Skip to content

Advertisement

  • Research
  • Open Access
  • Open Peer Review

A case control study investigating the effects of levels of physical activity at work as a risk factor for prostate cancer

  • 1, 5Email author,
  • 1,
  • 1, 2, 3,
  • 2, 3 and
  • 4
Environmental Health201413:64

https://doi.org/10.1186/1476-069X-13-64

  • Received: 14 January 2014
  • Accepted: 25 July 2014
  • Published:
Open Peer Review reports

Abstract

Background

A potential risk factor for prostate cancer is occupational physical activity. The occupational aetiology of prostate cancer remains unclear. The purpose of this research was to examine associations between the level of exposure to various measures of physical activity at work and the risk of Prostate Cancer.

Methods

Using the Finnish Job Exposure Matrix and the occupational history of 1,436 cases and 1,349 matched controls from an Australian case control study; we investigated five related exposure variables considered to be risk factors by comparing odds ratios.

Results

Modestly increasing odds ratios were detected with increasing levels of workload but there was no difference in this trend between moderate and high grade tumours. In regard to occupational physical workload no statistically significant association was observed overall but an increasing trend with level of exposure was observed for high grade compared with moderate grade tumours.

Conclusion

Both workload and physical workload merit further investigation, particularly for the latter in relation to grade of tumour.

Keywords

  • Manual handling of burdens
  • Occupational exposure
  • Physical activity
  • Physical workloads
  • Prostate cancer
  • Risk factors
  • Finish job exposure matrix

Background

Many past studies have investigated various occupational chemical and physical agents as likely causes of prostate cancer [1]. When investigating the causes of death after the diagnosis of prostate cancer it has also been previously found that men with low to moderate grade prostate cancer had a similar rate of death to men without prostate cancer [2]. There are very few well established risk factors of prostate cancer especially those that are potentially modifiable risk factors [3]. Therefore the rationale for this study is to investigate the likely association of some modifiable occupational risk factors and prostate cancer. Previous reported studies investigating the role that physical activity plays in the occupational environment, have described physical activity by various metrics [4, 5]. Ricciardi provided a model for the concept of Sedentarism that included attributes such as expending less than 10% daily energy in the performance of moderate and high-intensity activities in which the metabolic rate increases at least four times from baseline, or not engaging in physical activities five or more times per week or no leisure activity or no physical activity for up to 3hrs per week that increases the metabolic rate by four times from base [4]. In relation to Leisure Time Physical Activity (LTPA), Kirk found that those employed in occupations demanding long work hours and low Occupational Physical Activity (OPA) are at higher risk of inactivity.

Some authors [6] have demonstrated that men who participated in regular LTPA reduced their risk for clinical prostate cancer and in the workplace concluded that physical activity at work was also beneficial in reducing the risk of prostate cancer [710]. However, Bairati et al. suggested that socio-economic status was a probable confounder [10]. In two systematic reviews, it was found that few studies demonstrate a protective role of OPA, even though high levels of OPA and LTPA together seemed to reduce the risk of advanced prostate cancers [11, 12]. A further review concluded there was inconsistent evidence for an inverse association between OPA and prostate cancer [13]. Recently, a meta-analysis [14] found no clear evidence for an association between job strain and the risk of prostate cancer in relation to OPA. Discacciati et al., adds another dimension to the overall picture by concluding that obesity may have a dual effect on PCa by a decreased risk of low grade PCa and an increased risk of high grade PCa [15].

In contra distinction, a positive association was reported [3] between prostate cancer risk and the highest category of workplace physical activity, which is the opposite of what has been reported by most other studies [710] of physical activity and prostate cancer.

Our aim was to investigate whether an association existed between occupational physical activity exposures (assessed using FINJEM ) and prostate cancer, and, in order to address issues of possible detection bias, also to inspect whether such associations differed by grade of tumour. Occupational studies using job exposure matrices (JEMs) have reported some associations with prostate cancer risk, but these have not consistently been replicated by other studies [16]. Although leisure time physical activity may be a limitation and potential cause of bias due to misclassification, there is no reason to suggest that the profile of the cases are different to the controls. This article specifically discusses the reported OPA in men in relation to prostate cancer using the demographic profile of the sample and the odds ratios found in relation to five exposure variables (that are possible risk factors) measuring different forms of workplace physical activity, rather than LTPA. In this study we have used the term ‘occupational exposure’ to include both ergonomic and psychological variables.

Methods

Giles et al. [17] has reported on the Australian Prostate Cancer study elsewhere. Briefly, population based cancer registries in Melbourne, Sydney and Perth were utilized to recruit a random sample of 2,528 cases with prostate cancer diagnosed at age 39–80 and 3,125 controls which were considered eligible at the time of selection. For the purposes of this analysis the number of participants was reduced due to factors such as no access to patient records, refusal of controls, insufficient English skills, or moved address. Further analysis was restricted to 1,495 (65%) cases with prostate cancer diagnosed at age 39–70 and 1423 (46%) controls aged between 40 and 70 years. The final analysis for which there was sufficient information regarding occupational work histories included 1,436 cases (96%) and 1,349 (94%) controls, aged between 39 and 70 years.

Participants were also asked to complete a Lifetime Calender of residence and employment in order to prompt more complete answers when responding to the study questionnaires. The controls were matched through frequency based matching with age and were free of prostate cancer upon recruitment. Recruitment was stratified by age and all men under the age of 60 years were invited to participate. Initially, random samples of 50% of men aged 60–64 and 25% of men aged 65–70 were selected, with the proportions varying overtime to fit interview quotas.

Cases recruited in Melbourne, Sydney and Perth, Australia were diagnosed in the study period and notified to the population-based cancer registries with a histopathologically-confirmed diagnosis of adenocarcenoma of the prostate, and excluded tumours that were well-differentiated (defined as low grade tumours, that is, those with a Gleason score of less than five).

A major concern with prostate cancer is the diagnostic staging and whether any occupational exposures are associated with medium or high grade cancers. One approach to overcoming concerns regarding the inclusion of clinically unimportant tumours as cases is to select cases who are diagnosed in the study period and notified to the population-based cancer registries with a histopathologically confirmed diagnosis of adenocarcenoma of the prostate, excluding tumours that are well-differentiated (defined as low grade tumours i.e. those with a Gleason score less than five). This has been addressed in this study.

The self-reported data from the calendar and questionnaires that related to occupation were collated together with date-of-birth, location, children and their gender, and school/occupation and linked with other clinical data variables from study such as smoking, alcohol consumption and physical activity at work. This data was has been shown to be both valid and reliable in other analysis [17].

This further analysis of Giles et al. original study [17] was undertaken using de-identified data and is covered by the AVCC Institutional Ethics Committee permission (1992) from the Anti-Cancer Council of Victoria, and permission from the Chief Investigator of the Risk Factors for Prostate Cancer Case–control Study (2004).

Exposure assessment

For exposure assessment we used FINJEM [18], a community-based job exposure matrix, originally developed by the Finnish Institute of Occupational Health, for use in epidemiological studies. FINJEM covers a wide range of physical, chemical, microbiological, ergonomic and psychological exposures and is the only job exposure matrix that covers all the different types of radiation. FINJEM is coded for 311 classes of occupation, according to the Finnish occupation coding classification. The exposure is measured in line with the method described by Kauppinen at al. [19]. Each job or employment episode of the participants in this study was coded according to the Finnish occupation coding classification and a FINJEM code number (O-Code) was allocated. The coding facilitated the linkage between the occupational activity exposures and prostate cancer status. As an example of this linkage please see Table 1 for a list of the three top occupations with the highest levels of exposures for each exposure variable. It is noted that this list should be treated cautiously as quite a few men had more than one occupation during the course of their working life.
Table 1

A list of top three occupations with the highest levels of exposures for each exposure variable for men with high grade tumours

Manual handling of burdens

Sedentary work

Physical workload

Total cumulative activity-over-time

Work load

Cabinet maker

Taxi driver

Meat worker

Meat worker

Editor

Meat worker

Truck driver

Cabinet maker

Sailor

Travel consultant

Market gardener

Draftsman

Cleaner

Policeman

Ships officer

Each exposure variable has a specific definition and value and exposure is characterised by the proportion of exposed workers (P) and the mean level of exposure (L) and is given as P × L for each occupation. Cumulative exposure was calculated by P × L × Years exposed in the various exposed jobs reported by the participant. Occupational exposures that did not exceed the non-occupational background level were omitted for example, background radiation levels).

The cumulative exposure for the exposed participants was calculated in tertiles and quartiles. In the model used it included occupational exposure variable plus age, family history and the SEIFA index of economic resources, which is a measure of socio-economic disadvantage, first produced by the Australian Bureau of Statistics [20] following the 1971 census. Only occupational cumulative exposures from our model are presented in the results as it takes into account the level of economic and social disadvantage within the sample, as well as age and family history confounders.

The exposure variables

We investigated three exposure variables; manual handling of burdens, physical workloads, sedentary work, and two created variables of cumulative activity-over-time and workload by comparing odds ratios in tertiles and quartiles through analysis by binary logistic regression. Moderate and high grade tumours were compared using polytomous regression. The manual handling of burdens consists of lifting, and carrying of heavy burdens, and is an essential feature of the everyday work tasks. Physical workloads consist of tasks where the whole body is exerted by dynamic muscular work. Sedentary work consists of work done in seated posture [19]. Also, one of the variables were created, total cumulative activity-over-time is calculated by adding an individual’s total scores over their disclosed working life so that some comparison could be undertaken in regard to working in high to low activity jobs for a prolonged period. The psychological exposure variable Workload is a measurement of the overall psychological impact of perceived occupational load over the years of employment. If a subject considers they have been stressed from a high workload over the majority of their working history, it might indicate that, as a stressor, this could have a long-term harmful outcome. Workload is defined as a psychological factor in FINJEM [19] and is derived from the demand to work under tight schedules and time pressure, and to adjust conflicting demands from others subjective perceptions.

Results

The profile of the sample in Table 2 shows the ages of the participants were relatively evenly spread between cases and controls. In the 65–70 year age group, this group was slightly larger and consistent with the expected occurrence and diagnosis with the control group having the greater number of participants below the age of 55 years.
Table 2

Demographic description of characteristics of cases and controls

Variable

Cases (n = 1436)

No. of controls (n = 1349)

Age group

n

%

n

%

 39- 55

296

20.6

329

24.4

 55 - 59

328

22.8

217

16.1

 60 - 64

359

25.0

398

29.5

 65 - 70

453

31.5

405

30.0

Country of birth

    

 Australia

988

68.8

816

60.5

 Not-Australia

448

31.2

533

39.5

Educational level

    

 Primary only

97

6.8

135

10.0

 Secondary only

459

32.1

426

31.7

 Post-secondary training

633

44.2

581

43.2

 Tertiary

243

17.0

202

15.0

Family history

    

 No first degree relative affected

1180

82.6

1257

93.8

 At least one first degree relative affected

249

17.4

83

6.2

Marital status

    

 Married/de facto

1230

85.8

1140

85.2

 Once married

159

11.1

142

10.8

 Never married

45

3.1

65

4.0

State

    

 New South Wales

419

29.2

319

23.6

 Victoria

767

53.4

781

57.9

 Western Australia

250

17.4

249

18.5

There were ten percent more Australian born cases than controls. Educationally both groups were closely matched, but numerically the control group had a higher number of men with lower educational attainment. In regard to family history, cases had an 11.2% greater difference of at least one first degree relative being affected by prostate cancer. Marital status had a similar spread in both groups.

Table 3 shows the results of the binary logistic regression for the five exposure variables, manual handling of burdens, sedentary work, workload, cumulative activity-over-time and physical workloads. None of the three ergonomic factors, manual handling of burdens, physical workload and sedentary work were associated with prostate cancer risk, nor was the calculated variable of cumulative activity-over-time. The psychological variable of workload which measures the worker’s perceptions in relation to an occupational lifetime of high workload activity was the only statistically significant association and it showed a positive relationship with PCa risk.
Table 3

Cumulative occupational exposures for prostate cancer by binary logistic regression

Binary logistic regression

Exposure total

Exposure

Cumulative exposure

Controls

Cases

OR*

95% CI

p for trend (unadjusted model)

Manual handling of burdens (score)

1st Tertile

0

- ≤ 2.654

472

534

1.00

  

2nd Tertile

> 2.655

- ≤ 6.808

440

443

0.96

0.79 – 1.15

3rd Tertile

> 6.808

 

437

459

1.01

0.84 – 1.22

Continuous

  

1349

1436

  

0.82

Sedentary work (score)

Unexposed

0

 

843

912

1.00

  

1st Tertile

>0

– ≤ 1.6

158

177

0.96

0.78 – 1.25

2nd Tertile

>1.6

- ≤ 5.108

180

161

0.78

0.62- 0.99

3rd Tertile

>5.108

 

168

186

1.02

0.81- 1.29

Continuous

 

1349

1436

  

0.94

Work load (score)

1st Tertile

> 0

- ≤ 102

426

366

1.00

  

2nd Tertile

> 103

- ≤ 127

466

496

1.20

0.99 – 1.46

3rd Tertile

> 128

 

457

574

1.34

1.09 – 1.65

Continuous

 

1349

1436

  

0.001

Total cumulative activity (score)

1st Tertile

> 0

- ≤ 69

442

446

1.00

  

2nd Tertile

> 69

- ≤ 98

455

496

1.16

0.95 – 1.40

 

3rd Tertile

> 98

 

452

494

1.19

0.97 – 1.46

 

Continuous

 

1349

1436

  

0.11

Physical workload (score)

1st Tertile

0

- ≤ 2.656

456

486

1.00

  

2nd Tertile

> 2.656

- ≤ 7.132

448

463

1.06

0.87 – 1.28

 

3rd Tertile

> 7.132

 

445

487

1.15

0.95 -1.40

 

Continuous

  

1349

1436

  

0.13

*Odds ratios (OR) and associated 95% confidence intervals (95% CI) adjusted for Age, Family History and SEIFA Index of Economic Resources.

Table 4 describes the results of the Polytomous Logistic Regression comparing the associations between moderate and high grade prostate cancers for each of the exposures. No associations were observed for manual handling of Burdens or for Sedentary Work for either moderate or high grade prostate tumour risk.
Table 4

Cumulative occupational exposures for prostate cancer by polytomous logistic regression

Polytomous logistic regression

Exposure total

Exposure

Cumulative exposure

Moderate grade

High grade

p –value {heterogeneity

 

Cases

OR*

95% CI

Cases

OR*

95% CI

 

Manual handling of burdens (score)

1st Tertile

0

- ≤ 2.654

446

1.00

 

88

1.00

  

2nd Tertile

> 2.655

- ≤ 6.808

360

0.93

0.76 – 1.12

83

1.03

0.74 – 1.44

 

3rd Tertile

> 6.808

 

372

1.01

0.83 – 1.22

87

1.08

0.77 – 1.50

0.81

Sedentary work (score)

Unexposed

0

 

750

1.00

 

162

1.00

  

1st Tertile

> 0

- ≤ 1.6

153

1.04

0.81 – 1.31

24

0.77

0.48 – 1.22

 

2nd Tertile

>1.6

- ≤ 5.108

132

0.81

0.64 – 1.04

29

0.82

0.54 – 1.26

 

3rd Tertile

>5.108

 

143

0.98

0.77 – 1.29

43

1.30

0.89 – 1.89

0.21

Work load (score)

1st Tertile

> 0

- ≤ 102

312

1.00

 

54

1.00

  

2nd Tertile

> 103

- ≤ 127

409

1.21

0.99 – 1.48

87

1.36

0.94 – 1.96

 

3rd Tertile

> 128

 

457

1.35

1.09 – 1.57

117

1.66

1.14 – 2.41

0.57

Total cumulative activity (score)

1st Tertile

> 0

- ≤ 69

362

1.00

 

84

1.00

  

2nd Tertile

> 69

- ≤ 98

422

1.23

1.01 – 1.50

74

0.82

0.58 – 1.16

 

3rd Tertile

> 98

 

394

1.20

0.98 – 1.46

100

1.05

0.75 – 1.48

0.06

Physical workload (score)

1st Tertile

0

- ≤ 2.656

414

1.00

 

72

1.00

  

2nd Tertile

> 2.656

- ≤ 7.132

360

0.96

0.79 -1.17

103

1.46

1.05 – 2.04

 

3rd Tertile

> 7.132

 

404

1.14

0.94 -1.38

83

1.22

0.86 - 1.74

0.03

* Odds ratios (OR) and associated 95% confidence intervals (95% CI) adjusted for Age, Family History and SEIFA Index of Economic Resources.

For total cumulative activity a significant trend in increasing risk was observed for moderate grade but not high grade tumours (heterogeneity p = 0.06). For workload, both moderate and high grade tumours were positively associated with increasing exposure but were not significantly different in this regard. For physical workload, a statistically significant trend was observed with increasing levels of exposure but only for high grade tumour risk (heterogeneity p = 0.03).

In this study there were 16,331 reported jobs, providing good variability in job histories for application of FINJEM [21]. The occupational exposure OR’s across all of the variables did not vary substantially from the adjustment models for age, family history and SEIFA Index of Economic Resources in both the binary and polytomous logistic regressions.

Discussion

This study has found that workload is modestly associated with an increased risk of prostate cancer and for both moderate and high grade tumours. This is at odds with the findings of Heikkilä et al. [14] that work related psychological stress is unlikely to be an important factor for prostate cancer. We also found cumulative activity-over-time to be modestly associated with prostate cancer risk and showed a small trend with the moderate grade tumours. However, total cumulative activity-over-time appears to have a stronger association with the higher grade tumours. These results suggest that the greater the physical activity in the work place over a long period of time the greater the likelihood of the development of high grade prostate cancer.

In comparison with other studies, our finding regarding total cumulative activity-over-time is contrary to the findings of Bairati [10] where physical activity in the job had an inverse relationship with prostate cancer and they concluded that physical activity was beneficial. Bairati also found that sedentary/light work had a positive association with prostate cancer whereas our findings showed no associations or trends in regard to prostate cancer. The current analysis also does not support the earlier findings by Ricciardi [4] and Kirk & Rhodes [5] in relation to the combination of OPA and LPTA reduce the risk of advanced prostate cancers. It should be noted that Bairati’s study did not use a Job Exposure Matrix, but instead coded the data related to occupational activity using the five levels of physical activity described by the US Department of Labor.

The major strengths of our study are its sample size and the stratification of the subjects to reflect the population of men with and without prostate cancer. With the three main sites we are confident that the study may be generalizable to the population of men in Australia [21].

In Australia, the treatment of Prostate Cancer is very limited outside state capitals, so our sampling frame is unlikely to be a limitation and has not been compromised due to unrepresentative case ascertainment, even though we recruited cases and controls from three major metropolitan centres of the three selected states. The use of FINJEM in Australia has previously been found to be acceptable for various exposures when compared with expert assessment [18].

The principal weakness of our study is the use of a Job Exposure Matrices (JEMs) that can lead to non-differential misclassification of exposure [22], and we would expect non-differential misclassification to have occurred. This will almost certainly have led to a bias towards null-effect (OR = 1). Multiple hypothesis testing may produce chance positive results, but we did not find anything significant in this regard. The advantage of using a JEM is that it does not rely on self-reported exposure by subjects potentially leading to differential exposure bias from then cases, which is particularly important for more subjective exposure indices such as workload. Being an objective measure of OPA it overcomes the problem of criterion validity of questionnaires [23]. A second limitation is in not having access to BMI’s for the cases and controls, in order to confirm or deny other researchers conclusions [15].

Finally, there would seem to be two contradictory results. Firstly that manual handling of burdens which displayed a slight trend with high grade tumours did not show an association with total prostate cancer. However, Physical workloads did show a small association with total prostate cancer but no association with either moderate or high grade tumours. Therefore there is insufficient evidence to support any causal relationship and it must be concluded that OPA is unlikely to be beneficial in relation to protecting against PCa, even though other studies have suggested that the combination of OPA and LTPA strongly reduced risk [11, 12].

Conclusions

Our findings are in line with other authors recommendations that suggest further research might be merited in regard to workload and physical workload and prostate cancer risk. We recognize, however, that our findings may point to another more definable psychological agent related to stress in the workplace. Given the modest nature of the associations we describe, we provide little evidence to support any causal relationship and conclude that OPA is not proven to be beneficial in relation to protecting against prostate cancer.

Consent

A written informed consent was obtained from all participating participants in the original study and this further analysis was undertaken on de-identified data.

Declarations

Funding statement

The authors received no financial support for the research and/or authorship of this article. The original study of which this further analysis is based was funded by grants from the National Health and Medical Research Council (940394, 991129) and was further supported by funding from Tattersall’s and The Whitten Foundation as well as infrastructure provided by The Cancer Council Victoria.

Authors’ Affiliations

(1)
Department of Epidemiology & Preventive Medicine, Monash University, The Alfred Centre, The Alfred, Commercial Road, Melbourne, Victoria, 3004, Australia
(2)
Cancer Epidemiology Centre, Cancer Council Victoria, 615 St. Kilda Road, Melbourne, Victoria, 3004, Australia
(3)
Centre for Genetic Epidemiology, University of Melbourne, 200 Berkeley Street, Carlton, Victoria, 3053, Australia
(4)
Finnish Institute of Occupational Health, Topeliuksenkatu 41aA, FIN, 00250, Helsinki, Finland
(5)
P.O. Box 276, Trafalgar, Victoria, 3824, Australia

References

  1. Ilic M, Vlajinac H, Marinkovic J: Case–control study of risk factors for prostate cancer. Br J Cancer. 1996, 74: 1682-1686.View ArticleGoogle Scholar
  2. Ketchandji M, Kuo YF, Shahinian V, Goodwin JS: Cause of death in older men after the diagnosis of prostate cancer. J Am Geriatr Soc. 2009, 57: 24-30.View ArticleGoogle Scholar
  3. Sass-Kortsak AM, Purdham JT, Kreiger N, Darlington G, Lightfoot NE, Sass K: Occupational risk factors for prostate cancer. Am J Ind Med. 2007, 50: 568-576.View ArticleGoogle Scholar
  4. Ricciardi R: Sedentarism: a concept analysis. Nurs Forum. 2005, 40: 79-87.View ArticleGoogle Scholar
  5. Kirk MA, Rhodes RE: Occupation correlates of adults’ participation in leisure-time physical activity: a systematic review. Am J Prev Med. 2011, 40 (4): 476-485.View ArticleGoogle Scholar
  6. Barnard R, Shan-Leung P, Aronson WJ, Cohen P, Golding LA: A mechanism to explain how regular exercise might reduce the risk for clinical prostate cancer. Eur J Cancer Prev. 2007, 16: 415-421.View ArticleGoogle Scholar
  7. Lagiou A, Samoli E, Georgila C, Minaki P, Barbouni A, Tzonou A, Trichopoulos D, Lagiou P: Occupational physical activity in relation with prostate cancer and benign prostatic hyperplasia. Eur J Cancer Prev. 2008, 17: 336-339.View ArticleGoogle Scholar
  8. Strom SS, Yamamura Y, Flores-Sandoval FN, Pettaway CA, Lopez DS: Prostate cancer in Mexican-Americans: identification of risk factors. Prostate. 2008, 68: 563-570.View ArticleGoogle Scholar
  9. Krishnadasan A, Kennedy N, Zhao Y, Morgenstern H, Ritz B: Nested case–control study of occupational physical activity and prostate cancer among workers using a job exposure matrix. Cancer Causes Control. 2008, 19: 107-114.View ArticleGoogle Scholar
  10. Bairati I, Larouche R, Meyer F, Moore L, Fradet Y: Lifetime occupational physical activity and incidental prostate cancer (Canada). Cancer Causes Control. 2000, 11: 759-764.View ArticleGoogle Scholar
  11. Anzuini F, Battistella A, Izzotti A: A Physical activity and cancer prevention: a review of current evidence and biological mechanisms. J Prev Med Hyg. 2011, 52 (4): 174-180.Google Scholar
  12. Liu YP, Hu FL, Li DD, Wong F, ZHU L, Chen WY, Ge J, An R, Zhao Y: Does physical activity reduce the risk of prostate cancer? A systematic review and meta-analysis. Eur Urol. 2001, 60: 1029-1044.View ArticleGoogle Scholar
  13. Doolan G, Benke G, Giles GG: An update on occupation and prostate cancer. Asian Pac J Cancer Prev. 2014, 15: 501-516.View ArticleGoogle Scholar
  14. Heikkilä K, Nyberg S, Thorell T, Alfredsson L, Bjorner JB, Bonenfant S, Borritz M, Bouillon K, Burr H, Dragano N, Geuskens GA, Goldberg M, Hamer M, Hooftman WE, Houtman IL, Joensuu M, Knutsson A, Koskenvuo M, Koskinen A, Kouvonen A, Madsen IE, Magnusson Hanson LL, Marmot MG, Nielsen ML, Nordin M, Oksanen T, Pentti J, Salo P, Rugulies R, Steptoe A: Work stress and risk of cancer: meta-analysis of 5700 incident cancer events in 116,000 European men and women. BMJ. 2013, 346: 165-View ArticleGoogle Scholar
  15. Discacciati A, Orsini N, Wolk A: Body mass index and incidence of localized and advanced prostate cancer–a dose–response meta-analysis of prospective studies. Ann Oncol. 2012, 23 (7): 1665-1671.View ArticleGoogle Scholar
  16. Weston TL, Aronson KJ, Siemiatycki J, Howe GR, Nadon L: Cancer mortality among males in relation to exposures assessed through a job-exposure matrix. Int J Occup Environ Health. 2000, 6: 194-202.View ArticleGoogle Scholar
  17. Giles GG, Severi G, McCready MR, English DR, Johnson W, Hopper JL, Boyle P: Smoking and prostate cancer: findings from an Australian case–control study. Ann Oncol. 2001, 12: 761-765.View ArticleGoogle Scholar
  18. Benke G, Sim M, Fritschi L, Aldred G, Forbes A, Kauppinen T: Comparison of occupational exposure using three different methods: hygiene panel, job exposure matrix (JEM), and self reports. Appl Occup Environ Hyg. 2001, 16: 84-91.View ArticleGoogle Scholar
  19. Kauppinen T, Toikkanen J, Pukkala E: From cross-tabulations to multipurpose exposure information systems: a new job-exposure matrix. Am J Ind Med. 1998, 33: 409-417.View ArticleGoogle Scholar
  20. ABS: An Introduction to Socio-Economic Indexes for Areas (SEIFA) Australia 2006 (Preliminary). Australian Bureau of Statistics. 2006, Canberra, Australia: Commonwealth of AustraliaGoogle Scholar
  21. Breslow NE, Day NE: Statistical methods in cancer research: Vol II - The design and anallysis of cohort studies. 1987, Lyon, France: IARC Scientific Publications, 82-297.Google Scholar
  22. Bouyer J, Dardenne J, Hemon D: Performance of odds ratios obtained with a job-exposure matrix and individual exposure assessment with special reference to misclassification errors. Scand J Work Environ Health. 1995, 21: 265-271.View ArticleGoogle Scholar
  23. Kwak L, Proper KI, Hagstömer M, Sjöström M: The repeatability and validity of questionnaires assessing occupational physical activity-a systematic review. Scand J Work Environ Health. 2011, 37 (1): 6-29.View ArticleGoogle Scholar

Copyright

Comments

By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate. Please note that comments may be removed without notice if they are flagged by another user or do not comply with our community guidelines.

Advertisement