In this unique study of primarily local-stage prostate cancer, in which the presence or absence of prostate cancer was determined by prostate biopsy, there were no statistically significant associations of nutrient intake or dietary supplement use with prostate cancer overall. When results were stratified by disease grade (low- vs. high-grade disease (Gleason score 2–7 vs. 8–10)), there were several noteworthy associations. Polyunsaturated fat intake was positively associated with risk of high-grade cancer, and dietary calcium intake was positively associated with risk of low-grade cancer and inversely associated with risk of high-grade cancer. Based on a post-hoc analysis, there was evidence that dietary zinc intake beyond a relatively low threshold was associated with reduced risk of high-grade cancer. There was also some evidence that a high alcohol intake was associated with increased risk of high-grade disease; the associations of alcohol intake with cancer risk in the PCPT are complex and have been described previously (28
). Neither use of dietary supplements nor intake of antioxidants, folate, vitamin D, or long-chain n-3 fatty acids was significantly associated with low- or high-grade prostate cancer risk.
Investigators in many large case-control and cohort studies have reported that calcium intake from foods and/or supplements was associated with increased cancer risk (29
). Our finding of no association with total prostate cancer risk (odds ratios contrasting quartile 4 with quartile 1 were 1.16 (95% CI: 0.95, 1.43) and 1.09 (95% CI: 0.91, 1.32) for dietary intake and total intake, respectively) was consistent with the null findings from several other large cohort studies (35
). Our finding that calcium intake was inversely associated with high-grade cancer but positively associated with low-grade cancer is inconsistent with several other studies that found associations to be stronger or exclusive for high-grade or advanced-stage disease (29
); in particular, we found no evidence that very high dietary calcium intakes (>1,400 mg/day) were associated with increased risk of high-grade disease. Our findings are similar to those reported from the screening arm of the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial (33
). In both the PCPT and the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial, and in contrast to other studies, almost all prostate cancers were local-stage and screen-detected. It is possible that risk factors for screen-detected cancers are different from those diagnosed clinically. For example, if we assume that low-grade cancers develop into high-grade cancers, perhaps calcium decreases the rate at which low-grade cancers progress. However, lacking a strong biologic rationale, the calcium findings from both the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial and the PCPT should be considered provisional until they are replicated in studies that separate screen-detected cancers from clinically detected cancers.
Many investigators have studied associations of dietary fat with prostate cancer risk, and their findings are inconsistent. In a 2004 meta-analysis, Dennis et al. (39
) found a significantly increased risk associated with high fat consumption in case-control studies but no association in cohort studies; and in more recently published cohort studies, investigators have found either no associations (40
) or significant inverse associations for high-grade disease (43
). Study results differ somewhat when risk is examined separately by stage and/or grade and when fats are separated into polyunsaturated, monounsaturated, and saturated fats, but overall there is little support for associations of fat with risk. We know of no studies which have found that a high intake of polyunsaturated fat—more specifically, the substitution of polyunsaturated fat for saturated fat—was associated with increased risk of high-grade cancer; however, this finding is biologically plausible. The n-6 fatty acids, which constitute the majority of dietary polyunsaturated fats, are proinflammatory (44
), and inflammation may play an important role in prostate cancer pathogenesis (45
). A single study of heavy smokers and/or asbestos-exposed men found a substantially increased risk associated with high polyunsaturated fat consumption, which was restricted to the small subset of men with a family history of prostate cancer (41
). Nevertheless, our findings are generally inconsistent with those in the literature and require replication in studies of screen-detected cancer.
The most significant weakness in this study was the use of FFQs to measure nutrient intake. Recently, some investigators have questioned the validity of FFQs for dietary assessment (46
), and some scientists have challenged their continued use in epidemiologic research (48
), although this view is controversial. As demonstrated in studies of dietary fat and breast cancer risk (50
), there is a distinct possibility that moderate or weak associations of diet with cancer risk cannot be detected using FFQs but can be detected using multiple-day food records. We believe that strong associations will probably be detected across extreme intake categories, and our concern is that weak but meaningful associations may not be detected. We also chose not to follow several common practices used in nutritional epidemiology. First, we did not adjust model results for multiple dietary factors simultaneously, because most dietary covariates are highly correlated and poorly measured, and their use could therefore lead to unstable models with unpredictable results (52
). Second, we did not conduct multiple subgroup analyses—for example, examining results stratified by age or nutrients stratified by type of dietary exposure (e.g., folate from food vs. folate from supplements)— because, lacking a strong biologic rationale, this increases the likelihood of chance findings. It is possible that true, subgroup-specific or nutrient-adjusted associations were missed in our analyses. Our plan is to examine these more complex hypotheses in future analyses based on biomarkers of diet and then attempt to confirm the results using dietary intake data.
There are unique aspects of this study that both increase its quality and limit its generalizability. The most significant are that study participants had PSA levels less than 3 ng/mL at study entry, there was annual screening (PSA plus DRE) during the 7 years of the trial, and determination of the presence or absence of disease was based on endpoint biopsies. Thus, almost all of the cancers that were detected were local-stage, and while the use of endpoint biopsies to identify cancer cases and noncases minimized detection bias, it also identified cancers that would never have been detected by means of either screening or clinical symptoms. A second unique aspect of this study is the use of uniformly graded Gleason scores of 8–10 to define high-grade disease, in contrast to other studies that have used a mix of stage (often surgical and clinical) and grade, as well as long-term clinical outcomes, to define “aggressive” disease. Taken together, the mix of cancer phenotypes in the PCPT may differ markedly from the phenotype mixes in studies that are based on cancers detected by screening alone or by locally defined standards of clinical practice. Thus, risk factors for cancers identified in the PCPT could be quite different from those for clinically detected or advanced-stage disease. Nevertheless, a major strength of this study is the mitigation of the detection biases present in most observational cohort studies in which PSA levels and DREs affect the decision to perform a prostate biopsy. Use of PSA screening is probably associated with dietary patterns (53
), such that biases due to screening may have seriously confounded the results of previous studies.
In conclusion, in this unique sample of local-stage, biopsy-detected cancers, we found no evidence that dietary or supplemental intake of nutrients often proposed to prevent prostate cancer, including lycopene, n-3 fatty acids, vitamin D, vitamin E, and selenium, was associated with risk of low- or high-grade cancer. Our finding that polyunsaturated fat was associated with increased risk of high-grade prostate cancer suggests that further research into inflammation and other metabolic processes affected by these fats may be important in understanding prostate cancer etiology. Our finding of a positive association of calcium with low-grade disease and an inverse association with high-grade disease adds to the inconsistency of findings related to calcium, which may be important and may require further inquiry. The consistent and strong findings from ecologic studies that the adoption of a diet high in fat and animal products, characteristic of Western diets (54
), increases prostate cancer risk are perplexing. It is possible that these ecologic studies are yielding results that do not reflect individual-level cancer risk, that the specific aspects of diet affecting prostate cancer risk have not been adequately measured or identified, or that the association of a Western-style diet with prostate cancer risk cannot be reduced to studies of a single nutrient or set of nutrients.