The data from our nationwide prospective cohort study supports the hypothesis that a low level of occupational physical activity might increase the relative risk of prostate cancer. Men with sedentary jobs in both 1960 and 1970 had a small, excess relative risk compared to men who had very high or high levels of activity at work. There were no significant differences in relative risk estimates for prostate cancer occurring before or after 70 years of age.
The strengths of our study are the prospective design, the nationwide population-based cohorts, the large size of the cohorts, repeated measures which indicated stable, long-term exposure, and the long follow-up time with substantial numbers of observed prostate cancer cases in all cohorts. A shortcoming is the indirect assessment of physical activity, based on job titles. However, when our physical activity matrix was validated against self-reported level of occupational physical activity in the 1970s, we found good agreement. Moreover, the use of job titles may even be an advantage, since the choice of vocation preceded the onset of the prostate cancer by many years, thus giving little room for reversed causation. Nonetheless, considerable non-differential misclassification of the factual physical activity, introduced by the crude job title classification, may have substantially attenuated the true dose-response (Rothman and Greenland, 1998
). Another weakness with this study is the lack of information for individuals on possible confounders. However, there are no known strongly related risk factors for prostate cancer (Ross and Schottenfeld, 1996
) and when we adjusted for place of residence no major changes in relative risk estimates were observed. Still the observed small association might be explained by unknown confounding factors.
Both occupational activity and socio-economic status are independent risk factors for prostate cancer in this data. However, since they are both based on the same variable (job title) and thus correlated, it is difficult to clearly disentangle the effect of occupational activity from the effect of socio-economic status on prostate cancer risk. The relative risk estimates for physical activity were slightly attenuated when socio-economic status was included in the model, as was also observed in another study (Dosemeci et al, 1993
). Socio-economic status per se
can not be biologically associated with prostate cancer, but factors closely related to social status might be and this requires closer future investigating. However, in an earlier case–control study in Sweden, socio-economic status has not been associated with prostate cancer risk (Andersson et al, 1996
Ascertainment bias is a potential limitation of this study. The observed association between low occupational activity and increased prostate cancer risk could be due to the fact that manual labour is correlated with a low socio-economic status. Those men might be less likely to visit the physician and be diagnosed with prostate cancer. Regular visits to a physician in higher socio-economic groups and among older men might result in detecting latent prostate cancers among sedentary men, leading to an overestimation of relative risk for the lower activity group or among older men. However, we did not observe any significant differences in relative risk between older and younger men.
Information on total physical activity was not available in our study. Influence of recreational physical activity may be of concern if there is a systematic difference in recreational physical activity among men in different occupational activity categories. The association would be overestimated if men with sedentary jobs were less physically active during their leisure-time than men with strenuous occupations. However it is more likely that men with sedentary jobs would be more active during their leisure-time leading to an attenuation of the estimated relative risk.
Our findings are in agreement with a majority of previous studies showing a statistically significant inverse association between high level of recreational activity (Oliveria et al, 1996
; Albanes et al, 1989
; Hartman et al, 1998
; Severson et al, 1989
) and occupational activity (Brownson et al, 1991
; Dosemeci et al, 1993
; Hsing et al, 1994
; Bairati et al, 2000
; Clarke and Whittemore, 2001
) and prostate cancer risk, but in contrast to two studies showing significant positive association (Ilic et al, 1996
; Cerhan et al, 1997
). In a follow-up study by Health Professionals, an inverse association with recreational physical activity was observed for metastatic prostate cancer only (Giovannucci et al, 1998
). We observed a statistically significant inverse association between occupational activity and prostate cancer risk in agreement with the NHANES I study (Clarke and Whittemore, 2001
), but not with the majority of other cohort studies investigating the role of occupational activity and not showing significant associations (Vena et al, 1987
; Albanes et al, 1989
; Severson et al, 1989
; Thune and Lund, 1994
; Steenland et al, 1995
; Hartman et al, 1998
; Lund Nilsen et al, 2000
; Putnam et al, 2000
). This may be explained by that our cohort study is the largest performed to date, thus it has enough power to show even weak associations to be statistically significant. In a case–control study estimating lifetime occupational physical activity, the increased risk for prostate cancer was higher when a larger proportion of life was spent in sedentary jobs, which is consistent with our findings (Bairati et al, 2000
). Strenuous occupational activities in the mid-teens or early twenties have been reported to be inversely associated with prostate cancer risk (Villeneuve et al, 1999
). These findings may indicate that physical activity in different periods in life, or the cumulative life-long occupational activity is of importance for prostate cancer risk. The lack of significant inverse association with prostate cancer mortality observed in our study does not exclude a possibility of a weak association reported previously from a large US cohort (Vena et al, 1987
). Effect of potential confounding factors on our risk estimates might mask the true association of physical activity with prostate cancer mortality.
The association between prostate cancer and physical activity can be explained by at least three different possible mechanisms. First, high levels of bioavailable testosterone has been associated with increased risk of prostate cancer (Gann et al, 1996
; Wolk et al, 1997
). Several studies have reported lower levels of free testosterone (Tikkanen et al, 1998
; Tymchuk et al, 2001
) in athletically trained men relative to untrained men and immediately after exercise (Nindl et al, 2001
) as well as increased levels of sex hormone binding globuline (SHBG) immediately after exercise in older men (Zmuda et al, 1996
) and after an exercise programme (Caballero et al, 1992
; Tymchuk et al, 1998
). However, there are also studies showing increasing levels of testosterone with increasing levels of exercise (Zmuda et al, 1996
). Secondly, a higher level of circulating insulin-like growth factor I (IGF-I) has been associated with increased risk of prostate cancer both in prospective (Chan et al, 1998a
; Stattin et al, 2000
) and in case–control studies (Wolk et al, 1998
). IGF-I might also be associated with long-term or short-term physical activity levels, although observed associations are inconsistent (Bonnefoy et al, 1998
; Eliakim et al, 1998
; Tissandier et al, 2001
). Thirdly, the immune system, which is involved in regulating the susceptibility to both the initiation and promotion of tumours, can be modified by physical activity (Shephard and Shek, 1995
). Depending on intensity, duration and frequency of activity, immune functions can be suppressed or enhanced by physical activity, but moderate physical activity usually stimulates the immune response.
In summary, we have shown that work in sedentary and light physical activity occupations might be associated with increased risk of prostate cancer incidence but not mortality. The observed associations may have been confounded by factors related to socio-economic status. To better disentangle the effects of occupational physical activity from socio-economic status, future studies should include specific information on physical activity at work and during leisure-time, as well as detailed information on potential confounders.