter 1). Moreover, the findings are probably reasonably generalizable, as noise parameters found in this lumber mill setting -broadband frequency, and continuous noise, with a component of impact noise - are typical of many industrial settings. Thus there may be a considerable public health impact of noise-related coronary disease. Hearing protection may offer protection against the non-auditory as well as auditory effects of noise, and some work has shown that acute stress response to noise is lower in those wearing hearing protectors. However, the efficacy of hearing protection is widely questioned, due to problems of compliance, fit, comfort and inconsistent use. The pathophysiological mechanism for noise-related heart disease is not solely biomechanical, as it is in NIHL. Hence, some researchers have suggested that the role of HPD may be more complex than simply acting as a noise-control method, and may constitute a source of psychological stress. There is much opportunity for further research on this study population. Analyses will be extended to heart disease morbidity through linkage to the BC Linked Health Database (Chamberlayne et Where the attributable fraction = [relative risk - 1]/relative risk al, 1998). Data on hospital separations between 1985 and 1996 will allow us to investigate non-fatal diseases such as hypertension more fully, as well as provide additional power for analyses of more finely divided disease categories. Although these analyses will only be done on a subset of the full cohort, statistical power will be maintained because hospital discharge rates are roughly 5 times greater than mortality rates. Hospital separation data can also be used to develop disease "trajectories" that characterize disease progression prior to death. This will help characterize truly acute coronary events, and aid the understanding of mechanistic issues, such as those surrounding the high levels of risk for acute MI while still employed. Morbidity data may also present an opportunity to extend the investigation to women, who were excluded from analyses in this mortality study. Additional insight into some of the mechanistic aspects of noise-related cardiovascular disease may be obtained by reanalysis of the existing data using different categories of coding for cause of death. For example, examining mortality in terms of acute (acute MI and unstable angina) and chronic (Chronic ischemic heart disease and stable angina) coronary syndromes, sequelae of hypertension (stroke, HT disease, kidney disease and heart failure), and arrhythmias. Such coding may more closely represent the underlying pathophysiology. Further investigation of the elevated risk of acute MI while still employed and exposed to noise is warranted. As mentioned, characterizing the disease trajectory prior to death may help understand if these are truly acute events. In this study only chronic noise exposure was measured. Adopting a different study design, such as the case cross-over design (Mittleman, 1995) may permit a better understanding of the role of acute exposure to noise. It is believed that risk of heart disease may be increased through interaction between noise and other occupational stressors. We will investigate some of these interactions, using the components of the job-strain construct that have previously been assessed for all jobs in the lumber mill cohort. These include the degree of psychological and physical demand, as well as 'control' and social-support (Ostry, 1998). Other potential interacting factors, such as subjects' communication requirements, could be assessed based on job task requirements, and then interactions analyzed. We also intend to investigate other disease outcomes. Several studies have found evidence of increased rates of injury-accidents in those exposed to high levels of noise and those who had poor hearing. Lumber mills have very high injury rates, and improving our understanding of causative factors is vital in reducing injury incidence. In addition, other non-auditory stress-related effects such as psychiatric and immunological outcomes could be pursued. Furthermore, there is room for further improvements in exposure assessment. Besides using A-weighting, this study used C-weighting to explore the hypothesis that low-frequency noise was more harmful with respect to cardiovascular disease than broadband noise. However, other weighting schemes may be more appropriate, including novel schemes that weight low-frequency components differently or that weight different portions of the spectra, e.g. high frequencies. It may also be possible to improve the assessment of pre-1970 exposure levels if additional sources of noise-determinant data for these periods could be found. There are also other attributes of noise exposure considered potentially important in determining the level of effect, that we did not measure -e.g. impulse vs. continuous sound, predictability of the noise, and individuals' control over the noise source. Methods of measuring and integrating these into the exposure-response analyses should be explored. 127 Bibliography 1. AHA. 2001 Heart and Stroke Statistical Update [Web Page]. Accessed 2001 Sep 19. Available at: http://www.americanheart.org/statistics/hb.html. 2. al Absi, M. and Arnett, D. K. Adrenocortical responses to psychological stress and risk for hypertension. Biomedicine & Pharmacotherapy. 2000; 54(5):234-244. 3. Alberti, PW. Noise, the Most Ubiquitous Pollutant. Noise and Health. 1998; 1:3-5. 4. Andren, L.; Hansson, L.; Eggertsen, R.; Hedner, T., and Karlberg, B. E. Circulatory effects of noise. Acta Medica Scandinavica. 1983; 213(l):31-35. 5. Aro, S. Occupational Stress, Health Related Behaviour, and Blood Pressure: A 5-year Follow up. Preventive Medicine. 1984; 13:333-338. 6. Atherly G. and W. Noble. Occupational Deafness: The continuing Challenge of Early German and Scottish Research. American Journal of Industrial Medicine. 1985; 8:101-117. 7. Babisch, W. Epidemiological Studies of the Cardiovascular Effects of Noise - A Critical Appraisal. Noise and Health. 1998; 1:24-39. 8. Babisch, W.; Gallacher, J. E.; Elwood, P. C , and Ising, H. Traffic noise and cardiovascular risk. The Caerphilly study, first phase. Outdoor noise levels and risk factors. Archives of Environmental Health. 1988; 43(6):407-414. 9. Babisch, W; H Ising; B Kruppa, and D Weins. The Incidence of Mysocardial Infarction and its Relation to Road Traffic Noise - The Berlin Case Control Studies. Environment International. 1994; 20(4):469-474. 10. Babisch W; H Ising; JEJ Gallacher; PC Elwood; PM Sweetnam; JWG Yarnell; D Bainton, and IA Baker. Traffic Noise, Work Noise and Cardiovascular Risk Factors: The Caerphilly and Speedwell Collaborative Heart Disease Studies. Environment International. 1990; 16:425-435. 11. Babisch, W.; Ising, H.; Elwood, P. C ; Sharp, D. S., and Bainton, D. Traffic noise and cardiovascular risk: the Caerphilly and Speedwell studies, second phase. Risk estimation, prevalence, and incidence of ischemic heart disease. Archives of Environmental Health. 1993a; 48(6):406-413. 12. Babisch, W.; Ising, H.; Gallacher, J. E.; Sharp, D. S., and Baker, I. A. Traffic noise and cardiovascular risk: the Speedwell study, first phase. Outdoor noise levels and risk factors. Archives of Environmental Health. 1993b; 48(6):401-405. 13. Babisch, W.; Ising, H.; Gallacher, J. E.; Sweetnam, P. M., and Elwood, P. C. Traffic noise and cardiovascular risk: the Caerphilly and Speedwell studies, third phase—10-year follow up. Archives of Environmental Health. 1999; 54(3):210-216. 14. BC Ministry of Health. Tobacco Use in BC, 1997 [Web Page]. 1997; Accessed 2002 Jul 15. 128 Available at: http://www.hlth.gov.bc.ca/tobacrs/index.html. 15. Belli, S.; Sani, L.; Scarficcia, G., and Sorrentino, R. Arterial hypertension and noise: a cross-sectional study. American Journal of Industrial Medicine. 1984; 6(l):59-65. 16. Berger, EH. Hearing Protection Devices. Berger, EH; WD Ward; JC Morrill, and L H Royster, Ed's. Noise and Hearing Conservation Manual. 4th ed. Akron, OH: American Industrial Hygiene Association; 1986. 17. Berger EH; JR Franks, and F Lindgren. International Review of Field Studies of Hearing Protector Attenuation. Axlesson A; H Borchgrevink; RP Hamernik; P Hellstrom; D Henderson, and RJ Salvi, Ed.'s. Scientific Basis of Noise-Induced Hearing Loss. New York: Thieme Medical Publishers, Inc.; 1996; pp. 361-377. 18. Bjorntorp, P. Stress and cardiovascular disease. Acta Physiologica Scandinavica. Supplementum. 640:144-8, 1997. 19. BosmaH ; Marmot MG ; Hemingway H ; Nicholson AC ; Brunner E , and Stansfeld SA. Low job control and risk of coronary heart disease in Whitehall II (prospective cohort) study. BMJ. 1997;314:558-565. 20. Boyle CA and AJ Dobson. The Accuracy of Hosptial records and death certificates for acute myocardial infarction. AustNZ J Med. 1995; 25:316-323. 21. Breslow NE and NE Day. Statistical Methods in Cancer Research. Lyon: Int. Agency for Research on Cancer; 1987. 22. Brini D ; R Ratti; PA Toricelli, and A M Cirla. Epidemiological Study of the Prevalence of Arterial Hypertension in Subjects Exposed to Continuous and Impulse Noise. 4th International Congress in Noise as a Public Health Problem; Turin. International Congress on the Biological Effects of Noise; 1983. 23. Burstyn, I and K Teschke. Studying the Determinants of Exposure: A Review of Methods. American Industrial Hygiene Association Journal. 1999; 60:57-72. 24. Campbell, NRC; DW McKay; A. Chockalingam, and JG Fodor. Errors in Assessment of Blood Pressure: Blood Pressure Measuring Technique. Canadian Journal of Pubic Health. 1994; 85(2):S18-S21. 25. Cannon, WB. Stresses and Strains of Homeostasis. American Journal of Medical Science. 1935; 189:1. 26. Carter, NL. Heart-rate and blood-pressure response in medium-artillery gun crews. Medical Journal of Australia. 1988; 149(4): 185-189. 27. Carter, NL and HC Beh. The effect of intermittent noise on cardiovascular functioning during vigilance task performance. Psychophysiology. 1989; 26(5):548-559. 28. Cassinelli R; KJ Kock; K Steenland, and S Spaeth. User Documentation: PC LTAS (Life Table Analysis Ssytem for the PC). Cincinnati OH: NIOSH; 2001. 29. Cavatorta, A; M Falzoi; A Romanelli; F Cigala; M Ricco; G Bruschi; I Franchini, and A Borghetti. Adrenal response in the pathogenesis of arterial hypertension in workers exposed to high noise levels. Journal of Hypertension - Supplement. 1987; 5(5):S463-129 466. 30. Chamerlayne R; B Green; ML Barer; C Hertzman; WJ Lawrence, and SB Sheps. Creating a Population-Based Linked Health Database: A New Resource for Health Services Research. Can J Public Health. 1998; 89(4):270-273. 31. Checkoway H and EA Eisen. Developments in Occupational Cohort Studies. Epidemiologic Reviews. 1998; 20(1): 100-111. 32. Checkoway, H; N Pearce, and DJ Crawford-Brown. Research Methods in Epidemiology. New York: Oxford University Press; 1989. 33. Checkoway, H; NJ Heye; PA Demers, and NE Breslow. Mortality among workers in the datomaceous earth industry. British Journal of Industrial Medicine. 1994; 50:586-597. 34. Cherek, DR. Effetcs of Acute Exposure to Increased Levels of Background Industrial Noise on Cigarette Smoking Behaviour. Int Arch Occup Environ Health. 1985; 56(23-30). 35. Chrousos, GP. Stressors, Stress, and The Neuroendocrine Integrations of the Adaptive Response. Annals of the New York Academy of Sciences. 1998; 851:311-335. 36. Chrousos GP and PW Gold. The Concepts of Stress and Stress System Disorders. JAMA. 1992; 267. 37. Cohen S and N Weinstein. Nonauditory Effects of Noise on Behaviour and Health. Journal of Social Issues. 1986; 1:36-70. 38. Cook NR; J Cohen; PR Herbert; JO Talyor, and CH Hennekens. Implications of Small Reductions in Diastolic Blood Pressure for Primary Prevention. Arch Intern Med. 1995; 155:701-709. 39. CSA (Canadian Standards Association). Procedures for the Measurement of Occupational Noise Exposure. 1994; Z107.56-94. 40. Davis RR and WK Sieber. Trends in hearing protector usage in American manufacturing from 1972 to 1989. American Industrial Hygiene Association Journal. 1998; 59(10):715-722. 41. DeJoy, D. M. The nonauditory effects of noise: review and perspectives for research. Journal of Auditory Research. 1984; 24(2):123-150. 42. Delin C. Noisy Work and Hypertension. Lancet. 1984; 2(8408):931. 43. Dosemeci, M ; Chen, J. Q.; Hearl, F.; Chen, R. G.; McCawley, M.; Wu, Z.; McLaughlin, J. K.; Peng, K. L.; Chen, A. L., and Rexing, S. H. Estimating historical exposure to silica among mine and pottery workers in the People's Republic of China. American Journal of Industrial Medicine. 1993; 24(l):55-66. 44. Dost, WA. Sawmill Noise at Operating Level. Forest Products Journal. 1974; 24(8): 13-17. 45. Duncan. RC; CE Easterly; J Griffith, and TE Aldrich. The Effect of Chronic Environmental Noise on the Rate of Hypertension: A Meta Analysis. Environment International. 1993; 19:359-369. 46. Eggertsen, R.; Svensson, A.; Magnusson, M., and Andren, L. Hemodynamic effects of loud noise before and after central sympathetic nervous stimulation. Acta Medica Scandinavica. 1987; 221(2): 159-164. 47. Eisen EA; TJ Smith; DH Wegman; TA Louis, and J Froines. Estimation of Long Term Dust Exposure in Vermont Granite Sheds. Am Ind. Hyg. Assoc. J. 1984; 45(2):89-94. 48. Elliot, P. High Blood Pressure in the Community. Bulpitt, CJ. Handbook of Hypertension. Vol 20. Epidemiology of Hypertension. Amsterdam: Elsevier Scientific; 2000; pp. 1-18. 49. Engeland, WC; P Miller, and DS Gann. Pituitary-Adrenal and Adrenomedullary Responses to Noise in Awake Dogs. Am J Physiol. 1990; 27:R672-677. 50. Fairfax, RE. Noise Abatement Techniques in Southern Pine Sawmills and Planer Mills. American Industrial Hygiene Association. 1989; 50(12):634-638. 51. Fisher, L. D. and Tucker, D. C. Air jet noise exposure rapidly increases blood pressure in young borderline hypertensive rats. Journal of Hypertension. 1991; 9(3):275-282. 52. Fogari, R.; Zoppi, A.; Vanasia, A.; Marasi, G., and Villa, G. Occupational noise exposure and blood pressure. Journal of Hypertension. 1994; 12(4):475^479. 53. Folkow, B. Sympathetic nervous control of blood pressure. Role in primary hypertension. American Journal of Hypertension. 1989; 2(3 Pt 2):103S-11 IS. 54. Fouriaud, C; MC Jacquinet-Salord ; P Degoulet; F Aime ; T Lang ; J Laprugne ; J Main ; J Oeconoms ; J Phalente , and A Prades. Influence of Socioprofessional Conditions on Blood Pressure Levels and Hypertension Controls. American Journal of Epidemiology. 1984; 120(1 ):72-86. 55. Franks, J. Preventing Noise-Induced Hearing Loss: Perspective View form the next Millenium. Noise Effects '98; Sydney. International Congress on Noise as a Public Health Problem; 1998:p11. 56. Franks JR; MR Stephenson, and CJ Merry. Preventing Hearing Loss - A Practical Guide. Rev 10/96 ed.. Cincinnati, OH: NIOSH; 1996; DHHS (NIOSH) 96-110. 57. Friedman, M; SO Byers, and AE Brown. Plasma Lipid Response of Rats and Rabbits to an Auditory Stimulus. Am J. Physiol. 1967; 212(5): 1174-1178. 58. Friesen, MC; PA Demers; HW Davies; S Marion, and K Teschke. A Comparison of Fixed-Effects and Mixed-Effects Statistical Models for Predicting Dust Exposure in Sawmills. Annals of Occupational Hygiene. 2002; submitted. 59. Garcia, A. M. and Garcia, A. Occupational noise as a cardiovascular risk factor. Schriftenreihe Des Vereins Fur Wasser-, Boden-, Und Lufthygiene. 1993; 88:212-222. 60. Germano, G.; Damiani, S.; Milito, U.; Germano, U.; Giarrizzo, C , and Santucci, A. Noise stimulus in normal subjects: time-dependent blood pressure pattern assessment. Clinical Cardiology. 1991; 14(4):321-325. 61. Glass DC and JE Singer. Urban Stress: Experiments on Noise and Social Stressors. New York: Academic Press; 1972. 62. Green, M. S.; Schwartz, K.; Harari, G., and Najenson, T. Industrial noise exposure and 131 ambulatory blood pressure and heart rate. Journal of Occupational Medicine. 1991; 33(8):879-883. 63. Greenland, S. Meta-Analysis. Rothman, K and S Greenland. Modern Epidemiology. 2nd ed. Philidelphia: Lippincott Williams Wilkins; 1998. 64. Greenspan CA; R Moure-Eraso; DH Wegman, and LC Oliver. Occupational Hygiene Characterization of a Highway Construction Project: A Pilot Study. Appl. Occup. Enivron. Hyg. 1995; 10(l):50-58. 65. Hansen, DJ; WG Adams, and RA Hochberg. Comparison of Sampling Strategies Used During a Machine Shop Noise Survey. Appl. Ind. Hyg. 1989; 4(3):75-80. 66. Hayter, R; E Grass, and T Barnes. Labour Flexibility: A Tale of Two Mills. Tijdschrift Voor Economische En Social Geografie. 1994; 85(l):25-38. 67. Hedstrand, H.; B Drettner ; I Klockhoff, and A Svedberg. Noise and Blood-Pressure. Lancet. 1977; 2(8051): 1291. 68. Henry JP. Biological Basis of the Stress Response. News in Physiological Science. 1993; 8:69-73. 69. Hertzman, C ; Teschke, K.; Ostry, A.; Hershler, R.; Dimich-Ward, H.; Kelly, S.; Spinelli, J. J.; Gallagher, R. P.; McBride, M., and Marion, S. A. Mortality and cancer incidence among sawmill workers exposed to chlorophenate wood preservatives. American Journal of Public Health. 1997; 87(l):71-79. 70. Hessel, P. A. and Sluis-Cremer, G. K. Occupational noise exposure and blood pressure: longitudinal and cross-sectional observations in a group of underground miners. Archives of Environmental Health. 1994; 49(2): 128-134. 71. Hinkle, LE. Stress and Disease: The Concept after 50 Years. Soc. Sci. Med. 1987; 25(6):561-566. 72. Hirai, A.; Takata, M.; Mikawa, M.; Yasumoto, K.; Iida, H.; Sasayama, S., and Kagamimori, S. Prolonged exposure to industrial noise causes hearing loss but not high blood pressure: a study of 2124 factory laborers in Japan. Journal of Hypertension. 1991; 9(11):1069-1073. 73. Hornung, R. W.; Greife, A. L.; Stayner, L. T.; Steenland, N. K.; Herrick, R. F.; Elliott, L. J.; Ringenburg, V. L., and Morawetz, J. Statistical model for prediction of retrospective exposure to ethylene oxide in an occupational mortality study. American Journal of Industrial Medicine. 1994; 25(6):825-836. 74. HSF. The Changing Face of Heart Disease and Stroke in Canada 2000. Ottawa: Heart and Stroke Foundation of Canada; 1999. 75. Idzior-Walus, B. Coronary risk factors in men occupationally exposed to vibration and noise. European Heart Journal. 1987; 8(10): 1040-1046. 76. Ising, H.; Babisch, W.; Kruppa, B.; Lindthammer, A., and Wiens, D. Subjective work noise: a major risk factor in myocardial infarction. Sozial- Und Praventivmedizin. 1997; 42(4):216-222. 77. Job, RFS. The Influence of Subjective Reactions to Noise on Health Effects of the Noise. 132 Environment International. 1996; 22(1):93-104. 78. Johsson, A. and Hansson, L. Prolonged exposure to a stressful stimulus (noise) as a cause of raised blood-pressure in man. Lancet. 1977; l(8002):86-87. 79. Kent SJ ; HE von Gierke , and GD Tolan. Analysis of the Potential Association Between Noise-Induced Hearing Loss and Cardiovascular Disease in USAF Aircrew Members. Aviation, Space and Environmental Medicine. 1986; 57:348-361. 80. Kjellberg, A. Subjective, behavioural and pschophysiological effects of noise. Scand. J. Work Environ Health. 1990; 16(suppl l):29-38. 81. Knipschild, P. V. Medical effects of aircraft noise: community cardiovascular survey. International Archives of Occupational & Environmental Health. 1977a; 40(3):185-190. 82. —. VI. Medical effects of aircraft noise: general practice survey. International Archives of Occupational & Environmental Health. 1977b; 40(3):191-196. 83. Knipschild, P. and Salle, H. Road traffic noise and cardiovascular disease. A population study in The Netherlands. International Archives of Occupational & Environmental Health. 1979; 44(l):55-59. 84. Knuttson A; T Akerstedt; B Jonsson, and K Orth-Gomer. Increased Risk of Ischemic Heart Disease in Shift Workers. Lancet. 1986; (8498):89-92. 85. Koch M. The Neurobiology of the Startle. Progress in Neurobiology. 1999; 59:107-128. 86. Krantz DS and SE Raisen. Environmental Stress, Reactivity, and Ischemic Heart Disease. British J of Med Psychology. 1988;61:3-16. 87. Kristal-Boneh, E.; Melamed, S.; Harari, G., and Green, M. S. Acute and chronic effects of noise exposure on blood pressure and heart rate among industrial employees: the Cordis Study. Archives of Environmental Health. 1995; 50(4):298-304. 88. Kristensen, TS. Cardiovascular diseases and the work environment. A critical review of the epidemiologic literature on nonchemical factors. Scandinavian Journal of Work, Environment & Health. 1989; 15(3):165-179. 89. —. The Demand-Control-Support Model: Methodological Challenges for Future Research. Stress Medicine. 1995; 11:17-26. 90. Kryter, KD. The Effects of Noise on Man. Orlando: Academic Press; 1985. 91. Lang, T.; Fouriaud, C , and Jacquinet-Salord, M. C. Length of occupational noise exposure and blood pressure. International Archives of Occupational & Environmental Health. 1992; 63(6):369-372. 92. Ledesert, B; MJ Saurel-Cubizolles ; M. Bourgine ; M. Kaminski; A. Touranchet, and C Verger. Risk Factors for High Blood Pressure among Workers in French Poultry Slaughterhouses and Canneries. European Journal of Epidemiology. 1994; 10:609-620. 93. Lees REM; C Smith Romeril, and LD Wetherall. A Study of Stress Indicators in Workers Exposed to Industrial Noise. Canadian Journal of Public Health . 1980; 71:261-265. 133 94. Lees, REM and J Hatcher Roberts. Noise Induced Hearing Loss and Blood Pressure. Canadian Medical Journal. 1979; 120:1082-1084. 95. Lercher, P.; Hortnagl, J., and Kofler, W. W. Work noise annoyance and blood pressure: combined effects with stressful working conditions. International Archives of Occupational & Environmental Health. 1993; 65(l):23-28. 96. Lercher P; SA Stansfled, and SJ Thompson. Non-Auditory Health Effects of Noise: Review of the 1993-1998 Period. Noise Effects '98; Sydney. Sydney: International Congress on the Biological Effects of Noise; 1998: pp 213-220 . 97. Lloyd-Jones, D. M.; Martin, D. O.; Larson, M. G., and Levy, D. Accuracy of death certificates for coding coronary heart disease as the cause of death. Annals of Internal Medicine. 1998; 129(12):1020-1026. 98. Marmot, MG and JF Mustard. Coronary Heart Disease From a Population Perspective. Evans, RG; ML Barer, and TR Marmor, Eds. Why are Some People Healthy and Others not? New York: Aldine de Gruyter; 1994; p. 189. 99. Maschke, C ; Rupp, T., and Hecht, K. The influence of stressors on biochemical reactions—a review of present scientific findings with noise. International Journal of Hygiene & Environmental Health. 2000; 203(1 ):45-53. 100. McKeigue PM; GJ Miller, and MG Marmot. Coronary heart disease in south Asians overseas: a review. Journal of Clinical Epidemiology. 1989; 42(7):597-609. 101. Medoff HS and A M Bongiovanni. Blood Pressure in rats subject to audiogenic stimulation. Am. J. Physiol. 1945; 143:300-305. 102. Meerson, FZ. Stress-Induced Arrhythmic Disease of the Heart - Part I. Clin. Cardiol. 1994; 17:362-371. 103. Melamed, S. and Bruhis, S. The effects of chronic industrial noise exposure on urinary Cortisol, fatigue and irritability: a controlled field experiment. Journal of Occupational & Environmental Medicine. 1996; 38(3):252-256. 104. Merlo F; M Caostatini; G Reggiardo; M Ceppi, and R Puntoni. Lung Cancer Risk among Refractory Brick Workers Exposed to Crystalline Silica: A Retrospective Cohort Study. Epidemiology. 1991; 2(4):299-305. 105. Mittleman, MA; M Maclure ; JB Sherwood ; RP Mulry ; GH Toiler ; SC Jacobs ; R Friedman ; H Benson , and JE Muller. Triggering of Acute Myocardial Infarction Onset by Episodes of Anger. Circulation. 1995; 92(7): 1720-1725. 106. Muller, J. E. Circadian variation and triggering of acute coronary events. American Heart Journal. 1999; 137(4 Pt2):Sl-S8. 107. Neus, H.; Ruddel, H., and Schulte, W. Traffic noise and hypertension: an epidemiological study on the role of subjective reactions. International Archives of Occupational & Environmental Health. 1983; 51(3):223-229. 108. Nieuwenhuijsen MJ; MB Schenker; SJ Samuels; JA Farrar, and SS Green. Exposure to Dust, Noise and Pesticides, Their Determinants, and USe of Protective Equipment Among California Farm Operators. Appl. Occup. Environ. Hyg. 1996; 11(10):1217-134 1225. 109. NIOSH. Criteria for A Recommended Standard - Occupational Exposure to Noise. HSM 73-11001 ed.. Washington DC; 2000. 110. OSHA. Occupational Noise Exposure: Hearing Conservation Amendment. 1981; 46 Fed Reg, 4078-4179. 111. Ostry AS. Psychosocial Job Strain and Coronary Heart Disease in a cohort of Blue Collar Workers. Vancouver, BC: University of Bristish Columbia; 1998. 112. Ostry AS; SA Marion; PA Demers; R Hershler; S Kelly; K Teschke; C Mustard, and C Hertzman. A Comparison of Expert Rater Methods of Assessing Psychosocial Job Strain. Scand J Work Environ Health. 2001; 27(l):70-75. 113. Passchier-Vermeer, W. and Passchier, W. F. Noise exposure and public health. Environmental Health Perspectives. 2000; 108(Suppl 1):123-131. 114. Persson Waye, K; J Bengtsson; R Rylander; F Hucklebridge; P Evans, and A Chow. Low Frequency Noise Enhances Cortisol Among Noise Sensitive Subjects During Work Performance. Life Sciences. 2002; 70:745-758. 115. Peterson, E. A.; Augenstein, J. S.; Tanis, D. C , and Augenstein, D. G. Noise raises blood pressure without impairing auditory sensitivity. Science. 1981; 211(4489):1450-1452. 116. Pickering T. Cardiovascular pathways: socioeconomic status and stress effects on hypertension and cardiovascular function. Annals of the New York Academy of Sciences. 1999; 896:262-277. 117. Rossi, L; G. Oppliger, and E. Grandjean. Gli Effetti Neurovegetativi SuH'uomo Di Rumori Sovrapposti Ad Un Rumore Di Fondo. Med. Lavaro. 1959; 50(5):332-337. 118. Rothman, KJ and S. Greenland. Precision and Validity in Epidemiological Studies. Rothman, KJ and S. Greenland. Modern Epidemiology. 2nd ed. Philadelphia: Lippincott Williams and Wilkins; 1998; p. 127. 119. Rozanski, A; JA Blumenthal, and J Kaplan. Impact of Psychological Factors on the Pathogenesis of Cardiovascular Disease and the Implications for Therapy. Circulation. 1999; 99:2192-2217. 120. Schmidek M and P Carpenter. Intermittent Noise Exposure and Associated Damage Risk to Hearing of Chain Saw Operators. American Industrial Hygiene Journal. 1974; Mar: 152-158. 121. Schnall PL; PA Landsberis, and D Baker. Job Strain and Cardiovascular Disease. Annu Rev Public Health. 1994; 15:381-411. 122. Schnatter RA; JF Acquavella; FS Thompson; D Donaleski, and G Theriault. An Analysis of Death Ascertainment and Follow-Up through Canada's Mortality Database System. Can J Public Health. 1990; 81(60-65). 123. Schwarze S. and Thompson SJ. Reseach on Non-Auditory Physiological Effects of Noise Since 1988. Noise as a Public Health Problem; Nice. International Congress on Noise as a 135 Public Health Problem; 1993: p 252. 124. Sehdev AE and GM Hutchins. Problems With Proper Completion and Accuracy of The Cause of Death Statement. Arch Intern Med. 2001; 161:277-284. 125. Seixas, NS and H Checkoway. Exposure Assessment in Industry Specific Retrospective Occupational Epidemiology Studies. Occupational and Environmental Medicine. 1995; 52:625-633. 126. Selye, H. Stress and Cardiovascular Disease. Cardiovascular Medicine. 1979; (February): 183-192. 127. Singh, A. P.; Rai, R. M.; Bhatia, M. R., and Nayar, H. S. Effect of chronic and acute exposure to noise on physiological functions in man. International Archives of Occupational & Environmental Health. 1982; 50(2): 169-174. 128. Staessen, JA; ET O'brien; L Thijs, and RH Fagard. Modern Approaches to Blood Pressure Measurement. Occupational and Environmental Medicine. 2000; 57:510-520. 129. Stamler, J. Established Major Coronary Risk Factors . Marmot, M and P Elliott, Ed's. Coronary Heart Disease Epidemiology,from Aetiology to Public Health. New York: Oxford; 1992. 130. Stansfeld, SA. Psychological Medicine. Cambridge: Cambridge University Press; 1992. 131. Steenland, K. Methods of Control for Smoking in Occupational Cohort Mortality Studies. Scand J Work Environ Health. 1984; 10:143-149. 132. —. Risk Assessment for Heart Disease and Workplace ETS Exposure Among Non-Smokers. Environmental Health Perspectives. 1999; 107(suppl 6):859-863. 133. Stern FB; WE Halperin; RW Hornung; VL Ringenburg, and CS McCammon. Heart Disease Mortality Among Bridge and Tunnel Officers Exposed to Carbon Monoxide. American Journal of Epidemiology. 1988; 128:1276-1288. 134. Stewart, P. Challenges to retrospective exposure assessment. Scand J Work Environ Health. 1999; 25(6):505-510. 135. Stewart PA; PSJ Lees, and M Francis. Quantification of Historical Exposures in Occupational Studies. Scand J Work Environ Health. 1996; 22:405-414. 136. Stewart, PA and RF Herrick. Issues in Performing Retrospective Exposure Assessment. Appl. Occup. Environ. Hyg. 1991; 6(6):421-427. 137. Sudo, A.; Nguyen, A. L.; Jonai, H.; Matsuda, S.; Villanueva, M. B.; Sotoyama, M.; Nguyen, T. C ; Le, V. T.; Hoang, M. H.; Nguyen, D. T., and Nguyen, S. Effects of earplugs on catecholamine and Cortisol excretion in noise-exposed textile workers. Industrial Health. 1996; 34(3):279-286. 138. Svensson, A; R. Eggertsen; M. Magnusson, and and L. Andren. Persisting Cardiovascular Effects of Noise Exposure in Essential Hypertension. Journal of Hypertension. 1987; 5(suppl 5):S459-S461. 139. Talbott, E; J Helmkamp ; K Matthews ; L Kuller; E Cottington , and G Redmond. Occupational Noise Exposure, Noise Induced Hearing Loss, and The Epidemiology of High Blood Pressure. American Journal of Epidemiology. 1985; 121(4):501-514. 140. Talbott, E. O.; Gibson, L. B.; Burks, A.; Engberg, R., and McHugh, K. P. Evidence for a dose-response relationship between occupational noise and blood pressure. Archives of Environmental Health. 1999; 54(2):71-78. 141. Tarter, S. K. and Robins, T. G. Chronic noise exposure, high-frequency hearing loss, and hypertension among automotive assembly workers. Journal of Occupational Medicine. 1990; 32(8):685-689. 142. Teschke, K.; Ostry, A.; Hertzman, C ; Demers, P. A.; Barroetavena, M. C ; Davies, H. W.; Dimich-Ward, H.; Heacock, H., and Marion, S. A. Opportunities for a broader understanding of work and health: multiple uses of an occupational cohort database. Canadian Journal of Public Health. 1998; 89(2):132-136. 143. Theriault GP; CG Tremblay, and BG Armstrong. Risk of Ischemic Heart Disease Among Primary Aluminum Production Workers. American Journal of Industrial Medicine. 1988; 13:659-666. 144. Thompson, S. J. Review: extraaural health effects of chronic noise exposure in humans. Schriftenreihe Des Vereins Fur Wasser-, Boden-, Und Lufthygiene . 1993; 88:91-117. 145. Thompson, SJ. Epidemiology Feasibility Study: Effects of Noise on the cardiovascular System. Washington DC: US EPA; 1981; 68-01-6274. 146. Tomei, F.; Tomao, E.; Papaleo, B.; Baccolo, T. P., and Alfi, P. Study of some cardiovascular parameters after chronic exposure to noise. International Journal of Cardiology. 1991; 33(3):393-399. 147. Traven, ND; LH Kuller, and DG Ives, et al. Coronary Heart Disease Mortality and Sudden Death: Trends and Patterns in 35-44-Year old White Male, 1970-1990. Am. J. Epi. 1995; 142(l):45-52. 148. Tubbs, RL (NIOSH). Health Hazard Evaluation Report: Neiman Sawmills Inc. Department of Health and Human Services; 1991; HETA-88-030-2109. 149. Tupper, VJ (Workers' Compensation Board of British Columbia). Noise-Con 81; North Carolina State University, Raleigh. 1981: 329-334. 150. van Dijk, F. J.; Souman, A. M., and de Vries, F. F. Non-auditory effects of noise in industry. VI. A final field study in industry. International Archives of Occupational & Environmental Health. 1987a; 59(2): 133-145. 151. van Dijk, F. J.; Verbeek, J. H., and de Fries, F. F. Non-auditory effects of noise in industry. V. A field study in a shipyard. International Archives of Occupational & Environmental Health. 1987b; 59(l):55-62. 152. van Dijk, FJH. Epidemiological Research on Non-Auditory Effects of Occupational Noise Exposure. Environment International. 1990; 16:405-409. 153. Verbeek, J. H.; van Dijk, F. J., and de Vries, F. F. Non-auditory effects of noise in industry. IV. A field study on industrial noise and blood pressure. International Archives of Occupational & Environmental Health. 1987; 59(l):51-54. 137 154. Westman, J. C. and Walters, J. R. Noise and stress: a comprehensive approach. Environmental Health Perspectives. 41:291-309, 1981 Oct. 1981. 155. WHO (World Health Organization). Arterial Hypertension. Geneva: World Health Organization; 1978; Technical report Series No. 628. 156. Wielgosz AT and RP Nolan. Biobehavioural Factors in the Context of Ischemic Cardiovascular Disease. Journal of Psychosomatic Research. 2000; 48:339-345. 157. Wu, T. N.; Ko, Y. C , and Chang, P. Y. Study of noise exposure and high blood pressure in shipyard workers. American Journal of Industrial Medicine. 1987; 12(4):431-438. 158. Xu, X; T Niu ; DC Christiani; ST Weiss ; Y Zhou; C Chen; J Yang; Z Fang; Z Jinag; W Liang, and F Zhang. Environmental and Occupational Determinatns of Blood Pressure in Rural Communities in China. Ann Epidemiol. 1999; 7:95-106. 159. Zhao, Y. M.; Zhang, S. Z.; Selvin, S., and Spear, R. C. A dose response relation for noise induced hypertension. British Journal of Industrial Medicine. 1991; 48(3): 179-184. 138 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0076817/manifest

Comment

Related Items