UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Determinants of flour dust exposure in bakeries Burstyn, Igor 1996

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

Item Metadata

Download

Media
831-ubc_1996-0577.pdf [ 4.74MB ]
Metadata
JSON: 831-1.0087217.json
JSON-LD: 831-1.0087217-ld.json
RDF/XML (Pretty): 831-1.0087217-rdf.xml
RDF/JSON: 831-1.0087217-rdf.json
Turtle: 831-1.0087217-turtle.txt
N-Triples: 831-1.0087217-rdf-ntriples.txt
Original Record: 831-1.0087217-source.json
Full Text
831-1.0087217-fulltext.txt
Citation
831-1.0087217.ris

Full Text

D E T E R M I N A N T S O F F L O U R DUST E X P O S U R E IN B A K E R I E S by I G O R B U R S T Y N B.Sc , The University of British Columbia, 1993 A THESIS S U B M I T T E D FN P A R T I A L F U L F I L L M E N T O F 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 O F M A S T E R O F SCDXNCE in T H E F A C U L T Y O F G R A D U A T E STUDBES (Occupational Hygiene Programme) We accept this thesis as conforming to the required standard T H E UNIVERSITY OF BRITISH COLUMBIA October 1996 © Igor Burstyn, 1996 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 The University of Britisn Columbia Vancouver, Canada Date 0^/0 y) 996 DE-6 (2/88) Abstract This thesis presents the results of an investigation into factors that are associated with flour dust exposures in B.C. bakeries. We describe observed exposure levels and advise on control measures which may reduce an individual's exposure to flour dust. In this investigation, flour dust exposure was assesed using three measures: inhalable dust, wheat antigen and fungal a-amylase. Inhalable dust was measured gravimetrically. The wheat antigen and a-amylase content of the water soluble fraction of inhalable dust was assayed via ELISA. Ninety-six bakery workers, employed in seven different bakeries, participated in the investigation. Two side-by-side full-shift inhalable dust samples were obtained from each study participant on a single occasion. We were able to obtain useable exposure measurements for all three exposure measures from each study participant. During sampling, information on potential determinants of exposure was collected. It was used in analysis of variance and multiple regression to identify significant predictors of exposure (determinants). The exposure levels observed in this study indicated that the bakery workers were subject to a mean inhalable dust exposure of 8.2 mg/m3 (ranging from 0.1 to 110 mg/m3), a mean a-amylase exposure of 22.0 ng/m3 (ranging from below limit of detection of 0.1 to 307.1 ng/m3) and a mean wheat antigen exposure of 109 ug/m3 (ranging from below limit of detection of 1 to 1018 ug/m3). Seventeen percent of the measured inhalable dust exposures exceeded the regulatory exposure limit of 10 mg/m3. Some wheat antigen and a-amylase exposures showed levels capable of causing respiratory symptoms. Regression, models which explained 74 to 79% of variability in inhalable dust, a-amylase and wheat antigen exposures were constructed. According to the models, tasks such as weighing, pouring and operating dough forming machinery increased flour dust exposure, while fully automated forming, packing, catching and decorating decreased exposure. Bread, bun and puff pastry production lines were associated with increased exposure, while cake production was associated with decreased exposures. We describe a novel method of controlling the exposures in bakeries via the substitution of dusting with the use of the divider oil, and discuss which pieces of equipment appear to be associated with elevated exposures. ii Table of Contents Abstract • 1 1 Table of Contents iu" List of Tables... v List of Figures •• ™ List of Abbreviations Acknowledgments. vi" Chapter 1 Introduction 1 Chapter 2 Background 3 2.1 Bakeries..: 3 2.1.1 Disease overview 3 2.1.2 Exposure overview 4 2.1.3 Determinants of flour dust exposure in bakeries. 6 2.2 Methodological considerations in studies of determinants of exposure 7 Chapter 3 Methods 10 3.1 Sampling strategy 1° 3.1.1 Rationale and framework 10 3.1.2 Recruitment and sample development 10 3.1.3 Final form of sampling strategy. 12 3.2 Collection of personal inhalable dust samples 12 3.3 Gravimetric analysis 13 3.4 Wheat antigen and ct-amylase assays (ELISA) 13 3.5 Quality control 16 3.6 Collection of information on determinants of exposure 17 3.7 Data analysis 2 0 3.7.1 Calculation of personal exposure 20 3.7.2 Estimation of time spent per task 21 3.7.3 Description of variables 21 3.7.4 Analysis of variance and comparisons of means 22 3.7.5 Multiple regression 23 3.7.6 Building an empirical model of exposure 23 3.7.7 Regression models examining inhalable dust as a predictor of a-amylase and wheat antigen exposure 25 3.7.8 Statistical software 26 Chapter 4 Results 27 4.1 Bakers and bakeries 27 4.2 Description of potential determinants of exposure 28 4.3 Description of exposure measures 31 4.4 Effect of bakery, type of product manufactured and dough-forming techniques on flour dust exposure 50 4.5 Effect of use of divider oil on personal flour dust exposures 52 4.6 Effect of mixer type on personal flour dust exposures 53 4.5 Regression models of inhalable dust exposure 54 4.5.1 Building regression Models 1 and 2 of inhalable dust exposure 54 4.5.2 Verification of regression analysis assumptions in the model of inhalable dust exposure55 4.6 Regression models of wheat antigen exposure 57 iii 4.6.1 Building regression Models 1 and 2 of wheat antigen exposure 4.6.2 Verification of regression analysis assumptions in the final forms of model 1 and 2 of wheat antigen exposure 60 4.6.3 Inhalable dust as a predictor of wheat antigen exposure. 4.7 Regression models of a-amylase exposure 4.7.1 Building regression Models 1 and 2 of a-amylase exposure 4.7.2 Verification of regression analysis assumptions in the model of a-amylase exposure... 4.7.3 Inhalable dust as a predictor of a-amylase exposure Chapters Discussion 5.1 Inhalable dust exposure levels 5.2 Wheat antigen and a-amylase exposure levels 5.3 Determinants of exposure and recommendations 5.3.1 Production lines defined by products manufactured 5.3.2 Dusting vs. use of divider oil 5.3.3 Dough-forming and mixing equipment.... 5.3.4 Other task-related variables and amount and type of flour used. 5.3.5 Inhalable dust as a predictor of wheat antigen and a-amylase exposure levels 5.3.6 Comparison to other researchers' findings.... 5.4 Limitations 5.5 Future directions 5.6 General conclusions Bibliography Appendix A A correlation between a-amyalse and wheat antigen exposure Appendix B Selection of variables for analysis in multiple linear regression iv List of Tables Table 1 Measures of exposure to inhalable flour dust among workers and associated fractions of a-amylase, total protein and water soluble protein in flour (from Burdorf et al., 1994) 5 Table 2 Definition of tasks considered to be potential determinants of exposure 18 Table 3 Grouping of equipment used in forming dough" 18 Table 4 Comparison of bakeries enrolled in the study with those that were eligible, but declined to participate.... 28 Table 5 Summary of descriptive statistics of all potential determinants of exposure represented by continuous variables 30 Table 6 Summary of number of bakers for all potential determinants of exposure represented by categorical variables, by bakery 31 Table 7 Results of paired t-tests of exposure levels measured at different shoulders 33 Table 8 Summary of descriptive statistics of exposure measures for all bakeries 33 Table 9 Summary of descriptive statistics of inhalable dust exposure (mg/m3), stratified by bakery 46 Table 10 Summary of descriptive statistics of a-amylase exposure (ng/m3), stratified by bakery 46 Table 11 Summary of descriptive statistics of wheat antigen exposure (jig/m3), stratified by bakery 46 Table 12 Summary of descriptive statistics of inhalable dust exposure (mg/m3), stratified by product manufactured.47 Table 13 Summary of descriptive statistics of a-amylase exposure (ng/m3), stratified by product manufactured... 47 Table 14 Summary of descriptive statistics of wheat antigen exposure (fig/m3), stratified by product manufactured.48 Table 15 Summary of descriptive statistics of inhalable dust exposure (mg/m3), stratified by type of forming equipment 48 Table 16 Summary of descriptive statistics of a-amylase exposure (ng/m3), stratified by type of forming equipment 49 Table 17 Summary of descriptive statistics of wheat antigen exposure (ng/m3), stratified by type of forming equipment 49 Table 18 Summary of inferential statistics about geometric mean of flour dust exposure measures, stratified by bakery 50 Table 19 Summary of inferential statistics about geometric mean of flour dust exposure measures, stratified by product manufactured 51 Table 20 Summary of inferential statistics about geometric mean of flour dust exposure measures, stratified by type of forming equipment 52 Table 21 Effect of use of divider oil in bread and bun production on exposure in bakery workers3 52 Table 22 Effect of different mixer designs on exposure in bakery workers8 who performed mixing tasks 53 Table 23 Parameters of regression Models 1 of inhalable dust exposure at step I of variable selection process 54 Table 24 Parameters of inhalable dust exposure model in final form 55 Table 25 Parameters of regression Models 1 of wheat antigen exposure at step I of variable selection process 57 Table 26 Parameters of regression Models 1 of wheat antigen exposure at step II of variable selection process.... 58 Table 27 Parameters of regression Models 2 of wheat antigen exposure at step A of variable selection process.... 59 Table 28 Parameters of regression model of wheat antigen exposure as a function of inhalable dust exposure and production characteristics 63 Table 29 Parameters of regression Model 1 of a-amylase exposure at step I of variable selection process 64 Table 30 Parameters of regression Model 1(2) of a-amylase exposure at step 11(A) of variable selection process. 65 Table 31 Parameters of a-amylase exposure model in final form 66 Table 32 Parameters of regression model of wheat antigen exposure as a function of inhalable dust exposure and production characteristics 69 Table 33 Comparison of inhalable dust levels (in mg/m3) found in this study with those observed by other researchers 71 Table 34 Comparison of wheat antigen levels (in jig/m3) and a-amylase levels (in ng/m3) found in this study with those observed by other researchers 72 Table 35 Summary of important determinants of flour dust exposure, identified by this study 83 Table 36 Selection of treatment of predictor variables in regression analysis of natural logarithm of dust available for inhalation (dependent variable) .- 91 Table 37 Selection of treatment of predictor variables in regression analysis of natural logarithm of wheat antigen available for inhalation (dependent variable) 95 Table 38 Selection of treatment of predictor variables in regression analysis of natural logarithm of a-amylase available for inhalation (dependent variable) 99 v List of Figures Figure 1 Distribution of measured inhalable dust exposures (mg/m3) 34 Figure 2 Plots investigating assumption of normality of distribution of measured inhalable dust exposures (mg/m3) : 35 Figure 3 Distribution of natural logarithm of measured inhalable dust exposures (mg/m3) 36 Figure 4 Plots investigating assumption of normality of distribution of natural logarithm of measured inhalable dust exposures (mg/m3) 37 Figure 5 Distribution of measured wheat antigen exposure (ug/m3) 38 Figure 6 Plots investigating assumption of normality of measured wheat antigen exposure (ug/m3) 39 Figure 7 Distribution of natural logarithms of measured wheat antigen exposure (ng/m3) 40 Figure 8 Plots investigating assumption of normality of distribution of natural logarithms of measured wheat antigen exposure (ug/m3) 41 Figure 9 Distribution of measured a-amylase exposures (ng/m3) 42 Figure 10 Plots investigating assumption of normality of measured a-amylase exposure (ng/m3) 43 Figure 11 Distribution of natural logarithms of measured a-amylase exposure (ng/m3) 44 Figure 12 Plots investigating assumption of normality of distribution of natural logarithms of measured a-amylase exposure (ng/m3) 45 Figure 13 Plots investigating assumption of normality of residuals in model of inhalable dust exposure 56 Figure 14 Plot investigating assumption of linearity and homogeneity of variance in the model of inhalable dust exposure. 56 Figure 15 Plots investigating assumption of normality of residuals in model 1 of wheat antigen exposure 60 Figure 16 Plot investigating assumption of linearity and homogeneity of variance in the model 1 of wheat antigen exposure...... 61 Figure 17 Plots investigating assumption of normality of residuals in model 2 of wheat antigen exposure 61 Figure 18 Plot investigating assumption of linearity and homogeneity of variance in the model 1 of wheat antigen exposure 62 Figure 19 Linear correlation between natural logarithms of inhalable dust (mg/m3) and wheat antigen exposure (Ug/m3) 63 Figure 20 Plots investigating assumption of normality of residuals in model of a-amylase exposure 67 Figure 21 Plot investigating assumption of linearity and homogeneity of variance in the model of a-amylase exposure 67 Figure 22 Linear correaltion between natural logarithms of inhalable dust (mg/m3) and a-amylase exposure (ng/m3) 68 Figure 23 Linear correlation between natural logarithms of a-amylase and wheat antigen exposure (Spearman r between a-amylase and wheat antigen was calculated to be 0.78 (p = 0.000)) 90 vi List of Abbreviations a probability of type I error %2 Pearson Chi-square Ug/m3 micrograms per cubic meter urn micrometer lx| absolute value of X * multiplied by ± plus or minus / divided by ACGIH American Conference of Governmental Industrial Hygienists A M arithmetic mean B.C. British Columbia Dev deviation eg. for example F-statistic Fisher statistic g grams GM geometric mean GSD geometric standard deviation IgE immunoglobulin class E IgG4 immunoglobulin class G, subclass 4 L/min liter per minute lnX natural logarithm of X kg kilogram min minute Min minimum value Max maximum value mg/m3 milligrams per cubic meter mm millimeter n sample size nm nanometer ng/m3 nanograms per cubic meter P probability of statistical significance Q-Q expected quartiles versus observed quartiles r Pearson correlation coefficient R2 coefficient of determination Rsq coefficient of determination RSR regression studentized residual R adj coefficient of determination, adjusted for inflationary effects SD standard deviation SIC Standard Industrial Classification TLV Threshold Limit Value WCB Workers' Compensation Board X, Y o r Z any numbers or variables X' some function of X xy X taken to the power of y vii Acknowledgments I am grateful to Dr. Susan Kennedy and Dr. Kay Teschke for all their guidance and support in all aspects of the project, without which it would not have been possible. I also want to thank Dr. Steve Marion for his advice on statistical matters. Drs. Heederik, Doekes and Houba of Wageningen Agricultural University deserve acclamation for sharing their experience, derived from studies of bakeries and allowing us to use wheat antigen and a-amylase assays developed by them. Laboratory staff at Wageningen Agricultural University should be recognized for supplying materials for ELISA and help with performing the immunoassays. I also extend my gratitude to Karen Bartlett for performing ELISA and her patience in explaining its finer points to me. To Pamela Cruise goes my gratitude for moral support and superb editing job on this manuscript. All study participants, labor and management alike, deserve recognition for their cooperation in making sure that this study took place. My special thanks go to all those who agreed to partake in personal sampling: both for effort made in carrying the sampling devices and great company they provided me with during sampling shifts. Finally, I want to express gratitude to the B.C. Lung Association and the National Engineering and Sciences Research Council of Canada for funding this research. viii Chapter 1 Introduction Evidence documenting the condition known as 'baker's asthma' began with Roman slaves (millers and bakers) who had to wear masks and gloves when handling flour (Thiel and Ulmer, 1980). Baker's asthma was one of the occupational diseases described by Ramazzini (1713). Although bakers have been shown to develop asthma in response to a variety of food products, the major respiratory hazard of employment in bakeries appears to be exposure to flour dust and its components, leading to outcomes such as rhinitis, chronic bronchitis, and asthma. Even though it has been suggested that the hazard of developing respiratory symptoms due to flour dust exposure has decreased (Kallos and Kallos-Deffiier, 1971; Popa et al., 1970), in Quebec, flour was recently recognized as the second most common cause of occupational asthma (Lagier et al., 1990). A similar finding was obtained in the UK (Gannon and Burge, 1993). A fatality due to baker's asthma has also been recently reported (Ehrlich, 1994). The only investigation of respiratory and allergic disorders among B.C. bakers was conducted in 1983 by Block et al. The authors reported a total of 15 symptomatic bakery employees, of which 12 were identified in a survey of a workforce of 50. These findings are especially important, since it is generally recognized that baker's asthma is underreported (Gannon and Burge, 1993; De Zotti et al., 1994). Currently, no published data are available about the prevalence or incidence of baker's asthma in British Columbia. Five Workers' Compensation Board claims for respiratory irritation due to flour were accepted in British Columbia between 1989 and 1993: two in the food industry, and three in the restaurant-hotel industry (WCB, 1994). (In B.C. WCB claims statistics, occupational asthma claims are included in the "respiratory irritation" category). The 1993 Manufacturers' Directory lists 2,329 individuals employed in bakeries in B.C. As an estimate of the number of flour-exposed bakers in the province, this value may either be low, since the Directory does not include bakeries in large grocery stores, or high, since it is not restricted to bakery workers exposed to flour dust. If, however, we assume that this is a reasonable 1 estimate of the number of bakery employees exposed to flour dust, we can estimate the incidence of compensated baker's asthma (since it is likely that this is the only respiratory outcome that would be considered compensable under the "respiratory irritation" category in B.C.). This rate would be approximately: [(5 bakers with the disease) / (2329 bakers in B.C.) / (5 years)] * 106 = 429 per million per year. This value is very close to the estimated incidence of reported baker's asthma in the UK: 409 per million per year (Meredith et al., 1991). The true number of cases of occupational asthma is generally believed to be much higher than either the reported number or the compensated number (De Zotti et al., 1994). Although it is known that components of flour dust are involved in the etiology of baker's asthma, factors associated with increased exposure (and increased risk) are not well described. Studies measuring personal flour dust exposure in bakers have recently been performed (Masalin et al., 1988; Musk et al., 1989; Nieuwenhhuijsen, 1992; Burdorf et al., 1994; Houba, 1996a). No such studies have been done in British Columbia. Furthermore, the selection of effective control measures to prevent exposures requires an understanding of which factors in the work environment cause exposures. No studies of bakeries with such a primary focus are currently reported in the scientific literature. The purposes of this study were therefore to measure levels of flour dust exposure, define the determinants of exposure to flour dust in small B.C. bakeries and to propose control measures. 2 Chapter 2 Background 2.1 Bakeries 2.1.1 Disease overview The effects of exposure to flour dust depend on the degree of sensitization of an individual to allergens in flour dust and on the inhaled dose of these allergens. Baker's asthma is an IgE mediated asthmatic response to high molecular weight proteins present in flour dust (Blands et al., 1976; Jarvinen et al., 1979; Block et al., 1983). These proteins may be of cereal origin (e.g. wheat, rye, barley proteins), or from non-cereal additives or contaminants (see section 2.1.2, page 4). The population at greatest risk for baker's asthma includes those persons who handle flour on a day-to-day basis: bakers, pastry makers, cooks and animal feed manufacturers. Allergic responses to flour have been reported to develop in 10 to 30 % of all bakers (Schwartz, 1947; Herxheimer, 1973; Blands et al., 1976). In a large cross-sectional study of 318 bakers, 13% reported work-related chest symptoms (Musk et al., 1989). The prevalence of asthma among bakers has been reported to range from about 9 -10 % (Bjorkstein, 1982; Theil, 1980) to 21% (Venables, 1987). It has been suggested, however, that these estimates may be low, since there is evidence that individuals susceptible to flour allergy or predisposed to asthma leave the trade very early (Prichard et al., 1984, De Zotti et al., 1994). The incidence of the disease has been shown to increase with increased duration of employment in a bakery (Popa et al., 1970; Thiel and Ulmer, 1980; Prichard et al., 1984; Prichard et al., 1985). Both positive skin test response to wheat antigens and nonspecific bronchial hyperreactivity increase with duration of employment as a baker (Prichard et al., 1985). It has been reported that the prevalence of positive skin tests to wheat increased from 9% in baker recruits to 30% after five years of work as a baker (Herxheimer, 1973). Baker's asthma has been reported to have an average latency period of 4.2 years from commencement of employment in a bakery (Jarvinen et al., 1979), although this is highly variable. A 3 relationship has also been shown between exposure levels measured subjectively as perceived dustiness, and increased asthma prevalence (Musk et al., 1989). Bakery size appears to play a role, perhaps associated with exposure level. Skin sensitization occurred among 25% of bakers in large bakeries versus 44% in small bakeries (Von Dishoeck and Roux, 1939). Rowe has reported a decrease in forced vital capacity in bakers in small bakeries, while no such changes were observed in a large bakery's employees (Rowe, 1989). A longitudinal study of British bakery workers (Nieuwenhuijsen et al., 1994a) found an association between the geometric mean of flour allergen exposure and work-related symptoms (consistent with flour dust allergy) in individuals not previously exposed to flour in occupational settings. The duration of occupational exposure in that group was on the order of two years. A strong positive dose-response relationship was recently demonstrated between flour aeroallergen exposure (measured by a-amylase and wheat protein) and both sensitization and work-related symptoms among bakery workers (Houba, 1996a). 2.1.2 Exposure overview The specific allergens implicated in respiratory illnesses of flour handlers include cereal proteins such as wheat, soybean, rye, barley, rice, oat, corn and buckwheat proteins (Bonnevie, 1958; Herxheimer, 1973; Heidrick et al., 1976; Napolitano and Weiss, 1978; Thiel and Ulmer, 1980; Block et al., 1982; Block et al. 1984; Prichard et al., 1984; Prichard et al., 1985; Walsh et al., 1985; Popp et al., 1988); contaminants such as grain weevils (Frankloud and Luann 1965), rice-flour-beetles (Schultze-Werninghaus et al., 1991), spores oiAlternaria and Aspergillus organisms (Klaustermeyer et al., 1977), and storage mites (Popescu, 1981; Tee et al., 1992); and in recent decades, dough improvers and other additives, such as a-amylase and cellulase from Aspergillus oryzae (Popa et al., 1970; Baur et al., 1988; Wuthrich and Baur, 1990; Brisman and Belin, 1991; Quirce et al. 1992). 4 Most exposure studies have not focused on these specific antigens, but have simply measured total particulate exposure levels. Musk et al. (1989) measured personal flour dust exposure, and found 9 out of 79 measurements to be above the exposure limit for nuisance dust of 10 mg/m3 (ACGIH TLV, 1994). The authors concluded that "in a modern bakery control of dust exposure presents a continuous problem". Nieuwenhuijsen et al. (1992) reported personal total dust exposure in bakeries varying from 0.1 mg/m3 to 28.5 mg/m3, with an average below 5 mg/m3. A study of the confectionery industry (baking being the major activity) in Finland found 63% of total dust concentrations below 2.5 mg/m3, 25% between 2.5 and 5.0 mg/m3, 7% between 5.0 mg/m3 and 10 mg/m3, and 5% above 10 mg/m3 (Masalin et al., 1988). Burdorf et al. (1994) measured total protein and a-amylase to help isolate the allergenic components of the dust; their results are summarized in Table 1. The method used for a-amylase assay in their study is limited in that it only accounts for catalytically active enzymes. It has been suggested that the allergenic domains of the enzyme do not have to be accompanied by an active protein. Table 1 Measures of exposure to inhalable flour dust among workers and associated fractions of a-amylase, total protein and water soluble protein in flour (from Burdorf et al., 1994). Measure of exposure n A M GM GSD Range Inhalable dust (mg/m3) 129 3.83 2.48 2.77 0.01-16.90 Total protein as a fraction of inhalable dust (ug/mg) 40 104.90 90.00 1.76 22.00-340.00 Water-soluble protein as a fraction of inhalable dust (ug/mg) 3 27.30 23.15 2.03 12.00-49.00 a-Amylase as a fraction of inhalable dust (ug/mg) 61 0.36 0.25 2.52 0.03-1.67 n = number of measurements, AM = arithmetic mean, GM = geometric mean, GSD = geometric standard deviation Nieuwenhuijsen et al. (1994b) attempted to overcome this limitation by measuring specific flour allergens (wheat antigens) using rabbit polyclonal anti-flour sera in personal inhalable dust exposure samples. Observed geometric means of levels in different exposure groups ranged from 0.4 mg/m3 to 6.4 mg/m3 for inhalable dust and from 45.5 ng/m3 to 252.0 ug/m3 for wheat antigens. 5 Houba et al. used a similar approach in addressing the problem of heterogeneity of flour allergens. He employed pooled serum from bakery workers to detect wheat antigens in inhalable dust samples (Houba et al., 1996b) and rabbit polyclonal anti-ct-amylase sera to measure personal a-amylase exposure (Houba, 1996a). Geometric means of inhalable dust exposure in various job categories ranged from 0.4 mg/m3 to 4.5 mg/m3. Geometric means of wheat antigen exposure in different job categories varied between 0.036 ug/m3 and 15.5 ug/m3 and geometric means of a-amylase exposure varied between below the limit of detection (0.250 ng/m3) and 18.1 ng/m3. 2.1.3 Determinants of flour dust exposure in bakeries Nieuwenhuijsen et al. (1995c) found that exposures associated with specific tasks in bakeries exceeded work-shift average exposures containing these tasks, suggesting that exposures to flour dust and flour allergens occur in a series of peaks associated with specific tasks. Some of the cleaning tasks in bakeries were associated with the highest levels of exposure. Still, other cleaning tasks appeared to result in lower exposures than the flour dusting task group (bread and roll production, dough brake) (Nieuwenhuijsen et al. 1995c). In addition, a high correlation between the number of peak exposures and full-shift time-weighted averages was reported in a separate study (Jongedijk et al., 1995). This indicates that a closer investigation of task profiles of bakery workers may lead to better understanding of significant predictors of flour dust exposure levels. Nieuwenhuijsen et al. (1995b) reported that for flour dust exposure 'the largest variance component was the between-group component followed by the considerably smaller between-worker and within-worker components'. Since exposure groups were formed on task-specific criteria, this is in agreement with an earlier observation that task profile is the primary determinant of flour dust exposure among bakery workers. Burdorf et al. (1994) report that exposure profiles of bakery workers vary significantly with task groups. Dough making/bread forming task groups and oven workers/packers task groups were observed to form distinct exposure categories. Task-specific grouping accounted for 61 to 69 percent of total variability in flour dust exposure, making it the principal source of variance in the study. 6 Specific tasks such as mixing, packing (Brisman and Belin, 1991) and weighing (Jauhiainen et al., 1993) powdered flour additives have been associated with increased a-amylase exposures. Houba (1996a) found that tasks and products manufactured accounted for most of the variation in inhalable dust and wheat antigen exposure in bakeries. The same study demonstrated that a-amylase exposure occurred almost exclusively among bakers making dough on bread and crispbake production lines. However, a large between-worker variability in a-amylase exposure during bread production was difficult to explain from the data collected, introducing a possibility that the variability in tasks within each job title and production line designation may be an important factor in the variability of a-amylase exposure. 2.2 Methodological considerations in studies of determinants of exposure Experimental studies of factors influencing exposure (other than flour dust) have been reported in the literature (e.g. Andersson and Rosen, 1995; Archibald, 1995; Plinke et al., 1995; Thorpe and Brown, 1995; Heinonen et al., 1996). They provide detailed descriptions of the circumstances under which exposure was measured, and real-time exposure measurements. Such studies require a strong a priori hypothesis with respect to factors that cause exposure. If that is the case, it is possible to focus a study on a particular piece of equipment (e.g. Thorpe and Brown, 1995) or task (e.g. Andersson and Rosen, 1995; Archibald et al., 1995; Plinke et al., 1995; Heinonen et al., 1996). Experimental studies generally suffer from a lack of generalizability due to the small number of workers and working environments studied, but they are the design of choice if a task or a piece of equipment has been identified as a principal source of exposure. An advantage of experimental studies is the control a researcher has over the conditions under which exposure is measured. This enables the isolation of the effects of parameters which influence exposure levels. As can be expected, such focused studies can rarely be conducted in real workplaces. In many occupational settings, there is little information available on sources of exposure, especially if several such sources are conceivable. When faced with such a situation, a researcher might 7 choose to conduct an observational study, involving exposure measurement in a variety of workplace situations aimed at describing the circumstances under which exposure occurs. In the past, such observational studies have been conducted in a variety of settings, usually as the exposure assessment part of an epidemiological investigation. Since the primary objective of an epidemiological study is to establish a dose-response relationship, information collected on the determinants of exposure is useful in establishing the level at which workers are exposed. However, a good predictive model of exposure does not always provide information on causes of exposure (i.e., where efforts to control exposure should be focused). For example, Smid et al. (1992a), having grouped workers according to job category, was able to arrive at meaningful exposure categories in a cross-sectional study of respiratory effects. Nevertheless, due to the limited amount of information collected on causes of exposure, the author found it difficult to (a) explain causes of differences between exposures at facilities with different dust control measures and (b) discuss factors causing exposure within the job categories (Smid et al., 1992b). The author concluded that control 'measures ... will... be difficult to implement' (Smid et al., 1992b). Similar situations occurred in elucidating determinants of a-amylase exposure among doughmakers in bread factories (Houba, 1996a), where efficient grouping of workers by exposure level was achieved. However, the level of detail of the information obtained on determinants of a-amylase exposure was insufficient to suggest exposure control measures. Thus, information collected in epidemiological studies on determinants of exposure is not always useful for purposes of controlling exposures. Having said that, there are a number of epidemiological studies that provided detailed information on the determinants of exposure that can be used to develop exposure control measures (e.g. Preller, 1995; Teschke et al., 1995; Nieuwenhuijsen et al., 1995c). Recently, the use of empirical modeling exclusively for the development of exposure control strategies has become more common (e.g. Kromhout et al., 1994; Nieuwenhuijsen et al., 1995d; Scheeper et al., 1995; Zock et al., 1995; Kumagai et al., 1996). Regardless of the primary objective of an investigation, empirical modeling of determinants of exposure is characterized by detailed description of work environments and worker activities (which can be 8 altered by industrial hygiene control measures) during an exposure measurement period for large samples of workers. To obtain such detailed information, several approaches have been used. The most obvious one is to observe workers during exposure measurement (e.g. Teschke et al., 1995). The above approach is very labor-intensive, and consequently, task profile diaries (e.g. Nieuwenhuijsen et al., 1995d; Preller, 1995), worker interviews (e.g. Kromhout et al., 1994) or questionnaires at the end of sampling (e.g. Scheeper et al., 1995), and production characteristics (e.g. Zock et al., 1995; Kumagai et al., 1996) are used as an alternative. Another approach to identifying sources of exposure in an observational study is to measure exposure close to and away from a potential source (e.g. Scheeper et al., 1995). The latter method, unfortunately, underestimates personal exposure and ignores worker-machine interaction (Scheeper et al., 1995). If it is suspected that exposure occurs in a series of peaks associated with various activities, a task-specific sampling strategy can be applied (Nicas and Spear, 1993) providing the duration of the task is sufficiendy long to obtain a detectable exposure measurement (Preller, 1995; Teschke et al., 1995), as in the study by Nieuwenhuijsen et al. et al. (1995c). Thus, short-term or long-term sampling can be used in studies of the determinants of exposure, depending on measurement techniques and the nature of exposure. It is worth noting that short tasks connected to high exposure might not be found to be associated with a full-shift time-weighted average exposure, unless their duration is estimated (Kromhout et al., 1994). One of the difficulties with all observational studies is that they only allow the examination of effects of particular determinants that do vary in the studied population. Thus, in the study of Kromhout et al. (1994) it was not possible to examine the effect of some tasks' duration on exposure, since these tasks were performed for the same duration by all workers. 9 Chapter 3 Methods 3.1 Sampling strategy 3.1.1 Rationale and framework It was decided that the study would be focused on the empirical modeling of inter-individual exposure variability in the study population. A cross-sectional design in sampling strategy was considered to be the most consistent with that objective. Therefore, we set out to collect one exposure measurement from each individual in the recruited bakeries. Since types of work performed in each bakery varied with shifts and the particular product produced, sampling was repeated within each bakery to obtain a cross-section of process/product combinations. Such a sampling strategy ignores variability on the level of individual bakers (intra-personal variability) so as to include as many eligible bakery employees as possible. Assuming intra-personal variability is small compared to inter-personal sources (Nieuwenhuijsen et al. 1995b), this strategy provides for observing the largest number of distinct combinations of determinants of exposure, maximizing the efficiency of the study. This study focused on small to medium size bakeries for two reasons: a) the scientific literature revealed little information on exposures that occur in such bakeries, and b) with regard to the prevention of occurrences of baker's asthma, it appears reasonable to focus efforts on controlling exposures among bakery workers who appear to be at greatest risk for the disease (see section 2.1.1, page 3). 3.1.2 Recruitment and sample development The size distribution of bakeries that operate in British Columbia was examined using data available through the 1993 Manufacturer's Directory of British Columbia , the 1994 Vancouver Area Business Directory, and the 1994 Lower Mainland Business Directory. The majority employed between 6 and 25 individuals. This provided additional impetus to concentrate the study on small to medium size bakeries, making it relevant to the working conditions of the majority of bakers. According to studies done 10 by Burdorf et al. (1994) and De Zotti et al. (1994), these bakeries can be classified as small to medium in size. Only bakeries located in the cities of Vancouver, North Vancouver, Burnaby, Richmond, New Westminster or West Vancouver were considered in the following selection process. Step 1: Businesses listed under SIC 1072: "Bread and Other Bakery Products Industry" and under Products: "Bakery Products (Including Frozen but Excluding Pre-Cooked Frozen)" were selected if they employed 5 to 30 individuals (as reported by the 1993 B.C. Manufacturer's Directory). Step 2: The 1994 Lower Mainland Business Directory and the 1994 Vancouver Area Business Directory were consulted. The search was restricted to SIC 5462: "Retail Bakeries -Baking and Selling" and SIC 2051: "Bread Cake and Related Products". Bakeries with 6 to 25 employees were selected. Lists of bakeries considered eligible upon completion of steps 1 and 2 were combined, yielding a list of 123 bakeries. Step 3: A telephone survey of all bakeries selected upon completion of step 2 was conducted in June of 1995. The following questions were asked: Do you make your own dough? How many individuals are involved in handling flour/dough per shift? What are your major products? Who is your major flour supplier? Do you use dough improvers like a-amylase? Step 4: As it was observed that bakeries categorized by the Directories as having 6 to 10 employees had less than 5 individuals handling dough or flour per shift, the list obtained as a result of step 2 was further restricted to bakeries reported to have 11 to 25 employees, yielding a list of 45 bakeries. 11 Step 5: The list from step 4 was further reduced by excluding bakeries that were discovered, as a result of the telephone survey (step 3), to have fewer then 5 individuals handling dough or flour per shift, yielding a list of 26 bakeries. Step 6: All bakeries that do not make their own dough (four) were excluded. All eligible bakeries (22) were invited to participate in the study. A letter of invitation was followed by a phone call to the manager/owner. Written consent was obtained from each study participant on the day of sampling. This study was approved by UBC Behavioral Sciences Screening Committee. 3.1.3 Final form of sampling strategy All bakeries that agreed to participate in the study were visited by a researcher to confirm the findings of the telephone survey in terms of enumerating eligible bakers. Only individuals who spend most of their workday on the premises of the selected bakery were considered for personal exposure measurement. If six or fewer bakers were eligible to participate, all of them were planned to be sampled on the same day. If more than six workers per shift were eligible to participate and the bakery contained several production lines, sampling was restricted to one production line per shift. Due to logistical limitations, no more than six bakers could be sampled per shift (i.e. per sampling trip). Sampling was conducted until at least one set of samples was obtained from each process/product combination that occurred in recruited bakeries. Sampling trips were scheduled for randomly chosen days that fell between September 14, 1995 and March 27, 1996. Proposed days of sampling were censored to accommodate logistical considerations, such as equipment and personnel availability. Whenever necessary, a replacement random date was drawn until logistical requirements were satisfied. 3.2 Collection of personal inhalable dust samples The particulate samples were collected on 25-mm diameter 0.45 urn pore size Teflon filters with PVC support (Gelman Sciences # 66148). Seven-hole inhalable dust samplers were used to restrict the 12 aerosol size of the samples. The inlet has a 50% aerodynamic diameter cut-off of about 100 um and collects the inhalable fraction of particulate (i.e. all those captured from the nasal region down). A seven-hole inhalable dust sampler was chosen over other inhalable dust samplers because it provides an unbiased estimate of the "inhalable" dust curve at low air velocities (HSE, 1995). Air velocities defined as "low" occur in indoor environments (and hence, in bakeries). Standard personal sampling pumps (SKC model 224-PCXR4) were used to draw air through the filters. The pumps were calibrated to a flow rate of 2 L/min at each sampling site immediately prior to the commencement of sampling. Flow rates were re-measured at the end of sampling. Air flow though sampling apparatus was measured as an average of 10 readings by a frictionless piston (DC-1 Flow Calibrator, BIOS International). Because, prior to the study, it was thought that each inhalable dust sample could be subjected to only one immunoassay, two inhalable dust samples per individual per sampling event were collected (with samplers placed at either shoulder). However, this concern was not justified, due to the similarity of protein extraction procedures between the immunoassays. On sampling days, eligible bakers were asked to wear sampling devices for the duration of their work-shifts. The duration of sampling was recorded. 3.3 Gravimetric analysis Total inhalable dust was determined gravimetrically. The filters were equilibrated at a stable temperature and relative humidity (20°C ± 0.3°C and 50% ± 5% relative humidity) for 12 hours prior to both pre- and post-sampling triplicate weighings on a electrobalance readable to 10 ug (Mettler Analytical balance AE240). 3.4 Wheat antigen and a-amylase assays (ELISA) Wheat antigen and a-amylase exposure were determined from measurements of the protein content in the water soluble fraction of inhalable dust. For wheat antigen, inhibition enzyme-linked immunosorbent 13 assay (ELISA) was used; for a-amylase — sandwich ELISA. Each filter was subjected to both immunoassays. The antibodies used to detect wheat antigen and a-amylase were obtained from Wageningen Agricultural University. The anti-wheat antigen antibodies were derived from pooled serum from sensitized bakery workers and anti-a-amylase antibodies were obtained from rabbit sera (Houba, 1996a). This allowed us to measure wheat antigen exposure in terms of epitopes implicated in causing allergic reactions in bakers and a-amylase exposure in terms of antigenically active proteins. We followed the method for wheat antigen and a-amylase developed in Wageningen Agricultural University (Houba et al., 1996b; Houba et al., 1996c). BioRad ELISA plates were used. A Molecular Devices Thermomax® Kinetic Microplate Reader was used to read the ELISA plates. SOFTmax PRO software (Molecular Devices) was used to construct the standard curves and calculate the antigen concentartion in ELISA wells. Water soluble fractions of inhalable dust were extracted from the filters with 2.5 ml 0.15M Phosphate Buffered Saline (PBS, pH 7.4) in 10 ml centrifuge tubes. Supernatants were separated and stored at -20°C for up to 10 months. Wheat antigen and a-amylase standards were supplied by Wageningen Agricultural University and were identical to those used by Houba et al. (1996b; 1996c). They were stored at 4°C and diluted in PBS prior to use in ELISA. The protein concentration in the standards was measured via BCA-protein assay (BioRad Protein Assay Kit II). In the inhibition ELISA for wheat antigen, 100 ul of the water soluble fraction of the flour dust samples were mixed with known dilutions of the water soluble fraction of wheat antigen (100 ul with concentrations starting at 12.5 ug/ml protein in Phoshate Buffered Tween Gelantine (PBTG, pH 7.4)) and a limiting titre (100 ul of 1:800 diluted pooled serum) of IgG4 human anti-wheat antibodies (detection antibodies, derived for bakery workers, Houba et al., 1996b). Plates were washed with PBS-Tween and incubated with PBTG at 37°C for 30 minutes. Next, they were added to ELISA wells coated overnight at 4°C with wheat antigen at a concentration of 5 ug/ml protein in 0.15M PBS in volumes of 200 ul/well. Plates were incubated for 2 hours at 37°C. At this stage, detection antibodies that have not yet formed a 14 cx>mplex with the wheat antigen, originating from either the unknown or the antigen solution, bind wheat antigen coating ELISA wells. Only antigen-antibody complexes bound to the wells' walls remained in the wells after washing. Plates were washed with PBS-Tween. Visualizing antibodies (i.e. anti-human IgG4 antibodies conjugated with peroxidase), were added next, binding detection antibody-antigen complexes. Plates were incubated for 1 hour at 37°C. Subsequently a colorimetric reaction was triggered by incubation 30 minutes in the dark at 20°C with 200 pi o-phenylenediamine (2 mg/ml in 0.05 citrate/phosphate buffer, pH 5.5, containing 0.015% H 20 2) and stopped by adding 50 ul 2M HC1. Optical density was measured at 450 nm. As the proportion of wheat antigen in the unknown increases, relative to the dilutions of wheat antigen solution, the amount of detection antibody available to bind antigen on the well's walls decreases, resulting in a reduced optical density reading (this is referred to as a decrease in % inhibition). Hundred percent inhibition occurs when antigen in the unknown becomes non-detectable. Standard curves, spanning concentrations between 0.40 pg/ml to 12.50 ug/ml wheat antigen, were prepared on each ELISA plate analyzed. Each concentration on standard curve was prepared in duplicate. Four parameter calibration curves with known concentrations of wheat antigen were constructed for each ELISE plate analyzed. In the sandwich ELISA for a-amylase, ELISA wells were first coated overnight at 4°C with 2.35 pg/ml in 200 pi of detection antibodies (rabbit anti-a-amylase, described by Houba et al., 1996c). The detection antibodies were stored at -20°C. Part of these antibodies were biotinylated (Houba, 1996a). Next, plates were washed with PBS-Tween and incubated for 30 min with PBTG. Dilutions of 200 pi of the unknown in standard a-amylase preparation were added to the wells and incubated for 1 hour at 37°C. Plates were washed with PBS-Tween. Subsequently, 200 pi of 1:500 diluted biotinylated detection antibodies were added to each well and incubated for 1 hour at 37°C. After the excess of the biotinylated antibodies was washed away, 200 pi of 1:2000 diluted avidin-peroxidase conjugates were added to the wells. During 1 hour of incubation at 37°C, they formed bounds with biotinylated detection antibodies which were part of antigen-antibody complexes. A colorimetric reaction was triggered by incubation 30 15 minutes in the dark at 20°C with 200 ul o-phenylenediamine (2 mg/ml in 0.05 citrate/phosphate buffer, pH 5.5, containing 0.015% H202) and stopped by adding 50 ul 2M HC1. Optical density was measured at 450 nm. Increase in optical density corresponded to increase in a-amylase concentration. Standard curves, spanning concentrations between 0.035 ng/ml to 2.000 ng/ml of a-amylase, were prepared on each ELISA plate analyzed. Each concentration on standard curve was prepared in duplicate. Four parameter calibration curves with known concentrations of a-amylase were constructed for each ELISA plate analyzed. We followed the original methods described by Houba (1996a), with the following deviations. Optical density was determined at 450 nm, instead of 492 nm, due to differences in equipment. This deviation from the original protocol should have a negligible effect on the interpretation of the assays' results, since the difference in light absorption between these frequencies is negligible in the assays. Furthermore, standard curves in our study were also constructed at 450 nm. The substrate buffer was prepared with anhydrous sodium phosphate and hydrogen peroxide was added to it just prior to use. The final composition of the substrate buffer was identical to that of the original method 3.5 Quality control The analytic limit of detection of the gravimetric analysis was determined by calculation of 3 times the standard deviation of the means of triplicate weights for 19 laboratory filter blanks (10% of the total number of field samples). The analytical limit of detection for the ELISA was estimated at the concentration associated with 100% inhibition, for wheat antigen, and exposure corresponding to the lowest concentration on the standard curve, for a-amylase. Samples with optical density readings that fell above the range of the standard curve were re-analyzed at higher dilutions. For exposure measurements below the limit of detection, the analytic limit of detection was divided by the square root of 2 for data analysis (Hornung and Reed, 1990). 16 Nine water soluble extracts, representing entire range of inhalable dust concentrations and determinants of exposure encountered during collection of samples, were analyzed for a-amylase in both our laboratory and that of Dr. Gert Doekes at Wageningen Agricultural University. The results were compared in attempt to validate results of ELISA obtained in our laboratory. For every 10 field samples analyzed, two field filter blanks were analyzed in the same way. Field blanks detennined any contamination due to handling and only differed from field samples in that no air was drawn through them. A total of 17 field blanks were collected. They were compared to the analytic limits of detection. 3.6 Collection of information on determinants of exposure The activities of the sampled bakers were observed throughout the sampling shifts and recorded at fifteen-minute intervals. If a task was performed in any fifteen-minute interval, a check-mark was made on a checklist corresponding to that task and time interval. Tasks were defined as a result of walk-through surveys by observing bakery workers and consulting with them. They were classified according to the definitions in Table 2. 17 Table 2 Definition of tasks considered to be potential determinants of exposure. Task Definition decorating manipulation of icings and creams in manufacturing of cakes/pastries; application of materials other than dough or flour onto a formed (but not necessarily baked) product mixing operation of mixer; combining flour with other ingredients forming manipulation of dough's shape, except when completely automated cutting manual cutting of dough cleaning any cleaning of equipment or production area (sweeping, washing etc.) packing/catch-ing transfer of formed products, usually from conveyor into a box or onto a tray pouring pouring of flour (manual or automated), including dusting weighing determining the weight of ingredients in powder form via scale; determing the weight of dough (after cutting) via scale tray movement of racks of trays around production area; manipulation of trays with products (usually in conjunction with packing/catching) oven operation of oven; movement of product in and out of oven rest individual is in production area, but is not performing any tasks quality control inspection of products or machinery; start-up; consultation with recipes in production area break3 person is outside of production area (lunch, coffee, smoke and washroom breaks) a ~ used to estimate time spent on other tasks; was not considered as potential determinant of exposure per se as it provided no opportunity for exposures For the task described as 'forming' in Table 2, the type of equipment used in forming was recorded. Equipment used in forming was grouped according to the definitions in Table 3. Table 3 Grouping of equipment used in forming dough". Type of equipment Definition dough-braker operation of dough-braker reversible sheeter operation of reversible sheeter automated method of 'forming' which does not involve manual dough handling manual method of 'forming' which involves use of table a - groups are mutually exclusive, i.e. it is not possible to use several types of forming equipment simultaneously 18 Type of mixers used was recorded. We differentiated between vertical and horizontal mixers which differ in plane of rotation of the agitator shaft. Vertical mixers tend to have smaller capacity and require, generally, flour to be poured into them at lower height (i.e. further from the operator's breathing zone). Horizontal mixers are used only for making dough, while vertical mixers can be utilized to make icings as well. Matz (1972) has a more detailed description of the differences between the two types of mixers. At the end of each air sampling day, each sampled baker was questioned about the raw materials (types and quantities of flour and dough improvers) used on that day. Recipes and/or production records were consulted when possible. Amounts of flour used by a baker were assigned individually, i.e. if baker B personally handled X kg of coarse flour, Y kg of enriched flour and Z kg of pastry flour, for baker B, the total flour used was (X+Y+Z) kg and X kg — for coarse flour, Y kg — for enriched flour, (X+Y) kg for coarse and enriched flour and Z kg — for pastry flour. If other bakers, sampled on the same shift as baker B, did not handle flour directly, they were assigned 0 kg of flour used for all flour types. Since dough improvers (a-amylase in particular) are added to enriched flour prior to purchase by the bakeries, we sought irrformation on the composition of flours as purchased. For each bakery employee sampled, the number of years of bakery employment and which hand was primary (i.e. left vs. right) were recorded. The method of preventing dough adhesion to surfaces was documented (i.e. dusting vs. the use of divider oil). Divider oil was applied to surfaces manually via a spray bottle. A detailed description of divider oil is available in Matz (1972). Briefly, it is a white mineral oil solution with maximum content of petrolleum hydrocarbons (mostly paraffmic and naphthenic) not exceeding 0.15%. It may contain sulfur compounds and antioxidants. Any maintenance or malfunction of equipment on each sampling day was recorded. The environmental conditions inside each bakery, including temperature, pressure and humidity were recorded using the TSI thermoanemometer at the beginning, midpoint and end of sampling on each sampling day. 19 The types of products manufactured during sampling periods were recorded. 3.7 Data analysis 3.7.1 Calculation of personal exposure Let us define following variables: <J> = sample, takes on values of 1 or 2 Mi$j = arithmetic average of triplicate weights of a filter pre-sampling (i) for sample <J> of baker j (mg) ' ' Mf^ = arithmetic average of triplicate weights of a filter post-sampling (f) for sample of baker j (mg) Qi4,j = flow rate through a sampling assembly pre-sampling (i) for sample § of baker j (m3/min) Qf4,j = flow rate through a sampling assembly post-sampling (f) for sample <j) of baker j (mVmin) T j^ = duration of sampling for sample (j) of baker j (minutes). Inhalable dust exposure measured by sample (j) of baker j (referred to as 1^ ) was calculated as I^ j = (Um - M4i)/(((Qf*> + 0^/2)*^). Since two inhalable dust samples were collected for each baker, the inhalable dust exposure of baker j (Ij) was calculated as the average of 1^  values. Wheat antigen and a-amylase exposure associated with sample § of baker j were calculated as: A» = (CVVVWQ^ + Q , ^ ) * ^ ) , where V = elution volume used in extraction of water soluble flour dust fraction (ml) and = average concentration of antigen of interest in water soluble flour dust fraction (ng/ml for a-amylase and Ug/ml for wheat antigen), as reported by ELISA. Personal wheat antigen and a-amylase exposures for baker j were calculated as the average of the A^ values. 20 3.7.2 Estimation of time spent per task It was assumed that all tasks performed in a 15 minute interval were of equal duration. Then, time (in minutes) spent per task T was defined according to the following formula: £ 1 5 / i n all i in which task T was observed where /n, = total number of check-marks corresponding to all tasks in time interval i Tasks are defined in Table 2. Check-marks are described on page 17 (section 3.6). The above algorithm was executed in SAS version 6.1. It was possible to estimate time spent using dough-brakers, since no other dough-forming equipment was used by dough-braker operators. This resulted in time engaged in the 'forming' task from Table 2, being split into 'time spent forming dough with a dough-braker' and 'time spent forming dough a without dough-braker.' 'Time spent forming dough a with dough-braker' was calculated for each baker as a product of the time spent forming and a dummy variable representing the use of a dough-braker (1 = used dough-braker, 0 = did not use dough-braker). 'Time spent forming dough without a dough-braker' was calculated for each baker as the difference between time spent forming and 'time spent forming with dough-braker.' 3.7.3 Description of variables Data were summarized descriptively for each of the three exposure measures over all bakeries, by bakery, type of product produced and type of forming equipment (means, standard deviations, geometric means, geometric standard deviations, and ranges). The distribution of each of the exposure measures was examined using frequency histograms, and tested for normality using Q-Q plots. It was anticipated that log-normal distributions would be observed for the exposure measures, necessitating ln-transformation of the observed exposure measures prior to their entry into a regression model or ANOVA. 21 Categorical variables were constructed as dummy variables (coded as 0 or 1 in sets of variables). Each of the independent variables were also summarized descriptively (counts for categorical variables; means, standard deviations, medians, and ranges for continuous variables). Spearman correlation coeffients were calculated between the exposure measures, since they were expected to not to follow normal distribution. Paired t-tests were used to compare natural logarithms of exposure measured at different shoulders. 3.7.4 Analysis of variance and comparisons of means In examining the influence of different bakeries, products manufactured, use of divider oil and dough forming machinery, simple factorial analysis of variance was used, followed by a comparison of means via Bonferroni's test (if more than 2 factors were examined in the analysis of variance); or Students test (if only 2 factors were examined in the analysis of variance). A comparison of means was carried out only for normally distributed variables. The significance level (a) for comparison of means was 0.05. The equality of variance assumption of the Students test was ascertained via Levene's Test for Equality of Variances, with a 0.05 level of significance. Pooled variance estimates were used when this assumption held and separate variance estimates when it was not met. Since we have no a priori hypothesis as to the directionality of effect of any determinant of exposure, two-tailed tests were performed in all comparisons. The dependent variable used in analysis of variance had to satisfy the assumption of normality, which was achieved by transformations. Analysis of variance models were tested for the assumption of homogeneity of the variance of the dependent variable for each main effect. This was accomplished via Bartlett-Box F test with a significance level of 0.05, therefore the assumption of homoscedasticity was violated when the p-value for the test was less than 0.05. Analysis of variance is robust to assumptions of homoscedasticity only when sample sizes in groups for main effects are approximately equal and the dependent variable is normally distributed (Norusis, 1990). Thus, the Kruskal-Wallis test was considered 22 as an alternative to parametric ANOVA in the case of heteroscedasticity and unequal sample sizes (Norusis, 1990). 3.7.5 Multiple regression The natural logarithm of an exposure measures was used as a dependent variable in regression. Potential determinants of exposure were treated as predictor variables. All combinations of predictor variables were considered in selecting the regression equation with the best fit (i.e. the highest R 2 ^ value). The following random effects model with n determinants of exposure was considered: 1=1 where Y, = exposure in baker j po = logarithmic mean of background exposure Pi = regression coefficient associated with determinant of exposure i Xy = value of determinant of exposure / experienced by baker j Sj = random error in natural logarithm of exposure of baker j. 3.7.6 Building an empirical model of exposure The correlation matrix for all the predictor variables was examined, and where the bivariate Pearson correlation exceeded |0.7| and was statistically significant at a 0.05 level, a decision was made as to which of the correlated variables should be offered for entry into the regression model, based on ease of interpretation. Two-tailed inferential statistics were calculated due to the lack of an a priori hypothesis about the sign of correlation. Predictor variables that were narrowly distributed were excluded from analysis. Univariate linear regressions for each pair of predictor and dependent variable were performed. Only predictor variables which had regression coefficients different from zero at a 0.20 significance level 23 were considered for use in a multiple regression model. Variables that did not meet this criterion were excluded from further consideration. Two regression models were built for each dependent variable. Regression Model 1 was built in three steps. On step I, predictor variables with positive significant regression coefficients and those representing control measures were offered into the model. Variables included in Model 1 upon completion of step I were offered into the model on step II. In addition, remaining variables having significant negative regression coefficients in univariate linear regressions were offered into the model. Variables included in Model 1 upon completion of step II and dummy variables representing bakeries were offered into the model on step HI, in order to see if bakery identity could further explain the residual error. Regression Model 2 was built in two steps. On step A, all predictor variables with significant regression coefficients (both positive and negative) and those representing control measures were offered into the model. On step B, dummy variables representing bakeries were offered into Model 2 to see if they could further explain variability accounted for by variables selected in step A. Semi-partial Type II correlation coefficients were calculated for variables included in the models upon completion of step III and step B, to examine the relative contributions of the predictor variables to explaining total variability in the dependent variable (Howell, 1987). A total of no more than 20 predictor variables were used in the regression model at any step, reducing the probability of a predictor variable being a significant predictor of exposure by chance alone. The resulting regression models were tested for violation of the assumptions of regression analysis. Assumptions of homoscedasity and linearity were tested via graphical methods. The tolerance of each predictor variable in Models 1 and 2 was examined to test the assumption of independence. 24 3.7.7 Regression models examining inhalable dust as a predictor of a-amylase and wheat antigen exposure We constructed two regression models, one for a-amylase and one for wheat antigen exposure as a function of inhalable dust exposure and production characteristics. The natural logarithm of inhalable dust exposure was used as a predictor variable which was forced into each of the models. In addition, dichotomous variables describing production characteristics were offered into the models in a stepwise linear regression procedure. We intentionally excluded all continuous variables from the building of these models, since measuring these variables is a) expensive prospectively, requiring labor-intensive observation of bakery workers during sampling and b) impossible retrospectively. The dichotomous variables had to meet a 0.05 significance level to enter the model and 0.10 to remain in it. The assumptions of the regression analysis were tested via the procedure described in section 3.7.6. The following model with (n+1) determinants of exposure was considered: <=1 where 0; = natural logarithm of aeroallergen exposure in baker j Ij = natural logarithm of inhalable dust exposure in baker j 8 = regression coefficient associated with natural logarithm of inhalable dust exposure Po = mean of function natural logarithm of background aeroallergen exposure pi = regression coefficient associated with determinant of aeroallergen exposure i Xy = value of dichotomous determinant of aeroallergen exposure i experienced by baker j Sj = random error in Qj. 25 3.7.8 Statistical software SAS version 6.1, SPSS version 6.1.2, and Excel version 5.0 were used in data analysis. 26 Chapter 4 Results 4.1 Bakers and bakeries Twenty-two bakeries were found to be eligible to participate in the study. Seven bakeries agreed to enroll in it. Upon combining the findings of the walk-through surveys and the telephone survey, it was estimated that 240 eligible bakery workers could be enrolled. It was discovered in a walk-through survey that one of the recruited bakeries did not qualify as a small bakery. In fact, it was spread over two production plants, respectively employing 130 and 50 employees. Since bakery management had already consented to participate in the study and various production lines inside these large bakery plants resembled small bakeries, the large bakeries were kept in the study. Table 4 compares bakeries that refused to participate with those enrolled in the study. In Table 4, the number of bakeries that declined to participate does not appear as 15 because information on these bakeries as well as small bakeries participating in the study was based exclusively on the results of a telephone survey. One bakery manager declined to answer questions over the phone, but the bakery was found to be eligible to take part in the study and did participate. The bakery that was discovered during a walk-through survey to be two large bakeries is described separately in Table 4. Information about that bakery obtained over the phone is included in the column describing small bakeries that participated in the study. Eighteen sampling trips took place which corresponds to the number of process/product combinations found at the participating bakeries. During that time period, 96 bakery workers were sampled. For logistical reasons, not all eligible bakery workers took part in the study. Namely, the sampling strategy focused on process/product combinations that occurred in bakeries, therefore while several bakery workers were employed in the same production line on different shifts, only one shift of workers was sampled, excluding workers who were involved in the same processes on a different shift. Two bakers refused to participate in the study. For seven of the study participants, only partial shift samples were obtained. Average duration of sampling was 8 hours. 27 Table 4 Comparison of bakeries enrolled in the study with those that were eligible, but declined to participate Data from telephone survey Bakeries that refused Small bakeries Large bakeries to participate participating in the participating in the study study n 13 5a 2 average number of bakers 5.8 6.8 50b 130c product type (% major product or yes/no) bagel 0 20 no yes buns 15 40 no yes cakes 31 20 yes no cookies 8 0 no no pastries 63 40 yes yes pies 38 0 no no rye bread 46 40 no no white bread 44 40 no yes other products 23 40 no yes add dough improvers 15 20 no yes a — information in this column includes the two large bakeries as the one contacted in the telephone survey, and does not include 1 small bakery which participated in the study but refused to answer questions over the phone b — referred to as bakery G c ~ referred to as bakery D 4.2 Description of potential determinants of exposure A description of the variables considered as potential detenninants of flour dust exposure is presented in Table 5 and Table 6 Variables describing environmental conditions inside the bakeries showed symmetrical distributions according to their histograms, with a small deviation from normality according to Q-Q plots. All other variables presented in Table 5 exhibited left-skewed distributions according to their histograms, with deviation from normality indicated by Q-Q plots. Differences between arithmetic means and geometric 28 means for continuous variables support the above conclusions (i.e., A M > GM for non-normal left skewed distributions and A M « GM for normal distributions). Labels and manufacturers' specifications were used to group different brands of flour into enriched, pastry and coarse flours. Coarse and pastry flours did not contain fungal a-amylase, unlike enriched flours. It was suspected that the flours varied in particle size distributions. Gluten- and a-amylase-containing powdered dough improvers were handled in only one bakery (D) in bread and bun production on a single occasion. Thus, due to the small number of bakers who were exposed to dough improvers (one out of 96), the amount of dough improvers used per shift was excluded from analysis. The diversity of the characteristics of the bakeries is illustrated by Table 6. A combination of bread and buns, and cakes and pastries were produced by 3 out of 7 bakeries. The most common type of product at the bakeries was bread and buns (5/7), followed by cakes and pastries (4/7). The majority of bakeries were engaged in the manufacture of one or two types of products, the exception being bakery D, which manufactured 5 different products. Most of the products manufactured at bakery D were not produced by other bakeries in the study (cinnamon buns, croissants, crumpets and puff pastry). According to Table 6, manual dough-forming and the use of a table in forming dough were the most common dough-forming techniques used by the bakeries (4/7). Fully automated dough-forming and the use of reversible sheeters were a close second, being used by 3 bakeries. Dough-brakers were operated at bakery D only, exclusively in dough-brake on the puff-pastry production line. No other machinery was used to perform this the task in the study. Bakeries E and F utilized both manual dough-forming and reversible sheeters in bread and bun production. Bakery D showed the greatest diversity of dough-forming machinery, while bakeries A, B and C used either automated or manual dough-forming. Dough-forming was not performed at bakery G, since dough was not produced there. Instead, in making cakes at bakery G, batter was poured into baking forms. 29 Out of 33 bakers who operated mixers, only 3 used horizontal mixers, while the rest utilized vertical mixers. Horizontal mixers were used in bakery D only on the following production lines: croissants, bagels, bread and buns. Local exhaust systems were not found in any of the participating bakeries. Table 5 Summary of descriptive statistics of all potential determinants of exposure represented by continuous variables. variables n A M SD GM GSD Min Max environmental conditions atmospheric pressure (mmHg) 65 768 5 765 1 760 777 temperature (°C) 71 20.3 2.0 20.3 1 17 25 relative humidity (%) 71 52 8 51 1 38 63 Task-related ma variables (min/shift) ian decorating 96 23 52 0 0 233 mixing 96 24 55 0 0 276 forming 96 48 86 0 0 373 forming with dough-braker 96 12 49 0 0 325 forming without dough-braker 96 36 77 0 0 373 cutting 96 13 31 0 0 170 cleaning 96 20 39 6 0 253 packing/catching 96 80 134 16 0 562 pouring 96 36 54 4 0 207 weighing 96 11 23 0 0 109 tray 96 26 28 19 0 114 oven 96 12 33 0 0 185 rest 96 32 61 16 0 405 quality control 96 38 69 11 0 325 Mass of flour used (kg/baker/shift) Enriched flour 86 178 665 0 0 3534 Pastry flour 85 14 87 0 0 776 Coarse flour 83 27 92 0 0 558 Total flour 90 444 1140 0 0 4868 Baker-related Seniority (years work in bakery) 87 8 9 5 0 45 30 Table 6 Summary of number of bakers for all potential determinants of exposure represented by categorical variables, by bakery. variables number of bakers sampled bakery A bakery B bakery C bakery D bakery E bakery F bakery G All bakeries products manufactured bagels 3 0 0 6 0 0 0 9 bread and buns 0 3 5 4 2 3 0 17 cakes and pastries 0 1 0 0 4 8 12 25 cinnamon buns 0 0 0 6 0 0 0 6 croissants 0 0 0 8 0 0 0 8 crumpets 0 0 0 6 0 0 0 6 pizza or pita bread or tortilla 3 0 0 0 0 0 0 3 puff pastry 0 0 0 14 0 0 0 14 forming equipment dough-brake 0 0 0 6 0 0 0 6 reversible sheeter 0 0 0 1 3 6 0 10 automated 6 0 5 6 0 0 0 17 manual 0 3 0 4 5 6 0 18 other use of divider oil 0 0 5 0 0 0 0 5 left-handed 0 0 0 1 0 1 0 2 Total number of eligible bakers 24 8 5 130 8 15 50 240 Total number of bakers sampled 14 4 5 44 6 11 12 96 % eligible bakers sampled 58 50 100 34 75 73 24 40 4.3 Description of exposure measures A summary of the exposure measures obtained in the course of the study, for the entire study, is presented in Table 8. Examination of the ranges and standard deviations of all three exposure measures revealed considerable variability in the exposures (i.e., a 1000- to 3000- fold difference between the lowest and the highest values). Inhalable dust exposure had a left-skewed distribution according to the histogram (Figure 1) with deviation from normality indicated by Q-Q plots (Figure 2). logarithmic transformation 31 resulted in a substantial reduction in deviation from normality (Figure 3 and Figure 4), even though the detrended normal Q-Q plot still showed a pattern. Wheat antigen exposure had a left-skewed distribution according to the histograms (Figure 5) with deviation from normality indicated by Q-Q plots (Figure 6). The natural logarithm of wheat antigen exposure showed a bimodal distribution (Figure 7) with a substantial deviation from normality (Figure 8). ct-Amylase exposure had a left- skewed distribution according to the histogram (Figure 9) with deviation from normality indicated by Q-Q plots (Figure 10). Logarithmic transformation resulted in a substantial reduction in deviation from normality (Figure 11), even though the detrended normal Q-Q plot still showed a pattern. The highest values of aeroallergen exposure did not occur in conjunction with the use of dough improvers. Dough improvers were used only by one baker. There was a good agreement between a-amylase assay results reported by our laboratory and analyses performed in the laboratory of Dr. Gert Doekes at Wageningen Agricultural University. These analyses were performed on a representaive sample of inhalable dust extracts obtained in our study. Spearman correlation of 0.86 (p = 0.003) between the assays' results appeared to be driven by one point, removal of which did not result in large reduction in the correlation (Spearman correlation coefficient of 0.80 (p = 0.017)). Out of 192 inhalable dust samples collected, 12 were discarded due to problems with seven-hole samplers (e.g., poor fit of filter onto O-ring). Two more samples were excluded from analysis in immonassays due material loss during gravimetric analysis. Furthermore, two more samples were lost during the a-amylase assay and one more in the wheat antigen assay. Because we had two inhalable dust samples collected from each study participant, we were able to obtain at least one measure of inhalable dust, wheat antigen and a-amylase for all study participants. 32 There was not difference between exposure levels measured at the right and left shoulders of study participants. There was no consistent trend in differences between exposures measured at different shoulders. These result are illustrated in Table 7. Table 7 Results of paired t-tests of exposure levels measured at different shoulders. Exposure Measure number of pairs Correlation between natural logarithms of exposure measures between right and left shoulder (P) mean difference between natural logarithms of exposure measures (right - left) p(mean paired difference between natural logarithms of exposure measures is 0) Inhalable dust (mg/m3) 84 0.93 (0.00) -0.09 0.17 a-Amylase (ng/m3) 80 0.91 (0.00) -0.06 0.60 Wheat antigen (Ug/m3) 81 0.95 (0.00) +0.04 0.60 The analytic limit of detection was calculated to be 0.1 mg/m3 for inhalable dust, 0.1 ng/m3 for a-amylase and 1 ug/m3 for wheat antigen. Nine a-amylase personal exposure measures fell below the limit of detection. Nineteen wheat antigen personal exposure measurements were found to be below the limit of detection. All field blanks were below the limits of detection for all three exposure measures. Table 8 Summary of descriptive statistics of exposure measures for all bakeries. Exposure Measure number of bakery workers sampled A M SD GM GSD Min Max Inhalable 96 8.2 18.6 2.1 5.1 0.1 110 dust (mg/m3) a- 96 22.0 50.1 2.8 10.4 O . l 307.1 Amylase (ng/m3) Wheat 96 109 164 21 10 <1 1018 antigen (Ug/m3) 33 Figure 1 Distribution of measured inhalable dust exposures (mg/m3). (Solid line represents expected distribution under assumption of normality.) Histogram 10.0 30.0 50.0 70.0 90.0 110.0 dust available for inhalation per shift 34 Figure 2 Plots investigating assumption of normality of distribution of measured inhalable dust exposures (mg/m3). (Solid lines represent expected distributions under assumption of normality.) Normal Q-Q Plot of dust available for inhalation per shift 3 . _ ; 1 -40 2^0 0 20 40 60 80 100 120 Observed Value Detrended Normal Q-Q Plot of dust available for inhalation per shift 120 Observed Value 35 Figure 3 Distribution of natural logarithm of measured inhalable dust exposures (mg/m3). (Solid line represents expected distribution under assumption of normality.) Histogram 14n -2.00 -1.50 -1.00 -.50 0.00 .50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 ln(dust available for inhalation per shift) 36 Figure 4 Plots investigating assumption of normality of distribution of natural logarithm of measured inhalable dust exposures (mg/m3). (Solid lines represent expected distributions under assumption of normality.) Normal Q-Q Plot of ln(dust available for inhalation per shift) 3 . Observed Value Detrended Normal Q-Q Plot of ln(dust available for inhalation per shift) Observed Value 37 Figure 5 Distribution of measured wheat antigen exposure (ug/m3). (Solid line represents expected distribution under assumption of normality.) Histogram 50 T 100.0 300.0 500.0 700.0 900.0 wheat antigen available for inhalation per shift 38 Figure 6 Plots investigating assumption of normality of measured wheat antigen exposure (ug/m3). (Solid lines represent expected distributions under assumption of normality.) •a & -2 T J 8. (0 E k_ o z E 2 M — Q Normal Q-Q Plot of wheat antigen avaialble for inhalation per shift -400 -200 Observed Value 200 400 600 800 1000 1200 Detrended Normal Q-Q Plot of wheat antigen available for inahalation per shift -1 -200 0 Observed Value 200 400 600 800 1000 1200 39 Figure 7 Distribution of natural logarithms of measured wheat antigen exposure (ug/m3). (Solid line represents expected distribution under assumption of normality.) Histogram 2 0 1 -.50 0.00 .50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 5.00 5.50 6.00 6.50 7.00 ln(wheat antigen available for inahalation per shift) 40 Figure 8 Plots investigating assumption of normality of distribution of natural logarithms of measured wheat antigen exposure (ug/m3). (Solid lines represent expected distributions under assumption of normality.) 75 E E S > <D Q Normal Q-Q Plot of ln(wheat antigen) 15 -1 TJ - 4 - 2 0 2 Observed Value Detrended Normal Q-Q Plot of ln(wheat antigen) 1.0 .8 .6 .4 .2 0.0 -.2 -.4 -.6 -.8 -2 0 Observed Value 10 41 Figure 9 Distribution of measured a-amylase exposures (ng/m3). (Solid line represents expected distribution under assumption of normality.) Histogram 20.0 60.0 100.0 140.0 180.0 220.0 260.0 300.0 alpha-amylase available for inhalation per shift 42 Figure 10 Plots investigating assumption of normality of measured a-amylase exposure (ng/m3). (Solid lines represent expected distributions under assumption of normality.) o aj .2 aj o dl > Q Normal Q-Q Plot of alpha-amylase available for inahalation per shift -100 0 Observed Value 100 200 300 400 Detrended Normal Q-Q Plot of alpha-amylase available for inhalation per shift (0 E i— o 2 E 0 p -1 • • • 0 ° o o 0 a • • • • • • -100 0 Observed Value 100 200 300 400 43 Figure 11 Distribution of natural logarithms of measured a-amylase exposure (ng/m3). (Solid line represents expected distribution under assumption of normality.) Histogram -2.00 -1.00 0.00 1.00 2.00 3.00 4.00 5.00 In (alpha-amylase available for inhalation per shift) 44 Figure 12 Plots investigating assumption of normality of distribution of natural logarithms of measured a-amylase exposure (ng/m3). (Solid lines represent expected distributions under assumption of normality.) 00 E o z E 2 Q Normal Q-Q Plot of In (alpha-amylase) m -1 T J U 8. - 6 - 4 - 2 0 Observed Value Detrended Normal Q-Q Plot of In(alpha-amylase) 1.0 .4 .2 .0 -.2 -.4 -.6 O 0 a \ • -2 Observed Value Each exposure measured was stratified by bakery, product type and type of forming equipment. Table 9, Table 10 and Table 11 present a description of exposure measures for each bakery. There is a 45 considerable amount of variability in exposure levels measured at different bakeries. Bakeries C and G appear to have the lowest exposures, while bakeries D and F — the highest. Table 9 Summary of descriptive statistics of inhalable dust exposure (mg/m3), stratified by bakery. Bakery number of bakery workers sampled A M SD GM GSD Min Max A 14 2.1 2.1 1.3 2.7 0.3 6.7 B 4 5.3 4.4 2.9 5.2 0.3 10.8 C 5 0.6 0.5 0.43 2.1 0.3 1.4 D 44 12.8 21.2 3.9 5.1 0.2 104 E 6 3.7 1.9 2.7 3.1 0.3 5.4 F 11 12.2 32.6 1.9 6.0 0.2 110 G 12 1.1 1.4 0.51 3.5 0.1 4.7 Table 10 Summary of descriptive statistics of a-amylase exposure (ng/m3), stratified by bal Bakery number of bakery workers sampled A M SD GM GSD Min Max A 14 7.5 16.4 0.7 10.3 <0.1 59.2 B 4 6.4 4.9 3.8 4.4 0.4 11.4 C 5 0.6 0.7 0.3 4.4 <0.1 1.7 D 44 40.5 68.4 11.6 5.8 <0.1 307.1 E 6 2.7 5.2 0.9 4.8 <0.1 13.3 F 11 14.8 22.0 4.1 6.6 0.4 68.9 G 12 2.1 6.0 0.3 5.7 <0.1 21.0 Table 11 Summary of descriptive statistics of wheat antigen exposure (ug/m3 ), stratified bj Bakery number of bakery workers sampled A M SD GM GSD Min Max A 14 63 84 12 10 <1 268 B 4 104 83 51 7 3.4 181 C 5 1 1 <1 2 <1 2 D 44 147 189 49 7 <1 1018 E 6 160 141 65 10 <1 390 F 11 126 225 18 12 <1 708 G 12 26 46 4 8 <1 128 46 Table 12, Table 13 and Table 14 present a description of exposure measures for each product type. There is a considerable degree of variability in exposure levels associated with different product types. Bread, bun and puff pastry production seem to give rise to the highest exposures. Wheat antigen and a-amylase exposures associated with croissant production also appear to be high. Table 12 Summary of descriptive statistics of inhalable dust exposure (mg/m3), stratified by product manufactured. Product number of bakery workers8 A M SD GM GSD Min Max bagels 9 2.9 3.3 1.5 3.7 0.2 9.9 bread and 17 18 34 4.5 6.7 0.3 110 buns cakes and 25 1.4 1.6 0.7 3.4 0.1 5.3 pastries cinnamon 6 3.6 3.5 2.2 3.1 0.6 9.1 buns croissants 8 5.3 5.9 2.9 3.2 0.9 15 crumpets 6 2.0 3.2 1.0 3.1 0.3 8.5 pizza or 3 2.2 0.9 2.1 1.6 1.3 3.1 pita bread or tortilla puff pastry 14 23 23 12 3.8 1.8 66 a — it was not possible to assign product type for 8 bakery workers since they were removed from main production area and were packaging several different products Table 13 Summary of descriptive statistics of a-amylase exposure (ng/m3), stratified by product manufactured. Product number of bakery workers sampled3 A M SD GM GSD Min Max bagels 9 31.4 57.7 5.5 10.0 <0.1 176.6 bread and 17 44.0 82.7 6.0 13.6 O . l 307.1 buns cakes and 25 2.6 5.7 0.5 1.7 O . l 21.0 pastries cinnamon 6 12.3 8.8 9.5 2.3 3.2 23.1 buns croissants 8 38.1 68.1 19.2 2.7 9.7 206.5 crumpets 6 1.5 0.8 1.3 1.6 0.8 2.8 pizza or 3 1.5 1.1 1.2 2.3 0.5 2.8 pita bread or tortilla puff pastry 14 44.9 50.4 25.6 3.1 2.9 162.7 a — it was not possible to assign product type for 8 bakery workers since they were removed from main production area and were packaging several different products 47 Table 14 Summary of descriptive statistics of wheat antigen exposure (ug/m3), stratified by product manufactured. Product number of bakery workers sampled8 A M SD GM GSD Min Max bagels 9 83 109 21 8 <1 268 bread and 17 153 191 34 13 <1 706 buns cakes and 25 32 47 6 8 <1 169 pastries cinnamon 6 74 66 32 7 1 168 buns croissants 8 162 149 107 3 33 407 crumpets 6 35 76 7 6 2 191 pizza or 3 67 32 63 2 45 103 pita bread or tortilla puff pastry 14 280 256 173 4 4 1018 a — it was not possible to assign product type for 8 bakery workers since they were removed from main production area and were packaging several different products Table 15, Table 16 and Table 17 present description of the exposure measures for each type of dough-forming equipment. There is a broad range of exposures associated with each equipment type. The automated forming seems to be associated with reduced exposures. Exposures associated with use of dough-brakers have very little variability (as assessed by GSD). Table 15 Summary of descriptive statistics of inhalable dust exposure (mg/m3), stratified by type of forming equipment. Forming number A M SD GM GSD Min Max Equipment of bakery workers sampled8 dough-brake 6 46 14 44 1.4 30 66 reversible 10 15 34 3.8 5.2 0.2 110 sheeter automated 17 4.2 4.4 2.2 3.7 0.3 15 manual 18 13 25 5.3 3.8 0.2 110 a - not all bakery workers performed forming tasks 48 Table 16 Summary of descriptive statistics of a-amylase exposure (ng/m3), stratified by type of forming equipment Forming Equipment number of bakery workers sampled3 A M SD GM GSD Min Max dough-brake 6 74.5 66.4 44.1 3.6 6.0 162.7 reversible 10 14.5 23.2 3.6 6.4 0.4 69.0 sheeter automated 17 24.1 49.6 4.1 10.8 <0.1 206.5 manual 18 16.0 19.6 5.9 5.5 0.4 69.0 a — not all bakery workers performed forming tasks Table 17 Summary of descriptive statistics of wheat antigen exposure (ug/m3), stratified by type of forming equipment Forming Equipment number of bakery workers sampled" A M SD GM GSD Min Max dough-brake 6 345 118 325 1 158 479 reversible 10 164 222 66 6 <1 706 sheeter automated 17 121 127 31 12 <1 408 manual 18 173 175 97 4 <1 706 a — not all bakery workers performed forming tasks 49 4.4 Effect of bakery, type of product manufactured and dough-forming techniques on flour dust exposure Statistics describing the effect of bakery alone on differences in personal flour dust exposure are presented in Table 18. Table 18 Summary of inferential statistics about geometric mean of flour dust exposure measures, stratified by bakery. Bakery number of bakery workers sampled Inhalable dust (mg/m3) a-Amylase (ng/m3) Wheat antigen (ug/m3) GM (95%CI) GM (95%CI) G M (95%CI) A 14 1.3 0.7° 12 (0.8 - 2.4) (0.2 - 2.7) (3-45) B 4 2.9 3.8 51 (0.2 - 39.6) (0.4 - 40.9) (3 - 997) C 5 0.4a 0.3" <la c (0.2-1.1) ( O . l - 1.7) (<l-2) D 44 3.9ab n 6 a b e d 4 9 b c (2.4 - 6.4) (6.8 -19.8) (28 - 88) E 6 2.72 0.9d 65a (0.8 - 9.0) (0.2 - 4.4) (6 - 695) F 11 1.9 4.1' 18 (0.6 - 6.3) (1.1 -14.5) (3 - 96) G 12 0.5b 0.3* 4b (0.2 -1.2) ( O . l -0.8) (<1 - 13) ANOVA models' parameters pofBartlett-BoxF 0.26 0.86 0.21 F-statistic (p) • 4.38 (0.00) 10.87 (0.00) 5.43 (0.00) R 2 adi 0.18 0.38 0.22 a, b, c, d, e — pairs of statistically different means according to Bonferroni's test with significance level 0.05 50 Statistics describing the effect of type of product manufactured alone on differences in personal flour dust exposure are presented in Table 19. Table 19 Summary of inferential statistics about geometric mean of flour dust exposure measures, stratified by product manufactured. Product number Inhalable dust a-Amylase Wheat antigen of (mg/m3) (ng/m3) (ug/m3) bakery workers sampled GM (95%CI) GM (95%CI) GM (95%CI) puff pastry 14 11.7abc 25.6 173ab (5.4 - 25.2) (13.2-49.6) (81 - 366) bread and 17 4.50d 6.0 34 buns (1.7-11.8) (1.6-23.1) (9 - 124) croissants 8 2.9 19.2 107° (1.1-7.7) (8.4 - 43.9) (47 - 246) cinnamon 6 2.21 9.5 32 buns (0.7 - 7.2) (4.0 - 22.7) (4 - 236) pizza or pita 3 2.1 1.2 63 bread or (0.7 - 6.2) (0.2 - 9.7) (21 -187) tortilla bagels 9 1.5C 5.5 21 (0.6-4.1) (0.9-31.9) (4 -105) crumpets 6 1.0b 1.3 7b (0.3 - 3.2) (0.8 - 2.3) (1-42) cakes and 25 0.7ad 0.5 6ac pastries (0.4 - 1,2) (0.3-1.1) (2-14) ANOVA models' parameters pofBartlett-BoxF 0.33 0.00" 0.052 F-statistic or %2 (p) 6.28 (0.00) 37.0" (0.00) 5.04 (0.00) R adi 0.31 n/ae 0.25 a, b, c, d — pairs of statistically different means according to Bonferroni's test with significance level 0.05 e — assumptions of ANOVA were not met and we resorted to non-parametric equivalent of ANOVA (Kruskal-Wallis test), hence %2 was calculated and R 2 ^ was not applicable Statistics describing the effect of dough forming methods on flour dust exposure are presented in Table 2 0 . Since both reversible sheeter and manual forming were used by some bakers (see Table 6 ) , and only one baker used nothing but reversible sheeter in forming dough, the reversible sheeter and manual dough-forming groups were collapsed into one category. This produced three independent variables that could be entered into an analysis of variance. 51 Table 20 Summary of inferential statistics about geometric mean of flour dust exposure measures, stratified by type of forming equipment. Forming number Inhalable dust (mg/m3) a-Amylase (ng/m3) Wheat antigen (ug/m3) Equipment of bakery workers sampled GM (95%CI) GM (95%CI) GM (95%CI) dough-brake 6 44.2 44. l a 325 (32.1-61.0) (11.5-168.7) (213 -494) automated 17 2.23 4.1a 31 (1.1-4.4) (1.2-13.8) (9 - 108) reversible 19 5.3 5.9 96 sheeter (2.8 - 10.0) (2.7-13.1) (49 - 187) and/or manual ANOVA models' parameters pofBartlett-BoxF 0.025b 0.17 0.001" F-statistic or %2 (p) 16.5b (0.00) 3.40 (0.00) 9.27b (0.00) R adi n/ab 0.11 n/ab a - pair of statistically different means according to Bonferroni's test with significance level 0.05 b — assumptions of ANOVA were not met and we resorted to non-parametric equivalent of ANOVA (Kruskal-Wallis test), hence yl was calculated and R 2 ^ was not applicable 4.5 Effect of use of divider oil on personal flour dust exposures A method of preventing dough adhesion to surfaces that employs divider oil rather than dusting of flour was used during bread and bun production only in bakery C. Therefore the effect of divider oil use on the exposure was examined separately for bakery workers who participated in bread and bun production (Table 21). There were statistically significant differences between the exposure levels (p = 0.00 in cases of inhalable dust and wheat antigen and p = 0.01 for a-amylase). Table 21 Effect of use of divider oil in bread and bun production on exposure in bakery workers8. Method of preventing adhesion n Inhalable Dust (mg/m3) a-Amylase (ng/m3) Wheat antigen (ug/m3) Divider oil 5 0.43(0.17-1.11) 0.3 ( O . l - 1.7) l(<l-2) Dusting 12 11.95 (5.70-25.03) 22.1 (7.5-64.9) 153 (89 - 265) Levene's Test for Equality of Variances (p) 0.56 0.23 0.98 a — exposure is measured as geometric mean (95% confidence interval). 52 4.6 Effect of mixer type on personal flour dust exposures Due to low numbers of bakers who used horizontal mixers in our study we conducted only limited analysis of their effect on the exposure. The analysis was restricted to bakery workers who actually used mixers (Table 22). There were statistically significant differences between the exposure levels (p = 0.00 in cases of inhalable dust and a-amylase and p = 0.04 for wheat antigen). Dummy variables representing the type of mixer used were offered into the final forms of the regression models described below, but failed to improve their fit, except in the model described in section 4.8.3 on page 68. Table 22 Effect of different mixer designs on exposure in bakery workers9 who performed mixing tasks. Type of mixer n Inhalable Dust (mg/m3) a-Amylase (ng/m3) Wheat antigen (ug/m3) Horizontal 3 13.00 (7.23 - 23.37) 192.0 (157.9 - 233.6) 231 (50- 1076) Vertical 30 3.79 (2.29 - 6.26) 3.8(1.7-8.1) 67(31 - 145) Levene's Test for Equality of Variances 0.167 0.00 0.21 a — exposure is measured as geometric mean (95% confidence interval). 53 4.5 Regression models of inhalable dust exposure 4.5.1 Building regression Models 1 and 2 of inhalable dust exposure STEP I. In step I of building regression Model 1, a regression equation explaining 73% of the variability in inhalable dust exposure was obtained. Parameters of the resultant regression model are presented in Table 23. Table 23 Parameters of regression Models 1 of inhalable dust exposure at step I of variable selection process. SE* Tolerance semi-partial R 2 (%) factors increasing exposure forming with dough-braker (min) 0.010 0.002 0.52 4.5 forming without dough-braker (min) 0.005 0.001 0.83 4.2 forming with reversible sheeter (yes/no) 0.954 0.470 0.36 1.2 cutting dough (min) 0.006 0.003 0.66 1.1 pour (min) 0.014 0.003 0.32 6.7 weigh (min) 0.010 0.005 0.52 1.1 enriched and coarse flour (kg) 0.000ft 0.000 0.22 0.4 bread & bun (yes/no) 2.005 0.398 0.32 7.3 puff pastry (yes/no) 0.632 0.353 0.48 0.9 factors decreasing exposure divider oil (yes/no) -2.622 0.550 0.50 6.5 mixing (min) -0.005 0.003 0.22 0.7 manual forming (yes/no) -0.604ft 0.487 0.21 0.4 background exposure (constant) -0.514 0.131 t regression coefficient J standard error of regression coefficient f t - not different from zero (a > 0.15) STEP n and A. In step II of building regression Model 1, a regression equation explaining 79% of the variability in inhalable dust exposure was obtained. This constituted an increase in Model 1 fit from step I, as measured by R2adj. Step A of Model 2 construction resulted in the identical equation to step II in building of Model 1. STEP III and B. In steps III and B, identical regression equations were obtained for Models 1 and 2, since they converged on the previous step. No increase in fit was observed compared to results of steps II and A, as 54 measured by R2adj. Thus the model obtained on step 11(A) was adopted as a final form of regression for inhalable dust exposure. Parameters of the model are shown in Table 24. Table 24 Parameters of inhalable dust exposure model in final form. SE* Tolerance semi-partial R 2 (%) factors increasing exposure forming with dough-braker (min) 0.009 0.002 0.64 4.6 forming without dough-braker (min) 0.002 0.001 0.70 0.4 forming with reversible sheeter (yes/no) 0.869 0.277 0.80 2.2 pour (min) 0.011 0.002 0.37 5.3 weigh (min) 0.012 0.004 0.83 2.5 bread & bun (yes/no) 1.057 0.258 0.59 3.7 factors decreasing exposure divider oil (yes/no) -2.395 0.410 0.68 7.5 mixing (min) -0.006 0.002 0.48 1.7 cake & pastries (yes/no) -1.125 0.286 0.36 3.4 decorating (min) -0.004 0.002 0.63 0.8 pack/catch (min) -0.003 0.001 0.53 4.3 background exposure (constant) 0.639 0.219 t regression coefficient J standard error of regression coefficient 4.5.2 Verification of regression analysis assumptions in the model of inhalable dust exposure The following tests were used to examine the model of inhalable dust exposure for violation of the assumptions of regression analysis. The assumptions of normality (Figure 13) and homogeneity (Figure 14) of variance appear to be satisfied by the model. Figure 14 upholds the assumption of linearity of the relationship described by the model. Low tolerance values for some predictor variables indicate that some correlation exists among the predictor variables, albeit one not significant enough to violate the assumptions of regression analysis. 55 Figure 13 Plots investigating assumption of normality of residuals in model of inhalable dust exposure. (Solid lines represent expected distributions under assumption of normality.) D^ml«\*h l^r<clriaa^ fCTrh*a=riper*ifg -225 -1.75 - t Z -.75 -25 X .75 125 175 225 -200 -190 - u s -si am sa uoo taD 200 293 lipencfet^redelr(driaal*fCTirhrtoperstif9 QCD 25 a .75 too CteavedQjrmjEt^RtfcEklity Figure 14 Plot investigating assumption of linearity and homogeneity of variance in the model of inhalable dust exposure. (Solid line represents best linear fit.) Scatterplot Dependent Variable: ln(dust available for inhalation per shift) -2 -1 0 Regression Standardized Predicted Value 56 4.6 Regression models of wheat antigen exposure 4.6.1 Building regression Models 1 and 2 of wheat antigen exposure STEP I. In step I of building regression Model 1, a regression equation explaining 70% of the variability in wheat antigen exposure was obtained. Parameters of the resultant regression model are presented in Table 25. Table 25 Parameters of regression Models 1 of wheat antigen exposure at step I of variable selection process. B* SE* Tolerance semi-partial R 2 (%) factors increasing exposure forming with dough-braker (min) 0.006 0.004 0.53 1.0 forming without dough-braker (min) 0.004 0.002 0.71 1.3 forming with reversible sheeter (yes/no) 1.556 0.467 0.83 3.6 pour (min) 0.011 0.004 0.38 2.6 weigh (min) 0.018 0.006 0.80 2.5 enriched and coarse flour (Teg) 0.000 0.000 0.46 2.3 bread and buns (yes/no) 2.182 0.456 0.56 7.4 puff pastry (yes/no) 1.847 0.536 0.47 3.8 croissants (yes/no) 2.517 0.537 0.77 7.0 factors decreasing exposure divider oil (yes/no) background exposure (constant) -3.688 1.191 0.709 0.200 0.68 8.7 f regression coefficient J standard error of regression coefficient STEP II. As a result of step II of building regression Model 1, a regression equation explaining 74% of the variability in wheat antigen exposure was obtained. This constitutes an increase in Model 1 fit from step I, as measured by R 2^. Parameters of the resultant regression model are presented in Table 26. 57 Table 26 Parameters of regression Models 1 of wheat antigen exposure at step n of variable selection process. B* SE* Tolerance semi-partial R 2 (%) factors increasing exposure forming with dough-braker (min) 0.005 0.003 0.55 0.7 forming with reversible sheeter (yes/no) 1.857 0.445 0.78 4.8 pour (min) 0.007 0.004 0.38 1.0 weigh (min) 0.018 0.006 0.78 2.5 enriched and coarse flour (kg) 0.000ft 0.000 0.39 0.4 bread and buns (yes/no) 1.405 0.493 0.41 2.2 puff pastry (yes/no) 1.059 0.546 0.39 1.0 croissants (yes/no) 1.906 0.524 0.69 3.6 factors decreasing exposure divider oil (yes/no) -3.957 0.649 0.70 10.2 cakes and pastries (yes/no) -1.273 0.507 0.29 1.7 decorating (min) -0.006 0.003 0.63 1.0 packing/catching (min) -0.005 0.001 0.47 3.8 background exposure (constant) 2.746 0.401 f regression coefficient % standard error of regression coefficient f t - not different from zero (a > 0.15) STEP A. In step A of building regression Model 2, a regression equation explaining 74% of the variability in wheat antigen exposure was obtained. Parameters of the resultant regression model are presented in Table 27. 58 Table 27 Parameters of regression Models 2 of wheat antigen exposure at step A of variable selection process. B* SE* Tolerance semi-partial R 2 (%) factors increasing exposure forming with dough-braker (min) 0.005 0.003 0.55 0.6 forming with reversible sheeter (yes/no) 1.305 0.649 0.37 1.1 pour (min) 0.008 0.004 0.34 1.3 weigh (min) 0.014 0.007 0.59 1.1 enriched and coarse flour (kg) 0.000ft 0.000 0.39 0.4 bread and buns (yes/no) 0.971 0.616 0.26 0.7 puff pastry (yes/no) 0.938 0.555 0.38 0.8 croissants (yes/no) 1.935 0.524 0.69 3.7 manual forming (yes/no) 0.752ft 0.642 0.23 0.4 factors decreasing exposure divider oil (yes/no) -3.504 0.754 0.51 5.9 cakes and pastries (yes/no) -1.269 0.506 0.29 1.7 decorating (min) -0.006 0.003 0.63 1.1 packing/catching (min) -0.005 0.001 0.47 3.7 background exposure (constant) 2.746 0.401 f regression coefficient J standard error of regression coefficient f t - not different from zero (a > 0.15) STEP III and B. Variables representing bakeries were offered in all possible combinations with variables forming Model 1 on step II and Model 2 on step A. No substantial increase in model fit of was observed compared to the models on step II and A, as measured by R2adj (0.75 and 0.76 respectively). Thus, equations obtained on steps II and A were considered to be final forms of the models. The results obtained in building Model 2 of wheat antigen dust exposure were not identical to those of Model 1. Even though the two variable selection processes did not converge on an identical equation, the obtained equations had identical fit and differed by only one variable, which had a coefficient not significantly different from zero (manual forming). Thus, from now on, the two models will be discussed simultaneously, assuming that they provide the same information about the identity and relative importance of determinants of wheat antigen exposure. 59 4.6.2 Verification of regression analysis assumptions in the final forms of model 1 and 2 of wheat antigen exposure The following tests examined the models of wheat antigen exposure for violation of the assumptions of regression analysis. The assumptions of normality (Figure 15 and Figure 17) and homogeneity (Figure 16 and Figure 18) of variance appear to be satisfied by the models. Figure 16 and Figure 18 uphold assumption of linearity of relationship described by the models. Low tolerance values for some predictor variables indicate that some correlation exists among the predictor variables, albeit one not significant enough to violate the assumption of regression analysis. One outlier was found in both models, corresponding to baker on bread and bun production line in bakery D. There were no special circumstances observed in this baker's tasks, products manufactured or raw materials used to explain this result. Figure IS Plots investigating assumption of normality of residuals in model 1 of wheat antigen exposure. (Solid lines represent expected distributions under assumption of normality.) Hstojsm Nbrrrsl P-PRotof RegessOTSariiardzBd Residual -175 -75 25 125 225 3-25 0.00 25 .50 .75 1.00 RagEsacngtjdadaitJRBadtri ObGetved Currrrulative Probability 60 Figure 16 Plot investigating assumption of linearity and homogeneity of variance in the model 1 of wheat antigen exposure. (Solid line represents best linear fit.) Scatterplot Dependent Variable: In(ug/m3 wheat antigen) outlier 4 3 is 2 •g OJ or 1 " D a 1 o OJ T3 co - H CZ o 'co 2> 2 o> or - 3 •( a • a E 3 • • a - = H • a a DO • a • OP ° ° o • % D O D a • ° • ° ° D O 0 c a ™ a L = to a a a • • a a a CP" o H a a ° o 8 • u a a a • a a o • • a o -2 -1 0 Regression Standardized Predicted Value Figure 17 Plots investigating assumption of normality of residuals in model 2 of wheat antigen exposure. (Solid lines represent expected distributions under assumption of normality.) Hstccpam Isfarrrd PPRctof F^geBOTSbrtfarr^rtecW [iperrJsrt\^ n±te tr<L0h8vtGEta1)gsr^  1<D1 7| a m .25 D .75 CfceasedCUrmjaweRttatfiy 61 Figure 18 Plot investigating assumption of linearity and homogeneity of variance in the model 1 of wheat antigen exposure. (Solid line represents best linear fit.) Scatterplot Dependent Variable: In(ug/m3 wheat antigen) 41 outlier o: - 3 J _ — . . . J - 2 - 1 0 1 2 Regression Standardized Predicted Value 62 4.6.3 Inhalable dust as a predictor of wheat antigen exposure A correlation between natural logarithms of inhalable dust and wheat antigen exposure is illustrated in Figure 19. Spearman correlation coefficient between inhalable dust and wheat antigen levels was calculated to be 0.89 (p = 0.000). Figure 19 Linear correlation between natural logarithms of inhalable dust (mg/m3) and wheat antigen exposure (ug/m3). Rsq = 0.7336 .1 .3 .5 2 4 10 30 50 200 .2 .4 1 3 5 20 40 100 dust available for inhalation per shift Regression model that accounts for 79% of variability and satisfies the assumptions of regression analysis (plots not shown) was obtained. The model is presented in Table 28. No outliers were found in the model. Table 28 Parameters of regression model of wheat antigen exposure as a function of inhalable dust exposure and production characteristics. B* SE* p(B*0) ln(inhalable dust) 1.080 0.073 0.000 croissant (yes/no) 1.192 0.404 0.004 automated forming 1.049 0.348 0.003 (yes/no) manual forming (yes/no) 0.827 0.299 0.007 divider oil (yes/no) -2.191 0.594 0.000 constant 1.948 0.142 0.000 t regression coefficient J standard error of regression coefficient 63 4.7 Regression models of a-amylase exposure 4.7.1 Building regression Models 1 and 2 of a-amylase exposure STEP I. In step I of building regression Model 1, a regression equation explaining 62% of the variability in inhalable dust exposure was obtained. The parameters of the resultant regression model are presented in Table 29. Table 29 Parameters of regression Model 1 of a-amylase exposure at step I of variable selection process. B* SE* Tolerance semi-partial R 2 (%) factors increasing exposure 0.006tf forming with dough-braker (min) 0.004 0.52 0.7 forming without dough-braker (min) 0.003n 0.002 0.69 0.7 pour (min) 0.008 0.005 0.37 1.1 weigh (min) 0.014 0.007 0.80 1.5 enriched and coarse flour (kg) 0.001 0.000 0.23 2.6 bread and buns (yes/no) 2.900 0.530 0.53 11.9 puff pastry (yes/no) 2.392 0.604 0.48 . 6.2 croissants (yes/no) 2.982 0.614 0.75 9.4 cinnamon buns (yes/no) 2.278 0.656 0.86 4.8 factors decreasing exposure divider oil (yes/no) -3.555 0.802 0.68 7.8 mixing (min) -0.007tt 0.006 0.22 0.6 background exposure (constant) -0.790 0.224 f regression coefficient | standard error of regression coefficient tt - not different from zero (a > 0.15) STEP II and A. Step II of building Model 1 resulted in a regression equation explaining 64% of the variability in a-amylase exposure. This constituted an increase in Model 1 fit from step I, as measured by 11%. Step A of Model 2 construction resulted in the identical equation to step II in building of Model 1. The parameters of the regression equation obtained at this stage are illustrated in Table 30. 64 Table 30 Parameters of regression Model 1(2) of a-amylase exposure at step 11(A) of variable selection process. SE 1 Tolerance semi-partial R 2 (%) factors increasing exposure 0.005tf forming with dough-braker (min) 0.004 0.5 0.5 pour (min) 0.005n 0.004 0.37 0.5 weigh (min) 0.012 0.007 0.78 1.1 enriched and coarse flour (kg) 0.001 0.000 0.23 2.1 bread and buns (yes/no) 2.841 0.493 0.58 12.5 puff pastry (yes/no) 2.162 0.587 0.48 5.1 croissants (yes/no) 2.864 0.565 0.84 9.7 cinnamon buns (yes/no) 2.439 0.630 0.88 5.6 factors decreasing exposure divider oil (yes/no) -3.978 0.774 0.69 9.9 mixing (min) -0.007tf 0.006 0.22 0.6 packing/catching (min) -0.003 0.001 0.67 1.7 decorating (min) -0.005tf 0.003 0.76 0.8 cleaning (min) -0.009 0.004 0.87 1.8 background exposure (constant) 0.057ft 0.360 t regression coefficient t standard error of regression coefficient tt - not different from zero (a > 0.15) STEP III and B. In steps III and B, identical regression equations were obtained for Models 1 and 2, since they converged on the previous step. A regression equation, explaining 74% of variability in a-amylase exposure was obtained. This constitutes a noticeable increase in fit as compared to the results of steps II and A, as measured by R 2adj. Thus, the regression model obtained at this stage was adopted as the final form of the a-amylase exposure model. It was represented by a single equation, whose parameters are shown in Table 31. 65 Table 31 Parameters of a-amylase exposure model in final form. B* SE* Tolerance semi-partial R 2 (%) factors increasing exposure forming with dough-braker (min) 0.005 0.003 0.53 0.6 pour (min) 0.004tt 0.004 0.37 0.4 weigh (min) 0.012 0.006 0.78 1.1 enriched and coarse flour (kg) 0.001 0.000 0.22 2.6 bread and buns (yes/no) 2.580 0.426 0.57 10.1 puff pastry (yes/no) 1.226 0.569 0.37 1.3 croissants (yes/no) 1.933 0.552 0.65 3.4 cinnamon buns (yes/no) 1.396 0.604 0.70 1.5 bakery D 1.826 0.400 0.38 5.8 bakery G 1.834 0.436 0.78 4.9 factors decreasing exposure divider oil (yes/no) -2.780 0.698 0.63 4.3 mixing (min) -0.005ft 0.005 0.22 0.4 packing/catching (min) -0.003 0.001 0.65 1.6 decorating (min) -0.003n 0.003 0.71 0.4 cleaning (min) -0.004ft 0.004 0.81 0.3 background exposure (constant) 0.057ft 0.360 t regression coefficient t standard error of regression coefficient "ft - not different from zero (a > 0.15) 4.7.2 Verification of regression analysis assumptions in the model of a-amylase exposure The following tests were used to examine the model of a-amylase exposure for violation of the assumptions of regression analysis. The assumptions of normality (Figure 20) and homogeneity (Figure 21) of variance appear to be satisfied by the model. Figure 21 upholds the assumption of linearity of the relationship described by the model. Low tolerance values for some predictor variables indicate that some correlation exists among the predictor variables, albeit one not significant enough to violate the assumption of regression analysis. One outlier was found in the model. It originated from cake and pastry production in bakery G. There were no special circumstances observed in this baker's tasks, products manufactured or raw materials used to explain this result. 66 Figure 20 Plots investigating assumption of normality of residuals in model of a-amylase exposure. (Solid lines represent expected distributions under assumption of normality.) I fcjLyrfn CEpaid^ \Qiabislr(r0rr8a^ Barr}iass) LCD 200 150 233 tarrd PPRctcf r^ gres^SbrdardzedRssild 000 25 50 .75 1.00 CfceervedCumUEfteRdiebBy Figure 21 Plot investigating assumption of linearity and homogeneity of variance in the model of a-amylase exposure. (Solid line represents best linear fit.) Scatterplot Dependent Variable: In (ng/m3 alpha-amylase) outlier 4 ' V ) Si R S 2 . R B ° o • o " O D D • " a 1 ' o „ o ° • a • •• • °. r -• - . s - B 0 • ° D D D • V B . 0 B n 0 - 1 ' ° Oo° • ° o n° o 0 °" 1 O 0 DB D O O 0 -2 ' a • • -3 -2 -1 0 Regression Standardized Predicted Value 67 4.7.3 Inhalable dust as a predictor of a-amylase exposure A correlation between natural logarithms of inhalable dust and a-amylase exposure is illustrated in Figure 22. Spearman correlation coefficient between inhalable dust and a-amylase levels was calculated to be 0.78 (p = 0.000). Figure 22 Linear correaltion between natural logarithms of inhalable dust (mg/m3) and a-amylase exposure (ng/m3). Rsq = 0.6028 dust available for inhalation per shift 68 Regression model that accounts for 74% of the variability and satisfies the assumptions of regression analysis (plots not shown) was obtained. The model is presented in Table 33. No outliers were found in the model. Table 32 Parameters of regression model of wheat antigen exposure as a function of inhalable dust exposure and production characteristics. B* SE* P(B* 0 ) ln(inhalable dust) 0.726 0.110 0.000 croissant (yes/no) 2.650 0.496 0.000 bagel (yes/no) 1.715 0.449 0.000 bread and bun (yes/no) 1.668 0.486 b.ooi cinnamon bun (yes/no) 2.212 0.533 0.000 puff pastry (yes/no) 1.997 0.470 0.000 horizontal mixer (yes/no) 1.978 0.746 0.010 divider oil (yes/no) -1.842 0.730 0.013 constant -0.536 0.185 0.005 t regression coefficient X standard error of regression coefficient 69 Chapter 5 Discussion 5.1 Inhalable dust exposure levels This is the first report of inhalable dust exposure levels experienced by Canadian bakery workers. If it is assumed that an inhalable dust fraction is similar to total dust, then it can be suggested that the regulatory limit of 10 mg/m3 time weighted average for eight hours (B.C. Workers' Compensation Board, 1980 and ACGIH TLV, 1994) is exceeded, on average, by a factor of 2 among bakers involved in bread/bun production and puff-pastry production. Individual exposures exceeding the exposure limit were also observed during croissant manufacture. Some inhalable dust exposures obtained during the manufacture of other products come close to the exposure limit, indicating that some tasks in all manufacturing processes present a threat of high exposures. All forming tasks were associated with some exceedences of the regulatory limit by some bakers. Only automated dough-forming resulted in an average exposure below 10 mg/m3. For the entire study, the proportion of bakery workers who were overexposed is higher than that reported by other investigators (Table 33). This emphasizes the need for controlling dust exposures in B.C. bakeries. Inhalable dust exposures observed in this study are comparable to findings of other researchers, cited in Table 33. However, our study found somewhat higher average exposure levels than the majority of other investigations. 70 Table 33 Comparison of inhalable dust levels (in mg/m3) found in this study with those observed by other researchers. Study A M (SD) GM (GSD) Range % > 10 mg/m3 Masalin et al. (1988) NR. NR 0.1-48.8 5 Musk et al. (1989) NR 0.1 to 11.0 0.0-37.6 11 Nieuwenhuijsen et al. (1992) 1.1 to 3.2 0.6 to 1.8 0.1-28.5 NR Jauhiainen et al. (1993) 2.3 (0.9) to 4.6 NR 0.9-14.7 NR (3.6) Bohadana et al. (1994) 0.7 (0.2) to NR 0.5-98.1 11 41.3 (39.5) Kolopp-Sarda et al. (1994) 4.9(9.1) NR NR NR Burdorf etal. (1994) 3.8 2.5 (2.8) 0.01-16.9 NR Nieuwenhuijsen et al. (1994a) 0.5 to 9.0 0.4 (1.7) to 0.13-86.02 6.5 6.4 (1.8) Nieuwenhuijsen et al. (1995b) NR 1.5 (1.6) 0.03-33.7 NR Houba (1996a) 2.0 1.0 0.1-37.7 NR Burstyn (1996) 8.2 (18.6) 2.1 (5.1) 0.1 -110 17 NR - not reported 5.2 Wheat antigen and a-amylase exposure levels In this section, we compare wheat antigen and a-amylase levels found in our study to those reported by other researchers. To make such a comparison possible, we will only consider results of investigations in which wheat antigen or a-amylase were measured in immunoassays. Levels of wheat antigen found in our study were similar to those reported by Nieuwenhuijsen et al.(1994a, 1995b) and higher than those found by Houba et al. (1996a). There are no regulatory limits of exposure applicable to wheat antigen and a-amylase. Nevertheless, since flour aeroallergens in our study were assayed using antibodies developed by Houba (1996a), we can compare exposures observed by us to the levels at which Houba (1996a) saw health effects among bakery workers. He reported a dose-response relationship between flour aeroallergen levels and respiratory symptoms at flour aeroallergen exposure levels that were lower than those found in our 71 study. This emphasizes the need to reduce the exposures in B.C. bakeries, especially since Houba (1996a) found lower flour aeroallergen exposures than we did. Table 34 Comparison of wheat antigen levels (in ug/m3) and a-amylase levels (in ng/m3) found in this study with those observed by other researchers. Study Wheat antigen a-Amylase A M (SD) GM (GSD) Range A M (SD) GM (GSD) Range Nieuwenhuijsen et al. (1994a) 57 to 387* 46 (2) to 229 (2) 8-1913 not measured Nieuwenhuijsen et al. (1995b) NR 170 46 - 1899 not measured Houba (1996a) 5.4 0.7 (7) 0.03-252 3.3 0.3 (3.1) <0.3 -222 Burstyn (1996) 109 (164) 21 (10) <1 -1018 22 (50) 2.8 (10.4) <0.1 -307 NR — not reported * ~ arithmetic means were reported separately for different task groups 5.3 Determinants of exposure and recommendations 5.3.1 Production lines defined by products manufactured The following information has been derived from the ANOVA results. Bakers manufacturing puff-pastry were exposed to higher concentrations of inhalable dust than those involved in cake/pastry, bagel and crumpet production. Inhalable dust exposure during bread/bun manufacturing was higher than that in cake/pastry production. Wheat antigen exposures during puff-pastry production exceeded those arising in crumpet and cake/pastry production. Croissant production was associated with higher wheat antigen exposure than the manufacture of cakes and pastries. Puff-pastry, cinnamon bun and croissant production lines were associated with high a-amylase exposure, and cake/pastry and crumpet production with low exposure. These observations are supported by the results of the multiple regression analysis. Two product types, bread and bun increasing and cake and pastry decreasing exposure, made significant contributions (when adjusted for the effect of other variables in the model) to explaining the total variability in the 72 inhalable dust exposure: According to the signs and magnitudes of the regression coefficients, as well as the variables' individual contributions to explaining the total variability in the multiple regression model, croissant production had the strongest association with increased wheat antigen exposures, followed by breadVbun production and puff-pastry production. Judging by the same criteria, bread and bun, croissant, puff pastry and cinnamon bun production were the most important factors contributing to increases in a-amylase exposure. Let us now consider all three exposure measures in prioritizing production lines for implementation of exposure control measures. The puff-pastry production line should be given priority in this respect, since it was associated with the highest exposure levels for all three agents. If we apply the same reasoning, the bread and bun production lines should be next on the priority list if achieving compliance is the primary objective. Nevertheless, exposures to specific allergens were found to be low on that type of production line, compared to those described for croissant production. Therefore, croissant production, lines should be considered second priority for the implementation of control measures aimed at protecting workers' health. Cake/pastry and crumpet production lines should be the last to receive attention in this respect, as they appear to be associated with low exposures for all three agents. 5.3.2 Dusting vs. use of divider oil An approximately 28-fold decrease in inhalable dust exposures that occurred with the use of divider oil compared to dusting. Similarly, wheat antigen exposure observed during the use of divider oil was 150 times lower than that associated with dusting. The effect of divider oil use on decreases in a-amylase exposure was less dramatic (6-fold), but was still statistically significant. Thus, we can reasonably suggest that substituting divider oil for dusting in bread and bun making is an effective way to reduce flour dust exposure in bread and bun production. This finding is especially significant since regulatory limits of exposure were exceeded in bread and bun production when divider oil was not used. The importance of the use of divider oil in decreasing flour dust exposure is supported by its inclusion in multiple regression models, after adjustment for all other determinants of exposure. This is the first report 73 in occupational hygiene literature of substituting dusting with the use of the divider oil in bakeries. Potential health risks of the divider oil use, and the substitution's impact on product quality remains to be investigated. 5.3.3 Dough-forming and mixing equipment Dough forming techniques have an important bearing on exposure to inhalable dust, a-amylase and wheat antigen. Even though the statistical techniques employed do not allow us to rank all types of dough forming equipment for all exposure measures in a single comparison, the following ranking of the equipment in terms of contribution to levels of flour dust exposure appears to be justified (increasing from left to right): automated -» reversible sheeter -> manual -> dough-braker. The conclusions arising from the results of the ANOVA presented above are strengthened by the results of the multiple regression analysis. According to the signs and magnitudes of the regression coefficients in the multiple regression, the duration of forming and type of forming equipment were associated with increased inhalable dust exposures. Time spent forming with a dough-braker was one of the most important factors increasing wheat antigen and a-amylase exposure (when adjusted for the effect of other variables in the model) in explaining total variability in exposure. A strong association of reversible sheeter use with increased inhalable dust and wheat antigen exposure can be accounted for by suggesting that rapid changes in the direction of movement of the machine's belt-like surface can result in emission of flour particles. Duration of dough-braker use was associated with increased flour dust exposure, probably because the machine throws flour into the operator's face. Thus, exposure due to dough-braker use occurred in a series of peaks and its time-weighted average can be expected to increase the longer the machine is used. It appears reasonable to suggest that automation of production lines in bakeries will reduce flour dust exposure attributable to forming tasks. Automation reduces exposure by isolating the operator from the process. However, in automated dough-forming, high exposures were observed, indicating that 74 automation alone will not ininimize the exposures. If complete automation in puff-pastry production is not feasible, the use of dough-brakers should be avoided, especially since a reversible sheeter can be used instead, resulting in an approximately 20-fold decrease in inhalable dust exposure, a 12-fold decrease in a-amylase exposure and a 5-fold decrease in wheat antigen exposure. Another argument for eliminating the use of dough-brakers is that they present a safety hazard (Matz, 1972). The limitations of substituting dough-brakers with reversible sheeters are that the latter occupy more space, are slower (Matz, 1972) and are still associated with high exposures. Dough-braker operators observed in this study wore dust masks and should continue to do so until their dust and flour aeroallergens exposures are reduced. Manual forming might be difficult to avoid in small bakeries due to the diversity of tasks and products, calling for the use of personal protective equipment (such as dust masks) and minimizing the use of dusting in handling dough or substituting dusting with divider oil. Horizontal mixers might be associated with higher exposure levels than vertical ones. The two types of mixers differ in the way powdered ingredients are poured into them, which may account for the observed 'mixer' effect. Powdered ingredients were poured into horizontal mixers closer to the operator's breathing zone. It is not clear whether other methods of loading horizontal mixers exist, but they were not observed in our study. Substituting horizontal mixers with vertical ones might not be practical, since the two types of mixers have different capacities and performance characteristics. We recommend the issue be investigated fiirther. 5.3.4 Other task-related variables and amount and type of flour used 5.3.4.1 Inhalable dust Let us now examine what additional information was revealed by the multiple regression model about the factors that increase inhalable dust exposure in bakeries. Time spent pouring/dusting was one of the most important factors increasing exposure (when adjusted for the effect of other variables in the model) in explaining total variability in the exposure. The role of dusting as factor increasing exposure was established above in connection with the discussion of the divider oil use. The association between 75 weighing and increased exposure can be attributed to re-suspension of flour while placing dusted dough onto a scale. A negative association between time spent decorating, packing/catching, mixing and inhalable dust exposure was revealed. All factors associated with decreased exposure had substantial individual contributions to explaining total variability when adjusted for effects of other predictor variables in the model, except for time spent in decoration. Amounts and types of flour were found not to be associated with inhalable dust exposure. This contradicts bakery employees' reports that dustiness increased with the introduction of finer flours. Urrfortunately, the amounts of different types of flours used in this study were highly correlated, making it difficult to address the issue. Closer investigation of the issue is warranted, since material properties have been demonstrated to have a substantial impact on dust generation. Binding forces were shown to be the most significant factors in predicting dust generation rates, but particle size was demonstrated not to be very important factor in predicting dust generation (Plinke et al., 1995). 5.3.4.2 Wheat antigen Let us now examine what additional information was revealed by the multiple regression models about the factors that increase wheat antigen exposure in bakeries. According to the signs and magnitudes of the regression coefficients in multiple regression, time spent weighing and pouring/dusting had contributions to increases in wheat antigen exposure comparable to those arising from bread/bun and puff-pastry production. Mechanisms by which weighing and pouring/dusting can increase exposure were discussed above. Time spent decorating and packing/catching was associated with decreased wheat antigen exposure. Time committed to decorating had a lower individual contribution to explaining total variability than time devoted to packing/catching, probably due to a correlation with cake and pastry production, which was also included in the models. 76 5.3.4.3 a-Amylase The following additional information on determinants of a-amylase exposure was derived from the multiple regression model. The amount of enriched and coarse flour is an important factor associated with increased a-amylase exposure. This is not surprising, since a-amylase is added to enriched, but not pastry flour. The amount of enriched and coarse flour used has a significant individual contribution to explaining total variability in the model. Time spent weighing has a larger contribution to increasing exposure, probably through the mechanism described above. A number of tasks were negatively associated with a-amylase exposure, however, only one of them had a regression coefficient that was statistically different from zero, namely time spent packing and catching. 5.3.4.4 Interpretation of negative coefficients in multiple regression model There are several reasons why a variable can have a negative regression coefficient in the model. They will be illustrated by an examination of inhalable dust exposure model. It can represent activity for which exposures are lower than average background. This holds for cake and pastry production, associated with a geometric mean exposure of 0.73 mg/m3 (0.44 - 1.22 95%CI), compared to the background in the model of 1.89 mg/m3. A negative association between exposure and time spent decorating and packing/catching can be attributed to the same cause. Another suggested cause may be a low exposure activity that occurs in conjunction with high exposure activity. In such cases, the time occupied by low-exposure activity takes away from the opportunity to be highly exposed, becoming negatively associated with a time-weighted average of the exposure. Thus, mixing occurred in conjunction with pouring of powdered ingredients and, by comparison, was the task of lower exposure. This suggestion was supported by an observation that mixing was positively associated with exposure in univariate analysis, but changed the sign of association when adjusted for time spent pouring. 77 5.3.5 Inhalable dust as a predictor of wheat antigen and a-amylase exposure levels Inhalable dust exposure levels were significant predictors of specific aeroallergen exposure levels, implying that by controlling dustiness, one can also reduce exposure to specific aeroallergens. This suggestion is supported by the high degree of correlation between different exposure measures in this study and the fact that the some determinants of exposure were shared by two or even all three exposure measures. Our data shows that we can use inhalable dust exposure levels to predict exposure to specific aeroallergens, when inhalable dust exposure information is supplemented with knowledge of product and machinery types used by a baker when samples were collected. This can prove useful in exposure assessment in epidemiological studies since immunoassays, used to detenriine levels of specific aeroallergens, are expensive, relative to gravimetric analysis. The a-amylase model with inhalable dust had the same fit as the original model that included continuous predictor variables (time spent per task and amount of raw materials used). Similarly, the wheat antigen model based on inhalable dust levels and categorical predictor variables had a higher fit, when compared to the original model employing continuous (time spent per task and amount of raw materials used) and categorical predictor variables. 5.3.6 Comparison to other researchers' findings The importance of task profiles in predicting bakers' exposures was upheld, since examination of task profiles in regression allowed us to account for most of the variability in inhalable dust and wheat antigen exposure. The differences in inhalable dust and wheat antigen exposure between bakeries in our study were attributable to differences in tasks, equipment and products as was shown by the inability of bakery identity variables to further explain residual error when all of the above factors were considered. Not all variability in a-amylase exposure was explained by the task profiles, since bakery identity contributed significantly to explaining variability in exposure. A similar problem was encountered by Houba et al. (1996a). Our models of flour dust exposure for bakery workers have the best ability to explain variability in exposure (74% to 79%) compared to other investigators' models used to predict 78 inhalable dust exposure in bakeries. Burdorf et al. (1994) were able to explain 61 to 69 percent of variability, while Nieuwenhuijsen et al. (1994b) explained 31 to 46 percent. Our models compare favorably in this respect with other empirical studies of determinants of exposure for other compounds (e.g. 24% in the study of Teschke et al.(1995); 13% for the between-worker variance component in the study of Kromhout et al. (1994); 34% in the study of Preller (1995)). A limitation of our models is that they do not account for intra-individual components of variability. In addition, it is hard to interpret coefficients of log-linear models. Any occupational hygiene exposure data is expected to follow log-normal distribution regardless of mechanism by which exposure is generated (Esmen and Hamrnad, 1977) and a variety of mechanisms can result in log-normality (Koch, 1966; Koch 1969). As in any empirical model, the ability to examine a relationship between any factors depends on the presence of these factors in a sample. For example, among cleaning tasks (Nieuwenhuijsen et al., 1995a) found that mainly silo and bin cleaning tasks resulted in the highest inhalable dust and wheat antigen exposure. These cleaning tasks were not observed in this study. The absence of local exhaust ventilation in the studied bakeries maked it impossible to use our data in assessing local exhaust systems' effectiveness in controlling flour dust exposure. No descriptions of the use of local exhaust ventilation systems in bakeries are available in scientific literature. However, 'adequate' general ventilation and machinery were demonstrated to reduce the risk of sensitization of individuals to flour (Thiel and Ulmer, 1980), presumably by reducing flour aeroallergen exposures. Our findings agree with and expand on what has been reported about the causes of flour dust exposures in bakeries. The tipping task group (including pouring, weighing and mixing) was reported to be associated with elevated inhalable dust and wheat antigen exposures (Nieuwenhuijsen et al., 1995a). Our results point out that it is time spent weighing and/or pouring that might be responsible for this observation. Identification of dusting as a cause of exposure is in agreement with the fmding of Nieuwenhuijsen et al. (1995a) that the flour dusting task group in bread/roll (bun) production showed elevated inhalable dust and wheat antigen exposures in comparison to other tasks. It was reported that inhalable dust exposures 79 doubled among bakers who handled flour in a confectionery bakery, compared with those who did not (Nieuwenhuijsen et al., 1992; Nieuwenhuijsen et al., 1995b). In task-specific sampling, Nieuwenhuijsen et al. (1995a) found dough-braking to lead to lower inhalable dust and wheat antigen exposures than bread or roll (bun) production. However, the method of dough-braking was not described. Still, full-shift time-weighted averages of inhalable dust exposure in dough-braking exceeded those in bread/roll (bun) production (Nieuwenhuijsen et al., 1995b), as in our study. The dispensing/mixing exposure group described by Nieuwenhuijsen et al. (1995a) was defined too broadly to allow for the identification of specific tasks that resulted in elevated exposures. In general, dispensing/mixing and dough brake areas of bakeries were reported to result in the highest time-weighted averages of inhalable dust exposure (Nieuwenhuijsen et al., 1995b), which is in agreement with our findings. Kolopp-Sarda et al. (1994) found that making dough resulted in elevated inhalable dust exposures, while oven workers and those involved in c»llecting/packing products experienced lower exposure levels. Making dough and weighing flour additives (manual tasks) were reported by Jauhiaineu et al. (1993) to cause higher total dust and a-amylase exposures in bakers than the highly automated making of bread, which supports our conclusion that automation of production lines can reduce exposure. Houba (1996a) found that dough-makers and bread bakers were exposed to higher levels of inhalable dust than oven operators, slicers, packers, transport operators, production managers and maintenance/cleaning workers. In addition, bakers involved in bread production tended to be more exposed than those in confectioneries. This is in agreement with our finding that time spent on cleaning, oven operation, performing quality control duties and moving trays within the bakery were not associated with inhalable dust exposure in the multiple regression models and that time spent on packmg/catching, decorating and production of cakes and pastries were negatively associated with exposure. Houba (1996a) describes similar relationships-for wheat antigen and a-amylase. An association between a-amylase exposure and the amount of the raw materials containing a-amylase was implied by Houba (1996a) when he showed that crispbread was linked to elevated a-amylase exposure, because crispbread production involves larger 80 amount of a-amylase than any other product. In our study, an increase in a-amylase exposure was associated with the amount of a-amylase-containing ingredients used (enriched flour and coarse flour), supporting the idea presented above. The relationship between total particulate levels and specific allergen exposure has been explored by various researchers. Tee et al. (1992) observed a linear relationship between gravimetric dust measurements and flour allergen concentrations between 5 mg/m3 and 10 mg/m3, suggesting that gravimetric measurement of flour dust in that concentration range may be a suitable surrogate measure of flour allergens. Similarly, Nieuwenhuijsen et al. (1995a) found a Spearman correlation coefficient of 0.86 between personal average work-shift wheat allergen and inhalable dust exposure levels. Nevertheless, in an earlier paper the same author (Nieuwenhuijsen et al., 1994b) reported that the strength of the relationship varied greatly with the location within a bakery. Likewise, we found strong correlation between flour aeroallergens' and inhalable dust exposure levels. Correlation between inhalable dust and wheat antigen levels found in our study (Spearman correlation coeficient of 0.89) was almost identical to the one reported by Nieuwenhuijsen et al. (1994b). 5.4 Limitations Amounts of flours used had a very broad range, so that their effects in the regression could be driven by a small number of data points. It may have been more appropriate in analysis to re-code amounts of flour used into categorical variables, representing low, medium and high amounts of flour. Small sample sizes for some of the variables decrease the strength of our conclusions. It was not possible to examine the effect of dough improvers on exposure due to small sample sizes. Nor was it possible to investigate the effect of dough improvers added to flour prior to its use in bakeries due to insufficient level of detail in the information provided by the manufacturers about flour composition. A small number of observations impeded investigation into the role of mixer types and worker left-handedness in predicting exposure. Some other conclusions were weakened, since they were based on rather small 81 sample sizes. For example, only three bakers participated in pizza shell, pita bread and tortilla production. Similarly, the effect of divider oil was assessed on the basis of five observations. In addition to small sample size problems, some combinations of variables were not available for examination. This is similar to the problem of incomplete blocks in experimental design. For example, since only dough-brakers were used in puff-pastry production the question remains as to whether the use of a reversible sheeter in dough braking on puff-pastry production lines will reduce exposure. As we have mentioned earlier, in this study it was not possible to separate the effects of forming with the reversible sheeter from the effects of manual forming, which is another example of the problem described above. 5.5 Future directions Further validation of our regression models is needed. In addition to that, we shall attempt to develop predictive models of one aeroallergen, given information on inhalable dust and one other aeroallergen exposure, supplemented by information on production characteristics. This study points to a number of factors that can be determinants of flour dust exposure in bakeries. It seems to provide sufficient information to initiate experimental studies on tasks and equipment identified as determinants of flour dust exposure in this study, which should be able to overcome difficulties with incomplete blocks and small sample sizes, encountered in this study. Furthermore, the cost-effectiveness of the proposed control measures must be investigated, by looking, for example, at the effect of their implementation on labor productivity and product quality. 82 5.6 General conclusions Levels of flour dust reported in this study indicate that there is a need to reduce flour dust exposures in B.C. bakeries. Information derived in the course of this study on important determinants of flour dust exposure is summarized in Table 35. In addition, it appears that all three exposure measures are highly correlated, which implies that inhalable dust exposure can be a significant predictor of exposure to flour aeroallergens. Table 35 Summary of important determinants of flour dust exposure, identified by this study. Inhalable dust Wheat antigen a-Amylase factors increasing exposure bread/bun production puff-pastry production , time spent weighting time spent pouring time spent forming with dough-braker time spent forming with dough-braker/use of dough-braker croissant production use of reversible sheeter cinnamon bun production enriched and coarse flour factors decreasing exposure substitution of dusting with the divider oil use automation of forming tasks time spent packing anc catching cake and pastry production time spent decorating time spent mixing The findings of this study may be applicable to all small B.C. bakeries, given the similarities between participating bakeries and those excluded from the study. However, it is not possible to eliminate the possibility that the participating bakeries were not representative of all small B.C. bakeries, since the telephone survey proved to be an inaccurate tool in characterizing bakeries and because there may be other differences not measured by our survey. Agreement between our findings and those of earlier investigators in other countries increases confidence that our results can be used in controlling flour dust exposures in other bakeries. 83 Bibliography ACGIH (1994) 1994-1995 Threshold Limit Values for Chemical Substances and Physical Agents and Biological Exposure Indices. American Conference of Governmental and Industrial Hygienists. Technical Affairs Office, Kemper Woods Center, 1330 Kemper Meadows Drive, Cincinnati, OH 45240 Andersson I-G, Rosen G (1995) Detailed work analysis for control of exposure to airborne contaminants in workplace. Appl Occup Environ Hyg 10(6): 537 - 544 Archibald BA, Solomon KR, Stephenson GR (1995) Estimation of pesticide exposure to greenhouse applicators using video imaging and other assesment techniques. Am J Ind Hyg Assoc J 56: 226 - 235 Baur X, Sauer W, Weiss W (1988) Baking additives as new allergens in baker's asthma. Respiration 54: 70 - 72 Bjorkstein F (1982) Immunological Methods in the identification of occupational asthma. Eur J Respir Dis 63 (Suppl 123): 21 Blands J, Diamant B, Kailos P et al (1976) Flour allergy in bakers. Int Arch Allergy Appl Immunol 52: 392 - 406 Block G, Kijek K, Chan H, et al (1982) Pathogenic mechanisms in baker's asthma (abstract). Am Rev Respir Dis 125: 74 Block G, Tse KS, Kijek K Chan H, Chan-Yeung M (1983) Baker's asthma: Clinical and immunological studies. Clin Allergy 13: 359-370 Block G, Tse KS, Kijek K, Chan H, Chang - Yeung M (1984) Baker's asthma. Studies of cross - antigenicity between different cereal grains. Clin Allergy 14: 177 - 185 Bohadana AB, Massin N, Wild P, Kolopp M-N, Toamain J-P (1994) Respiratory symptoms and airway responsiveness in apparently healthy workers exposed to flour dust. Eur Respir J 7: 1070 - 1076 Bonnevie P (1958) Occupational allergy in bakeries. In Stenfort - Kroese (ed): Occupational Allergy. Springfield, IL, Charles C Thomas, 1958, pp 161 - 168 Brisman J and Belin L (1991) Clinical and immunological responses to occupational exposure to a-amylase in the baking industry. Br J Ind Med 48: 604 - 608 Burdorf A, Lillienberg L, Brisman J (1994) Characterization of exposure to inhalable flour dust in Swedish bakeries. Ann Occup Hyg 38: 67 - 78 84 De Zotti R, Larese F, Bovenzi M, Negro C, Molinari S (1994) Allergic airway disease in Italian bakers and pastry makers. Occup Environ Med 51: 548 - 552 Ehrilch IR (1994) Fatal Asthma in a Baker: A Case Report. Am J Ind Med 26: 799-802 Esmen and Hammad (1977) Log-normality of envorinmental sampling data. J Environ Sci Health A12 (1&2): 29 -41 Frankloud AW, Luann JA (1965) Asthma caused by the grain weevil. Br J Ind Med 22: 157 - 159 Gannon PF, Burge PS (1993) The SHIELD scheme in the West Midlands Region, United Kingdom. Midland Thoracic Society Research Group. Br J Ind Med 50: 791 - 796 Heinonen K, Kulrnala I, Saamanen A (1996) Local ventilation for powder handling ~ combination of local supply and exhaust air. Am Ind Hyg Assoc J 57: 356 - 364 Hendrick DJ, Davies RJ, Pepys J (1976) Baker's asthma. Clin Allergy 6: 241 - 250 Herxheimer H (1973) The skin sensitivity to flour of baker's apprentices. Acta Allergologica 28: 42 Hornung RW, Reed LD (1990) Estimation of average concentration in presence of nondetectable values. Appl Occup Environ Hyg 5 (1): 46 - 51 Houba R, Heederik D & PEM van Run (1994) Baker's asthma: sensitization to wheat flour and alpha-amylase and relation to allergen exposure. ATS Conference 1994, Boston. USA. American Journal of Respiratory and Clinical Care Medicine 140: A849 Houba R (1996a) Occuptional respiratory allergy in bakery workers — relationship with wheat and fungal a-amylase aeroallergen exposure. Thesis Landbouwuniversiteit Wageningen. ISBN 90-5485-527-4, pp. 25 -120 Houba R, van Run P, Heederik D, Doekes G (1996b) Wheat antigen exposure assessment for epidemiologic studies in bakeries using personal sampling and inhibition ELISA. Clinical and Experimental Allergy 26: 154 -163 Houba R, van Run, Doekes G, Heederik D, Spithoven (1996c) Airborne levels of a-amylase in bakeries. Accepted for publication by J of Allergy and Clinical Immunology. HSE (1995) Pilot study of CEN protocols for the performance testing of workplace aerosol sampling instruments. EC contract MAT1-CT92-0047, Report prepared by LC Kenny (project coordinator), Health and Safety Executive Health and Safety Laboratory, Broad Lane, Sheffield S3 7HQ, UK 85 Jarvinen KAJ, Pirila V, Bjorkstein F, et al (1979) Unsuitability of bakery work for a person with atopy: A study of 234 bakery workers. Ann Allergy 42: 192 - 195 Jauhiainen A, Louhelainen K, Linnainmaa M (1993) Exposure to dust and a-amylase in bakeries. Appl Occup Environ Hyg 8: 721 -725 Jongedijk T, Meijler M, Houba R, Heederik D (1995) Tijdstudies en vergelijkende piekblootstellingsmetingen in ambachtelijke bakkerijen. Tijidschrift voor toegepaste Arbowetenschap 1: 1 - 8, as quoted by Houba R (1996) Occuptional respiratory allergy in bakery workers ~ relationship with wheat and fungal a-amylase aeroallergen exposure. Thesis Landbouwuniversiteit Wageningen. ISBN 90-5485-527-4, p. 127 Kallos P & Kallos-Deffher L (1971) Flour allergy in bakers. 75th Congress of the Trade Union of Swedish Food Workers, Stockholm, 182 Klaustermeyer WB, Bardana EJ, Hale FC (1977) Pulmonary hypersensitivity to Alternaria and Aspergillus in baker's asthma. Clin Allergy 7: 227 - 233 Koch AL (1966) The logarithm in biology. I. Mechanism of generating the log-normal distribution exactly. J TheoretBiol 12: 276-290 Koch AL (1969) The logarithm in biology. II. Distributions simulating the log-normal. J Theoret Biol 23: 251 -268 Kolopp-Sarda MN, Massin N, Gobert B, Wild P, Moulin J-J, Bene MC, Faure GC (1994) Humoral immune responce of workers occupationally exposed to wheat flour. Am J Ind Med 26: 671-679 Kromhout H, Swuste P, Boleij JSM (1994) Empirical modelling of chemical exposure in the rubber manufacturing industry. Ann Occup Hyg 38 (1): 3 - 22 Kumagai S, Kusaka Y, Goto S (1996) Cobalt exposure level and variablity in hard metal industry of Japan. Am Ind Hyg Assoc J 57: 365 - 369 Lagier F, Cartier A, Malo JL (1990) Statistiques medico-legales sur Fasthme professionel au Quebec de 1986 a 1988. Rev Mai Pespir 7: 337 - 341 Lower Mainland Business Directory (1994) Contact Business Information, a division of Contact Marketing Inc., 201-460 Nanaimo St., Vancouver, B.C., V5L 4W3 Manufacturers' Directory (1993) Province of British Columbia, Ministry of Goverment SendcesManufacturers' Directory (1993) British Columbia p393 & 555 86 Masalin KE, Degerth RK, Murtomaa HT (1988) Airborne sugar and flour dust in the Finnish confectionery industry. Appl Ind Hyg 3: 231 - 235. Matz SA (1972) Bakery technology and engineering. Westport, Connecticut, The Avi Publishing Company Inc. Meredith SK, Taylor VM, McDonald JC (1991) Occupational respiratory disease in the United Kingdom 1989: a report to the British Thoracic Society and Society of Occupational Medicine by the SWORD project group. Br J Ind Med 48: 292 - 298 Musk AW, Venables KM, Crook B et al (1989) Respiratory symptoms, lung function, and sensitization to flour in a British bakery. Br J Ind Med 46: 636-642 Napolitano J, Weiss NS (1978) Occupational asthma of bakers. Ann Allergy 40: 258 - 261 Nicas M, Spear RC (1993) A task-based statistical model of a worker's exposure distribution. Part 1, Description of the model. Am Ind Hyg Assoc J 54: 211 - 220 Nieuwenhuijsen MJ, Gordon S, Harris JM, Tee RD, Venables K M , Taylor AJN (1995d) Variation in rat urinary aeroallergen levels explained by differences in site, task and exposure group. Ann Occup Hyg 39 (6): 819 -825 Nieuwenhuijsen MJ, Lowson D, Venables KM, Newman Taylor AJ (1995a) Correlation between measures of exposure in a cohort of bakery workers and flour millers. Ann Occup Hyg 39(3): 291 - 298 Nieuwenhuijsen MJ, Lowson D, Venables KM, Newman Taylor AJ (1995b) Flour dust exposure variability in flour mills and bakeries. Ann Occup Hyg 39(3): 299 - 305 Nieuwenhuijsen MJ, Sandiford C, Lowson D, Tee RD, Venables KM, Newman Taylor AJ (1995c) Peak exposure concentrations of dust and flour aeroallergen in flour mills and bakeries. Ann Occup Hyg 39(2): 193 -201 Nieuwenhuijsen MJ, Sandiford CP, Lowson D, Tee RD, Venables KM, McDonald JC, Newman Taylor AJ (1994a) Dust and flour aeroallergen exposure in flour mills and bakeries. Occup Environ Med 51: 579 - 583 Nieuwenhuijsen MJ, Sandiford CP, Lowson D, Tee RD, Venables KM, McDonald JC, Newman Taylor AJ (1994b) Dust and flour aeroallergen exposure in flour mills and bakeries. Occup Environ Med 51: 584 - 588 Nieuwenhuijsen MJ, White R, Olliver S, Venables KM, Newman Taylor AJ (1992) Exposure variability within and between workers and particle size distribution in a British flour mill and bakery. 1992 BOHS Annual Conference, Edinburgh, Book of Abstracts, p 4 87 NoruSis MJ (1990) SPSS/PC+ Statistics™4.0 for the IBM PC/XT/AT and PS/2. SPSS Inc. 444 N. Michigan Ave., Chicago, Illinois 60611, p. B-29 Plinke MAE, Leith D, Boundy MG, Loffler F (1995) Dust generation from handling powders in industry. Am Ind Hyg Assoc J 56: 251 -257 Popa V, George SAL, Gavanescu O (1970) Occupational and non - occupational respiratory allergy in bakers. Acta Allergologica 25: 159 - 177 Popp W, Zwick H, Rauscher H (1988) Short - term sensitizing antibodies in bakers' asthma. Int Arch Appl Immunol 86: 215-219 Preller L (1995) Respiratory health effects in pig farmers: assessment of expoure and epidemiological studies of risk factors. Thesis Landbouwuniversiteit Wageningen. ISBN 90 - 5485-427-8, pp. 55-71 Prichard MG, Ryan G, Musk AW (1984) Wheat flour sensitization and airway disease in urban bakers. Br J Ind Med 41: 450-454 Prichard MG, Ryan G, Walsh BJ, et al (1985) Skin test and RAST responses to wheat and common allergens and respiratory disease in bakers. Clin Allergy 15: 203 - 210 Quirce S, Cuevas M, Diez - Gomez ML, Fernandez Rivas M, Hinojosa M (1992) Respiratory allergy to Aspergillus - derived enzymes in bakers' asthma J Allergy Clin Immunol 90: 970 - 978 Ramazzini B (1713) Disease of bakers and millers. In De Morbis Artificum (Diseases of Workers; Latin text of 1713 revised and translated by Wilmer Cave Wright). Chicago, University of Chicago Press, 1940, pp 235 -252 Rowe KW (1989) Difference in worker lung function values between large and small commercial bakeries. In: Advances in Industrial Ergonomics and Safety I, Antil Mital (ed), Taylor & Francis, pp. 81 - 87 Scheeper B, Kromhout H, Boleij JSM (1995) Wood-dust exposure during wood-working processes. Ann Occup Hyg 39 (2): 141 - 154 Schultze-Werninghaus G, Zachgo W, Rotermund H, Weiwrodt R, Merget R, Wahl R (1991) Tribolium confusum (Confused Flour Beetle, Rice Flour Beetle) - An occupational allergen in bakers: Determination of IgE antibodies. Int Arch Allergy Appl Immunol 94: 371 - 372 Schwartz M (1947) Flour allergy. J Allergy Clin Immunol 18: 341 88 Smid T, Heederik D, Houba R, Quanjer PH (1992a) Dust - and endotoxin related respiratory effects in the animal feed industry. Am Rev Respir Dis 146: 1474 - 1479 Smid T, Heederik D, Mensink G, Houba R, Boleij JSM (1992b) Exposure to dust, endotoxin and fungi in the animal feed industry. Am Ind Hyg Assoc J 53: 362 - 368 Tee RD, Gordon DJ, Gordon S, Crook B, Nunn AJ, Musk AW et al. (1992) Immune response to flour and dust mites in a United Kingdom bakery Br J Ind Med 49: 581 - 587 Teschke K, Marion SA, van Zuylen MJ, Kennedy SM (1995) Maintenance of stellite and tungsten carbide saw tips: determinants of exposure to cobalt and chromium. Am Ind Hyg Assoc J 53: 661 - 669 Thiel H, Ulmer WT (1980) Bakers' asthma: Development and possibility for treatment. Chest 78: 400 Thorpe A, Brown RC (1995) Factors influencing the production of dust during the hand sanding of wood. Am Ind Hyg Assoc J 56: 236 - 242 Vancouver Area Business Directory (1994) Contact Business Information, a division of Contact Marketing Inc., 201-460 Nanaimo St., Vancouver, B.C., V5L 4W3 Venables K M (1987) Epidemiology and prevention of occupational asthma. Br J Ind Med 44: 73 - 75 Von Dishoeck HAE, Roux DJ (1939) Sensitization to flour and flour illnesses among flour workers. J Hyg 39: 674 Walsh BJ, Wrigley CW, Musk AW (1984) A comparison of the binding of IgE in sera of patients with baker's asthma to soluble and insoluble wheat - grain proteins. J Allergy Clin Immunol 76: 23 - 28 WCB (1980) Industrial Health and Safety Regulations. Workers' Compensation Board of British Columbia WCB (1994) Occupational Disease in British Columbia 1989-1993, August 1994, Workers' Compensation Board of British Columbia Wuthrich B, Baur X (1990) [ Baking ingredients, especially a-amylase, as occupational inhalation allergens in the baking industry ] [ in German ] Schweiz Med Wochenschr 120: 446 - 450 Zock J-P, Heederik D, Kromhout H (1995) Exposure to dust, endotoxin and microorganisms in the potato processing industry. Ann Occup Hyg 39 (6): 841 - 854 89 Appendix A A correlation between a-amyalse and wheat antigen exposure. Figure 23 Linear correlation between natural logarithms of a-amylase and wheat antigen exposure (Spearman r between a-amylase and wheat antigen was calculated to be 0.78 (p = 0.000)). • 1 1 Jf^  ° ° . . Rsq = 0.6489 l~=D • • • •—• • • • •—• • • • •—• a • ^ 1 3 5 20 40 100 300 500 2000 2 4 10 30 50 200 400 1000 ug/m3 wheat antigen 90 Appendix B Selection of variables for analysis in multiple linear regression. Table 36 Selection of treatment of predictor variables in regression analysis of natural logarithm of dust available for inhalation (dependent variable) Variable name Range (Min-Max) Independent variables with significant bivariate Pearson correlation coefficient fr/> 0.70, p < 0.05 Regression coefficient (p-value) of simple linear regression with the dependent variable Sign of slope of simple linear regression if p < 0.20 (+ or -) Variable treatment in regression analysis bakery offered in step III and B atmospheric pressure (mmHg) 760-777 -0.011 (0.78) exclude due to narrow range and lack of linear association with the dependent variable temperature (°C) 17-25 -0.090 (0.37) excluded due to lack of linear association with the dependent variable and narrow range relative humidity 38-63 -0.005 (0.80) excluded due to lack of linear association with the dependent variable time spent on a task (min /shift) decorating 0-233 -0.010 (0.00) offered in step II and A due to significant negative linear association with the dependent variable mixing 0-276 0.006 (0.04) + offered in step I and A due to significant positive linear association with the dependent variable forming 0-373 forming without dough-braker 0.008 (0.00) + exclude in favour of its covariate, which is a more exact measure of tasks performed by virtue of being more narrowly defined forming with dough-braker 0-325 dough-braker use 0.015 (0.00) + offered in step I and A due to significant positive linear association with the dependent variable forming without dough-braker 0-373 forming 0.004 (0.05) + offered in step I and A due to significant positive linear association with the dependent variable cleaning 0-252 -0.003 (0.43) excluded due to lack of linear association with the dependent variable cutting 0-170 0.007 (0.20) + offered in step I and A due to significant positive linear association with the dependent variable pack/catch 0-562 -0.005 (0.00) - offered in step II and A due to significant negative linear 91 Variable name Range (Min-Max) Independent variables with significant bivariate Pearson correlation coefficient M> 0.70, p < 0.05 Regression coefficient (p-value) of simple linear regression with the dependent variable Sign of slope of simple linear regression if p < 0.20 (* or -) Variable treatment in regression analysis association with the dependent variable pour 0-207 0.019 (0.00) + offered in step I and A due to significant positive linear association with the dependent variable weigh 0-109 0.022 (0.00) + offered in step I and A due to significant positive linear association with the dependent variable tray 0-114 0.004 (0.54) excluded due to lack of linear association with the dependent variable oven 0-185 0.005 (0.34) excluded due to lack of linear association with the dependent variable rest 0-405 -0.000 (0.89) excluded due to lack of linear association with the dependent variable quality control 0-325 -0.001 (0.59) excluded due to lack of linear association with the dependent variable kg materials used per shift enriched flour 0-3534 coarse flour total flour 0.000 (0.02) + create new variable: enriched and coarse flour, since it was not always possible to differentiate between amounts of enriched and coarse flour, however their total mass was known coarse flour 0-558 enriched flour 0.007 (0.00) + create new variable: enriched and coarse flour, since it was not always possible to differentiate between amounts of enriched and coarse flour, however their total mass was known enriched and coarse flour 0 - 4092 0.000 (0.03) + offered in step I and A due to significant positive linear association with the dependent variable pastry flour 0-776 0.002 (0.39) excluded due to lack of linear association with the dependent variable total flour 0 - 4868 enriched flour, pastry and coarse flour excluded due to inclusion of its correlate which is easier to interpret dichotomous variables 92 Variable name Range (Min-Max) Independent variables with significant bivariate Pearson correlation coefficient M> 0.70, p < 0.05 Regression coefficient (p-value) of simple linear regression with the dependent variable Sign of slope of simple linear regression if p < 0.20 (* or -) Variable treatment in regression analysis automated forming 0-1 0.099 (0.82) excluded due to lack of linear association with the dependent variable forming with reversible sheeter 0-1 0.692 (0.20) + offered in step I and A due to significant positive linear association with the dependent variable forming with dough-braker 0-1 time forming with dough-braker 3.274 (0.00) + exclude in favour of its covariate, giving preference to continuous variable as a more exact description of task manual forming 0-1 1.161 (0.01) + offered in step I and A due to significant positive linear. association with the dependent variable divider oil 0-1 -1.645 (0.03) - offered in step I and A by virtue of being an exposure control measure bagels 0-1 -0.344 (0.55) excluded due to lack of linear association with the dependent variable tortilla 0-1 pizza or pita combined into one variable (TPP) with 'pizza or pita' due to perfect correlation pizza or pita 0-1 tortilla combined into one variable (TPP) with 'tortilla' due to perfect correlation TTP 0-1 pizza or pita tortilla 0.014 (0.99) excluded due to lack of linear association with the dependent variable bread & buns 0-1 0.952 (0.03) + offered in step I and A due to significant positive linear association with the dependent variable cinnamon buns 0-1 0.077 (0.91) excluded due to lack of linear association with the dependent variable crumpet 0-1 -0.775 (0.26) excluded due to lack of linear association with the dependent variable croissant 0-1 0.366 (0.54) excluded due to lack of linear association with the dependent variable puff pastry 0-1 2.030 (0.00) + offered in step I and A due to significant positive linear association with the dependent variable 93 Variable name Range (Min-Max) Independent variables with significant bivariate Pearson correlation coefficient /r/> 0.70, p < 0.05 Regression coefficient (p-value) of simple linear regression with the dependent variable Sign of slope of simple linear regression if p < 0.20 (* or -) Variable treatment in regression analysis cake & pastries 0-1 -1.392 (0.00) offered in step II and A due to significant negative linear association with the dependent variable left vs. right handed 0-1 exclude since only 2 out of 96 bakers are left-handed other variables seniority (years) 0-45 0.002 (0.94) excluded due to lack of linear association with the dependent variable 94 Table 37 Selection of treatment of predictor variables in regression analysis of natural logarithm of wheat antigen available for inhalation (dependent variable). Variable name Range (Min-Max) Independent variables with significant bivariate Pearson correlation coefficient /r/> 0.70, p < 0.05 Regression coefficient (p-value) of simple linear regression with the dependent variable Sign of slope of simple linear regression if p < 0.20 (+ or -) Variable treatment in regression analysis bakery offered in step III and B atmospheric pressure (mmHg) 760-777 0.02 (0.72) exclude due to narrow range and lack of linear association with the dependent variable temperature (°C) 17-25 -0.16(0.25) excluded due to lack of linear association with the dependent variable and narrow range relative humidity 38-63 0.02 (0.51) excluded due to lack of linear association with the dependent variable time spent on a task (min /shift) decorating 0-233 -0.01 (0.00) offered in step II and A due to significant negative linear association with the dependent variable mixing 0-276 0.01 (0.00) + offered in step I and A due to significant positive linear association with the dependent variable forming 0-373 forming without dough-braker exclude in favour of its covariate, which is a more exact measure of tasks performed by virtue of being more narrowly defined forming with dough-braker 0-325 dough-braker use 0.01 (0.00) + offered in step I and A due to significant positive linear association with the dependent variable forming without dough-braker 0-373 forming 0.01 (0.02) + offered in step I and A due to significant positive linear association with the dependent variable cleaning 0-252 -0.00 (0.52) excluded due to lack of linear association with the dependent variable cutting 0-170 0.01 (0.19) + offered in step I and A due to significant positive linear association with the dependent variable pack/catch 0-562 -0.01 (0.00) offered in step II and A due to significant negative linear association with the dependent variable pour 0-207 0.03 (0.00) + offered in step I and A due to 95 Variable name Range (Min-Max) Independent variables with significant bivariate Pearson correlation coefficient /r/> 0.70, p < 0.05 Regression coefficient (p-value) of simple linear regression with the dependent variable Sign of slope of simple linear regression if p< 0.20 (* or -) Variable treatment in regression analysis significant positive linear association with the dependent variable weigh 0-109 0.04 (0.00) + offered in step I and A due to significant positive linear association with the dependent variable tray 0-114 0.01 (0.37) excluded due to lack of linear association with the dependent variable oven 0-185 0.01 (0.32) excluded due to lack of linear association with the dependent variable rest 0-405 0.00 (0.62) excluded due to lack of linear association with the dependent variable quality control 0-325 -0.00 (0.88) excluded due to lack of linear association with the dependent variable kg materials used per shift enriched flour 0-3534 coarse flour total flour create new variable: enriched and coarse flour, since it was not always possible to differentiate between amounts of enriched and coarse flour, however their total mass was known coarse flour 0-558 enriched flour create new variable: enriched and coarse flour, since it was not always possible to differentiate between amounts of enriched and coarse flour, however their total mass was known enriched and coarse flour 0-4092 0.00 (0.00) + offered in step I and A due to significant positive linear association with the dependent variable pastry flour 0-776 0.01 (0.04) offered in step I and A due to significant positive linear association with the dependent variable total flour 0-4868 enriched flour, pastry and coarse flour excluded due to inclusion of its correlate which is easier to interpret dichotomous variables automated forming 0-1 0.46 (0.46) excluded due to lack of linear association with the dependent 96 Variable name Range (Min-Max) Independent variables with significant bivariate Pearson correlation coefficient /r/> 0.70, p < 0.05 Regression coefficient (p-value) of simple linear regression with the dependent variable Sign of slope of simple linear regression if p < 0.20 (+ or -) Variable treatment in regression analysis variable forming with reversible sheeter 0-1 1.27 (0.01) + offered in step I and A due to significant positive linear association with the dependent variable forming with dough-braker 0-1 time forming with dough-braker exclude in favour of its covariate, giving preference to continuous variable as a more exact description of task manual forming 0-1 1.87 (0.00) + offered in step I and A due to significant positive linear association with the dependent variable divider oil 0-1 -3.33 (0.00) - offered in step I and A by virtue of being an exposure control measure bagels 0-1 -0.03 (0.97) excluded due to lack of linear association with the dependent variable tortilla 0-1 pizza or pita combined into one variable (TPP) with 'pizza or pita' due to perfect correlation pizza or pita 0-1 tortilla combined into one variable (TPP) with 'tortilla' due to perfect correlation TTP 0-1 pizza or pita tortilla 1.13 (0.41) excluded due to lack of linear association with the dependent variable bread & buns 0-1 0.57 (0.36) excluded due to lack of linear association with the dependent variable cinnamon buns 0-1 0.44 (0.66) excluded due to lack of linear association with the dependent variable crumpet 0-1 -1.17(0.23) excluded due to lack of linear association with the dependent variable croissant 0-1 1.77 (0.04) + offered in step I and A due to significant positive linear association with the dependent variable puff pastry 0-1 2.46 (0.00) + offered in step I and A due to significant positive linear association with the dependent variable cake & pastries 0-1 -1.73 (0.00) - offered in step II and A due to significant negative linear 97 Variable name Range (Min-Max) Independent variables with significant bivariate Pearson correlation coefficient M> 0.70, p < 0.05 Regression coefficient (p-value) of simple linear regression with the dependent variable Sign of slope of simple linear regression if p < 0.20 (+ or -) Variable treatment in regression analysis association with the dependent variable left vs. right handed 0-1 exclude since only 2 out of 96 bakers are left-handed other variables seniority (years) 0-45 0.02 (0.44) excluded due to lack of linear association with the dependent variable 98 Table 38 Selection of treatment of predictor variables in regression analysis of natural logarithm of a-amylase available for inhalation (dependent variable). Variable name Range (Min-Max) Independent variables with significant bivariate Pearson correlation coefficient /r/> 0.70, p < 0.05 Regression coefficient (p-value) of simple linear regression with the dependent variable Sign of slope of simple linear regression if p < 0.20 (+ or -) Variable treatment in regression analysis bakery offered in step III and B atmospheric pressure (mmHg) 760-777 -0.04 (0.45) exclude due to narrow range and lack of linear association with the dependent variable temperature (°Q 17-25 -0.22(0.12) - excluded due to narrow range relative humidity 38-63 0.06 (0.07) + offered in step I and A due to significant positive linear association with the dependent variable time spent on a task (min./task/shift) decorating 0-233 -0.01 (0.00) offered in step II and A due to significant negative linear association with the dependent variable mixing 0-276 0.01 (0.03) + offered in step I and A due to significant positive linear association with the dependent variable forming 0-373 forming without dough-braker exclude in favour of its covariate, which is a more exact measure of tasks performed by virtue of being more narrowly defined forming with dough-braker 0-325 dough-braker use 0.01 (0.00) + offered in step I and A due to significant positive linear association with the dependent variable forming without dough-braker 0-373 forming 0.01 (0.00) + offered in step I and A due to significant positive linear association with the dependent variable cleaning 0-252 -0.01 (0.07) offered in step II and A due to significant negative linear association with the dependent variable cutting 0-170 0.01 (0.50) excluded due to lack of linear association with the dependent variable pack/catch 0-562 -0.01 (0.00) offered in step II and A due to significant negative linear association with the dependent variable 99 Variable name Range (Min-Max) Independent variables with significant bivariate Pearson correlation coefficient /r/> 0.70, p < 0.05 Regression coefficient (p-value) of simple linear regression with the dependent variable Sign of slope of simple linear regression if p < 0.20 (+ or -) Variable treatment in regression analysis pour 0-207 0.02 (0.00) + offered in step I and A due to significant positive linear association with the dependent variable weigh 0-109 0.03 (0.00) + offered in step I and A due to significant positive linear association with the dependent variable tray 0-114 0.00 (0.76) excluded due to lack of linear association with the dependent variable oven 0-185 -0.01 (0.44) excluded due to lack of linear association with the dependent variable rest 0-405 0.00 (0.33) excluded due to lack of linear association with the dependent variable quality control 0-325 -0.00 (0.76) excluded due to lack of linear association with the dependent variable kg materials used per shift enriched flour 0-3534 coarse flour total flour offered in step I and A due to significant positive linear association with the dependent variable coarse flour 0-558 enriched flour offered in step I and A due to significant positive linear association with the dependent variable enriched and coarse flour 0-4092 0.00 (0.01) + offered in step I and A due to significant positive linear association with the dependent variable pastry flour 0-776 0.00 (0.21) excluded due to lack of linear association with the dependent variable total flour 0-4868 enriched flour, pastry and coarse flour excluded due to inclusion of its correlate which is easier to interpret dichotomous variables automated forming 0-1 0.43 (0.49) excluded due to lack of linear association with the dependent variable forming with reversible 0-1 0.27 (0.74) excluded due to lack of linear association with the dependent variable 100 Variable name Range (Min-Max) Independent variables with significant bivariate Pearson correlation coefficient /r/> 0.70, p < 0.05 Regression coefficient (p-value) of simple linear regression with the dependent variable Sign of slope of simple linear regression if p < 0.20 (+ or -) Variable treatment in regression analysis sheeter forming with dough-braker 0-1 time forming with dough-braker exclude in favour of its covariate, giving preference to continuous variable as a more exact description of task manual forming 0-1 0.90(0.14) + offered in step I and A due to significant positive linear association with the dependent variable divider oil 0-1 -2.49 (0.02) - offered in step I and A by virtue of being an exposure control measure bagels 0-1 0.72 (0.38) excluded due to lack of linear association with the dependent variable tortilla 0-1 pizza or pita combined into one variable (TPP) with 'pizza or pita' due to perfect correlation pizza or pita 0-1 tortilla combined into one variable (TPP) with 'tortilla' due to perfect correlation TTP 0-1 pizza or pita tortilla -0.88 (0.53) excluded due to lack of linear association with the dependent variable bread & buns 0-1 0.92(0.14) + offered in step I and A due to significant positive linear association with the dependent variable cinnamon buns 0-1 1.29(0.19) + offered in step I and A due to significant positive linear association with the dependent variable crumpet 0-1 -0.79 (0.42) excluded due to lack of linear association with the dependent variable croissant 0-1 2.08 (0.02) + offered in step I and A due to significant positive linear association with the dependent variable puff pastry 0-1 2.57 (0.00) + offered in step I and A due to significant positive linear association with the dependent variable cake & pastries 0-1 -2.23 (0.00) offered in step II and A due to significant negative linear association with the dependent variable 101 Variable name Range (Min-Max) Independent variables with significant bivariate Pearson correlation coefficient /r/> 0.70, p < 0.05 Regression coefficient (p-value) of simple linear regression with the dependent variable Sign of slope of simple linear regression if p < 0.20 (+ or -) Variable treatment in regression analysis left vs. right handed 0-1 exclude since only 2 out of 96 bakers are left-handed other variables seniority (years) 0-45 0.01 (0.68) excluded due to lack of linear association with the dependent variable 102 

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.831.1-0087217/manifest

Comment

Related Items