Physical activity and risk of prostate cancer in the European Prospective Investigation into Cancer and Nutrition (EPIC) cohort

Abstract

The evidence concerning the possible association between physical activity and the risk of prostate cancer is inconsistent and additional data are needed. We examined the association between risk of prostate cancer and physical activity at work and in leisure time in the European Prospective Investigation into Cancer and Nutrition (EPIC) cohort. In our study, including 127,923 men aged 20–97 years from 8 European countries, 2,458 cases of prostate cancer were identified during 8.5 years of followup. Using the Cox proportional hazards model, we investigated the associations between prostate cancer incidence rate and occupational activity and leisure time activity in terms of participation in sports, cycling, walking and gardening; a metabolic equivalent (MET) score based on weekly time spent on the 4 activities; and a physical activity index. MET hours per week of leisure time activity, higher score in the physical activity index, participation in any of the 4 leisure time activities, and the number of leisure time activities in which the participants were active were not associated with prostate cancer incidence. However, higher level of occupational physical activity was associated with lower risk of advanced stage prostate cancer (p_trend = 0.024). In conclusion, our data support the hypothesis of an inverse association between advanced prostate cancer risk and occupational physical activity, but we found no support for an association between prostate cancer risk and leisure time physical activity. © 2009 UICC

Prostate cancer is a leading cause of morbidity and mortality in men,1 but the causes of prostate cancer are essentially unknown and identification of modifiable risk factors is urgently needed. Physical activity might protect against prostate cancer by lowering the level of testosterone and IGF-1 and improving energy balance, immune function and antioxidative defense systems.2 However, the epidemiological literature on physical activity and prostate cancer is inconsistent, and in a new report from World Cancer Research Fund and American Institute for Cancer Research the expert panel concludes that the evidence for an association between physical activity and prostate cancer is too limited for a formal judgment.3

Of the 22 published cohort studies, 9 found an inverse association of prostate cancer risk with cardiorespiratory fitness,4 occupational activity,5, 6 leisure time activity7–11 or both leisure time activity and occupational activity.12 One study found an increased risk of prostate cancer among college athletes13 and 12 studies found no significant association.14–25 However, the studies have been impeded by difficulties in assessing type, frequency, duration and intensity of physical activity; limited knowledge about the relevant exposure period of life; or limited access to information about stage and grade of disease.

Prostate cancer incidence shows large geographical variation, which may be used in the search for causal or protective factors for prostate cancer.26 In a large multinational cohort in Europe with detailed information about health and lifestyle and with data on stage and grade of disease, we investigated the associations between occupational activity and 4 types of leisure time activity and the incidence of prostate cancer.

Material and methods

The EPIC cohort is a large multicenter prospective study designed to investigate the associations between diet and lifestyle and the incidence of different cancers. The rationale and the methods of the EPIC study have previously been described.27, 28 The total cohort includes participants from 23 centers in 10 countries in Europe: Denmark, France, Germany, Greece, Italy, The Netherlands, Norway, Spain, Sweden and the United Kingdom. In our study, we present data for men from 18 centers in 8 of the countries: Denmark, Germany, Greece, Italy, The Netherlands, Spain, Sweden and the United Kingdom. France, Norway, Naples (Italy) and Utrecht (The Netherlands), only recruited women.

All participants gave their written informed consent. The study was approved by the ethical review boards of the International Agency for Research on Cancer and by all the local ethical committees in the participating countries.

Study design and population

Cohort participants were 20- to 97-year-old men and women recruited between 1992 and 2000 from the general population in defined geographic areas in each of the countries, except for participants in the Oxford cohort, who were recruited throughout the United Kingdom in order to enroll a large number of vegetarians. Participants from Ragusa and Turin in Italy, and from the Spanish centers were recruited among blood donors. A total of 153,455 men were enrolled in the study.

Men who had previously been diagnosed with prostate cancer or any other cancer at the time of completing 2 baseline questionnaires about diet and lifestyle were not eligible for these analyses, as were men with unknown vital status or missing data on physical activity (including all subjects from Umeå in Sweden, 16,617 in total) or dates of diagnosis or follow-up. A total of 25,532 men were excluded, leaving 127,923 men available for analysis.

Endpoint assessment

Incident prostate cancer cases were identified through population cancer registries in 6 of the participating countries (Denmark, Italy, The Netherlands, Spain, Sweden and the United Kingdom). In Germany and Greece, cases were identified via self-reported questionnaires, health insurance records, contact with study participants or their next of kin and verified through medical records.

Prostate cancer was defined as ICD-10 code C61. The end of follow-up was defined as dates for end of follow-up for prostate cancer incidence and vital status received at IARC by the end of December 2006. End of follow-up in the 8 countries varied from December 2003 to December 2006.

Data on TNM stage and Gleason grade of the prostate cancers were collected from each center, where all centers contributed with some, but incomplete data on stage and grade. Consequently, data on stage were available for 1,402 cases (57%). Of these, 914 were classified as localized (TNM staging score of T0/T1/T2 and N0/NX and M0, or stage coded in the recruitment center as localized), 488 were classified as advanced (T3/T4 and/or N1/N2/N3 and/or M1, or stage coded in the recruitment center as advanced or metastatic). Data on grade were available for 1,414 (57%) cases and of these, 832 were classified as low-grade (Gleason score <7, or coded in the recruitment center as well-differentiated or moderately differentiated) and 582 were classified as high-grade (Gleason score 7+, or tumors coded as poorly differentiated or undifferentiated).

Exposure assessment

Physical activity was assessed by face-to-face interviews or by a self-administered standardized lifestyle questionnaire as described previously.29, 30 The performance of the physical activity questions was assessed among 126 men and women using an extensive physical activity questionnaire administered 3 times (at baseline and after 5 and 11 months) and the questionnaire was found to rank participants satisfactorily with regard to their physical activity level.31

Data on physical activity included questions about the number of hours spent in a typical week, during summer and winter, in the past year, on 4 types of leisure time physical activities: sport(fitness, aerobics, swimming, jogging, tennis, etc.), cycling (cycling to work, shopping and leisure time), walking (walking to work, shopping and leisure time) and gardening. In Malmö (Sweden) physical activity was inquired in a different way (categorical durations of different activities instead of continuous) and these categorical data were assigned the midpoint value of each category and subsequently combined with data from the other EPIC centers. Data on current occupational activity included employment status and intensity of occupational activity in 5 categories: nonworking, sedentary, standing, manual or heavy manual work. As very few participants reported heavy manual work, the 2 categories manual and heavy manual work were later combined to 1 category, manual work.

Statistical methods

Age-standardized incidence rates for the common age band 50–69 years were estimated using direct standardization based on the European Standard Population in 5-year-intervals.32

Incidence rate ratios were estimated using Cox proportional hazards models with age as the underlying time scale.33 Subjects were followed from the age at which they completed the questionnaires to their age at exit, defined as the age of diagnosis of any cancer (except non-melanoma skin cancer), death, emigration, loss to follow-up or end of follow-up, whichever came first. The hazard rate was allowed to change with time under study34 with a boundary at 1 year after entry to the cohort. Furthermore, the underlying hazard was stratified according to study centre to control for differences in questionnaire design, follow-up procedures and diagnosis or screening procedures.

The analyses were adjusted for height (continuous), weight (continuous), marital status (categorical) and education (categorical). Other potentially confounding variables including smoking, alcohol intake and dietary factors like dairy products and dietary fat were either not associated with prostate cancer risk or did not change the associations materially and were not included in the final confounder package.

Quantitative variables were included linearly in the Cox model after linearity had been evaluated by linear spline models.35 No deviation from linearity was observed for any of the variables.

The implicitly assumed homogeneity of the effects of time spent on physical activity was evaluated among the 8 countries, as well as among the 4 types of physical activity, and subsequently, it was tested whether or not the effect of participation (defined as more than zero hours of activity per week) in each of the 4 activities on prostate cancer differed between the countries. Furthermore, it was tested whether the effect of occupational activity differed between the 8 countries. No significant heterogeneity was found among activities or among countries. Consequently, the hours of activity could be added across activities and across countries.

Some studies have indicated that an apparent dose-response relationship could be caused by a threshold effect attributable to a higher risk among men who were completely or relatively inactive compared to rest of the population,4, 7, 10, 36 and to address this issue we evaluated the effect of participation in each of the 4 activities (active versus non-active in each activity) as well as the effect of participation in at least one of the 4 leisure time activities. Furthermore, a linear effect of the number of activities in which the participants were active was investigated.

According to the Compendium of Physical Activities,37 a MET value, reflecting the intensity of a particular activity, was assigned to each of the 4 types of leisure time activity. A MET is defined as the ratio of work metabolic rate to a standard metabolic rate of 1 (resting metabolic rate). The MET values assigned to the different types of activity were 6.0 for sport, 6.0 for cycling, 3.0 for walking and 4.0 for gardening. The number of hours per week spent on each of the 4 leisure time activities during summer and winter were averaged and multiplied by the appropriate MET values to obtain a MET score of the total leisure time activity.

The MET score was categorized into quartiles and to maximize statistical power, cut-points were based on the distribution among cases. Trend tests of the MET score were based on a continuous variable (per 10 units) while allowing a separate effect for the inactive in each type of leisure time activity. For occupational activity the trend was only based on the 3 categories sedentary, standing and manual activity, because the nonworking was considered a heterogeneous group of retirees, disabled, ill and temporarily unemployed.

We also investigated the Cambridge Physical Activity Index, which is based on a cross-tabulation of occupational activity and time per week spent on sports and cycling and has been validated against heart rate monitoring and cardiorespiratory fitness tests30 (Appendix Table I).

A possible interaction between occupational activity and leisure time activity was investigated, corresponding to the hypothesis that leisure time activity may be particularly protective for those with sedentary work. Similarly, a possible interaction between BMI (cut-points: 18.5, 25 and 30) and leisure time activity was conducted to investigate whether overweight individuals may benefit more from physical activity than lean individuals. Lastly, we investigated a possible interaction with age (using 60 and 65 years as cut-points) to examine whether older men may particularly benefit from leisure time physical activity.

Stage and grade of disease were analyzed using a competing risks model. Likelihood ratio tests were performed to compare associations between stage- and grade-specific outcomes and the exposure variables.38

Two-sided 95% confidence intervals (CI) for the incidence rate ratios (IRR) were estimated based on Wald's test of the Cox regression parameter on the log(IRR) scale.

The SAS procedure PHREG on Unix platform was used for the statistical analyses (SAS version 9.1; SAS Institute, North Carolina).

Results

A description of participants from the 8 countries is provided in Table I. Complete data on physical activity and follow-up were available for 127,923 men, contributing 1.08 million person-years in total and with a median age at recruitment of 54 years. During a median follow-up of 8.5 years, 2,458 incident cases of prostate cancer were identified. In our study of European men, the standardized (European Standard) incidence rates of prostate cancer for the common age band of 50–69 years varied considerably, that is, from 78 per 100,000 in Greece to 680 per 100,000 in Sweden.

Table I. Cohort Size, Time Variables and Age-Standardized Incidence Rates by Country in the EPIC Study

Country	Cohort size (N)	Cases (N)	Median age at recruitment (y)	Median follow-up (y)	Person-years (y)	Incidence rates per 1,000 person-years1
1 Age-standardized (European standard population) prostate cancer incidence rates for each country in the EPIC cohort computed as a weighted average of incidence in 5-year intervals for the common age band of 50–69 years.
Denmark	26,700	376	56	7.7	204,607	264
Germany	21,386	408	53	8.4	180,690	421
Greece	9,468	38	53	7.6	72,426	78
Italy	14,184	146	50	8.8	124,345	278
The Netherlands	6,959	37	44	8.0	55,348	182
Spain	15,453	209	50	10.6	163,521	261
Sweden	10,221	723	59	10.6	108,134	680
United Kingdom	23,552	521	53	8.6	203,117	322
Total	127,923	2,458	54	8.5	1,084 327	–

Baseline characteristics of cases and the cohort are shown in Table II. The proportion of nonworking cases was higher than the proportion of nonworkers in the cohort. Slightly fewer cases participated in sports and walking compared to the general cohort. In contrast, cases were slightly more active in gardening. There were no large differences between the cases and the cohort with regard to the time per week spent on any of the 4 leisure time activities.

Table II. Baseline Characteristics by Case-Cohort Status in the EPIC Study

	Cases (n = 2,458)		Cohort (n = 127,923)
	N (%)1	Median (5, 95%)2	N (%)1	Median (5, 95%)2
1 For leisure time physical activity N (%) refers to the number and percentage of active participants in each activity, where active is defined as reporting more than 0 hours of activity per week. 2 For leisure time physical activity Median (5, 95%) refers to the median and 5, 95% percentiles in the distribution of hours per week spent on each type of leisure time activity among active participants. 3 Numbers do not add up to 100% because of missing values.
Age		61 (51, 73)		53 (36, 68)
Height		174 (163, 186)		174 (162, 186)
Weight		79 (63, 100)		80 (63, 102)
Smoking status3
Never	777 (32)	–	39,480 (31)	–
Former	1,123 (46)	–	48,599 (38)	–
Current	543 (22)	–	38,843 (30)	–
Education3
None	89 (3)	–	5,308 (4)	–
Primary school	818 (33)	–	34,884 (27)	–
Technical school	541 (22)	–	31,090 (24)	–
Secondary school	272 (11)	–	18,638 (15)	–
University	633 (26)	–	34,410 (27)	–
Marital status3
Married	1,564 (64)	–	68,071 (53)	–
Not married	289 (12)	–	16,008 (13)	–
Occupational activity
Nonworker	1,132 (46)	–	30,285 (24)	–
Sedentary	669 (27)	–	46,894 (37)	–
Standing	382 (16)	–	25,971 (20)	–
Manual	275 (11)	–	24,773 (19)	–
Leisure time activity (hour/week)
Sports	978 (40)	2 (0.5, 8)	58,097 (45)	2 (0.5, 8.5)
Walking	2,161 (88)	4.5 (1, 18)	119,032 (93)	4.5 (1, 20)
Cycling	1,102 (45)	2 (0.5, 10)	60,464 (47)	2 (0.5, 10)
Gardening	1,658 (67)	3 (0.5, 16)	80,579 (63)	2.5 (0.5, 14)

Table III shows the incidence rates ratios for prostate cancer in relation to different measures of physical activity. First, we investigated the associations of risk with participation in the 4 leisure time activities sports, walking, cycling and gardening by comparing the active with the nonactive men (that is, those who reported more than 0 hour of each activity with those reporting 0 hour of activity per week), but there was no significant association with participation in any of the 4 specific activities. The association with participation in leisure time activity in general was estimated by comparing those who participated in at least one of the 4 activities with those not participating in any leisure time activity, and we found no effect of participation in leisure time activity, independent of the type of activity. Also, a linear effect of the number of leisure time activities in which the participants were active was investigated, and we found no association with participation in more leisure time activities. Adjustment for potential confounders did not change any of these results.

Table III. Incidence Rate Ratios for Prostate Cancer in Relation to Physical Activity In the EPIC Study (2458 Prostate Cancer Cases)

Type of physical activity activity		IRR (95% CI)2	IRRadjusted (95% CI)2^,3
1 Participation defined as reporting more than 0 hour of a specific activity or of any activity (at least one of the specific activities), respectively. 2 Analyses on participation in specific leisure time activities, participation in at least one leisure time activity, and number of activities participated in are adjusted for occupational activity; analyses on MET hour/week of leisure time activity and occupational activity mutually adjusted and adjusted for participation in specific leisure time activities; Cambridge Index adjusted for participation in specific leisure time activities. 3 Further adjustment for height, weight, marital status, and education.
Participation in leisure time activity1	Sports	1.05 (0.97–1.15)	1.03 (0.95–1.13)
	Walking	1.01 (0.88–1.15)	1.01 (0.88–1.15)
	Cycling	0.97 (0.89–1.06)	0.97 (0.89–1.06)
	Gardening	1.01 (0.92–1.11)	1.00 (0.91–1.09)
	At least one leisure time activity	1.06 (0.87–1.28)	1.05 (0.87–1.28)
	Number of leisure time activities (per extra activity)	1.01 (0.97–1.05)	1.00 (0.96–1.04)
MET hours/week of leisure time activity	Q1 (≤25)	1	1
	Q2 (25–43)	1.02 (0.90–1.15)	1.02 (0.91–1.15)
	Q3 (43–71)	1.02 (0.90–1.16)	1.03 (0.91–1.17)
	Q4 (>71)	1.00 (0.87–1.14)	1.01 (0.88–1.16)
	p for trend over quartiles	0.97	0.94
	Linear trend (per 10 MET hours/week)	1.00 (0.99–1.01)	1.00 (0.99–1.01)
Occupational activity	Nonworking	0.84 (0.75–0.94)	0.89 (0.79–0.99)
	Sitting	1	1
	Standing	0.90 (0.80–1.03)	0.94 (0.83–1.07)
	Manual	0.84 (0.73–0.97)	0.90 (0.77–1.04)
	p for trend over the last three categories (sitting, standing and manual work)	0.03	0.18
Cambridge index	Inactive	1	1
	Moderately inactive	1.00 (0.89–1.12)	1.02 (0.91–1.14)
	Moderately active	0.96 (0.86–1.07)	0.99 (0.89–1.11)
	Active	0.92 (0.78–1.08)	0.98 (0.83–1.15)
	p for trend over categories	0.18	0.69

Second, we investigated the associations between the total leisure time activity MET score and prostate cancer risk. Men in the 3 upper quartiles did not have a risk of prostate cancer that was significantly different from men in the lowest quartile of total leisure time activity, and there was no clear pattern over the quartiles. The linear trend also showed no overall effect of leisure time activity and adjustment for the potential confounders did not change this result.

Last, when we investigated the association between occupational activity and prostate cancer incidence rate, men with standing or manual work had a slightly lower risk of prostate cancer (IRR (95% CI) = 0.90 (0.80–1.03) and 0.84 (0.73–0.97) for standing and manual occupation, respectively) than men with a sitting occupation (reference). A trend based on men with sedentary, standing and manual work confirmed this tendency (p = 0.03) but adjustment for potential confounders weakened the association (p = 0.18). However, nonworking men had a significantly lower risk of prostate cancer (IRR (95% CI) = 0.84 (0.75–0.94)) than men with a sitting occupation and the strength of this association was weakened, but persisted, after adjustment for the potential confounders (IRR (95% CI) = 0.89 (0.79–0.99)).

Further adjustment for variables like alcohol, smoking, dairy products, and dietary fat only changed the association for nonworking men, where the association became even weaker (IRR (95% CI) = 0.98 (0.87–1.11), 0.95 (0.84–1.09), 0.90 (0.77–1.05) for nonworking men, men with standing occupation and men with manual occupation, respectively, compared to men with a sitting occupation (reference)).

In a cross-tabulation of occupational activity and hours per week of sports and cycling (the Cambridge Physical Activity Index, Table III), there was a weak inverse association of prostate cancer with higher activity level (p = 0.18), but adjustment for potential confounders further weakened this result (p = 0.69). Stratification according to age (<65 or ≥65 years) did not reveal different effects for age subgroups.

There were no significant interactions between BMI and leisure time activity (p = 0.18), between occupational activity and leisure time activity (p = 0.08), or between age and leisure time activity (cut-point 60 years: p = 0.43, cut-point 65 years: p = 0.31).

The analyses on occupational activity and MET hours per week of leisure time activity were also conducted according to stage and grade of disease (Table IV) and we found a significant trend of lower risk of advanced prostate cancer with higher occupational activity level (p = 0.024, IRR (95% CI) = 0.75 (0.53–1.05) for men primarily standing at work, IRR (95% CI) = 0.79 (0.58–1.07) for men with manual work, both compared to men primarily sitting at work). There was also a tendency towards lower risk of high-grade prostate cancer with higher occupational activity level, but this trend did not reach statistical significance (p = 0.11). However, the competing risk test between advanced or localized stage of disease was not significant (p = 0.11), and nor was the test for difference between high-grade and low-grade disease (p = 0.31). Leisure time activity was not associated with risk of advanced, localized, high-grade or low-grade prostate cancer and competing risk tests were not significant (p = 0.35 for stage, p = 0.74 for grade).

Table IV. Incidence Rate Ratios for Prostate Cancer in Relation to Physical Activity According to Stage and Grade of Disease in the EPIC Study

Type of physical activity activity		Stage		Grade
		Advanced1 (n = 488) IRR (95% CI)	Localized1 (n = 914) IRR (95% CI)	High1 (n = 582) IRR (95% CI)	Low1 (n = 832) IRR (95% CI)
1 Leisure time activity and occupational activity mutually adjusted, and further adjusted for height, weight, marital status and education.
MET hours of leisure time activity per week	Q1 (≤25)	1	1	1	1
	Q2 (25–43)	1.00 (0.76–1.32)	0.96 (0.79–1.17)	1.08 (0.84–1.38)	0.99 (0.80–1.22)
	Q3 (43–71)	1.04 (0.78–1.38)	0.99 (0.81–1.21)	1.00 (0.77–1.30)	0.98 (0.79–1.21)
	Q4 (>71)	1.23 (0.91–1.65)	0.89 (0.71–1.11)	1.20 (0.91–1.57)	0.94 (0.75–1.19)
	p for trend over categories	0.16	0.37	0.27	0.61
	Linear trend (per 10 MET hours/week)	1.02 (0.99–1.04)	1.00 (0.98–1.02)	1.01 (0.99–1.04)	1.00 (0.98–1.02)
Occupational activity	Nonworking	0.92 (0.70–1.20)	1.07 (0.88–1.31)	0.97 (0.76–1.25)	1.07 (0.87–1.31)
	Sitting	1	1	1	1
	Standing	0.79 (0.58–1.07)	1.08 (0.88–1.31)	0.84 (0.64–1.12)	1.12 (0.90–1.39)
	Manual	0.75 (0.53–1.05)	0.91 (0.72–1.15)	0.76 (0.56–1.04)	0.99 (0.77–1.29)
	p for trend over the last three categories (sitting, standing and manual work)	0.024	0.30	0.11	0.99

Figure 1 shows the adjusted IRRs for prostate cancer in relation to hours of leisure time activity in individual countries and for all 8 countries together. Unadjusted values were very close to adjusted values. For all countries together the IRR was 1.00 for 10hour/week increase in leisure time activity.

Figure 1. IRRs (95% CI) for prostate cancer according to hours per week spent on leisure time activity in the EPIC cohort (IRR per 10 hour).

Discussion

In this large prospective study of European men, we found a trend towards lower risk of prostate cancer with more physically demanding occupation—a trend that was most pronounced for advanced prostate cancer. Adjustment for the potential confounders (weight, height, marital status and education) attenuated this association in the overall analysis, but the association remained significant for advanced prostate cancer. We found no associations between incident prostate cancer and participation in or energy (METs) spent on leisure time activity. The lack of associations between leisure time activity and prostate cancer was observed consistently across subgroups of men defined by age or BMI, and leisure time activity was not more strongly associated with advanced or high-grade disease.

Overall, our study supports the hypothesis that a higher level of occupational physical activity may reduce prostate cancer risk, particularly advanced stage prostate cancer risk.

The major strengths of our study include standardized data collection methods for physical activity and potential confounders and the large number of cases.

The study also has some limitations in that the information on physical activity was based on a self-administered questionnaire, which is vulnerable to misclassification.39 This misclassification could be nondifferential and cause attenuation of results, but could also be differential and potentially bias risk estimates, as could be the case if, for example, cases with a low fitness tended to report more time spent on activity because they perceived it as longer or simply because the same activity would take longer than for a fit, healthy person. It seems as if regular activities of higher intensity are more likely to be accurately recalled.40 In accordance with this, validation of the lifestyle questionnaire used in the present study showed that sport and cycling were closely associated with energy expenditure, whereas the 4 other low-intensity activities were weakly associated with energy expenditure.30 Although the physical activity questions have performed well in validation,31 some misclassification may still be present. Physical activity was assessed only at baseline and the reported physical activity may not reflect physical activity earlier in life nor changes after baseline, and furthermore, any protective effect of physical activity may require lifelong exposure. Consequently, the relevance of a single measurement at a given point in time depends on the degree to which the exposure tracks over time. Moreover, retirement or the physician's advice of more physical activity to inactive elderly persons with various health conditions may change these men's activity patterns. This situation could attenuate results towards a null result; however, associations have been found for other cancer sites in the EPIC study.41–43

There is also a possibility that health-conscious men, who spend more time on leisure time physical activity, are exposed to higher diagnostic intensity, that is, PSA testing, digital rectal exam, ultrasound, biopsy etc. This could lead to (over-)diagnosis of subclinical prostate cancer (diagnostic bias) and attenuation of the association between prostate cancer incidence and leisure time physical activity. Data on PSA testing were not available in our study, but the rates of PSA testing across Europe seems to be low (6% in England and Wales,44 7% in the Netherlands,45 about 10% in Spain46 and 16% in Italy47) compared to US rates of 57%.48

The study was based on men of relatively old age, which increases the risk of prevalent disease. To evaluate possible bias from prevalent disease, the incidence rate ratios were allowed to depend on time in the study. Furthermore, the results from analyses of the first 2 years of follow-up were compared to the rest of the follow-up, but we found no differences between results from the 2 time periods.

Our result of no overall inverse association between occupational activity and prostate cancer is consistent with most,12, 16, 19–25 but not all,5–7, 13 previous studies. That physical activity is more strongly, inversely related to risk of advanced, high-grade or aggressive prostate cancer has, however, been indicated previously.8, 10, 36 Our study supports this with regard to occupational activity.

We found no protective association with leisure time activity using the MET values. Although there is a risk of adding random variation by applying an estimated intensity to the included activities, the MET score has major strengths in being an accepted standard measure of physical activity and in facilitating comparisons of studies with data on different types of activities as well as permitting data to be combined into a single measure of physical activity. Six cohort studies4, 8–10, 17, 22 presented their data using MET values and, as we did, based it on activities of relatively high intensity. None of these studies found a significant overall association with physical activity, and even though at least 5 of these studies included participants who were considerably less active than participants in our study, 3 studies found an association in subgroups. In one study, the subgroups were defined by age (inverse association for men aged ≥65 years)8; in the second study by age (inverse association for men aged ≥65 years), BMI (positive association for obese men) and screening history (inverse association for men not screened 2 years before baseline)9; and in the third study by aggressiveness of the tumor (inverse association for men aged ≥65 years with aggressive tumors).10 We also examined the association between physical activity and prostate cancer by using the hours of activity instead of MET hours per week of activity, but did not observe any difference in the strength or direction of the results.

In conclusion, these data are compatible with the possibility that a high level of occupational physical activity may slow progression of prostate cancer. Future studies should be designed to address the aspect of etiologically relevant time periods where physical activity may be operating, preferably by measuring both types and total volume of physical activity throughout lifetimes.

Acknowledgements

The authors thank all the participants in the EPIC study, and Bertrand Hémon at the IARC and Katja Boll at the Danish Cancer Society for their expertise in the handling of data.

APPENDIX

Table AI. Cambridge Physical Activity Index as Cross–Tabulation of Occupational Activity and Time Per Week Spent on Sports and Cycling

Work activity	Leisure time physical activity (Duration of sport and cycling in hour/week)
	No	≤3.5	>3.5 and ≤7.0	>7.0
Sedentary	Inactive	Moderately inactive	Moderately active	Active
Standing or nonworking	Moderately inactive	Moderately active	Active	Active
Manual	Moderately active	Active	Active	Active