There was little evidence of an association between sunlight exposure during leisure-time and PCa risk in our data. Men never exposed to sunlight during leisure-time appeared to be at slightly higher risk of PCa than those who had been exposed at some point during adulthood, but none of the estimates achieved statistical significance.
The epidemiological evidence available to date on the association between sunlight exposure and PCa risk remains contentious. Some previous studies have looked at this association in terms of PCa mortality [29–34]. Since the latter can reflect both survival and etiology, findings from these are not expected to necessarily align with our own, which focuses on PCa incidence.
Of studies assessing the risk of incident PCa [6, 9–20, 35], many were in line with a protective effect of sunlight. However, the majority used an ecological approach, with confounding remaining of major concern , especially when studying sunlight exposure and cancer risk .
To our knowledge, only eight analytical studies of incident PCa have been conducted to date on this issue [14–20, 35]. Study design, size and sunlight assessment protocols have varied widely. Five of these support a protective effect of sunlight. Two were case–control studies, one  assessing exposure to sunlight residential, work and recreational settings as part of interviews (450 cases, 455 controls), the other  assessing cumulative sunlight exposure per year and sunbathing frequency (453 cases, 312 controls). Three cohort studies also reported a protective effect of sunlight. The first was based on 153 PCa cases, and estimated residential sunlight ambient levels . The second, including 161 PCa cases, assessed solar radiation levels at birth address and reports of sunlight exposure during leisure and at work . More recently, a third and much larger cohort study accrued over 21,000 incident PCa cases during the period of follow-up . This investigation assessed ambient ultraviolet radiation levels at the residence of subjects at baseline. However, individual information on time spent outdoors, which can vary greatly between individuals and over time, was not available.
Another large investigation, a nested case–control study of 1,020 cases and 5,044 controls conducted in the UK, observed no association overall between reported time spent outside during childhood and adulthood, and PCa risk . An inverse relationship was observed between PCa and intense sun exposure 2 years prior to the index date (all cancers), while time spent outside was associated with higher risks of advanced PCa.
Finally, two case–control studies documented an excess risk of PCa among individuals who had spent the greatest number of hours outdoor either recently , or at the ages of 30 and 50 years . One  was conducted in Singapore (240 cases, 268 controls) while the other , in New South Wales, Australia (1084 cases and 234 controls). In both studies, positive associations were apparent for all cancers, as well as for aggressive ones . It has been brought forward that a U-shaped relationship might reconcile apparent contradictions across studies, i.e., inverse or positive associations between sunlight exposure and PCa observed in low or high solar UV environments, respectively . In our study, none of the sunlight exposure levels showed statistically significant associations with PCa and confidence intervals overlapped between exposure categories. However, men in the middle category tended to be at lesser risk of PCa compared to those in other categories. Tests assessing the linearity of the exposure variables or the presence of a U-shape association were negative (data not shown).
Previous studies present methodological differences with our own. One salient difference across previous investigations relates to the sunlight exposure assessment methods applied. Geographical-based sunlight exposure indices were used in ecological studies [6, 9–12] as well as in some analytical studies [14–17]. The downside of such approaches is that ambient solar levels are assumed to reflect individual exposures, which may result in important misclassification of exposure. Individual-based exposure assessment methods used in previous studies include a self-reported recreational sun exposure level , a sun exposure index based on skin pigmentation , lifetime or adulthood frequencies of outdoor leisure activities in hours per week [14, 19] and a sum of hours of sunlight exposure during summer weekends at two age points . Although our sunlight exposure assessment may not be as precise as that used in some of the aforementioned studies, it was based on detailed questionnaire data eliciting participation in outdoor leisure activities. Changes in engagement in the different activities over the years were factored in, so were variations across seasons for most study subjects. However, we did not collect information on acute sunlight exposures (e.g. sunbathing), on sunlight exposure while traveling to work or elsewhere, or on the specific number of hours involved in each outdoor activity. Nevertheless, when we applied typical leisure activity event durations to differentiate between typically shorter and longer exposure events it had little impact on risk estimates as compared to the cumulative number of events. We had information on the reported overall sunlight exposure during work-time. Albeit resulting in a crude exposure index, workplace exposure had relatively low prevalence in our population and did not influence the association between leisure-time exposure and PCa.
Our study was based on the population living in Montreal, Canada, at a latitude entailing low solar radiation intensity . It may be that our findings are not readily comparable with those observed among subjects living in different latitudes. Another difference relates to the ancestries represented in the different study populations. Our study predominantly included subjects of European ancestry, whereas, for instance, the Singapore study focused on Asians, a population known to have lower risks of PCa .
As is the case for any study evaluating risks or benefits associated with sunlight, exposure misclassification likely occurred to some extent in our data. However, it is believed to have been largely non-differential, which tends to result in conservative estimates . Questions on leisure activities were primarily formulated to assess energy expenditure and subjective judgment needed to be applied in a number of occasions when assigning them as indoor or outdoor activities. Physical activity is potentially related with PCa [40, 41], and we adjusted for it in our analyses.
Reporting bias based on case–control status was possible, but unlikely. Sunlight exposure was not the primary focus of the PROtEuS study. Moreover, there is no widespread belief in the population that sunlight exposure is associated with PCa.
The vitamin D mechanism is most often called upon to explain a possible protective effect of sunlight exposure on PCa. Different pathways may be involved, such as decreased cell proliferation, cell cycle regulation, cell invasiveness and angiogenesis and increased cellular differentiation and apoptosis [42, 43]. Although the evidence for an anti-cancer mechanism of vitamin D is strongly suggested by biological studies, it has not been consistently supported by epidemiological findings . It is thus possible that other mechanisms exist between sunlight exposure and PCa. One such mechanism, less documented in sunlight exposure studies, could be UV-induced nitric oxide . At high concentrations, inhibitions of prostate cancer cell proliferation, of metastases and of epithelial-mesenchymal transition have been observed, and at low concentrations angiogenesis is promoted.
Should a relation between sunlight and PCa truly exist, and should this association involve vitamin D synthesis, then a number of factors would need to be considered for valid exposure assessment, such as skin coverage by clothing, skin type, geographical location and meteorological conditions . First, the amount of skin exposed to the sun can determine the rate of vitamin synthesis and darker skins tend to allow for less vitamin D production . However, previous data indicate that after a certain threshold, a greater exposed surface and a higher dose of UVB will not significantly increase vitamin D levels . This could be explained by the negative feedback loop triggered by high levels of vitamin D, which involves its inactivation by the CYP24A1 hydroxylase . Second, the amount of solar radiation received has been shown to differ between geographical regions, even when doing the same activity for the same amount of time . Third, meteorological conditions such as the presence of direct, diffuse and reflected radiations, the time of the day and the atmospheric conditions  also affect the amount of solar radiation received. In addition, diet and supplementation play a role in the level of serum vitamin D, especially in the winter months when, at northern latitudes, no important vitamin D synthesis is triggered . Information for these factors was not available for our study, so each recreational sunlight event was assumed here to be equal in terms of sunlight exposure dose.
Our study presents several important strengths. This is the largest study to date to assess the association between sunlight exposure, using an assessment protocol that takes into account individual sunlight-related behavior, and PCa. Moreover, the exposure assessment covered the entire adulthood period, something rarely achieved in the past. Cases were histologically confirmed, and information was collected as part of face-to-face interviews. Participation rates in the study were relatively good, compared to those often observed in similar studies. We were able to compare socio-demographic characteristics of eligible men who declined to participate in the study, to those of study subjects; only slight differences were observed, suggesting that strong selection bias was not at play. We collected information on a wide range of potential confounders and considered these in our analyses. This included the use of solar protection, which few studies have been able to take into account [18, 20].