Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Depression as a risk factor for mortality and disease progression in cancer patients : a meta-analysis Satin, Jillian Robyn 2007

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

Item Metadata

Download

Media
831-ubc_2007-0582.pdf [ 2.69MB ]
Metadata
JSON: 831-1.0101214.json
JSON-LD: 831-1.0101214-ld.json
RDF/XML (Pretty): 831-1.0101214-rdf.xml
RDF/JSON: 831-1.0101214-rdf.json
Turtle: 831-1.0101214-turtle.txt
N-Triples: 831-1.0101214-rdf-ntriples.txt
Original Record: 831-1.0101214-source.json
Full Text
831-1.0101214-fulltext.txt
Citation
831-1.0101214.ris

Full Text

D E P R E S S I O N A S A R I S K F A C T O R F O R M O R T A L I T Y A N D D I S E A S E P R O G R E S S I O N I N C A N C E R P A T I E N T S : A M E T A - A N A L Y S I S by J I L L I A N R O B Y N S A T I N B . A . (Honours) M c G i l l Univers i ty , 2005 A T H E S I S S U B M I T T E D IN 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 A R T S in T H E F A C U L T Y O F G R A D U A T E S T U D I E S (Psychology) T H E U N I V E R S I T Y O F B R I T I S H C O L U M B I A A U G U S T 2007 © J i l l i an R o b y n Satin, 2007 ABSTRACT Depress ion is the most c o m m o n psychologica l problem affecting cancer patients. W h i l e there is a g rowing body o f research l i n k i n g depression to phys io log ica l mediators o f cancer progression, evidence for the effect o f depression on cancer mortal i ty and progression is m i x e d . Researchers continue to investigate the impact o f psycho log ica l intervention on phys ica l outcomes o f cancer before reso lv ing whether depression affects these outcomes in the first place. The present meta-analysis synthesizes the body o f work that has examined the effect o f depression on cancer mortal i ty and progression. U s i n g two databases (Psyc lnfo and Med l ine ) , a search was performed to identify articles that examined the effect o f depression on cancer mortal i ty or progression. The references o f the identified articles were then examined for other relevant articles. The inc lus ion cri ter ia were met by 42 publicat ions reporting on 41 studies and 14 publ icat ions reporting on 13 studies measuring progression. U s i n g a random-effects mode l , odds ratios ( O R ) , risk ratios ( R R ) , and hazard ratios ( H R ) are presented as estimates o f the combined overal l effect o f depression. There is some support for the effect o f depression on mortal i ty in cancer patients, ( O R = 1.281;CI, 1.077-1.523, p=.005; H R = 1.095; 1.027,1.027-1.167, p = .006), but not on progression (OR=1.043; C I , .860-1.265, p = .668; H R = 1.038; C I , .999-1.078, p = .053). There is a significant effect o f depression on cancer mortal i ty, though the magnitude is smal l . Dif f icul t ies in interpretation are discussed. Th i s f ind ing fuels future work to identify subgroups o f patients for w h o m depression has a greater effect on physical outcomes in cancer. TABLE OF CONTENTS Abstract ii Table of Contents iii List of Tables v List of Figures vi Acknowledgements vii Dedication viii 1 Introduction 1 1.1 Depress ion as a c o m m o n prob lem among cancer patients 2 1.2 Depression as most the most frequently studied psychologica l var iable . 3 1.3 Ev idence o f phys io log ica l mechanisms 4 1.4 Depression as frequent target in psycho log ica l treatment 6 2 Methods 10 2.1 Identification o f studies 10 2.2 Inclusion criteria 10 2.3 E x c l u s i o n criteria 11 2.4 Statistical analysis 11 2.5 Types o f effect sizes 12 2.6 H a n d l i n g miss ing data 13 2.7 T y p e o f statistical analysis 13 2.8 A n a l y t i c a l strategies 14 3 Results 15 3.1 Descr ip t ion o f studies 15 3.1.1 M o r t a l i t y 15 3.1.2 Progression 15 3.2 T y p e o f participants 16 3.2.1 Mor t a l i t y 16 3.2.2 Progression 16 3.3 Measurement o f depression 17 3.3.1 M o r t a l i t y 17 3.3.2 Progression 18 3.4 T i m e o f measurement 18 i i i 3.5 Study outcome 18 3.5.1 Mor ta l i t y .. 18 3.5.2 Progression 19 3.6 Statistics 19 3.7 Effect o f depression 20 3.7.1 Mor ta l i t y 20 3.7.2 Progression 23 4 Discussion ..25 4.1 Interpretation o f total effect sizes 25 4.2 Interpretation o f analysis o f moderators 26 4.3 L imi ta t ions 30 4.4 Strengths 31 4.5 Implicat ions o f f indings and suggestions for future studies 31 R e f e r e n c e s 33 Appendix 41 LIST OF TABLES Table 1. N u m b e r o f studies measuring effect o f psycholog ica l variables on cancer mortal i ty 42 Table 2. N u m b e r o f studies measuring the effect o f each psycholog ica l variable on cancer progression 43 Table 3. Studies examin ing the effect o f depression on cancer mortal i ty , study characteristics 44 Table 4. Studies examin ing the effect o f depression on cancer progression, study characteristics 47 Table 5. S u m m a r y o f combined effects, mortal i ty 48 Table 6. A n a l y s i s o f moderator variables, mortal i ty , O R , conservative 49 Table 7. A n a l y s i s o f moderator variables, mortal i ty , O R 50 Table 8. A n a l y s i s o f moderator variables, mortal i ty H R 51 Table 9. S u m m a r y o f combined effects, progression 52 Table 10. A n a l y s i s o f moderator variables, progression, O R 53 Table 11. Statistical s ignif icance across a l l inc luded studies 54 LIST OF FIGURES Figure 1. M o d e l o f phys io log ica l pathways, based on A n t o n i et a l . , 2006 55 Figure 2. F lowchar t o f search 56 Figure 3. Forest plot o f effect o f depression on mortal i ty, O R , conservative 57 Figure 4. Funnel plot corresponding to mortal i ty , O R , conservative 58 Figure 5. Forest plot o f effect o f depression on mortal i ty , O R 59 Figure 6. Forest plot o f effect o f depression on mortal i ty, R R 60 Figure 7. Forest plot o f effect o f depression on mortal i ty, H R 61 Figure 8. Funnel plot corresponding to mortal i ty , H R 62 Figure 9. Forest plot o f effect o f depression on progression, O R 63 Figure 10. Forest plot o f effect o f depression on progression, H R 64 ACKNOWLEDGEMENTS I w o u l d l ike to acknowledge and sincerely thank the volunteers w h o contributed to this project: Sara A s s a d , Paul Ruescher, and A l e n a Talbot . I w o u l d also l ike to extend a wa rm thank y o u to our lab manager, M e l a n i e Ph i l l i p s , for her conceptual and practical help. It has been an honour to be supervised by D r . W o l f g a n g L i n d e n , whose generosity and w i s d o m made this thesis possible. vn This work is dedicated to my uncle, the late Jules Fleischer, who will forever inspire me. 1 INTRODUCTION Cancer k i l l e d an estimated 70,400 Canadians i n 2006, ranking the disease as the leading cause o f death i n the country. A l t h o u g h the diagnosis o f cancer is no longer a death sentence, the l i fet ime probabi l i ty o f developing cancer is increasing, w i t h a probabi l i ty o f 4 4 % for males, and 38 .9% for females i n 2006 (Cancer Canadian Statistics, 2007). Cancer and its treatment are associated w i t h considerable psycho log ica l sequelae. The psycho log ica l issues most c o m m o n l y reported b y cancer patients are depression, anxiety and general distress (van ' t Spi jker , Tr i j sburg , & Du ivenvoorden , 1997). Ev idence is accumulat ing supporting the theory that psycho log ica l stress, and depression more speci f ica l ly , negat ively impact health behaviours (Andersen, K ieco l t -G la se r , & Glaser , 1994), and phys io log ica l processes i nvo lved i n cancer ( A n t o n i et a l . , 2006). Th i s fuels the popular be l ie f that psycho log ica l variables, namely depression, contribute to d imin i shed health outcomes i n cancer. In response, researchers have been attempting to show that psycho log ica l intervention can improve both psycho log ica l and phys ica l outcomes i n cancer patients. There is meta-analytic evidence that psycho log ica l interventions improve psycho log ica l outcomes, but the evidence for phys ica l improvement is poor (Rogers et a l . , 2005). In order to understand w h y these interventions often fa i l to produce effects, it is important to assess whether the psycho log ica l variables that the interventions attempt to manipulate are themselves related to cancer prognosis . Unfortunately, the studies that have measured these relationships present m i x e d results and, to date, there is no quantitative rev iew avai lable to evaluate the strength o f these effects. The present study examines the effect o f depression o n cancer mortal i ty and progression by synthesiz ing the results o f exis t ing studies meta-analyt ical ly. Support for the focus o n depression is p rov ided i n the f o l l o w i n g subsections: 1. Depress ion is a c o m m o n p rob lem among 1 cancer patients; 2. Depress ion is frequently studied w i t h respect to cancer prognosis; 3. There is evidence o f phys io log ica l mechanisms supporting a relationship between depression and cancer outcomes; 4. P sycho log i ca l intervention regular ly focuses on depression as a target. 1.1 Depression as the common problem among cancer patients Sadness is a normal human reaction to the serious stressor o f a cancer diagnosis . S o m e patients, though not a l l , w i l l develop c l i n i c a l depression, and others w i l l experience depressive symptoms. In a meta-analytic rev iew, c l i n i c a l depression stood out as the on ly psycho log ica l p rob lem that is more c o m m o n among cancer patients than among the general popula t ion (van' t Spi jker et a l . , 1997). The same rev iew found that depression is characterized by stable occurrence throughout the course o f the disease, whereas other psycho log ica l problems occur shortly after diagnosis but decrease thereafter. Estimates o f c l i n i c a l depression i n cancer vary but are as h igh as 4 6 % (Mass ie , 2004). The estimates o f the occurrence o f the more prevalent depressive spectrum disorder are as h igh as 58%. The differences i n the rev iewed studies are attributable to differences i n measurement and cutoff decisions, d i f f icul ty i n teasing apart the somatic symptoms o f cancer and those o f depression, and var ia t ion i n patient samples (i.e. disease type, stage). P a i n is s ignif icant ly associated w i t h c l i n i c a l depression, w i t h evidence supporting the causal role o f levels o f pa in i n increasing the incidence o f c l i n i c a l depression (Spiegel , Sands, & K o o p m a n , 1994). The l i k e l i h o o d o f depression also varies by cancer type. The highest prevalence is found among patients suffering f rom oropharyngeal , pancreatic, breast, and lung cancer, w h i l e c o l o n cancer, gyneco log ica l cancer, and l y m p h o m a are associated w i t h part icular ly lower rates o f depression (Mass ie , 2004). 2 1.2 Depression as most frequently studied psychological variable In preparation for the present study, a literature search was performed i n order to determine w h i c h psycho log ica l variables have been studied w i t h respect to cancer mortal i ty and progression. P s y c l n f o and M E D L I N E onl ine databases were used to identify relevant articles. A list o f psycho log ica l variables was c om pi l ed that was meant to be exhaustive, w h i c h is appended to this document (Append ix ) . In order to identify the articles examin ing the effect on mortal i ty , these terms were crossed w i t h 'cancer ' and 'morta l i ty or su rv iva l ' . In order to identify studies examin ing the effect on progression, the terms 'progression, relapse, recurrence, or metastasis' were used. Depress ion emerged as the variable most often studied in both the mortal i ty and cancer progression literature. Ou t o f the total o f 89 studies examin ing the effect o f psycho log ica l variables, 35 examined the effect o f depression on cancer mortal i ty. 14 out o f 35 studies examined the effect o f depression on cancer progression. Deta i led results o f this search w i l l not be presented here, but a b reakdown o f the number o f studies by psycho log ica l variable for the effect o n mortal i ty and progression is p rov ided in Table 1 and Table 2, respectively. O f note, other variables frequently studied are socia l support and quali ty o f l i fe . It w i l l be helpful to conduct a meta-analysis o f the effects o f these variables i n the future. A l s o wor th not ing is the large number o f variables studied (25 for mortal i ty and 18 for progression), and the broad range o f these variables. A d d i t i o n a l l y , there is a need for theory as the majority o f studies examin ing the effects o f psycho log ica l variables on cancer outcomes do not test mechanisms (De Boer , M c C o r m i c k , P ruyn , R y c k m a n , & van den Borne , 1999). The f inding that depression is the most frequently studied variable i n this literature provides further jus t i f ica t ion for selecting depression as the variable o f interest i n the present 3 study. T h e number o f avai lable studies makes it possible to analyze the effect meta-analyt ical ly, w h i c h has not been done previously . Depress ion has been shown by meta-analysis to have a smal l , but significant effect on cancer incidence i n the general popula t ion ( M c G e e , W i l l i a m s , & E l w o o d , 1994). The present meta-analysis goes beyond the level o f populat ion health i n order to assess i f cancer patients are par t icular ly at r isk for poor health outcomes due to depression. 1.3 Evidence of physiological mechanisms A t the same t ime that researchers have been examin ing the effect o f psycho log ica l factors on cancer outcomes, potential mechanisms have been studied i n paral le l . The least controversial route by w h i c h depression cou ld affect cancer outcomes is through health behaviours, i nc lud ing sleep, lack o f exercise, poor diet, a lcohol and drug use, smok ing and poor treatment adherence. Depress ion is k n o w n to worsen a l l o f these harmful behaviours. F o r instance, a p rev ious ly publ i shed meta-analysis reveals that the odds o f treatment non-compl iance w i t h med ica l advice is three times greater i n depressed medica l patients compared to their non-depressed counterparts for a range o f med ica l condit ions (D iMa t t eo , Lepper , & Croghan , 2000). There is also evidence that depression cou ld affect cancer outcomes through the dysregulat ion o f the immune system. Th i s evidence is largely based on the examinat ion o f v i ra l ly-media ted cancers (eg. ce rv ica l cancer, l iver cancer, lymphomas) , w h i c h account for approximately 15% of cancers w o r l d w i d e (zur Hausen, 1991). A meta-analysis o f depression and the immune system found that depression is associated w i t h an increase o f neutrophils, and a decrease o f Natura l K i l l e r ( N K ) cel ls , T and B lymphocytes and both helper and suppressor cy to toxic T cells (Herbert & C o h e n , 1993). Furthermore, depression has been suggested to increase tumour progression through the inh ib i t ion o f the expression o f class-I and class-II major his tocompatabi l i ty complex ( M H C ) molecules , by reducing N K act ivi ty and also by increasing 4 D N A damage, impa i r ing D N A repair and inh ib i t ing apoptosis (Reiche, Nunes & M o r i m o t o , 2004); a l l o f w h i c h facilitate uncontrol led cel lu lar growth, w h i c h is the foundation o f cancer. These studies are generally based o n animal models . There are shared pathways o f chronic stress and depression as contributors to immune dysfunct ion (Herbert & C o h e n , 1993) and chronic stress is predominant ly tested i n animals models suggesting a consistent l inkage, whereas the l i n k w i t h depression is more l i k e l y studied i n humans, where, however , the evidence o f the effect o f depression on the immune system is less clear due to heterogeneity o f methods across studies ( M c D a n i e l , M u s s l e m a n , Porter, Reed , & Nemeroff , 1995). W h i l e v i r a l cancers have received the most attention for their potential suscept ibi l i ty to psycho log ica l impact through the immune system, evidence has accumulated supporting the effects o f psycho log ica l stress on other types o f cancers. In a rev iew o f mechanisms by w h i c h behavioural stress can influence tumour b io logy , the endocrine system and immune system are impl ica ted i n several different processes affecting tumour growth ( A n t o n i , et a l . , 2006). Stress has been shown to affect the immune system and subsequently increase metastasis, tumour growth, and number, size and density o f tumours i n mice and rats across a number o f different studies ( B e n - E l i y a h u , Page, Y i r m i y a , & Shakhar, 1999; Pa lermo-Neto , de O l i v i e r a , & Robespierre de Souza , 2003; Sau l , et a l . , 2005). Injections o f 0 - Adrenerg ic agonists have also been shown i n one study to have these effects, conf i rming that the act ivat ion o f the sympathetic nervous system ( S N S ) is responsible (Basu et a l . , 2001). Note that the types o f cancer i n these experiments are non-v i ra l . A mode l o f the phys io log ica l effect o f depression on cancer progression, based o n that o f A n t o n i et a l . (2006) is presented i n F igure 1. A series o f studies by Thaker et a l . (2006) outlines a complete pathway by w h i c h psycho log ica l stress influences tumour growth i n mice . In this mode l , stress is associated w i t h an 5 increase i n catecholamines, w h i c h activates adrenergic 02 receptors o n ovar ian ca rc inoma cel ls , and subsequently increases the expression o f the Vascu la r Endo the l i a l G r o w t h Factor ( V E G F ) gene, result ing i n enhanced tumour vascular izat ion, more aggressive tumour growth, and increased spreading o f malignant ce l ls . T h e experimenters elegantly ver i f ied the mechan i sm b y showing that the effect disappeared when the mice were g iven p-antagonists pr ior to exposure to the stressor. Taken together, an imal research supports that chronic stress may increase cancer progression and mortal i ty through a variety o f potential mechanisms i n v o l v i n g the immune and endocrine systems. There is lit t le evidence to date that the same is true for humans. G i v e n that the vast majority o f the studies to be inc luded i n the meta-analysis are atheoretical and not designed to test mechanisms, the present study does not a i m to answer questions regarding causal pathways. 1.4 Depression as frequent target in psychological treatment for cancer patients F o l l o w i n g the advent o f the b iopsychosoc ia l mode l (Enge l , 1977), a g rowing number o f researchers argue that psycho log ica l treatment should be integrated into standard med ica l care, inc lud ing the care o f cancer (Bot tomly , 1998). Access to psycho log ica l services is at best l imi ted , especial ly i n rural areas, and not p rov id ing these services on a routine basis has been deemed unethical g iven its effectiveness (Cunn igham, 2000). Howeve r , w h i l e psycho log ica l interventions have been shown to improve some psycho log ica l outcomes, the evidence is not w h o l l y conv inc ing . A meta-analysis o f the effects o f psychosoc ia l intervention w i t h cancer patients revealed a significant decrease i n the psycho log ica l endpoints o f emot ional adjustment, inc lud ing depression, but the reported effect size o f Cohen ' s d = -.24 is smal l for a psycho log ica l intervention study ( M e y e r & M a r k , 1995). Note , however , that c l i n i c a l trials w i th psycho log ica l 6 interventions rout inely fa i l to screen for actual elevations i n levels o f distress and thereby weaken obtainable reductions i n distress (L inden & Sat in, 2007). Ano the r meta-analysis examined the effect o f psychosoc ia l interventions i n cancer patients focusing on qual i ty o f l i fe as the endpoint o f interest. Th is analysis revealed a more impressive effect o f a d = .65, w h i c h can be considered moderate (Rehse & Pukrop , 2003). Perhaps because distress reduct ion is not persuasive enough for funds to be al located to a l low a fu l l integration o f psycholog ica l treatment into standard cancer care, researchers have attempted to show that psycho log ica l treatment can affect the harder outcomes o f cancer mortal i ty and progression. Meta-analyses have demonstrated, however , that the results o f these studies have been disappoint ing. These meta-analyses examined the effects o f psychosoc ia l intervention on medica l outcomes and found no significant effects. The effect remained n u l l when l o o k i n g at specific fo l low-up lengths o f one and four years ( C h o w , Tsao, & Har th , 2004; M e y e r & M a r k , 2005; Smeds lund & R i n g d a l , 2004). In a rev iew that summar ized the effect o f psychosocia l intervention in cancer and cardiovascular disease patient populat ions, psychosoc ia l interventions were found to contribute to smal l or moderate improvements i n psycho log ica l outcomes, but no improvement i n phys ica l outcomes i n cancer patients was demonstrated (Rogers et a l . , 2005). In contrast, psychosocia l intervention was shown to improve both psycholog ica l outcomes and phys ica l outcomes i n cardiovascular patients. F r o m this review, it appears that psychosoc ia l intervention provides a health benefit to patients suffering f rom cardiovascular disease, but not to those suffering f rom cancer. In t ry ing to understand the failure o f psychosoc ia l interventions to affect cancer mortal i ty and progression, it is helpful to consider the theory behind these interventions. The types o f interventions inc luded i n Rogers et al . 's (2005) review, for example , are var ied , i nc lud ing 7 psychoeducation, relaxat ion, and psychotherapy. A commona l i ty among various treatments is that they are designed to manipulate psycho log ica l variables l ike depression, anxiety and general distress. The under ly ing assumption, therefore, is that depression is indeed related to these health outcomes. It seems, however , that the study o f the effect o f psycho log ica l interventions on phys ica l outcome has been hasty, because the effect o f depression i tself has not yet been consistently demonstrated. R e v i e w s report that the evidence o f these effects is inconsistent and unconv inc ing (De Boer , M c C o r m i c k , P ruyn , R y c k m a n , & van den Borne , 1999; F o x , 1995). T h e literature is miss ing a clear demonstration that i nd iv idua l differences i n depression can predict cancer mortal i ty and/or progression. Unless depression is causal ly l i nked to phys ica l outcomes i n cancer, it is un l ike ly that changes to these variables w i l l affect patients' prognosis . R e v i e w s have been conducted to assess the effect o f depression o n cancer prognosis , but qualitative reviews must be interpreted w i t h caut ion as they are l imi ted by subjective bias. In the most recent rev iew o f studies that examine the relationship between depression and cancer surv iva l , M i l o (2003) conc luded that the majori ty (n = 16/21) o f studies showed that depression is associated wi th s ignif icant ly decreased surv iva l . U p o n closer inspection, however , this rev iew is unrepresentative as it is mis s ing relevant studies, and has inc luded studies that examine the effect o f related but different constructs such as helplessness and j o y . In a rev iew o f the effect o f depression on cancer progression, the Spiegel & Gie se -Dav i s (2003) conclude that the majority o f studies show that depression is a r isk factor for progression. The authors point out that the average sample size was twice as large i n the studies that fai led to f ind a significant difference than i n the ones that demonstrated an effect. The present meta-analysis provides a more objective rev iew o f the evidence to date. The m a i n hypothesis tested i n this meta-analysis is that depression increases mortal i ty and progression. The meta-analysis also a imed to identify 8 variables that moderate the relationship between depression and cancer outcomes in order to better identify those patients that are most affected. 9 2 METHODS 2.1 Identification of Studies A literature search was conducted to identify studies examin ing the effect o f depression on cancer mortal i ty and progression. Psyc ln fo (1887-1997) and M E D L I N E (1950-2007) onl ine databases were both used to identify relevant articles for the meta-analysis. 'Depress ion or depressed' was crossed wi th 'cancer ' and 'mor ta l i ty or su r v i va l ' to identify articles that examine the effect on cancer mortal i ty. 'Progress ion , relapse, recurrence or metastasis' was used instead to identify the articles examin ing the effect on cancer progression. A s a next stage in the meta-analytic process, the titles o f the search results were reviewed and the abstracts o f potential ly relevant articles were read in order to identify the articles to be included in the meta-analysis. The reference sections o f relevant articles were then examined for other relevant articles; a process referred to as the ancestry method. The search was duplicated by one o f four other investigators and search results were compared in order to m a x i m i z e comprehensiveness. For studies gauging mortal i ty, this search strategy yie lded 645 results in M E D L I N E and 148 results in Psyc ln fo . Thirty-three articles were identif ied through the search o f databases, and an addit ional 11 articles were identified through the ancestry method. The search cri ter ia h igh l igh t ing progression y ie lded 460 pre l iminary results in M E D L I N E and a further 105 results in Psyc lnfo . E leven articles were identified through the databases, and three addi t ional articles were identif ied through references. A flowchart o f the search strategy is p rov ided in Figure 2. 2.2 Inclusion criteria The ini t ia l inc lus ion cri teria required that the identified articles spec i f ica l ly examined the effect o f a psycho log ica l variable on cancer mortal i ty or progression in human participants 10 diagnosed wi th cancer. Foreign language articles were translated for potential inc lus ion , provided that they included an E n g l i s h abstract. Furthermore, dissertations located through the search were included in addi t ion to articles publ ished in peer-reviewed journals . T h i s decis ion was made in order to reduce publ ica t ion bias. 2.3 Exclusion criteria Studies that included an intervention were excluded; however , most studies d id not ment ion whether or not the patients sought psycho log ica l intervention independently, w h i c h is a potential confound that cannot be ameliorated. Studies w i t h outcome variables measuring progression other than relapse, recurrence or metastasis were excluded (eg., cytokines ; Cos tanzo , et a l . , 2005; Lutgendorf, Johnson, Cooper , Ande r son . Sorosky, Bu l l e r et a l . , 2002). A d d i t i o n a l l y , studies that prospect ively analyzed the effect o f depression on cancer mortal i ty and progression in participants who had not yet been diagnosed wi th cancer were excluded as these studies are confounded by cancer incidence. W h i l e the authors o f a meta-analysis o f the effect o f depression on cancer incidence conclude that the practical importance o f this association is negl ig ib le ( M c G e e , W i l l i a m s , & E l w o o d , 1994), it is nonetheless more rigorous to keep these research questions separate. 2.4 Statistical analysis Meta-analys is is a statistical tool that synthesizes a body o f research more object ively than a quali tat ive review. The Comprehens ive M e t a A n a l y s i s V e r s i o n 2.0 software package was used in this analysis (Borenstein, Hedges, H i g g i n s , & Rothstein, 2005). Th i s program uses a spreadsheet interface to a l l o w the user to compute effect sizes automatical ly in order to avoid error. 11 2.5 Types of effect sizes Effect sizes were computed from the data provided in the or iginal articles. In the present meta-analysis, three types o f effect sizes were computed. The first is the Odds Rat io ( O R ) statistic. The formula to compute an O R is: (p I (\-p)) I (QI ( 1 - 0 ) , wherep is the number o f events (e.g., deaths) occur r ing in one group (e.g., depressed), and Q is the number o f events occur r ing in another group (e.g., not depressed). A s the O R approaches one, the difference in the effects o f the variable o f interest between groups decreases (i.e. O R = l ; no difference between groups). The O R can be posit ive, indicat ing increased risk in the affected group, or negative, indicat ing decreased risk, assuming that the effect being measured is adverse (mortali ty or progression). It is appropriate to use the O R statistic when the event being measured is assessed at one fo l low-up t ime point. The second effect size statistic is the Relat ive R i sk or R i s k Rat io ( R R ) , w h i c h is also used when assessing the occurrence o f an event at one t ime point. The formula for R R is: Pressed / Plum-depressed- where P is the proport ion o f number o f deaths to total sample size o f the group. The relationship between odds ratios and risk ratios is often confused in the literature, as researchers use these terms interchangeably. W h i l e the O R compares the relative odds o f occurrence in two groups, (he R R compares the probabil i t ies o f occurrence compared to the popula t ion . Hence, the difference is subtle, but not t r iv ia l . The two statistics w i l l approximate each other on ly when the rate o f mortal i ty o f the non-depressed group is smal l . B o t h effect sizes can be converted from one type to the other only i f the rate o f mortal i ty for the non-depressed group is g iven . These data were not provided in the included studies, and the effect sizes are therefore reported separately. 12 The third effect size statistic, the Haza rd Rat io ( H R ) , is employed in studies in w h i c h researchers have access to t ime o f death o f each patient, rather than assessing one t ime point. Whenever possible, it is best to calculate the H R , as this effect size amalgamates mul t ip le t ime points. The H R can be reported as a crude (non-adjusted) H R or as an adjusted H R , w h i c h takes into considerat ion k n o w n c l in i ca l prognosticators. Since depression can be confounded by medical variables such as level o f pain or disease severity, it is valuable to separate these effects. Adjusted effect sizes were therefore compared wi th crude effect sizes. 2.6 Handling missing data In the case that authors report that results are non-significant, but do not provide necessary data to compute an effect size, an effect size o f O R = 1.00 was assigned, w h i c h indicates no relationship. Th i s method is conservative, because it is rare that there w o u l d be absolutely no relat ionship between variables. O n the other hand, i f these studies were s i m p l y omitted due to insufficient data, the sample o f inc luded studies w o u l d be biased in an over ly liberal d i rect ion. Therefore, the data were combined wi th and without these studies included and the true effect is assumed to lie in between the obtained effect sizes. 2.7 Type of statistical model Because the studies included ranged in terms o f cancer types and stages, the random-effects model was chosen, as opposed to the fixed-effects model . The random-effects model makes fewer assumptions about the populat ion, and is more conservative (Borenstein, 2007) . In both models , studies are weighted by their inverse variance, w h i c h a l lows for better est imation o f true effects, but the random-effects model assigns s l ight ly less weight to large sample studies, wh ich have less variance. W h e n using the random effects mode l , the summary effect size can be conceptual ized as the average effect, rather than a synthesis o f the total effect, as w o u l d be 13 reflected us ing the fixed-effects model (Borenste in , 2007). A c o m m o n c r i t i c i sm o f meta-analysis is that it mixes apples and oranges. Schwarzer (1989) c lever ly pointed out that one way to evade this problem is to s i m p l y make conclus ions about fruit. 2.8 A n a l y t i c a l s trategies Summary statistics o f the effect o f depression on both mortal i ty and progression are reported as O R , R R , and H R . Unfortunately, researchers reported O R s sometimes as adjusted and sometimes as unadjusted figures w h i c h adds measurement error to the poo l ing o f result ing O R S ; wherever possible, the adjusted O R was used. In addi t ion to these pooled results, studies are analyzed by potential moderating variables. These include publ ica t ion date, mean age, gender, cancer type, assessment o f c l i n i ca l depression versus depressive symptoms, type o f measure o f depression, t ime o f measurement o f depression, and length o f fo l low-up . A l l effects are reported wi th a confidence interval o f 9 5 % . 14 3 RESULTS 3.1 Description of the studies 3.1.1 Mortality Forty-four articles examined the effect o f depression on cancer mortal i ty. For ty- two o f these articles represent independent studies. These studies were publ ished between the years 1979 and 2007. The ma in characteristics o f the studies are presented in Table 3. N o t e that the effect sizes o f the two articles that shared the same populat ion were averaged (Watson , H a v i l a n d , Greer . Dav idson , & B l i s s , 1999; Watson , H o m e w o o d , Hav i l and , & B l i s s , 2005) . N o t e also that three studies (Desa i , 1997; M a i n i o et a l . , 2005; M a i n i o et a l . , 2006) each y i e l d two separate effect sizes, because they provide separate analyses o f patients w i th different types o f cancer. The studies included in the meta-analysis were performed internationally. Twen ty articles were conducted in the Uni ted States o f A m e r i c a , and three were from Canada. The European studies were conducted in the Denmark (one), F in land (four), Ge rmany (two), Italy (one), the Netherlands (two), N o r w a y (one), Spain (one), and the Uni ted K i n g d o m (five). One study was performed in Aus t ra l i a , two in Japan and one i n Israel. The number o f participants in each study ranged from 35 to 20,593. A total o f 46,582 patients participated in the inc luded studies. 3.1.2 Progression Fourteen articles examined the effect o f depression on cancer progression. These articles are based on fourteen independent studies. The characteristics o f these articles are presented in Table 4. The articles were publ ished between the years 1979 and 2006. N o t e again that the effect sizes o f two studies that examined the same populat ion were averaged (Watson et a l . , 1999; Watson et a l . , 2005). These articles were also performed internationally. T w o were conducted in the Uni ted States o f A m e r i c a , and one in Canada. The European studies were conducted in 15 B e l g i u m (one), Italy (one), the Netherlands (one), Sweden (one), and the U n i t e d K i n g d o m (six). One study was performed in Israel. The sample sizes range from 39 to 578. A total o f 2054 patients participated in the included studies. 3.2 Type of* participants 3.2.1 Mortality A l l o f the studies included patients who have been diagnosed w i t h cancer. O f the articles examin ing the effect o f depression on cancer mortal i ty , 13 exc lus ive ly inc luded breast cancer. Other cancers exc lus ive ly studied include brain cancer (three), head and neck cancer (one), l eukemia (one), melanoma (one), lung cancer (five), pancreatic cancer (one), and testicular cancer (one). T w e l v e studies included patients wi th m i x e d types o f cancers. The mean age was not reported in 15 articles. O f the articles that d id report mean age, the mean age ranged from 24.9 to 78. The grand mean is 50.07. Fourteen studies included on ly female patients, one study included only males, and 30 studies included both males and females. 3.2.2 Progression O f the articles examin ing the effect o f depression on cancer progression, nine include patients exc lus ive ly diagnosed wi th breast cancer. Other cancer types exc lus ive ly studied include bladder cancer (one), head and neck cancer (one), l eukemia (one), me lanoma (one) and m i x e d types o f cancer (one). For studies look ing at mortal i ty and progression, stage and h i s to log ica l grade var ied but cou ld not be documented or analyzed because too few studies report this data adequately. The mean age was not reported in f ive articles. O f the articles that d id report mean age, the mean age ranged from 44 to 63 years. The grand mean is 54.59 years. N i n e studies included on ly female patients, no studies included on ly males, and five studies included both -males and females. 16 3.3 Measurement of depression 3.3.1 Mortality The measurement o f depression var ied considerably across studies. O f the studies examin ing the effect o f depression on mortal i ty, three studies used medica l records to determine the presence o f past or current diagnoses o f c l in i ca l depression. Ten studies used interviews to assess M a j o r Depressive Disorder . The content o f the in terview was not a lways descr ibed, but was often based on either the Diagnost ic and Statistical M a n u a l , 3 r d vers ion (DSM-11I) or on the D S M - I V . The D S M - I I 1 cri teria y ie ld a higher incidence o f diagnosis than the revised D S M - I V , wh ich added the requirement o f significant distress or impairment in order to reach a diagnosis . Twenty-n ine studies used standardized self-report questionnaires, and two used unstandardized self-report questionnaires that were face-val id . The questionnaires generally measured depressive symptoms, and d iv ided subjects into high scorers and l o w scorers, based on cut-off scores or median-spli ts . The most c o m m o n l y used diagnostic instrument was the B e c k Depression Inventory ( B D I ) , w h i c h was employed in seven studies. The wide variety o f other standardized instruments included (in alphabetical order): The Affec t Balance Scale ( A B S ) , the Center for E p i d e m i o l o g i c Studies Depression Scale ( C E S - D ) , the Depress ion Scale ( D E P S ) , the Edmonton S y m p t o m Assessment System ( E S A S ) , the Hami l ton Depression Ra t ing Scale ( H A M -D) , the Hospi ta l A n x i e t y and Depress ion Scale ( H A D S ) , the Men ta l Componen t Summary-61 depression ( M C S - 6 1 ) , the Minneso ta M u l t i p h a s i c Personali ty Inventory ( M M P I ) , the Prof i le o f M o o d States ( P O M S ) , the Symptom Check l i s t -90 -Rev i sed ( S C L 9 0 - R ) , the S C L - 9 0 / B r i e f S y m p t o m Inventory ( S C L 9 0 / B S 1 ) , the Wakef i e ld Depression Inventory ( W D I ) , and the Z u n g Se l f -Ra t ing Depress ion Scale (SPS) . 17 3.3.2 Progression The measurement o f depression in the articles gauging the effect o f depression on cancer progression also var ied. T w o studies inc luded interviews to diagnose M a j o r Depress ive Disorder , and twelve used standardized instruments to assess depressive symptoms. These instruments included: the B D I , the B r i e f S y m p t o m Inventory (BS1), the Depress ion Adjec t ive Check l i s t , the C E S - D , the H A D S , the H A M - D , the Psychia t r ic S y m p t o m Index (PSI) , and the W D I . 3.4 Time of measurement It is often diff icul t to discern the length o f t ime after diagnosis at w h i c h patients participate. It was nonetheless deemed important to test i f a significant difference exists between studies that assess depression shortly after diagnosis and those that assess depression later. Depression is highest in the first month after diagnosis (Aass , Fossa, D a h l , & M o e , 1997), and may represent a normal reaction to diagnosis w h i c h dissipates over t ime. In the analysis that reported O R s , w h i c h pooled the greatest number o f studies, four studies were identif ied as measuring depression wi th in one month after cancer diagnosis . It is possible that other studies measured depression wi th in this t ime frame, but it was unclear from study descript ions. There were not enough studies o f this k ind in the other analyses to warrant compar ison . 3.5 Study outcome 3.5.1 Mortality The majority o f studies measuring the effect o f depression on cancer mortal i ty inc luded all-cause mortal i ty as the end-point, rather than teasing out cancer-specific mortal i ty . Because this information was not provided in a sufficient number o f studies, it is not meaningful to report these effects separately due to lack o f statistical power. The length o f fo l low-up varies from eight months to 10 years. For the purpose o f compar ison , the studies have been d icho tomized into 18 those that f o l l o w patients for less than five years and those that f o l l o w patients for more than five years, as l ive years is considered a ' g o l d standard' in measuring survival in the cancer literature. Twen ty - s ix studies fo l lowed patients for less than five years, w h i l e 19 studies fo l l owed patients for f ive years or longer. 3.5.2 P r o g r e s s i o n Studies that examined the effect o f depression on cancer progression either used metastasis, or recurrence as measures o f progression. Metastasis is the spread o f pr imary cancer to other places in the body. Recurrence is the return o f cancer cel ls to the same place or a different place in the body after a period o f remiss ion. The length o f fo l low-up ranged from three years to 10 years. Seven studies fo l lowed patients for less than five years, and seven studies fo l lowed patients for f ive years or longer. 3.6 Sta t i s t ics A research assistant ver if ied 2 5 % o f computed effect sizes and 100% agreement was reached. In order to measure the effect o f depression on mortal i ty, odds ratios ( O R ) were computed for 28 effects, from 26 independent studies. N i n e studies reported that the effect o f depression was not significant, but either reported no data or not enough data to compute effect sizes. In these cases, an O R o f 1.00 was assigned. A s has been previously discussed, this is a very conservative method, and the total effect is therefore presented wi th and without the inc lus ion o f these studies. R i sk ratios were computed for four studies, in cases when there was not enough information to report an odds ratio. Hazard ratios ( H R ) were reported for 12 studies. T h i s statistic cannot be combined wi th the O R or R R effect size statistics and was therefore analyzed separately. Three studies reported crude H R s , wh i l e 12 studies reported adjusted H R s . W h i l e the studies often c l a i m to adjust for 19 " k n o w n c l i n i c a l prognosticators", studies control for va ry ing factors. These factors include, in va ry ing combinat ions across studies: age, cancer type, gender, h is tologic grade, K a r n o f k y performance status, number o f posi t ive l y m p h nodes, pathologic stage, preoperative percentage forced expiratory v o l u m e in one second, serum a lbumin level , smok ing status, and treatment status. O f the studies that examined the effect o f depression on progression, 10 are presented as O R s , and four are presented as H R s . T o determine i f variables moderate the effects o f depression, the O statistic, associated degrees o f freedom, and level o f s ignif icance are presented. The Q statistic captures the degree o f heterogeneity between studies relative to the heterogeneity expected from the w i t h i n studies. If effects are homogenous across different levels o f a variable, the ^-s ta t is t ic is equal to the degrees o f freedom, w h i c h is the number o f inc luded studies minus one. T o determine the percentage o f variance attributable to true differences between studies, the / ' inconsistency statistic is reported. The fo rmula for J2 is (Q-df)/Q * 100, w h i c h represents the true variance over the total variance. T o interpret the meaning o f 7 2 the benchmarks o f 25, 50 and 75 have been proposed to indicate l o w , moderate and h igh degrees o f homogenei ty (Higg ins , Thompson , Deeks, & A l t m a n , 2001). 3.7 Effec t o f dep res s ion 3.7.1 M o r t a l i t y The results o f the combined effects o f depression on mortal i ty and progression are presented in Table 5. W h e n inc lud ing those studies that have been assigned O R = 1.00, the combined O R is based on 28 effect sizes, f rom 25 independent studies, and 2977 total participants. The combined O R is 1.281 (CI , 1.077 - 1.523), Z = 2.796,/? = .005. The ind iv idua l 20 effect sizes range from O R = .227 (CI , .062- .824), Z = -2.254, p = .024 to O R = 31.776, (C I , .986- 1024.477), Z = 1.952, p = .051. Figure 3 presents a forest plot d i sp lay ing the ind iv idua l point estimates o f effects, surrounded by confidence intervals, and inc lud ing the relative weight g iven to each study. The combined effect s ize is based on homogeneous resu Its (p = .097; /" = 26.83). The analyses o f moderator variables y ie lded no significant differences in effect sizes by moderator var iable . These results are displayed in Table 6. Modera tor variables tested inc luded: type o f cancer (breast versus other), gender (female versus both genders), type o f measure (questionnaire versus in terview versus records), depression classif icat ion (depressive symptoms versus M a j o r Depressive Disorder) , fo l low-up (less than five years versus f ive years or longer) , and t ime o f measurement (wi th in one month after diagnosis , one month or later after diagnosis) . A g e (slope = .012, S E = .008, (CI , - .028- .004), Z = -1.44,/? = .149) and publ ica t ion year (slope = .018, S E = .013, (C I , - .002- .038), Z = 1.74, p = .081) were also assessed for their effect, but y ie lded non-signif icant results in regression analyses. T o assess publ ica t ion bias, a funnel plot can be v iewed in Figure 4, w h i c h v i sua l ly represents the relationship between the effect sizes found and the precis ion o f the studies. Since smaller sample sizes are less precise, they are scattered at the bottom o f the plot, forming an inverse funnel shape. If the effects are plotted symmetr ica l ly around the point estimate o f the effect, publ ica t ion bias is not a concern (Borenste in , 2007). In this case, there is asymmetry. The fail-safe N is 55 , w h i c h is the number o f nu l l effects that w o u l d need to exist for the combined effect to become non-significant. Therefore, two unpublished nu l l effects w o u l d have to exist for every included study. Rosenthal (1991) has suggested that the tolerance level for the f i le drawer statistic is 5 A" + 10, where K represents the number o f studies inc luded in the meta-analysis. 21 A c c o r d i n g to this heuristic, there is publ ica t ion bias. Egger ' s test o f the intercept is another test to assess publ ica t ion bias, w h i c h uses precis ion (the inverse o f the standard error) to predict the standardized effect (effect size d iv ided by standard error). In this case, the intercept is .18 (CI , -.679-1.038), / = .358, df= 26, p = .723, w h i c h does not conf i rm publ ica t ion bias, al though it does not preclude the poss ib i l i ty that there is publ ica t ion bias. T o assess the impact o f outl iers, one extreme effect o f O R =31.78 ( L o w grade, M a i n i o et a l . , 2005) was removed from the analysis. Af ter its removal , the total effect size is O R = 1.272, (CI , 1.076-1.502), Z = 2.827, p= .005. The combined effect remains significant and the magnitude o f the effect remains s imi lar . Furthermore, when each study is removed ind iv idua l l y , the overal l effect a lways remains significant, and the point estimate o f O R ranges from 1.233 to 1.312. W h e n exc lud ing those studies that have been assigned O R = 1.00, the combined O R is 1.536, (C I , 1.190- 1.982), Z = 3.297, p = .001. Results are displayed in Figure 5. The combined effect is based on 19 effects from 16 studies, i nc lud ing 1795 participants. The combined effect size is based on homogeneous results (p = .077; I2- 33.656). A g a i n , the analyses o f moderators y ie lded no significant differences. The results are d isplayed in Table 7. The analysis o f R R s y ie lded a combined R R = 1.146, ( C I , .976-1.345), Z - 1.668, p = .095. Results are d isplayed in F igure 6. Th i s effect is based on four effects f rom four independent studies, and is based on 21,094 subjects. The standard error is .082. The combined effect is based on heterogeneous results (p =.000, / ' = 87.019), w h i c h is considered a h igh degree o f heterogeneity. A search for moderator variables was not conducted as there were too few studies to make meaningful comparisons. 22 The analysis o f H R s yie lded a combined H R = 1.095, (CI , 1.027-1.167), Z = 2 .771, p = .006. Results are presented in Table 7. The combined effect is y ie lded from 12 effects, from 12 independent studies, and included 22,511 participants. The combined effect size is based on moderately heterogeneous results (p - .001; I" = 64.194), w h i c h just if ies a search for moderators. Results are d isplayed in Table 8. No te that the type o f measure (Q value = 10.335, df= 3, p = .016), depressive symptoms versus M a j o r Depress ive Disorder (Q value = 8.542, df = 2,p = .014), and length o f fo l low-up (Q value = 7.848, df = 2,p = .020) s ignif icant ly moderated the effect o f depression on mortal i ty. A separate analysis was performed for those H R s that were adjusted for c l i n i c a l factors and those that d i d not. The combined adjusted H R = 1.105, (CI= 1.019- 1.199), Z = 2.409, p = .016. There was no significant difference between the effects y ie lded from the adjusted versus unadjusted H R s , (Q value = 0.00, df = l,p = .993). The funnel plot is presented in Figure 8. 3.7.2 P r o g r e s s i o n The results o f the effect o f depression on cancer progression in O R s are presented in Figure 9. W h e n inc lud ing those studies that have been assigned O R = 1.00, the combined O R is based on nine effect sizes, from nine independent studies, and 1392 total participants. The combined O R is 1.043 (CI , .860- 1.265), Z = .428, p = .668. The combined effect size is based on homogeneous results (p = .938; I2 = 0.00). The analyses o f moderator variables y ie lded no significant differences in effect sizes by moderator variable, but results are d i sp layed in Table 10. Since results were not significant, publ ica t ion bias is not a relevant concern. W h e n exc lud ing those studies that have been assigned O R = 1.00, the combined O R remains non-signif icant , O R = 1.074 (CI , 8.35- 1.381), Z = .558, p = .577. The combined homogenous effect is based on 5 effects i nc lud ing 804 participants, (p = .487; I = 0.00). 23 The analysis o f H R s yie lded a combined H R = 1.038, (CI , .999- 1.078), Z = 1.933, p = .053, as d isp layed in Figure 10. The combined effect is y ie lded from 3 effects and included 896 participants. T w o o f the three H R s were adjusted for c l i n i c a l factors. The combined effect s ize is based on results that were fair ly homogeneous, (p = .224; I2 = 33.091). M e a n i n g f u l comparisons cou ld not be made wi th so few studies. 24 4 Discussion 4.1 Interpretation of total effect sizes The present meta-analysis indicates that depression is indeed a significant r isk factor for mortal i ty in cancer patients. The magnitude o f this elevated r isk is generally considered o f modest c l i n i c a l importance. The analysis o f O R s indicates that depressed cancer patients have a 28-54% greater risk o f death compared to non-depressed patients, depending on conservativeness, w h i l e the analysis o f H R s presents a 9 .5% greater risk o f death. The difference between the sample o f studies for wh ich O R s were reported and the sample o f studies for w h i c h H R s were reported is that the majority o f the H R s are adjusted for c l in i ca l factors, w h i l e the O R s are not. Th i s is a very important dis t inct ion when consider ing that the rate o f depression increases in the presence o f metastasis and recurrence (Aass et a l . , 1997) and higher levels o f pain (Spiegel , Sands, & Kxiopman, 1994). In studies that do not adjust for c l i n i c a l factors, the association between depression and mortal i ty cou ld reflect the presence o f more aggressive cancer, rather than demonstrating that depression is an independent risk factor for mortal i ty . Th i s hypothesis was tested by compar ing the crude and adjusted H R s , but no significant difference was found. The abi l i ty to find a difference may have been l imi ted by the smal l number o f studies reporting crude H R s . There are no other apparent differences between the sample o f studies that report H R s and the sample o f studies that report O R s . Therefore, the difference in effect seems to be dr iven on ly by differences in calculat ion. I f the H R is a more sophisticated measure o f effect size wh ich takes mul t ip le endpoints into account, perhaps it is a better reflection o f the true state o f affairs. The total effect from the studies reporting R R s was not significant, but the most l ike ly interpretation is that the analysis lacked power, as it included few studies. 25 The meta-analyses o f the effect o f depression on cancer progression also fai led to f ind an effect. T h i s is surpris ing because patients who show progression are typ ica l ly aware o f their poor prognosis. G i v e n the few avai lable studies lack o f statistical power could prevent f ind ing a significant effect, but this does not appear to apply , because the O R s and R R s are so close to 1.00, i.e. a nu l l effect, that even a much larger poo l o f studies w o u l d not reach s ignif icance. A second poss ib i l i ty is that the defini t ion o f progression is h igh ly variable across studies thereby increasing measurement error. 4.2 Interpretation of analysis of moderators The experience o f cancer is different for every ind iv idua l . Cancer type is a major contributor to this var iabi l i ty , since each type differs in treatment options, prognosis, and presents unique issues such as loss o f function and disfigurement. Cancer type is also heterogeneous wi th respect to the involvement o f the immune system. For example , cancers that are caused by chemica l carcinogens may be less affected by depression than those associated wi th viruses due to the involvement o f the cel lu lar immune response (Reiche , Nunes & M o r i m o t o , 2004). It has been suggested that very early and very advanced tumours as w e l l as cancers wi th virulent ce l l histopathology (e.g., lung or pancreatic cancer) rarely deviate from their expected course, and are therefore less l i k e l y to be affected by psycho log ica l factors ( L e v y & Roberts, 1992). Instead, cancers that have been shown to be affected by hormonal and immune processes (e.g., breast and melanoma) merit more attention wi th regards to psycholog ica l influence. In the present meta-analysis, breast cancer was the only type o f cancer that was evaluated in enough studies to facilitate meaningful comparisons. W h e n breast cancer was compared wi th other cancer types clustered together, cancer type did not emerge as a significant 26 moderator o f the effect o f depression on mortal i ty or progression in any analysis . Gender , w h i c h is confounded by cancer type, also d id not s ignif icant ly impact the effect o f depression. Cancer stage (i.e., 1-IV) is another potential moderat ing variable, but this informat ion was not avai lable in most studies and the moderat ing effect cou ld thus not be tested. It is a reasonable assumption that depression w o u l d have a greater effect in earlier stages o f disease, before the cancer has progressed. In contrast to this assumption, physical vulnerabi l i ty has been found to increase the effect o f stress on immune change. Therefore, it is possible that later cancer stage w o u l d actually increase the effect o f depression on cancer outcomes (Segerstrom & M i l l e r , 2004). T h i s is s t i l l an open question. A d d i t i o n a l l y , one study included in the meta-analysis ( M a i n i o et a l . , 2005) d iv ided patients by cancer grade, w h i c h measures the aggressiveness o f the cancer. That study found that depression strongly and signif icant ly predicted short surv iva l t ime in patients wi th low-grade g l ioma , but not in patients w i t h high-grade g l ioma . Grade is therefore a potential moderator that should be examined, but cou ld not be assessed due to the insufficient number o f studies that separate or report this variable. A g e was analyzed as a potential moderator, but was not found to s ignif icant ly impact the effect o f depression. Th i s analysis was l imi ted , because several studies d id not report mean values, and also because most studies include patients o f a l l ages. It w o u l d be more meaningful to compare the effect o f depression in samples o f younger cancer patients w i th the effect in samples o f older cancer patients; however, these studies are rarely conducted. The dis t inct ion between a diagnosis o f c l i n i ca l depression and depressive symptoms is important. It is important both for research and theory to determine whether patients w h o do not meet cri teria for c l i n i c a l depression but exhibi t l o w m o o d or other subcl in ica l symptoms are at greater health risk than their less depressed counterparts. In this meta-analysis, both c l i n i ca l 27 depression and depressive symptoms s ignif icant ly placed patients at risk for mortal i ty . A s can be seen in f ab l e 9, there is a significant difference between the samples o f studies that measure depressive symptoms versus those that measure c l i n i ca l depression. T h i s difference, however , is l ike ly dr iven by the large difference in sample sizes. The measurement o f depression is another var iable that varies greatly across studies. A l t h o u g h several standardized instruments are considered to be ' g o l d standards' , researchers use a wide variety o f measures. A l l questionnaires were grouped together for compar ison . The method o f us ing standardized instruments was compared wi th in terview and wi th rev iew o f medica l records. T h i s compar ison overlaps wi th the compar ison o f depressive symptoms and c l in i ca l depression, as questionnaires assess the former, wh i l e in terview and records assess the latter. A n important difference here is that c l i n i ca l depression reflects worse symptomatology than a median-spl i t o f depressive symptoms, for example . A d d i t i o n a l l y , a diagnosis o f c l i n i c a l depression is based on a longer period o f t ime, w h i c h is more l ike ly to have phys io log ica l effects. Despite this hypothesis, the measurement o f depression d id not emerge as a significant moderat ing var iable in analyses other than the analysis reporting H R s for mortal i ty . T h i s obtained difference is l i ke ly dr iven by sample size. W i t h so few studies, def ini t ive conclus ions cannot be drawn. That no significant effects o f potential moderators emerged i n the analysis reporting O R s supports the notion that va ry ing sample size is d r iv ing these differences. The evidence suggests that there is little or no dose-dependent effect o f depression on cancer mortali ty or progression. The t ime o f measurement was compared on ly in the largest analysis ( O R s ) and was compared between the studies that measured depression in the first month after diagnosis versus 28 those that measured depression later. N o significant effect was found, but this is inconc lus ive due to an insufficient number o f studies. The t ime o f measurement o f depression varies considerably across studies. D u e to incomplete report ing and var iabi l i ty , measurement t ime could not be extensively analyzed. However , the few studies that clearly measured depression in the first month after diagnosis were compared w i t h other studies, and no effect was found. T i m e o f measurement remains an important var iable to consider. Depress ion is highest at in i t ia l t ime o f diagnosis (Ferrel l et a l . , 2003), and can be considered a normal reaction to the stressor o f diagnosis i f it shor t - l ived. F i n a l l y , length o f fo l low-up was compared between studies that fo l l owed patients for less than l ive years and those that fo l lowed patients for f ive years or more. A g a i n , this effect was significant on ly in the H R analysis o f mortal i ty, and is l i ke ly dr iven by sample size. In a study included in the meta-analysis that fo l lowed cancer patients over a 10-year per iod, higher baseline depression scores s ignif icant ly predicted lower surv iva l rates from five months post-diagnosis, but the effect increased substantially after 15-25 months ( B r o w n , L e v y , Rosberger , & Edgar 2003). T h i s t ime-delayed effect was also observed in a study examin ing the effect o f depression in medical inpatients (Hermann et a l . , 1998). The impl ica t ions o f these f indings are that longer fo l low-ups reveal stronger effects and are therefore worth pursuing when possible , and that the measure o f depression at one t ime point may be powerful enough to change the course o f disease over a long period o f t ime. However , a r ev iew o f the effect o f depression on cancer provides counter-evidence by showing that the average fo l low-up length for posi t ive f indings was five years, w h i l e the average length o f fo l low-up for negative f indings was 10 years (Spiegel & G i e s e - D a v i s , 2003). In this meta-analysis, it is possible that the H R s y ie lded a smaller total effect than d id the O R s due to the H R s inc lus ion o f earlier t ime points; however , this is speculative. T o 29 better understand the impact o f length o f fo l l ow-up , outcomes should be studied at mul t ip le t ime points. O v e r a l l , no variables were identif ied that s ignif icant ly moderated the effect o f depression on cancer mortal i ty or progression. 4.3 L i m i t a t i o n s Unfortunately, it is diff icul t to appreciate the meaning o f the combined effect sizes, as there are no established benchmarks indica t ing when values represent l o w , m e d i u m , or large effects, as has been done wi th C o h e n ' s d, for example . For sake o f compar ison , it is helpful to note that a comparable meta-analysis found that depressed patients wi th coronary heart disease have a two-t imes greater risk o f mortal i ty than non-depressed patients (Bar th et a l . , 2004) . W h i l e the effect in the present meta-analysis is not as strong by compar ison, it w o u l d be a disservice to conclude that it is negl ig ib le . There were a number o f obstacles to obta ining significant effects. Power was a problem beginning at the level o f pr imary studies. B y d i c h o t o m i z i n g continuous data (depressive symptoms) , those who are extremely depressed are combined wi th less depressed patients, w h i c h can result in failure to f ind a true effect. The total effect size l i ke ly underestimates the true effect. The total effect sizes that d id not emerge as significant were almost certainly due to lack o f power as they inc luded less studies than the significant effects. Ano the r l imi ta t ion o f the study is that it attempts to investigate whether morta l i ty is affected by an interaction between depression and cancer; however , because many studies report all-cause mortal i ty instead o f cancer-specific mortal i ty , conclus ions are somewhat weakened. Because depression has been shown to increase mortal i ty in the general popula t ion (Cuijpers & Smit , 2002), cancer-specif ic mortal i ty must be teased out to appreciate the effect o f depression on cancer outcomes. 30 4.4 Strengths The present meta-analysis synthesizes an impressive body o f research. The search strategy for this meta-analysis was thorough, and the analysis includes more studies than any review o f the topic to date. A d d i t i o n a l l y the meta-analysis does not leave out studies that d id not provide enough data, but instead assigns conservative values to a l l o w for inc lus ion . B y report ing the total effect w i th and without these values, the true effect can be better estimated. Meta-analys is goes beyond the examinat ion o f statistical s ignif icance at the level o f ind iv idua l studies to better estimate true effects. Table 11 classifies a l l inc luded effects accord ing to statistical s ignif icance. Note that the conc lus ion to be drawn from this s imple procedure is that depression does not seem to affect cancer prognosis , though evidence is m i x e d . Th rough meta-analysis, a significant effect was revealed that may have otherwise been over looked . 4.5 Implications of findings and suggestions for future research The f ind ing that depressive symptoms and c l i n i ca l depression place cancer patients at increased risk o f death highlights the need to continue this line o f research. H o w e v e r , it is important to acknowledge that the risk, on average, is relat ively smal l . A take-home message for patients is that they need not feel responsible for their o w n mortali ty i f they become depressed. It has become accepted in popular culture that cancer patients need to maintain a posi t ive attitude in order to hero ica l ly defeat cancer; a recommendat ion w h i c h G ie se -Dav i s and Spiegel (2003) have termed an 'emot ional straightjacket'. The magnitude o f the effect o f depression on mortal i ty does not seem to warrant the assignment o f responsibi l i ty and blame to cancer patients. Cons ide r ing the l imi ted effect o f depression, it is not surprising that studies assessing the impact o f psycholog ica l treatment often fai l to find significant effects on cancer mortal i ty . I f the psycholog ica l treatment is proposed to affect mortal i ty by amel iorat ing depression, its effect can 31 only impact mortal i ty to the extent that depression is a risk factor for mortal i ty and that the treatment successfully reduces this risk. T h i s mediator model needs to be tested; a crucia l step that is often omitted (L inden & Satin, 2007). Psycho log ica l treatment should be avai lable to cancer patients for ethical reasons, but an impressive improvement in surv iva l is un l ike ly , unless a subgroup is identif ied that cou ld benefit more than others. The obtained effect sizes may not be large enough to persuade the government to allot massive funds to psycho log ica l treatment; however , I posit that they are important enough to merit further study. Suggestions for i m p r o v i n g future research include: adjusting for c l i n i c a l factors, report ing cancer-specific mortal i ty separately, and testing moderators outright w i t h large enough samples. These moderators include age, gender, cancer type, stage, and grade. Depression should be measured at different t ime points, and not on ly w i th in the first month o f diagnosis . W h i l e there was no evidence o f differential effects in the meta-analysis, examin ing the effect o f c l i n i ca l depression as opposed to d i c h o t o m i z i n g depressive symptoms should add power. The f inding that depression s ignif icant ly affects cancer mortal i ty, despite methodologica l l imitat ions, fuels the need to better understand the mechanisms behind the effect and to identify potential subgroups o f cancer patients whose outcomes are part icularly affected by depression. 32 REFERENCES References marked wi th an asterisk indicate articles inc luded in the meta-analysis Aass , N . , Fossa. S. D . , D a h l , A . A . , & M o e , T . J . (1997). Prevalence o f anxiety and depression in cancer patients seen at the N o r w e g i a n R a d i u m Hospi ta l . European Journal of Cancer, 35,1597-1604. Andersen , B . L . , K ieco l t , -G la se r , .1. K . , & Glaser , R. (1994). A biobehavioral mode l o f cancer stress and disease course. American Psychologist, 49(5), 389-404. * A n d r y k o w s k i , M . , A . , Brady , M . J . , & Hens lee -Downey , P . J . (1994). Psychosoc ia l factors predict ive o f surv iva l after a l logeneic bone mar row transplantation for l eukemia . Psychosomatic Medicine, 56(5), 432-439. A n t o n i , M . H . , Lutgendorf , S. K . , C o l e , S. W . , Dhabhar, F. S., Sephton, S. E . , M c D o n a l d , P. G . , et a l . (2006). The influence o f bio-behavioural factors on tumour b io logy : pathways and mechanisms. Nature Rev i ews Cancer , 6, 240-248. A s h b u r y , F. D . , F i n d l a y , FI., Reynolds , B . , & M c K e r r a c h e r , K . (1998). A Canad ian Survey o f Cancer Patients ' Experiences: A r e their needs being met? Journal of Pain and Symptom Management, 16(5), 298-306. *Bar rac lough , J . , Pinder, P. , Cruddas, M . , O s m o n d , C , Tay lor , 1., & Perry, M . (1992). L i f e vents and breast cancer prognosis. British Medical Journal, 304, 1078-1082. Barth, .1., Schumacher , M . , F le rmann-Lingen , C . (2004). Depression as a risk factor for mortal i ty in patients w i th coronary heart disease: a meta-analysis. Psychosomatic Medicine 66, 802-813. Basu , S., N a g y , J . A . , Pa l , S., V a s i l e , E . , Eckelhoefer , I. A . , B l i s s S., et a l . (2001). The neurotransmitter dopamine inhibi ts angiogenesis induced by vascular permeabi l i ty factor/vascular endothelial growth factor. Nature Medicine, 7, 569-574. B e n - E l i y a h u , S., Page, G . G . , Y i r m i y a , R. , Shakhar, G . (1999). Ev idence that stress and surgical interventions promote tumor development by suppressing natural k i l l e r ce l l act ivi ty . International Journal of cancer, 80(6), 880-888. Borenst ien, M . A short course in meta-analysis [e-book]. h t t p : / / w w w . M e t a - A n a l y s i s . c o m (30 January 2007). Borenstein M , Hedges L , Fl iggins J , & Rothstein H . (2005). Comprehensive Meta-analysis. V e r s i o n 2. E n g l e w o o d , N J : Biostat , Inc. Bot tomly , A . (1998). Depression in cancer patients: A literature review. European Journal o f Cancer Care , 7(3), 181-191. *Broers , S., Hengeve ld , M . W . , Kap t e in , A . A . , L e Cessie , S., van de L o o , F . , & de V r i e s , T . (1998). A r e pretransplant psycholog ica l variables related to surv iva l after bone mar row transplantation? A prospective study o f 123 consecutive patients. Journal of Psychosomatic Research, 45(4), 341-51 . * B r o w n , K . W . , L e v y , A . R . , Rosberger, Z . , & Edgar , L . (2003). Psycho log ica l distress and cancer su rv iva l : a fo l low-up 10 years after diagnosis . Psychosomatic Medicine, 65(A), 636-643. * B u c c h e r i , G . (1998). Depressive reactions to lung cancer are c o m m o n and often fo l lowed by a poor outcome. European Respiratory Journal, / / ( l ) , 173-178. Car l son , L . E . , & B u l t z , B . D . (2004). Eff icacy and medica l cost offset o f psychosoc ia l interventions in cancer care: M a k i n g the case for economic analyses. Psycho-Oncology, 13, 837-849. *Chang , G . , Orav , E . J. , M c N a m a r a , T. , L o n g , M . Y . , & A n t i n , J. H . (2004). Depress ion , cigarette smok ing , and hematopoiet ic stem ce l l transplantation outcome. Cancer, 101(A), 782-789. C h o w E . Tsao M N . Harth T. (2004). Does psychosoc ia l intervention improve surviva l in cancer? A meta-analysis. Palliative Medicine, I8(\), 25-31. * C o l o n , E . A . , C a l l i e s , A . L . , P o p k i n , M . K.., M c G l a v e , P . B . (1991). Depressed m o o d and other variables related to bone marrow transplantation survival in acute l eukemia . Psychosomatics, 32(A), 420-425. Cos tanzo , E . S., Lutgendorf, S. K . , Sood , A . K . , Ande r son , B . , Sorosky, J. and Lubaroff , D . M . (2005). Psychosoc ia l factors and Interleukin-6 among w o m e n wi th advanced ovarian cancer. Cancer, 104(2), 305-313. C u n n i n g h a m , A . J. (2000). Adjuvant psycho log ica l therapy for cancer patients: Put t ing it on the same foot ing as adjunctive medical therapies. Psycho-Oncology, 9(5), 367-371. De Boer , M . F . , M c C o r m i c k , L . K . , Pruyn , .1. F. , R y c k m a n , R. M . , & van den Borne , B . W . (1999). Phys ica l and psychosocia l correlates o f head and neck cancer: a r ev iew o f the literature. Otolaryngology and Head and Neck Surgery, 120(3), 427-436. *De Brabander, B . , & Geri ts , P. (1999). C h r o n i c and acute stress as predictors o f relapse in pr imary breast cancer patients. Patient Education & Counseling, 37(3), 265-272. 34 *de GraelT, A . , de L e e u w , J . R. , Ros , W . J . , Hord i jk , G . J . , B l i j h a m , G . H . , & Winnubs t , J . A . (2001). Soc iodemographic factors and qual i ty o f life as prognostic indicators in head and neck cancer. European Journal o / C a n c e r , 37(3), 332-339. *Derogatis , L . R . , Abe lo f f , M . D . , & Mel isara tos , N . (1979). Psycho log ica l cop ing mechanisms and surviva l t ime in metastatic breast cancer. Journal of the American Medical Association, 242(14), 1504-1508. *Desa i , M . M . (1997). The effects o f psychiatr ic history on cancer outcomes: Long i tud ina l evidence from a communi ty sample o f w o m e n . Dissertation Abstracts International: Section B: The Sciences and Engineering, 58 (4-B) , 1828. D i M a t t e o , M . R. , Lepper , H . S., & Croghan , T . W . (2000). Depress ion is a risk factor for noncompl iance wi th medica l treatment. Archives of Internal Medicine, 160, 2101 -2107 . *Edwards , J . , D i C l e m e n t e , C . , & Samuels , M . L . (1985). Psycho log ica l Characteris t ics: A pretreatment surv iva l marker o f patients w i th testicular cancer. Journal of Psychosocial Oncology, 3(1), 78-94. *Ehlers , S. L . (2003). Effects o f depression and substance use on head and neck cancer mortal i ty . Dissertat ion Abstracts International: Sect ion B : The Sciences and Engineer ing , 63 (7 -B) , 3470. Enge l , G . L . (1977) . The need for a new medica l mode l : A challenge for b iomedic ine . Science, 796(4286), 129-136. T a l l e r , FL , & Schmidt , M . (2004). Prognost ic value o f depressive cop ing and depression in surv iva l o f lung cancer patients. Psycho-Oncology, 13(5), 359-63. *Fal ler , FL, Bu lzebruck , H . , Dr ings , P. , & L a n g , FI. (1999). C o p i n g , distress, and surv iva l among patients w i th lung cancer. Archives of General Psychiatry, -5(5(8), 756-762. F o x , B . FL (1999). A hypothesis about Spiegel et al.'s 1989 paper on psychosocia l intervention and breast cancer surv iva l . Psycho-Oncology), 7(5), 361-370. G ie se -Dav i s . , J . , and Spiegel , D . (2003). Emot iona l expression o f cancer progressin. In R . J . Dav idson , K.. R. Scherer, & H . H . G o l d s m i t h (Eds.) . Handbook of Affective Sciences (pp. 1053-1082). N e w Y o r k , N Y : O x f o r d Un ive r s i t y Press. * G i l b a r , O . (1996). The connect ion between the psycholog ica l condi t ion o f breast cancer patients and surv iva l . A fo l low-up after eight years. General hlospilal Psychiatry, 18(4), 266-270. 35 * G i r a l d i , T . , R o d a n i , M . G . , Car te i , G . , & Grass i , L . (1997). Psychosoc ia l factors and breast cancer: a 6-year Italian fo l low-up study. Psychotherapy & Psychosomatics, 66(5), 229-236. * G o o d w i n , J . S., Zhang , D . D . , & Ostir , G . V . (2004). Effect o f depression on diagnosis , treatment, and surv iva l o f older w o m e n wi th breast cancer. Journal of the American Geriatrics Society, 52(1), 106-111. *Graham, J . , Rami rez , A . , L o v e , S., Richards , M . , & Burgess , C . (2002). Stressful l ife experiences and risk o f relapse o f breast cancer: observational cohort study. British Journal of Medicine, 324(735}), 1420. *Greer, S., M o r r i s , T. , Pettingale, K . . W . (1979). Psycho log ica l response to breast cancer: effect on outcome. Lancet, 2(8146), 785-787. Higg ins , . ! . P. T . , Thompson , S. G . , Deeks, J . J . , & A l t m a n , D . G . (2001). M e a s u r i n g inconsistency in meta-analyses. British Medical Journal, 2,: 3625-3633. * H i s l o p . T . G . , W a x i e r , N . E . , C o l d m a n , A . J . , E l w o o d , J . M . , & K a n , L . (1987) . The prognostic s ignif icance o f psychosocia l factors in w o m e n wi th breast cancer. Journal of Chronic Diseases, 40(7), 729-735. *Hje r l , K . , Andersen , E . W . , K e i d i n g , N . , M o u r i d s e n , H . T. , Mortensen P . B . , & Jorgensen, T . (2003). Depress ion as a prognostic factor for breast cancer mortal i ty. Psychosomatics, 44(\), 24-30. Herbert, T . B . , C o h e n , S. (1993). Depression and immuni ty : a meta-analytic review. Psychological Bulletin, 113, 472-86. * Jamison , R. N . , B u r i s h , T . G . , & Wal l s ton , K . A . (1987). Psychogenic factors in predic t ing surviva l o f breast cancer patients. Journal of Clinical Oncology, 5(5), 768-772. *Lehto , U . S., Ojanen, M . , D y b a , T. , A r o m a a , A . , & K e l l o k u m p u - L e h t i n e n , P. (2007). Basel ine psychosocia l predictors o f surv iva l in loca l ized melanoma. Journal of Psychosomatic Research, 63, 9-15. * L e i g h , 1-L, Percarpio, B . , Opsah l , C , & Ungerer , J . (1987). Psycholog ica l predictors o f surv iva l in cancer patients undergoing radiation therapy. Psychotherapy & Psychosomatics, 47(2), 65-73. L e v y , S. M . , and Roberts, D . C . (1992). C l i n i c a l s ignif icance o f psychneuro immunulogy : Predic t ion o f cancer outcomes. In N . Schneiderman, P. M c C a b e , P. , & B a u m , A . (Eds.) . Perspectives in Behavioural Medcine: Stress and disease processes, (pp. 165-174). H i l l sda l e , N J : Lawrence E r lbaum Associa tes . 36 Life t ime probabi l i ty o f developing and d y i n g from cancer and the probabi l i ty o f deve lop ing by cancer by age, Canada. Retr ieved on M a r c h 19 l h , 2007, from ht tp : / /www.cancer .ca/vgn/ images/por ta l /c i t_86751114/36/15/1816216925cw_200 7stats_en.pt//' L i n d e n , W . & Satin, .1. R. (2007). A v o i d a b l e pitfal ls in behavioural medic ine outcome research. Annals of Behavioural Medicine, 33(2), 143-147. * L i t o f k s y . N . S., Farace. E . , Ander son , Jr., F . , Meye r s , C . A . , Huang , W . , and L a w s , Jr., E . R . (2004). Depression in patients w i th high-grade g l ioma : Results o f the g l i o m a outcomes project. Neurosurgery, 54, 358-367. * L o b e r i z a , F . R , Jr. R i z z o , J . D . , Bredeson, C . N . , A n t i n , J . H . , H o r o w i t z , M . M . , Weeks , J . C , & Lee , S.J . (2002). Assoc ia t ion o f depressive syndrome and early deaths among patients after s tem-cell transplantation for malignant diseases. Journal of Clinical Oncology, 20(8), 2118-2126. Lutgendorf, S. K . , Johnson, E . L . , Cooper , B . , Ande r son , B . , Sorosky, J . I., Bu l l e r , R . E . , et a l . (2002). Vascu la r Endothel ia l G r o w t h Factor and social support in patients w i t h ovarian cancer. Cancer, 95(4), 808-815. Mass ie (2004). Prevalence o f depression in patients w i th cancer. Journal of (he National Cancer Institute Monographs, 32, 57-71. * M a i n i o , A . , H a k k o , H . , T i m o n e n , M . , N i e m e l a , A . , Ko ivukangas , J . , & Rasanen, P . (2005). Depress ion in relation to survival among neurosurgical patients w i th a pr imary brain tumor: a 5-year fo l low-up study. Neurosurgery, 5676),1234-1242. * M a i n i o , A . , Tuunanen, S., He l ina , FL, N i e m e l a , A . , Ko ivukangas , & Rasanen, P. (2006). Decreased qual i ty o f life and depression as predictors for shorter surv iva l among patients wi th low-grade g l iomas: a fo l low-up from 1990 to 2003.European Archives of Psychiatry and Clinical Neuroscience, 256(8), 516-521. * M a i s e y , N . R. , N o r m a n , A . , Watson , M . , A l l e n , M . J . , H i l l , M . E . , & C u n n i n g h a m , D . (2002). Basel ine quali ty o f life predicts surv iva l in patients w i th advanced colorectal cancer. European Journal of Cancer, 35(10), 1351-1357. M c D a n i e l , S. J . , M u s s l e m a n , D . L . , Porter, M . R . , Reed, D . A . , & Nemeroff , C . B. (1995) . Depress ion in patients w i th cancer: Diagnos i s , b io logy , and treatment. Archives of General Psychiatry, 52 (2), 89-99. M c G e e , R . , W i l l i a m s , S., E l w o o d , M . (1994). Depression and the development o f cancer: a meta-analysis. Social Science and Medicine, 38 (1),187-192. 37 Meyer , T . .1., & M a r k , M . M . (1995). Effects o f psychosocia l interventions wi th adult cancer patients: a meta-analysis o f randomized experiments. Health Psychology. 14(2), 101-108. M i l o , K . M . (2003). The relationship between depression and cancer su rv iva l . Medicine and Health, 86(8), 249-250. M o a d e l , A . , M o r g a n , C , Fatone, A . , Grennan, J . , Carter, J . , Laruffa , G . , S k u m m y , A . , & Dutcher, .1. (1999). Seek ing meaning and hope: Self-reported spiri tual and existential needs among an ethnical ly-diverse cancer patient populat ion. Psycho-Oncology, 8, 378-385. * M o r r i s , T . , Pettingale, K . , & Haybi t t le , J . (1992). Psycho log ica l response to cancer diagnosis and disease outcome in patients w i th breast cancer and l y m p h o m a . Psycho-Oncology, 1(2), 105-114. * M u r p h y , K . C , Jenkins , P. L . , & Whit taker , J . A . (1996). Psychosoc ia l morb id i ty and surv iva l in adult bone marrow transplant recipients—a fo l low-up study. Bone Marrow Transplantation, 18(\), 199-201. *Nakaya , N . , Sa i to -Nakaya , K . , A k i z u k i , N . , Y o s h i k a w a , E . , K o b a y a k a w a , M . , F u j i m o r i , M . , N a g a i , K . , N i s h i w a k i , Y . , Fukudo, S., Tsubono, Y . , & U c h i t o m i , Y . (2006). Depress ion and surv iva l in patients w i th non-smal l ce l l lung cancer after curative resection: a pre l iminary study. Cancer Science, 97(3), 199-205. *Naughton , M . . ) . , Herndon, J . E , 2nd. , Shumaker , S. A . , M i l l e r , A . A . , K o r n b l i t h , A . B . , Chao , D . , & H o l l a n d , J . (2002). The health-related quali ty o f life and surviva l o f sma l l - ce l l lung cancer patients: results o f a compan ion study to C A L G B 9033. Quality of Life Research, 11(3), 235-248. *Osborne, R. H . , S a l i , A . , Aaronson , N . K . , E l swor th , G . R. , M d z e w s k i , B . , & S inc la i r , A . J . (2004). Immune function and adjustment style: do they predict surv iva l in breast cancer?. Psycho-Oncology, 13(3), 199-210. *Palapattu, G . S., Bast ian, P. J . , S lavney, P. R. , Ha i s f i e ld -Wol fe , M . E . , Wa lke r , J . M . , Br in tzenhofeszoc, K . , T rock , B . , Zabora , J . , & Schoenberg, M . P. (2004). Preoperative somatic symptoms are associated wi th disease progression in patients wi th bladder ca rc inoma after cystectomy. Cancer, 70/ (10) , 2209-2213. Pa le rmo-Neto , .1., de O l i v i e r a Massoco , C , & Robespierre de Souza , W . (2003). Effects o f phys ica l and psycholog ica l stressors on behavior, macrophage act ivi ty , and E h r l i c h tumor growth . Brain, Behavior, and Immunity, 17, 43-54. * Palmer, .1. L . , & F i sch , M . J . (2005). A s s o c i a t i o n between symptom distress and surv iva l in outpatients seen in a pal l iat ive care cancer center. Journal of Pain & Symptom Management, 29(6), 565-571. 38 *Prieto, J . M . , A t a l a , J . , B l a n c h , J . , Carreras, E . , R o v i r a , M . , C i re ra , E . , Esp ina l , A . , & Gasto , C . (2005). Ro l e o f depression as a predictor o f mortal i ty among cancer patients after stem-cel l transplantation. Journal of Clinical Oncology, 23(25), 6063-6071. Rehse, B . , & Pukrop , R. (2003). Effects o f psychosocia l interventions on quali ty o f life in adult cancer patients: M e t a analysis o f 37 publ ished control led outcome studies, Patient Education and Counseling, 50(2), 179-186. Re iche , E . M . V . , Nunes , S. O . V . , & M o r i m o t o , H . K . (2004) Stress, depression, the immune system, and cancer. The Lancet Oncology!, .5(10), 617-625. * Richardson , J . L . , Zarnegar, Z . , B i s n o , B . , & Lev ine A . (1990). Psychosoc ia l status at ini t ia t ion o f cancer treatment and surv iva l . Journal of Psychosomatic Research, 34(2), 189-201. • R i n g d a l , G . I . , Gotes tam, K . G . , Kaasa , S., K v i n n s l a n d , S., & R ingda l , K . (1996). Prognost ic factors and surviva l in a heterogeneous sample o f cancer patients. British Journal of Cancer , 73(12), 1594-1599. Rogers, M . , Fayter, D . , Richardson , G . , R i t ch i e , G . , L e w i n , R. , & Sowden , A . J . (2005). The effects o f psychosocia l interventions in cancer and heart disease: A rev iew o f systematic reviews. Centre for Rev i ews and Dissemina t ion , Un ive r s i ty o f Y o r k , U . K . Rosenthal , R. (1991). Meta-analyt ic procedures for social research: Rev ised Ed i t i on . N e w b u r y Park, C A : Sage. *Sa i to -Nakaya , K.., N a k a y a , N . , F u j i m o r i , M . , A k i z u k i , N . , Y o s h i k a w a , E . , K o b a y a k a w a , M . , N a g a i , K... N i s h i w a k i , Y . , Tsubono, Y . , & U c h i t o m i , Y . (2006). M a r i t a l status, social support and surviva l after curative resection in non-smal l -ce l l lung cancer. Cancer Science, 97(3), 206-213. Sau l , A . N . , O b e r y s z y n , T . M . , Daugherty, C , Kusewi t t , D . , Jones, S., J ewe l l , S. et a l . (2005). C h r o n i c stress and susceptibi l i ty to sk in cancer. Journal of the National Cancer Institute, 97(23), 1760-1767. * S c h u l z , R. , B o o k w a l a , J . , Knapp , J . E . , Scheier, M . , & W i l l i a m s o n , G . M . (1996). Pess imism, age, and cancer mortal i ty . Psychology & Aging, 11(2), 304-309. Schwarzer , R. (1989). Meta-analys is programs. Behavior Research Methods, Instruments, & Computers, 20(3), 338. Segerstrom, S. C , & M i l l e r , G . E . (2004). Psycho log ica l stress and the human immune system: a meta-analytic study o f 30 years o f inqui ry . Psychological Bulletin, 130(4), 601-30. 39 *She iban i -Rad , S., & V e l a n o v i c h , V . (2006). Effects o f depression on the surv iva l o f pancreatic adenocarcinoma. Pancreas, 32(1), 58-61. Smeds lund , G . , & R i n g d a l , G . I. (2004) Meta-ana lys is o f the effects o f psychosoc ia l interventions on survival t ime in cancer patients. Journal of Psychosomatic Research, 57, 123-131. Sol lner , W . , Z i n g g - S c h i r , M . , R u m p o l d , G . , & Fr i t sch , P. (1997). At t i tude toward alternative therapy, compl iance wi th standard treatment, and need for emotional support in patients w i th melanoma. Archives of Dermatology, 133(3), 379-80. Spiegel , D . & G i e s e - D a v i s , J . (2003). Depress ion and cancer: M e c h a n i s m s and disease progression. Biological Psychiatry, 54, 269-282. Spiegel , D . , Sands, S., & K o o p m a n , C . (1994). P a i n and depression in patients w i th cancer. Cancer, 74, 2570 - 2578 . *Stein, S., L i n n , M . W . , & Stein, E . M . (1989). Psycho log ica l correlates o f surv iva l in nursing home cancer patients. Gerontologist, 29(2), 224-228. * S t o m m e l , M . , G i v e n , B . A . , & G i v e n , C . W . (2002). Depress ion and functional status as predictors o f death among cancer patients. Cancer, 94(\0), 2719-2727'. Thaker , P. M . , Han , L . Y . , Kamat , A . A . , A r e v a l o , J . M . , Takahashi , R . , L u , C . , et a l . (2006). C h r o n i c stress promotes tumor growth and angiogenesis in a mouse mode l o f ovarian carc inoma. Nature M e d i c i n e , /2(8) , 939-944. van ' t Spi jker , A . , Tr i jsburg, R. W . , & Duivenvoorden , H . J . (1996). Psycho log ica l sequelae o f cancer diagnosis: A meta-analytical rev iew o f 58 studies after 1980, Psychosomatic Medicine, 59(3), 280-293. * V i g a n o , A . , Bruera , E . , Jhangri , G . S., N e w m a n , S. C , F ie lds , A . L . , Sua rez -Almazo r , M . E . (2000). C l i n i c a l survival predictors in patients w i t h advanced cancer. Archives of Internal Medicine, 160(6), 861-868. * Watson , M . , H a v i l a n d , J . S., Greer, S., Dav idson , J . , & B l i s s , J . M . (1999). Influence o f psycho log ica l response on survival in breast cancer: a population-based cohort study. Lancet, 354(9187), 1331-1336. * Watson , M . , H o m e w o o d , J . , Hav i l and , J . , & B l i s s , J . M . (2005). Influence o f psycho log ica l response on breast cancer surv iva l : 10-year fo l low-up o f a population-based cohort. European Journal of Cancer , 41(\2), 1710-1714. Z u r Hausen, H . (1991). Vi ruses in human cancers. Science, 254, 1167-1173. 40 A P P E N D I X : Psychological Variables Examined in Cancer Patient Research Literatu 1. A n g e r 2. A n x i e t y 3. Benefi t f ind ing 4. B l a m e 5. C o p i n g 6. Confus ion 7. Crea t iv i ty 8. Depress ion 9. Emot iona l suppression 10. Faith 1 1. F igh t ing spirit 12. Helplessness 13. Hope 14. Hopelessness 15. Hos t i l i ty 16. H u m o r , humour 17. Internal control , self-control , locus o f control 1 8. L i f e event 19. M e a n i n g 20. Men ta l disorder 21 . Men ta l health 22. Men ta l i l lness 23. O p t i m i s m 24. Personali ty 25. Psycho log ica l disorder, psychiatr ic disorder 26. Re l ig ios i ty 27. Repress ion 28. Self-eff icacy 29. Sense o f coherence 30. Sense o f purpose 31 . Soc ia l support 32. Spi r i tua l i ty 33. Stress 34. Tens ion 35. Qua l i ty o f l ife 36. W e l l - b e i n g Table 1: Number of studies measuring the effect of each psychological variable on cancer mortality Psychological Variables No. Studies Depression 35 Socia l Support 22 Qual i ty o f L i f e 22 Con t ro l 15 A n x i e t y 12 C o p i n g 13 Distress 11 Adjustment 9 M o o d / A f f e c t 8 Personali ty 5 Stress 5 Self-esteem 5 Re l ig ios i ty 4 Psycho log ica l Status 3 Hopelessness 3 L i f e Events 4 W e l l - b e i n g At t i tude towards disease 2 O p t i m i s m 2 Suppression 3 A n g e r 1 Cogn i t i ve Func t ion ing 1 Host i l i ty 1 L i f e Satisfaction 1 M e a n i n g 1 42 Table 2: Number of studies measuring the effect of each psychological variable on cancer progression Psycho log ica l Var iab les N o . Studies Depression 14 Soc ia l Support 9 Qual i ty o f L i f e 8 C o p i n g 6 Adjustment 6 Con t ro l 6 L i f e Events 6 A n x i e t y 5 Distress 4 M o o d / Affec t 4 Stress 3 Suppression 3 Hopelessness 2 L i fe Satisfaction 2 Personali ty 2 Psycho log ica l Status 2 Hos t i l i ty 1 Self-Esteem 1 43 Table 3: Study characteristics for articles examining the effect of depression on cancer mortality Study Year N Cancer type Gender Mean Age Symptoms /Clinical Depression Measure Length of follow- up Cause of death Andrykowski et al. 1994 42 Leukemia Both 34 Symptoms P O M S < 5 years Al l-cause Beresford et al . 2006 86 M ixed Both 55.5 Symptoms Beck Depression Inventory (BDI) < 5 years Al l -cause Broers et al. 1998 123 Leukemia Both 35.4 Symptoms Symptom checklist 90-R (Dutch version) < 5 years Al i-cause Brown et al. 2003 205 M ixed Both 56.3 Symptoms C E S - D < 5 years Al l -cause Buccheri 1998 133 Lung Both 65 Symptoms Zung's Self-Rating Depression Scale (SDS) < 5 years Al l-cause Chang et al. 2004. 114 Leukemia Both 46 Symptoms BDI < 5 years Al l -cause Colon et al. 1991 100 Leukemia Both 30 Both Psychiatric consultation < 5 years Al l-cause de Graeff et al . 2000 208 Head and neck Both 60 Symptoms C E S - D scale < 5 years Al l -cause Deragotis et al . 1979 35 Breast Female 55.2 Symptoms Affect Balance Scale ( A B S ) < 5 years Al l -cause Desai 1997 -72 -118 -Breast -M ixed Female 68.01 Cl in ical depression, current and lifetime Diagnostic Interview Schedule > 5 years Al l-cause Edwards et al. 1985 26 Testicular Male 24.9 Symptoms M M P I > 5 years Al l-cause Ehlers 2003 131 Head and neck Both 60 Symptoms BDI > 5 years Al l-cause Faller & Schmidt 2004 57 Lung Both 65 Symptoms H A D S > 5 years Al l-cause Faller et al. 1999 103 Lung Both 59 Symptoms Depression Scale (D-S) > 5 years Al l -cause Gi lbar 1996 40 Breast Female 50.06 Symptoms Brief Symptom Inventory (BSI) > 5 years Al l-cause Goodwin et al. 2004 24,6 96 Breast Female 75.85 Cl in ical depression Medicare < 5 years Cancer-specific Greer et al. 1979 47 Breast Female Symptoms Hamilton Scale > 5 years Cancer-4^ Study Year N Cancer type Gender Mean Age Symptoms /Clinical Depression Measure Length of follow- up Cause of death specific Hislop et al. 1987 133 Breast Female Symptoms Psychiatric symptom index < 5 years Al l -cause Hjerl et al. 2003 20,5 93 Breast Female Cl in ical depression Records from Danish Psychiatric Register > 5 years Natural causes and all-cause separate Jamison, et al. 1987 49 Breast Female 50.6 Symptoms Zung depression scale < 5 years Cancer-specific Jenkins et al. 1994 30 Leukemia Both 34.7 Symptoms, clinical depression Composite international diagnostic interview (CIDI), H A D S < 5 years Al l -cause Lehto et al. 2006 101 Breast Female 54.2 Symptoms Depression scale (DEPS) > 5 years Al l -cause Lehto et al. 2007 59 Melanoma Both 53.8 Symptoms Depression scale' (DEPS) > 5 years Al l -cause Leigh, et al. 1987 101 M ixed . Both 58.31 Symptoms BD1 < 5 years Al l -cause Loberiza et al. 2002 193 Leukemia Both Symptoms Not standardized < 5 years Al l -cause Main io et al. 2005 75 Brain Both 46.83 Symptoms BDI > 5 years Al l -cause Main io et al . 2006 101 Brain Both 49.03 Symptoms BDI > 5 years Al l -cause Morr is et al. 1992 138 - Breast - Lymphoma (Hodgkin 's and non-Hodgkin 's) , Both Symptoms Wakefield self assessment depression inventory > 5 years Al l -cause Murphy, Jenkins, & Whittaker 1996 56 Leukemia Both 35.4 Cl in ica l depression CIDI > 5 years Al l -cause Nakaya et al. 2005 -226 -229 Lung Both Symptoms, cl inical depression SCID, P O M S > 5 years Al l -cause Naughton, et al. , 2002 2002 45 Lung Both Symptoms C E S - D < 5 years Al l -cause LSI Study Year N Cancer type Gender Mean Age Symptoms /Clinical Depression Measure Length of follow- up Cause of death Osborne et al. 2003 61 Breast Female 55.48 Symptoms H A D S > 5 years Cancer-specific Palmer & Fisch 2005 225 Mixed Both Symptoms Anderson Symptom Assessment System < 5 years Al l -cause Prieto et al. 2005 199 Leukemia Both Cl in ical depression Memorial Sloan-Kettering Cancer Center < 5 years / > 5 years Al l -cause Richardson et al. 1990 139 Rec ta l , hematologic Both Symptoms B D I , Zung self-rating depression scale < 5 years Al l -cause Ringdal , et al. 1996 239 M ixed , lymphomas, gastrointestinal , prostate, lung Both 57 Symptoms H A D S < 5 years Al l -cause Saito-Nakaya et al. 2005 238 Lung Both Cl in ical depression S C I D Al l -cause Schulz et al. 1996 238 M ixed Both Symptoms C E S - D < 5 years Al l -cause Sheibani-Rad and Velanovich 2005 258 Pancreatic Both Cl in ical depression Medical records < 5 years Al l -cause Stein et al. 1989 90 M ixed Both 78 Symptoms P O M S < 5 years Al l -cause Stommel et al . 2002 871 M ixed - Both 70.1 Symptoms C E S - D < 5 years Al l -cause Vigano et al. 2000 227 Mixed Both 62 Symptoms Edmonton Symptom Assessment Scale < 5 years Al l -cause Watson et al . 1999 578 Breast Female 55 Symptoms H A D S > 5 years Al l-cause Watson et al. 2005 578 Breast Female 55 Symptoms H A D S > 5 years Al l -cause Note : E l l ipses indicate informat ion was not avai lable . 4 ^ ON Table 4: Studies examining the effect of depression on cancer progression and associated study characteristics Study Publication year N Cancer type Gender Mean Age Symptoms vs. clinical depression Measure Length of follow up Barraclough et al. 1992 204 Breast Female 54.3 Clinical depression Interview < 5 years Bergenmar et al. 2004 436 Melanoma Both Symptoms H A D S < 5 years Chang et al. 2004 114 Leukemia Both 44 Symptoms BDI < 5 years De Brabander & Gerits 1999 39 Breast Female 59.03 Symptoms Depression Adjective Checklist < 5 years de Graeff et al. 2000 208 Head and neck Both 60 Symptoms C E S - D scale < 5 years Gilbar 1996 40 Breast Female 50.06 Symptoms Brief Symptom Inventory (BSI) > 5 years Giraldi et al 1997 95 Breast Female 51 Symptoms C E S - D > 5 years Graham et al. 2002 171 Breast Female Cl inical depression Interview < 5 years Greer et al . 1979 57 Breast Cancer Female Symptoms Hamilton Scale > 5 years Hislop et al. 1987 133 Breast Female Symptoms Psychiatric symptom index < 5 years Lehto et al. 2006 101 Breast Female 54.2 Symptoms Depression scale (DEPS) > 5 years Morris et al. 1992 138 Breast, lymphoma Both Symptoms Wakefield self assessment depression inventory > 5 years Palapattu et al. 2004 65 Bladder Both 62.99 Symptoms BSI-18 < 5 years Watson et al. 1999 578 Breast Female 55 Symptoms H A D S > 5 years Watson et al. 2005 578 Breast Female 55 Symptoms H A D S > 5 years Note : E l l ipses indicate information was not avai lable . Table 5: Summary of total effects, Mortality 95% CI Significance Samp e Size Homogeneity Effect size Effect Lower Upper Z P k N P I2 O R Conservative 1.281 1.077 1.523 2.796 .005 28 2977 .097 26.83 O R 1.536 1.190 1.982 3.297 .001 19 1795 .077 33.656 R R 1.146 .976 1.345 1.668 .095 4 21094 .000 87.019 H R 1.095 1.027 1.167 2.771 .006 12 225111 .001 64.194 Table 6: Analysis of moderator variables, Mortality (OR, conservative) 95% CI Significance Sample Size Homogeneity Moderator O R Lower Upper Z P K n Q df P Type Breast .988 .644 1.516 -.055 .957 7 479 Other 1.358 1.130 1.631 3.269 .001 21 2498 2.493 1 .114 Gender Female 1.009 .698 1.459 .047 .963 8 597 Male/Both 1.366 1.128 1.655 3.192 .001 20 2380 2.472 1 .116 Measure3 Questionnaire 1.318 .932 1.864 1.563 .118 23 2631 Interview 1.524 .892 2.603 1.542 .123 4 346 Records .885 2 .643 Depression Symptoms 1.296 1.051 ' 1.599 2.424 .015 24 2391 Cl in ica l 1.256 .929 1.698 1.483 .138 4 586 .021 1 .885 Follow-up < 5 years 1.242 1.002 1.540 1.977 .048 15 2200 > 5 years 1.372 1.005 1.873 1.989 .047 13 777 .252 1 . .615 Time of measure < 1 month 1.168 0.785 1.739 0.767 0.44 3 369 >1 month 1.297 1.067 1.577 2.606 .009 25 2608 .235 1 .628 Measures analysis exc in terview) . udes one study (L i to f sky et a l . , 2004), because it combines across 2 levels o f measure (questionnaire and Table 7: Analysis of moderator variables, Mortality (OR) 95% CI Significance Sample Size Homogeneity Moderator OR Lower Upper Z P k N Q df P Type Breast .978 .348 2.754 -.041 .967 4 196 Other 1.647 1.317 2.058 4.380 .000 15 1599 2.467 l .116 Gender Female 1.017 .473 2.189 .044 .965 5 314 Male/Both 1.678 1.324 2.126 4.286 .000 14 1481 2.809 l .094 Measure* Questionnaire 1.576 1.104 2.249 2.505 .012 4 346 Interview 1.524 1.160 1.820 3.250 .001 14 1109 Records 0 1.224 2 .542 Depression Symptoms 1.667 1.194 2.327 3.000 .003 15 1209 Cl in ica l 1.256 .929 1.698 1.483 .138 4 586 2.976 1 .084 Follow-up < 5 years 1.422 1.035 1.954 2.175 .030 11 1315 > 5 years 1.846 1.192 2.861 2.745 .006 9 480 1.026 1 .311 a Measures analysis excludes one study ( L i t o questionnaire) sky et a l . , 2004), because the it combines effects across 2 levels o f measure ( interview, o Table 8: Analysis of moderator variables, Mortality (HR) 95% CI Significance Sample Size Homogeneity Moderator HR Lower Upper Z P k N Q df P Type Breast 1.326 .920 1.912 1.511 .131 -» 20284 Other 1.066 1.006 1.136 2.159 .031 9 2227 1.759 l .185 Gender Female 1.326 .920 1.912 1.511 .131 3 20284 Male/Both 1.066 1.006 1.136 2.159 .031 9 2227 1.759 l .185 Measure" Questionnaire 1.054 1.002 1.109 2.050 .040 7 1945 Interview 1.614 .533 4.887 .847 .397 2 437 Records 1.222 .869 1.717 1.154 .248 2 19903 10.335 .016 * Depression Symptoms 1.054 1.002 1.109 2.050 .040 7 1945 Cl in ica l 1.350 .935 1.948 1.602 .109 4 20340 8.542 2 .014* Follow-upb < 5 years 1.077 1.003 1.157 2.031 .042 5 21093 > 5 years 1.107 .967 1.266 1.476 .140 6 1219 7.848 2 .020 * Adjusted Yes 1.105 1.019 1.199 2.408 .016 9 21391 N o 1.843 .8333 4.080 1.508 .131 -» J 921 o:ooo 1 .993 * p < . 0 5 a Measures analysis excludes one study (Nakay , 2005) . because it combines mul t ip le levels (interview, measure). b F o l l o w - u p length analysis excludes one study (Prieto, 2005), because it combines mult iple levels (less than 5 years, greater than 5 years) Table 9: Summary of total effects, Progression 95 % CI Significance Sample Size Homogeneity Effect size Effect Lower Upper Z P k N P I O R 1.043 .860 1.265 .428 .668 9 1392 .938 1.00 H R 1.038 .999 1.078 1.933 .053 -I j 896 .224 33.091 Table 10: Analysis of moderator variables, Progression (HR) 95% CI Significance Sample Size Homogeneity Moderator O R Lower Upper Z P k TV Q df P Type Breast 1.150 0.810 1.633 0.784 0.433 5 414 Other 0.999 .779 1.282 -0.005 . .996 4 845 .411 i .521 Gender Female 1.150 0.810 1.633 0.784 0.433 5 414 Male/Both 0.999 .779 1.282 -0.005 .996 4 845 .411 l .521 Measure Questionnaire 1.041 0.834 1.300 0.358 0.721 8 1088 Interview 1.082 0.653 1.791 0.306 0.760 1 171 Records 0 .018 l .892 Depression Symptoms 1.041 0.834 1.300 0.358 0.721 8 1088 Clinical 1.082 0.653 1.791 0.306 0.756 1 171 .018 i .892 Follow-up < 5 years 1.053 0.831 1.333 0.427 0.669 5 917 > 5 years 1.034 .694 1.540 .163 .870 4 342 .006 i .938 Table 11: Statistical significance of included studies Outcome Mortality Progression Significant increase 12 0 Not significant 31 14 Significant decrease 1 0 Figure .1: Model of physiological pathways based on Antoni et al., 2006 Depression * Health behaviours E g : Sleep, phys ica l act ivi ty , diet, sexual act ivi ty, substance use, treatment adherence Neuroendocrine Immune response E g . Catecholamines , E g . T-cells, N K cel ls g lucocor t ico ids , prolac t in , oxy toc in Tumour growth E g : Apop tos i s , invas ion , angiogenesis, i m m u n o l o g i c a l escape Metastasis E g : E m b o l i s m , attachment, proliferat ion, angiogenesis, invas ion , migra t ion Remission/progression E g : G r o w t h support for m i n i m a l residual disease, immune survei l lance Figure 2: Search strategy for inclusion of depression-specific articles Potentially relevant articles (mortality) identified from: • MEDLINE n = 645 • Psyclnfo n = 148 Potentially relevant articles (Progression) identified from: • MEDLINE n = 460 • Psyclnfo n = 115 Abstracts reviewed for inclusion n = 50 Abstracts reviewed for inclusion n = 28 Articles meeting full criteria: n = 33 Articles meeting full criteria: n = 10 Articles identified in references: n = 9 Articles identified in references: n = 4 Articles included: n = 42 Articles included n = 14 Figure 3: Forest plot of effect of depression on mortality (OR, conservative) Effect of depression on mortality, conservative, OR S t u d y n a m e S t a t i s t i c s for e a c h s tudy O d d s ra t io a n d 9 5 % C l O d d s L o w e r U p p e r ra t i o l imi t l im i t p -Va lue And rykowsk i 1 0 8 2 0 .347 3 .380 0 .892 B e r e s f o r d et a l . 1.367 0 .395 4 . 7 3 3 0 .622 B u c c h e r i 1 .739 0 .822 3.681 0 148 C o l o n et a l . 2 . 7 8 0 1.313 5 .886 0 .008 de Graef f et a l . 1 .000 0 .609 1.643 1.000 D e r o g a t i s et a l . 0 .227 0 .062 0 .824 0 .024 D e s a i b reas t 0 9 8 2 0 . 2 7 0 3 . 5 6 5 0 .978 D e s a i o the r 1 .170 0 .437 3 .135 0 . 7 5 5 E d w a r d s 1.000 0 .227 4 .404 1 0 0 0 E h l e r s 2 . 6 6 0 1.387 5 .099 0 .003 F a l l e r et a l . 1 .000 0.491 2 .036 1.000 G i l b a r e t a l . 3 .576 0 .853 14 .984 0.081 G r e e r et a l . 1 .000 0 .380 2.631 1.000 H i s l o p et a l . 1.000 0 .536 1.866 1.000 J a m i s o n et a l . 1 .230 0 .445 3.401 0 . 6 8 9 L e i g h et a l . 0 . 808 0 .389 1.678 0 .568 L i t o f kye t a l . 2 0 0 4 1.318 0 .932 1.864 0 .118 L o b e r i z a et a l . 1.854 1.099 3 .129 0.021 Ma in io et a l . 2 0 0 5 (A) 3 1 . 7 7 6 0 .986 1024 .477 0.051 Ma in io et a l . 2 0 0 5 (B) 2 . 3 1 4 0 .230 2 3 . 2 5 8 0 .476 M a i n i o et a l . 2 0 0 6 (A) 3 .795 0 .483 2 9 . 7 9 9 0 .205 M a i n i o et a l . 2 0 0 6 (B) 0.571 0 .025 1 3 . 0 1 5 0 .725 Morris et a l . 1 0 0 0 0 .542 1.844 1.000 Murphy et a l . 1.057 0 .398 2 .807 0 9 1 2 P a l m e r et a l . 1.000 0.621 1.612 1.000 S c h u i z et a l . 2 . 2 0 8 1.309 3 .727 0 .003 S t e i n et a l . 1 .000 0 .467 2 . 1 4 3 1.000 V t g a n o et a l . 1.000 0 .622 1.608 1.000 1.281 1.077 1.523 0 .005 D e c r e a s e Increase Figure 4: Funnel plot corresponding to mortality (OR, conservative) Funnel Plot of Standard Error by Log odds ratio - 4 - 3 - 2 - 1 0 1 2 3 4 Log odds ratio 58 Figure 5: Forest plot of effect of depression on mortality (OR) Effect of d e p r e s s i o n on mortal i ty , O R S t u d y n a m e S ta t is t i cs for e a c h s tudy O d d s ratio a n d 9 5 % Cf O d d s L o w e r U p p e r rat io l imit limit p - V a l u e A n d r y k o w s k i 1 0 3 2 0 . 3 4 7 3 3 8 0 0 . 8 9 2 B e n e s f o r d e t a l . 1 3 6 7 0 . 3 9 5 4 7 3 3 0 . 6 2 2 B u c c h e r i 1 7 3 9 0 . 8 2 2 3 6 8 1 0 . 1 4 8 C o l o n e t a l . 2 7 8 0 1 . 3 1 3 5 8 8 6 0 . 0 0 8 D e r o g a t t s e t a l . 0 2 2 7 0 . 0 6 2 0 8 2 4 0 . 0 2 4 D e s a i b r e a s t 0 9 8 2 0 . 2 7 0 3 5 6 5 0 . 9 7 8 D e s a i o t h e r 1 1 7 0 0 . 4 3 7 3 1 3 5 0 . 7 5 5 E h i e r s 2 6 6 0 1 . 3 8 7 5 0 9 9 0 . 0 0 3 G i l b a r e t a ! . 3 5 7 6 0 8 5 3 14 9 8 4 0 . 0 8 1 J a m i s o n e t a l . 1 2 3 0 0 . 4 4 5 3 4 0 1 0 . 6 8 9 L e i g h e t a l . 0 8 0 8 0 . 3 8 9 1 6 7 8 0 . 5 6 8 L i t o f k y e t a l . 2 0 0 4 1 3 1 8 0 . 9 3 2 1 8 6 4 0 . 1 1 8 L o b e r i z a e t a l . 1 8 5 4 1 . 0 9 9 3 1 2 9 0 . 0 2 1 M a i n i o e t a l . 2 0 0 5 (A) 31 7 7 6 0 . 9 8 6 1 0 2 4 4 7 7 0 . 0 5 1 M a i n i o e t a l . 2 0 0 5 (B) 2 3 1 4 0 . 2 3 0 2 3 2 5 8 0 . 4 7 6 M a i n i o e t a l . 2 0 0 6 (A) 3 7 9 5 0 4 8 3 2 9 7 9 9 0 . 2 0 5 M a i n i o e t a l . 2 0 0 6 ( B ) 0 5 7 1 0 . 0 2 5 1 3 0 1 5 0 . 7 2 5 M u r p h y e t a ! . 1 0 5 7 0 . 3 9 8 2 8 0 7 0 . 9 1 2 S c h u l z e t a l . 2 2 0 8 1 . 3 0 9 3 7 2 7 0 . 0 0 3 1 5 3 6 . 1 . 1 9 0 1 9 8 2 0 . 0 0 1 0.1 0.2 0.6 1 2 5 10 D e c r e a s e I n c r e a s e Figure 6: Forest plot of effect of depression on mortality (RR) Effect of d e p r e s s i o n on mortality, RR Study name Statistics for each study Risk ratio and 95% CI Risk Lower U pper ratio lim it lim it p-Value B r o e r s et a l . 0.990 0.960 1.020 0.516 H jer l et a l . 1.310 1.121 1.531 0.001 R i c h a r d s o n et a l . 2.150 0.728 6.351 0.166 R i n g d a l et a l . 1.160 1.060 1.270 0.001 1.146 0.976 1.345 0.095 0.1 0.2 0.5 1 2 5 10 Decrease Increase Figure 7: Forest plot of effect of depression on mortality (HR) Effect of d e p r e s s i o n on mortality, HR S t u d y n a m e S t a t i s t i c s f o r e a c h s t u d y H a z a r d r a t i o a n d 9 5 % C l H a z a r d L o w e r U p p e r r a t i o l i m i t l i m i t p - V a l ue B r o w n et a ! . 1 .030 1 . 0 0 0 1 . 0 6 0 0 . 0 4 7 C h a n g e t a l 1 .036 0 . 9 9 7 1 . 0 7 6 0 . 0 6 9 F a l l e r a n d S c h m i d t 1 .050 0 . 9 7 8 1 . 1 2 7 0 . 1 7 9 G o o d w i n e t a l 1 .420 1 .1 28 1.787 0 . 0 0 3 L e h t o et a l . 2 0 0 7 1.360 0 . 9 8 2 1 .884 0 . 0 6 4 N a k a y a e t a l . 1 .755 0 . 7 4 5 4 . 1 3 7 0 . 1 9 9 O s b o r n e et a l . 1 .020 0 . 8 9 5 1 . 1 6 3 0 . 7 6 7 P r i e t o et a l . 2 . 6 3 2 1.351 5 . 1 2 7 0 , 0 0 4 S a i t o - N a k a y a 0 . 8 4 0 0 . 2 8 5 2 . 4 7 3 0 . 7 5 2 S h e i b a n i - R a d 1 .000 0 . 7 0 7 1.414 1 . 0 0 0 S t o m m e l e t a ! . 1 .660 1 .161 2 . 3 7 3 0 . 0 0 5 W a t s o n e t a l . 9 9 / 2 0 0 5 2 . 9 5 4 1.161 7.51 1 0 . 0 2 3 1 .095 1 .027 1 . 1 6 7 0 . 0 0 6 0.1 0.2 0.5 1 2 5 10 Decrease Increase ON Figure 8: Funnel plot corresponding to mortality (HR) Log hazard ratio 62 Figure 9: Forest plot of effect of depression on progression (OR) Effect of d e p r e s s i o n on p r o g r e s s i o n , O R S t u d y n a m e S t a t i s t i c s f o r each s t u d y O d d s ra t io a n d 95% CI O d d s L o w e r U p p e r rat io l imi t l imi t p - V a l u e Be rgenmar et a l , 2004 0.944 0.670 1.329 0.741 D e Brabander and Ger i t s 2.855 0.780 10.451 0.113 D e Graef f e t a l . 1.000 0.604 1.655 1.000 G i lba r 1.000 0.311 3.218 1.000 G i ra ld i et a l . 1.122 0.534 2.355 0.762 G r a h a m e t a l . 1.082 0.653 1.791 0.760 G r e e r et a l . 1.000 0.380 2.631 1.000 H i s l o p e t a l . 1.000 0.536 1.866 1.000 M o r r i s et a l . 1.000 0.542 1.844 1.000 Palapattu et a l . 1.664 0.596 4.648 0.331 1.043 0.860 1.265 0.668 1 0.2 0.5 1 2 5 10 Decrease Increase Figure 10: Forest plot of effect of depression on progression (HR) Effect of d e p r e s s i o n on p r o g r e s s i o n , HR Study name Statistics for each study Hazard ratio and 95% CI H azard Lower U pper ratio limit limit p-Value Barraclough et al. 1.260 0.488 3.250 0.633 Chang et al. 1.036 0.997 1.076 0.069 Watson etal. 99/05 1.740 0.700 4.328 0.233 Watson etal. 99/05 2.020 0.727 5.615 0.178 1.046 0.964 1.135 0.277 0.1 0.2 0.5 1 2 5 10 Decrease Increase 

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:
https://iiif.library.ubc.ca/presentation/dsp.831.1-0101214/manifest

Comment

Related Items