|Home | About | Journals | Submit | Contact Us | Français|
History of diabetes may be associated with decreased prostate cancer (PCa) risk. Published studies have not always accounted for time since diabetes diagnosis or confounding and effect modification by lifestyle factors. The authors investigated the relationship between diabetes and PCa risk in men in the Health Professionals Follow-Up Study from 1986–2004. During that time, 4,511 new PCa cases were identified. Multivariate hazard ratios (HR) were estimated using Cox regression. The HR of PCa comparing men with versus without diabetes was 0.83 and 95% confidence interval (CI): 0.74,0.94. PCa risk was not reduced in the first year after diabetes diagnosis (HR: 1.30, CI: 0.97, 1.72), was lower for men diagnosed for 1–6 years (HR: 0.82, CI: 0.66, 1.02), and was even lower for men who had been diagnosed for 6–15 (HR: 0.75, CI: 0.61, 0.93) or >15 years (HR: 0.78, CI: 0.63, 0.96). Reduced PCa risk was stronger in men diagnosed before 1994 (pre-PSA era) versus after 1994. The authors also demonstrated that obese and diabetic men had a lower HR for PCa than those who were either not obese and diabetic or obese and non-diabetic. Results are consistent with the hypothesis that diabetes is associated with reduced PCa risk. Potential biological mechanisms are discussed.
Type 2 diabetes mellitus (DM) is a metabolic disease that has been positively associated with an increased risk of cancers of some organs such as pancreas, liver, biliary tract, endometrium, kidney, colon, and esophagus 1–4. In contrast, the association between DM and prostate cancer (PCa) may be inverse. This inverse association was illustrated in two separate meta-analyses of studies published between 1971–2002 and between 1971–2006 5, 6. Bonovas et al. examined risk in approximately nine thousand prostate cancer patients and demonstrated that diabetic patients have a statistically significant 9% decrease in risk of developing PCa 5. The second meta-analysis included approximately twelve thousand additional prostate cancer cases 7–12 and the overall effect estimate of all relevant studies published through 2006 suggests an inverse association (HR: 0.84, 95% CI: 0.76, 0.93) 6. Additionally, in the past year several more studies examining this topic have also reported both significant and non-significant inverse relationships between DM and PCa 13–16. The biological basis for this inverse relationship may be routed in differential insulin, insulin-like growth factor-1, and/or bioavailable testosterone levels17 and if these two diseases are, in fact, inversely associated, this relationship may further our understanding of prostate cancer biology.
We investigated the relationship between DM and PCa in the Health Professionals Follow-Up Study (HPFS), with 1,369 PCa cases diagnosed during follow-up from 1986 to 1994, and demonstrated an inverse association 18. We now examine this relationship with further follow-up to 2004, with 3,142 additional cases of prostate cancer reported and 237 more prostate cancer cases among diabetics reported. Since only some studies evaluated the extent to which diabetic men are protected from PCa changes as time since diabetes diagnosis increases 10–12, 16, 18–20, we examined the temporal relationships between these two diseases. Further, the implementation of Prostate-Specific Antigen (PSA) testing may have changed the scope of disease that is diagnosed as prostate cancer. Prior to screening, disease was detected mostly at the clinical level, often as an advanced cancer that had a high potential for metastasis. PCa screening is able to detect disease at a much earlier localized stage, and it is not yet clear whether all of these PSA-detected cancers have the potential to evolve into clinically detectable disease. Thus, we examined whether PSA screening influenced the relationship between DM and PCa, and we examined the relationship according to grade and stage of prostate cancer. Finally, we examined whether body mass index (BMI) and physical activity levels, strong determinants of DM, modified the relationship between DM and PCa.
The HPFS, an ongoing prospective cohort study of the causes of cancer and heart disease in men, consists of 51,529 United States male dentists, optometrists, osteopaths, podiatrists, pharmacists, and veterinarians, who were 40 to 75 years old at baseline. The men responded to a mailed baseline questionnaire in 1986, which elicited information on age, marital status, height and weight, ancestry, medications, disease history (including diabetes mellitus and cancer), physical activity, history of PSA examination, and diet. Follow-up questionnaires were sent in biannually from 1988–2004 to ascertain new cases of a variety of diseases and to update exposure information.
On baseline and all follow-up questionnaires, participants were asked if and when they had been diagnosed with insulin- or non-insulin-dependent DM. Based on standards from the National Diabetes Data Group using participants medical records, we classified a man as diabetic if he met one of the criteria: (i) report of one or more classic symptoms (thirst, polyuria, weight loss, hunger, or pruritus), plus fasting plasma glucose at least 140 mg/dl or random plasma glucose at least 200 mg/dl; or (ii) at least two elevated plasma glucose concentrations on different occasions (fasting at least 140 mg/dl or random at least 200 mg/dl or concentrations at least 200 mg/dl after two or more hours on oral glucose tolerance testing); or (iii) treatment with hypoglycemic medication (insulin or oral hypoglycemic agent) 21. After the return of the baseline questionnaire, additional questionnaires were sent to those who reported new cases in order to confirm diagnosis at which time 94% of new cases were confirmed 22. The validity of self-reported diabetes using the same supplementary questionnaire was examined in the Nurses’ Health Study, a similar population of female health professionals 23. Medical records were obtained from sixty-two randomly selected women who self-reported as diabetic. The records confirmed that sixty-one out of sixty-two were diabetic. Since the self-reports were highly accurate, we considered self-reported diabetes as the exposure.
On the follow-up questionnaires, participants were also asked to report new diagnoses of various cancers including that of the prostate. For each new report of prostate cancer, permission was requested from all study participants (or next-of-kin in the event of death) to obtain hospital and pathology reports, which were reviewed by blinded study investigators. We documented approximately 90% of the cases using medical records and pathology reports. Participants provided information regarding their diagnosis for most of the remaining 10% of cases. Staging was done using information from the questionnaires and the tumor-node-metastasis system was used 24. In the event of death, medical records were reviewed by a study physician. If there was evidence of metastatic prostate cancer and no other apparent cause of death, the death was counted as due to prostate cancer.
We excluded 2,174 men who reported cancer at baseline (other than non-melanoma skin cancer), and 1,574 men because no they did not return the food frequency questionnaire or they did not adequately complete it (70 or more items left blank). After exclusion of 3,748 individuals, each of the remaining 47,781 participants contributed follow-up time beginning on the month that they returned the baseline questionnaire and ending on the month of diagnosis of prostate cancer, month of death from other causes, or the end of the study period—January 31, 2004.
Covariates include age, level of physical activity, current BMI, BMI at 21 years old, height, ancestry, smoking habits, family history, and intakes of calories, bacon, tomato sauce, alpha-linolenic acid, calcium, fish, and vitamin E supplementation.
For PCa-diagnosis, an important factor is PSA screening. We assess previous PSA screening on each biennial questionnaire. Thus, we updated and controlled for prior history of PSA screening (e.g. frequency of screening) at each two-year period and conducted secondary analyses limited to the sub-cohort from 1994 onwards in which most men had a PSA test. Using 1994 as the cutoff for after which time most men were screened for PSA, we used stratification to evaluate if screening-history influenced the HR between DM and PCa.
We calculated incidence rates of prostate cancer for men diagnosed with diabetes, and by years elapsed since initial diagnosis of diabetes, by dividing the number of incident cases by the number of person-years. The HR was computed as the rate among men with diabetes divided by the rate among non-diabetics. We used Cox proportional-hazards regression to estimate HRs when controlling simultaneously for the covariates mentioned above.
In this study, we examined the relationship between DM and total PCa as well as between DM and PCa’s of varying grades and stages. We used the tumor-node-metastasis system for stage and Gleason grade 25. Because stage A1 lesions are relatively benign and highly prone to detection bias, we limited our primary analysis to the non-stage A1 cases. We conducted analyses excluding cases of prostate cancer diagnosed in the first year after the diagnosis of diabetes since newly diagnosed diabetics may receive heightened medical surveillance possibly resulting in the diagnosis of latent prostate cancers.
Table 1 presents means or percentages of demographic, anthropometric, and health-related characteristics for men in our study by status of DM. Men with DM reported that intakes of alcohol and fructose were lower but that total fat, lycopene, calcium, and protein were higher than in the non-diabetic cohort members. As expected, average BMI was higher in diabetics than in non-diabetics. We noted, however, that with increasing time since DM-diagnosis, BMIs decreased and in the group of men who were diagnosed more than 15 years before the time of study, the average BMI was slightly lower than in the non-diabetics. Men with diabetes tended to smoke currently more than non-diabetic men. Table 1 also illustrates that men who described themselves as African-American or Asian-American were more likely to be diagnosed with DM than Caucasian men. Men who were diagnosed with DM were slightly less likely overall to have had a PSA-exam by 1994. When examining time since DM diagnosis, however, those recently diagnosed had higher rates of PSA-exam than men whose DM diagnosis was further in the past. With increasing time since DM-diagnosis, the rate of PSA-exam decreased and was lower in those diagnosed > 15 years ago than in the non-diabetics.
In the multivariate adjusted analysis, we found that a history of DM-diagnosis was associated with a 17% reduced risk of PCa, a 28% reduced risk for localized PCa, a 31% reduced risk for high-grade, and a 24% reduced risk for low-grade PCa (Table 2). For advanced prostate cancer, we also observed an inverse association but the result was not statistically significant. Age-adjusted results were almost identical to the values in the multivariate-adjusted models. For example, for diabetics, the HR(95% CI) for total PCa was 0.82 (0.73, 0.93), for advanced PCa it was 0.90 (0.65, 1.25), for non-advanced PCa it was 0.70 (0.60, 0.81), for high-grade PCa it was 0.70 (0.55, 0.87), and for low-grade PCa it was 0.72 (0.59, 0.87).
Also illustrated in Table 2, as time since DM-diagnosis increased, there was a significant trend of a decrease in risk for total PCa. Although there was no significant trend over time for any of the sub-groups of PCa (possibly due to small sample sizes), it is noteworthy that the risk of being diagnosed with prostate cancer for men >15 years since DM-diagnosis is lower than for those in <1 year since DM-diagnosis in all prostate cancer subtypes.
A recent paper 20 reported that the inverse association between DM and PCa is limited to diabetics who were diagnosed before age 30. We therefore investigated the relationship between PCa and DM and limited the analysis to diabetics who were diagnosed either before or after age 30. We found that there was an inverse relationship in both analyses. The HR for the subgroup limited to men who were diagnosed at or before age 30 was inverse with borderline significance (HR: 0.55, 95%CI: 0.30, 1.03). The relationship in the group of men diagnosed after age 30 was also inverse (HR: 0.85, 95%CI: 0.75, 0.95). In the subgroup of men diagnosed at or before age 30, there was a stronger inverse association than in the group diagnosed after age 30 but due to small sample size the confidence intervals were large.
Table 3 demonstrates that total PCa risk appears to be lower among men who were diagnosed with DM in the pre-PSA versus the PSA era. Similar to the trend observed for total PCa cases, as time since DM-diagnosis increased, PCa risk decreased. In the pre-PSA era subgroup, HRs (95%CI) for PCa were 1.55(0.95, 2.55), 0.83(0.55, 1.27), 0.58(0.37, 0.89), and 0.72(0.48, 1.07) (P for trend=0.12) for ≤1 year since DM-diagnosis, >1–6 years since DM-diagnosis, >6–15 years since DM-diagnosis, and >15 years since DM-diagnosis, respectively. In the PSA-era subgroup, the HRs for the same groups were 1.20(0.85, 1.70), 0.82(0.64, 1.05), 0.83(0.65, 1.06), and 0.81(0.63, 1.04) (P for trend=0.10). Additionally, in the pre-PSA era, the inverse relationship between DM and PCA seems to be stronger in high-grade as opposed to low-grade PCa.
In Table 4, we present data for analyses stratified by study characteristics including BMI, age, and exercise levels. The data in this table suggests that the inverse relationship between DM and prostate cancer may be stronger in men who exercise less, are older, and heavier currently and at age 21. Time-trends (since DM-diagnosis) are also presented. New variables were created for the interactions between diabetes status and the following variables—age, BMI, BMI at age 21, and exercise level. From an analysis of maximum likelihood estimates, P-values were determined. The P-value for the age interaction was 0.07, for the exercise interaction it was 0.50, and for the BMI variables, the P-values were 0.89 and 0.17 for current and age 21, respectively.
We examined whether the effect of DM was modified by BMI and found that the inverse relationship between DM and PCa was stronger in men with BMI greater than or equal to both 25 and 30 (Figure 1). We also investigated this relationship for BMI at age 21 and found stronger inverse relationship in the higher BMI category (Figure 1).
The study presented here confirms an association between DM and risk of PCa. Several studies have investigated this question previously, including an earlier study also in the HPFS cohort. The current study had a number of strengths. First, with more than 4,000 PCa cases, it is one of the largest studies investigating this relationship conducted to date. Second, this study examined the temporal relationships between these two diseases. Third, we investigated any potential modifying role of PSA screening and whether the association differed by sub-types of prostate cancer based on grade and stage. We also examined whether there was a modifying role for BMI, physical activity, and age.
This study demonstrated a temporal association between DM and PCa and specifically, for a man with DM, the risk of being diagnosed with PCa decreases as time since DM-diagnosis increases. This finding is supported by previous research which found a similar trend 10, 12, 16, 18. This association suggests that metabolic and hormonal changes consistent with DM appear to create a less carcinogenic environment for the prostate, and as time passes, this hormonal milieu becomes more prominent. It is possible that the presence of genetic factors such as TCF2 variants, which are associated with both PCa and DM risk, are the true basis of our finding. However, the finding that there is a temporal relationship between DM diagnosis and PCa risk (as time since DM-diagnosis increases, risk of PCa decreases), strongly argues against this.
We found that DM was associated similarly with lower risk of high-grade and low-grade PCa, and the association was slightly stronger in the pre-PSA era. In both pre-PSA and PSA eras, we observed similar association for non-advanced PCa. For advanced PCa, an inverse (nonsignificant) association was observed in the pre-PSA, but not in the PSA era, though the number of cases was small and the confidence intervals wide. Overall, these findings, especially the stronger inverse association between DM and high-grade cancer in the pre-PSA era, argue against detection bias being the primary cause of the inverse association between DM and PCa.
We investigated several other population subgroups of diabetic men in order to determine if the association was modified by BMI and physical activity, two important determinants of DM. Our study demonstrated that among men with higher BMIs currently and at age 21, the inverse association between DM and PCa was stronger than among leaner men. A similar pattern was observed for less physically active men, although the interaction was not significant. Although their results for a trend were not significant, Calton et al. recently reported that in the NIH-AARP cohort, for BMI at 18 years of age, men in the highest BMI category (≥30) had a particularly strong reduced risk for PCa (0.42(0.19, 0.89)). In the AARP cohort, other measures of adiposity did not reveal a similar finding 16. Others have found no difference for PCa risk between BMI-groups 7, 14. Higher BMI is a strong risk factor for DM, and our results suggests that DM, particularly that which is associated with higher BMI, is associated with a reduced risk of PCa.
Our results also illustrate different HRs for younger and older men. We observed that the decreased risk of PCa among diabetics is stronger in older men. This observation may explain why a recent study, with the 77.4% of cases diagnosed before age 65, reported that DM status was not significantly associated with PCa risk 20. The same publication also reported that the association was limited to those with early-onset diabetes. We observed an inverse association between DM and PCa in both subgroups (those diagnosed at or before age 30 as well as for those diagnosed after age 30). The relationship was stronger in the subgroup diagnosed at or before age 30 but due to small sample size, it is difficult to make any conclusions at this time.
The mechanism by which DM may reduce a man’s risk of PCa is not yet known but the metabolic and hormonal environment of a diabetic man is arguably consistent with protection from prostate cancer. One possible explanation is that with worsening DM, decreased levels of testosterone (T) provide an environment that is not conducive to proliferation of PCa cells. Published literature contains data both supporting and refuting this hypothesis. Some prospective studies suggest that increased T may be a risk factor for prostate cancer 26 and others, including a recent pooled analysis, report either no or small associations between serum sex-hormone levels, including T, and subsequent prostate cancer risk 27–33. Several studies suggest that men with low T levels may be at increased risk for more poorly-differentiated PCa 34–36. Although the current evidence may not suggest a strong relationship between adult T levels and PCa risk, it is possible that T levels earlier in life or levels of T in long-term diabetics (especially obese men) may reach levels below a threshold required to lower risk of PCa. In addition to changes in T, low levels of SHBG have been shown to be associated with increased risk of prostate cancer 26. Some animal and human studies support our hypothesis that in diabetics, there is less bioavailable T in the serum. Jackson and Hutson showed that diabetic rats have reduced T levels 37. A recent analysis of hormonal profiles of diabetic men demonstrated that as time since DM-diagnosis increased, the ratio of T:SHBG decreased a decrease in bioavailable T 17. Similarly, studies in men show that as blood glucose levels increase, T levels decrease 38, 39.
While patients with DM phenotypically presents with hyperglycemia, they all present with a relative hypoinsulinemia 40. Insulin has been shown to be a growth factor for prostatic epithelium in vitro 41, to stimulate growth of a rat prostate cancer cell line in vitro 42, and is associated with both higher risk 43 and recurrence 44 of the prostate cancer. Therefore, decreased insulin may have a growth-inhibitory effect on these cells and if long-term diabetic patients experience reduced levels of circulating insulin, they may be at a reduced risk of developing prostate cancer. Additionally, there is evidence that higher serum insulin levels are associated with poor outcome in prostate cancer 45. These observations strengthen the hypothesis that there is an inverse association between diabetes and prostate cancer. While some prospective studies demonstrate that there may be an association between hyperinsulinemia and prostate cancer risk 17, 46, there are also several studies that do not support a role for insulin in prostate cancer risk and thus it continues to be an area of investigation47, 48.
A proposed hypothesis for how hypoinsulinemia may decrease prostate carcinogenesis is by limiting the bioavailability of insulin-like growth factor I (IGF-I) which has been shown to be an important risk factor for PCa 48–51. IGF-I is a growth regulator that has been shown to be associated with carcinogenesis 52, and activation of IGF-I receptors stimulates the proliferation and inhibits apoptosis of PCa cells in an IGF-I-dose-dependent manner 50, 53. IGF-I circulates bound to binding proteins such as IGF-binding protein-1 (IGFBP-1) and IGF-binding protein-3 (IGFBP-3). In an insulin-deficient environment, IGFBP-1 is up-regulated, presumably resulting in less bioavailable IGF-I 54–56. Additionally, when bound to IGF-1, IGFBP-3 may decrease the bioavailability of IGF-1. It is possible that in diabetic men, IGF-1 levels may decrease over time since DM diagnosis and that this could be mediated through changes in IGFBP-3 levels. Although recent studies suggest that IGF-1 levels are not significantly different between diabetics and controls17, future studies should examine this hypothesis as well as the role that IGFBP-1 plays in prostate carcinogenesis.
A significant amount of research has determined risk factors that may contribute to prostate cancer. Factors include age, race, geographic location, diet, family history, and possibly DM. The overall evidence for an inverse association between DM and prostate cancer continues to grow and studying these biological clues will continue to provide insight into the metabolic and hormonal changes behind prostatic cancer. The absolute health effect of DM is not beneficial but the data presented here provides us further insight into our understanding of the process of prostate cells becoming malignant.
This work was supported by the National Cancer Institute [P01CA055075]; and the Department of Defense [W81XWH-06-1-0188]. The sponsors had no role in the design and conduct of the study; collection, management, analyses, and interpretation of the data; or preparation, review, or approval of the manuscript. The content is solely the responsibility of the authors and should not be constituted to represent the official views of the National Cancer Institute, the National Institutes of Health, or the Department of Defense.
Paper novelty and impact:
In this manuscript, we report the results of an investigation into the relationship between diabetes mellitus (DM) and prostate cancer (PCa) risk in men enrolled in the Health Professionals Follow-Up Study (HPFS) from 1986–2004. This analysis investigates changes in PCa risk as time since DM diagnosis increases and we also examine confounding and effect modification by lifestyle factors including PSA-examination status and BMI.
This manuscript provides deeper insight into the pathophysiology of PCa. Studying these biological clues will provide insight into the genetic, metabolic, and hormonal changes behind PCa. The overall health effect of DM is not beneficial but these data provide us further insight into our understanding of the etiology of PCa.