Our findings suggest that there is no strong association between dietary GI, GL, II, fiber or whole grains intake in relation to prostate cancer. We observed a weak positive association between dietary GI and low-grade prostate cancer, which was slightly stronger after restriction to men who have had a PSA test. In contrast, the observed significant positive association between dietary intake of whole grains and prostate cancer did not persist when the cohort was restricted to PSA-screened men.
Few studies have related dietary factors that are likely specifically to influence insulin levels to prostate cancer incidence. Our observation of no overall association between GI or GL and risk of total prostate cancer is in agreement with results from a large cohort study of 262,642 men aged 50–71 years at baseline (15,949 prostate cancer cases) [12
]. However, one case-control study reported significant positive associations with both GI and GL in relation to prostate cancer [11
]. Neither of these two studies investigated subgroups of prostate cancer as defined by stage or grade. We observed a weak positive association between GI but not GL and risk of low-grade prostate cancer, which persisted after restriction to men with PSA-screening. Although this might be a chance finding, it is note-worthy that in several nested case-control studies [29
], including one within HPFS [30
], IGF-1 was more strongly associated with low-grade than with high-grade prostate cancer.
Dietary II was unrelated to risk of total prostate cancer and subgroups of prostate cancer. In most studies relating dietary GI or GL to cancer risk the underlying hypothesis was modulation of the insulin and IGF axes. The dietary GI is based on the postprandial increase in blood glucose levels, while the II is directly based on the postprandial insulin response. Since the increase in blood glucose levels is not always proportional to the insulin response, the II seems to be the more relevant measure to test this hypothesis. The II of mixed meals predicted the insulin response better than GL in a small group (n=11) of lean young healthy men [10
]. At the current stage of research, further studies on the validity, reproducibility and usefulness of the food II with respect to clinical settings as well as to epidemiological research are warranted. Our results suggest that long-term exposure to high postprandial insulin responses is not associated with risk of prostate cancer incidence or progression. This finding needs to be confirmed in other studies.
Dietary fiber was not associated with total prostate cancer in our study, which is in agreement with several case-control studies [13
] and one cohort study [16
]. In contrast, a recent case-control study found a significant inverse association between dietary fiber and prostate cancer risk [15
]. We observed a significant positive association between fiber intake and high-grade prostate cancer, which, however, disappeared after restriction to men with PSA-screening.
We found whole grain intake to be significantly positively associated with total prostate cancer, but this finding was largely attenuated in the analysis limited to men who had PSA screening. A significant positive association between grain intake and prostate cancer has been observed in an case-control study [15
]. In this study, however, the definition of grains did not distinguish between whole grains and refined grains, so that it remains unclear whether this result is driven by whole grains. Another case-control study observed no association between consumption of whole grain foods (essentially whole grain bread or pasta) and prostate cancer [17
In light of the overall null results of this study, the ability to validly estimate dietary exposures by means of semi-quantitative FFQs is crucial. The repeated measurement of dietary intake by FFQs every 4 years reduces measurement error and is suitable to best reflect long-term dietary intake. Validation studies relating dietary intake assessed by FFQ to metabolic biomarkers, as surrogate markers to test for validity have been performed [33
].In the Nurses’ Health Study (NHS), high dietary GI/GL assessed by FFQ were associated with lower plasma HDL cholesterol (percentage difference between highest and lowest quintile of both GI and GL minus 11%) and higher fasting plasma triacylglycerol (percentage difference plus 75% for GI and plus 18% for GL) [34
] and were positively associated with plasma C-peptide (percentage difference 16% for GL) [35
], a marker of insulin secretion. In addition, cereal fiber (in HPFS, percentage difference minus 15.6%) [35
] as well as whole grain intakes (in HPFS and NHS, percentage difference minus 14%) [33
] were inversely associated with plasma C-peptide. The dietary II was positively associated with fasting plasma triacylglycerol in a subgroup of the current study population (percentage difference plus 26%, unpublished data). Dietary GI and GL have been positively associated with type 2 diabetes [26
] and inverse associations have been observed between dietary intake of fiber [26
] and whole grains [36
], and type 2 diabetes. These studies demonstrate the ability to measure the here investigated dietary exposure variables by means of FFQ. The dietary variables investigated here tended to be associated with a healthier lifestyle reflected by less smoking and higher vigorous physical activity. Although it is possible that measurement error in exposure or residual confounding may have contributed to the observed null-associations, it is rather unlikely that a substantial association would have been missed. However, we cannot rule out a small association.
Detection bias due to PSA screening is critical when participation in PSA-screening is associated with the exposure of interest. This was not the case for GI, GL and II, but PSA-screening seemed to be slightly more frequent in men with high intake of fiber and whole grains in our study population. Thus, the positive association between dietary fiber and high-grade prostate cancer, as well as whole grain intake and total prostate cancer observed in the total cohort may be biased by more frequent PSA-screening among men with high intake of fiber or whole grains, respectively. Supporting this assumption, these positive associations disappeared in a cohort restricted to men who reported to have participated in at least one PSA-test by 2004. In contrast, results for GI, GL, and II remained unchanged after this restriction for PSA screening.
We had hypothesized that the group of dietary exposure variables investigated in this study might be related to incidence of prostate cancer via alterations in fasting circulating insulin levels. Insulin may affect cancer development by both exerting mitogenic effects itself and by decreasing IGF binding proteins (IGFBPs) thereby increasing the bioactivity of IGF-1, which is a generally stronger mitogen than insulin [37
]. Although not entirely consistent, several studies have shown that fasting insulin [8
], or plasma C-peptide [40
] were associated with prostate cancer incidence or mortality. In addition, a large body of epidemiological evidence is in favor of a moderate positive association between circulating IGF-1 and risk of prostate cancer, which is supported by two recent meta-analyses [6
]. The fact that we did not observe any strong associations between dietary factors that are likely to influence insulin response, and risk of prostate cancer does not necessarily negate the insulin hypothesis, because absolute long-term insulin exposure is additionally influenced by anthropometric factors, physical activity, genetic and epigenetic factors, dietary factors influencing insulin resistance (which differ from those that influence secretion), and beta-cell depletion. However, our findings suggest that in the context of the many factors that influence insulin levels, dietary factors specifically related to the insulin response are unlikely to be major determinants of prostate cancer risk.
A meta-analysis suggested an overall direct positive association between GI and GL and colorectal and endometrial cancer [42
]. The present study, however, suggests that for prostate cancer, GI and GL are not strong predictors of incidence or mortality. It is notable that BMI, which causes hyperinsulinemia, is a strong risk factor for those cancers, whereas it is not for total prostate cancer; hence our data are consistent with these observations. In summary, the findings of this study do not support the hypothesis that dietary GI, GL, II, fiber and whole grains are related to the risk of prostate cancer.