Diabetes and Risk of Prostate Cancer in a Prospective Cohort of US Men

Transcription

1 American Journal of Epidemiology Copyright 2005 by the Johns Hopkins Bloomberg School of Public Health All rights reserved Vol. 161, No. 2 Printed in U.S.A. DOI: /aje/kwh334 Diabetes and Risk of Prostate Cancer in a Prospective Cohort of US Men Carmen Rodriguez, Alpa V. Patel, Alison M. Mondul, Eric J. Jacobs, Michael J. Thun, and Eugenia E. Calle From the Department of Epidemiology and Surveillance Research, American Cancer Society, Atlanta, GA. Received for publication April 12, 2004; accepted for publication June 29, One previous study has suggested that diabetes may decrease risk of prostate cancer but only several years after diagnosis of diabetes. The authors examined the role of timing of diabetes diagnosis in relation to risk of prostate cancer among men in the Cancer Prevention Study II Nutrition Cohort. Participants in the study completed a mailed questionnaire including information on diabetes at enrollment in 1992 and at follow-up questionnaires in 1997 and Historical information on diabetes was also available from a previous study in The authors documented 5,318 cases of incident prostate cancer through August 31, 2001, among 72,670 men. Results from Cox proportional hazards models showed that diabetes was associated with a lower incidence of prostate cancer (rate ratio (RR) = 0.67, 95% confidence interval (CI): 0.60, 0.75). This association differed significantly by time since diagnosis of diabetes (p < ); risk of prostate cancer was slightly increased during the first 3 years after diagnosis of diabetes (RR = 1.23, 95% CI: 0.92, 1.65) but was reduced among men diagnosed 4 or more years before (RR = 0.63, 95% CI: 0.56, 0.71). Study results are consistent with the hypothesis that diabetes is associated with reduced risk of prostate cancer but only several years after diagnosis of diabetes. cohort studies; diabetes mellitus; prostatic neoplasms Abbreviations: CI, confidence interval; CPS, Cancer Prevention Study; RR, rate ratio. Type 2 diabetes is a complex metabolic disease characterized initially by insulin resistance and hyperinsulinemia. However, with increasing duration of the disease, the pancreas loses its ability to create insulin because of damage to the pancreatic B cells, and circulating levels of insulin decrease. Risk of prostate cancer has been associated with high circulating levels of insulin and insulin resistance (1 3). As a result, it has been hypothesized that risk may initially be increased among men recently diagnosed with diabetes, but risk may later be reduced as insulin levels decrease (4). The association between diabetes and prostate cancer risk has been studied in several epidemiologic studies. Four (5 8) of the five largest studies (5 9) found reductions in risk ranging from approximately 10 to 40 percent. Results from several smaller studies (less than 50 prostate cancer cases among diabetic men), however, have been mixed (10 17). The hypothesis that the relation between diabetes and prostate cancer may differ by time since diabetes diagnosis has been examined with sufficient power in only three studies (7 9). In the Health Professionals Follow-up Study cohort, prostate cancer incidence was higher immediately after diabetes diagnosis and lower among men with diagnosis of diabetes years earlier (7). The time since diabetes diagnosis did not modify the association between diabetes and prostate cancer in the large Physicians Health Study (8). In contrast, in the Cancer Prevention Study I cohort (9), prostate cancer incidence was significantly increased among men with diagnosis of diabetes for 5 or more years, although this result was based on only 22 cases. To clarify the potentially complex relation between timing of diabetes mellitus diagnosis and prostate cancer, we examined the association between self-reported diabetes and prostate cancer incidence in the Cancer Prevention Study II Nutrition Cohort. Because participants in this cohort had reported diabetes status at several time points, we were able to examine whether the risk of prostate cancer differed by time since diagnosis of diabetes. Reprint requests to Dr. Carmen Rodriguez, Department of Epidemiology and Surveillance Research, American Cancer Society, 1599 Clifton Road, NE, Atlanta, GA ( crodrigu@cancer.org). 147

2 148 Rodriguez et al. MATERIALS AND METHODS Men in this study were selected from the 86,404 male participants in the Cancer Prevention Study (CPS) II Nutrition Cohort (hereafter referred to simply as the Nutrition Cohort), a prospective study of cancer incidence and mortality among 184,192 US men and women. The Nutrition Cohort is a subgroup of the approximately 1.2 million participants in the CPS-II, a prospective mortality study established by the American Cancer Society in 1982 (18). Members of the CPS-II cohort who resided in 21 states with population-based state cancer registries and were years of age in 1992 were invited to participate in the Nutrition Cohort by completing a mailed questionnaire. The recruitment and characteristics of Nutrition Cohort participants are described in detail elsewhere (19). At enrollment in 1992, participants completed a selfadministered mailed questionnaire that included demographic, lifestyle, and medical factors. Follow-up questionnaires were sent to cohort members in 1997, 1999, and again in The response rate among living participants for all of the follow-up questionnaires (after multiple mailings) was at least 90 percent. For the present study, the follow-up period ended August 31, 2001, just prior to the first mailing of the 2001 questionnaire. In each Nutrition Cohort questionnaire, participants were asked whether they had ever been diagnosed with diabetes and whether they used insulin. Men were classified as diabetic if they responded yes to either of those questions. In addition, on the 1997 questionnaire, men were asked when they were diagnosed with diabetes (before 1992, , , and 1996 or after). Because all Nutrition Cohort participants were also participants in the original CPS-II cohort, information on diabetes was also available from the 1982 CPS-II questionnaire. We excluded from this analysis men who were lost to follow-up from baseline through August 31, 2001 (n = 3,431), and who reported any prevalent cancer (except nonmelanoma skin cancer) at baseline (n = 9,004). We also excluded men whose reported prostate cancer could not be confirmed (n = 464), men with stage I prostate cancer (n = 48), and men with uninterpretable responses on diabetes status (n = 691). In addition, men with inconsistent information on their diabetes status between 1992, 1997, 1999, and 2001 were censored at the time of their previous report, men with any incident of verified cancer were censored at the date of cancer report, and men lost to follow-up were censored at the time of their last questionnaire. In an attempt to limit the analysis to type 2 diabetes, we excluded men (n = 96) who reported having been diagnosed with diabetes at an age earlier than 30 years. After these exclusions, the analytical cohort consisted of 72,670 men. There were a total of 5,318 incident cases of fatal or nonfatal prostate cancer. Incident cases of prostate cancer were identified through a self-report of cancer on any of the questionnaires (n = 4,907) and subsequently verified by medical records (n = 4,097) or from linkage with state cancer registries (n = 810). An additional 110 prostate cancer cases were also identified if recorded as the underlying cause of death on a TABLE 1. Demographic characteristics of men according to diabetic status, Cancer Prevention Study II Nutrition Cohort, Ever diagnosed with diabetes No (n = 62,617) (age-adjusted %* or mean) Yes (n = 10,053) (age-adjusted %* or mean) Age group (years) < Race White Black Other/missing Educational level Less than high school High school graduate Some college College graduate Graduate school Body mass index (kg/m 2 ) < < Family history of prostate cancer No Yes Lycopene intake (µg) Mean 4, ,904.4 Total fat intake (g) Mean Total calcium (mg) Mean Total calorie intake (kcal) Mean 1, ,821.0 * Percentages and means are directly adjusted to the age distribution of the entire study population. Percentages may not sum to 100 because of missing data. death certificate through December 31, 2000, among cohort members who did not report cancer at enrollment (20). Seventy-six of the 110 deaths were subsequently verified through linkage with cancer registries. Finally, 301 cases of prostate cancer were not reported as prostate cancer but were identified during confirmation of another reported cancer. For analysis of aggressive prostate cancer, we included prostate cancer cases verified by medical records with stages III and IV, those with a Gleason score of 8 or higher or grade 3 4, prostate cancer cases verified by the cancer state registry and classified as regional or distant, and prostate

3 Diabetes and Prostate Cancer 149 TABLE 2. Prostate cancer incidence by diabetes status, Cancer Prevention Study II Nutrition Cohort, cancer deaths. A total of 1,306 aggressive prostate cancer cases were included in this analysis. We used Cox proportional hazards modeling (21) to examine the association between diabetes and incident prostate cancer while adjusting for other potential confounding factors. Exposure information was calculated as a timedependent variable based on the diabetes status reported in the 1992, 1997, 1999, or 2001 questionnaire. All Cox models were stratified on single year of age at enrollment and were adjusted for race, education, family history of prostate cancer, body mass index, and quintiles of total energy intake, energy-adjusted total fat intake, lycopene, and total calcium intake. Nondietary variables, other than age, were modeled as dummy variables using the categories shown in table 1. Information on prostate-specific antigen testing was first collected in the 1997 questionnaire and was modeled as a time-dependent variable. We examined the association between diabetes and prostate cancer incidence by time since diagnosis using information from the 1982 CPS-II questionnaire and the 1992, 1997, and 1999 Nutrition Cohort questionnaires. Men with a diagnosis of diabetes were classified in mutually exclusive groups based on time since diagnosis as shown in tables 2 and 3. Time since diabetes diagnosis was modeled as a timedependent variable. Men who reported being diagnosed with diabetes in the 1992 but not in the 1982 questionnaire were initially assigned 5 years since diagnosis. All p values presented for homogeneity of rate ratios are two sided and calculated using the Wald test. RESULTS Cases * Adjusted for age at entry, race, education, body mass index, family history of prostate cancer, prostate-specific antigen, quintiles of total fat, and quintiles of energy-adjusted intakes of the following: lycopene, total calcium, and total fat. A history of physician-diagnosed diabetes was reported by 10,053 men (13.8 percent), and 1,643 (16.3 percent) of those Person-years Rate ratio* 95% confidence interval* Diabetes No 4, , Yes , , 0.75 Time since diabetes diagnosis Not diabetic 4, , <4 years 46 5, , years , , 0.71 p homogeneity <7 years 79 11, , <10 years 82 12, , <13 years 85 11, , years 51 7, , 0.78 p trend 0.98 reported using insulin injections. The age-adjusted percentage of distribution of potential prostate cancer risk factors varied with diabetes status (table 1). Men who reported diabetes were more likely to be older, to be less educated or heavier, and to consume a diet higher in lycopene. In addition, men who reported having been diagnosed with diabetes were more likely to be Black. We also examined the prevalence of prostate-specific antigen testing among men in this analysis from information collected in the 1997 and 1999 follow-up questionnaires. A total of 81.9 percent of diabetic men and 85.5 percent of nondiabetic men reported having received a prostate-specific antigen test in any of these questionnaires. Self-reported diabetes was associated with lower prostate cancer incidence rates after adjustment for age, race, education, and prostate-specific antigen testing (rate ratio (RR) = 0.67, 95 percent confidence interval (CI): 0.60, 0.75) (table 2). The association between diabetes and risk of prostate cancer differed significantly by time since diagnosis (p homogeneity = ); prostate cancer rates among men diagnosed with diabetes within the last 3 years were higher than among nondiabetic men (RR = 1.23, 95 percent CI: 0.92, 1.65). Men who had had diabetes 4 or more years had a lower rate of prostate cancer than did men without diabetes (RR = 0.63, 95 percent CI: 0.56, 0.71), although no significant trend of decreasing prostate cancer risk with increasing time since diagnosis after the initial 3 years was observed (table 2). Results from models adjusting only for age and race were similar to the multivariate-adjusted rate ratios (data not shown). Overall, the association between diabetes and prostate cancer was similar in a stratified analysis by stage/grade of prostate cancer at diagnosis (table 3). The pattern of increased prostate cancer risk shortly after diabetes diag-

4 150 Rodriguez et al. TABLE 3. Prostate cancer incidence by diabetes status according to histologic presentation at diagnosis, Cancer Prevention Study II Nutrition Cohort, Cases * Adjusted for age at entry, race, education, body mass index, family history of prostate cancer, prostate-specific antigen, quintiles of total fat, and quintiles of energy-adjusted intakes of the following: lycopene, total calcium, and total fat. Includes prostate cancer stages I and II with a Gleason score of less than 8. Includes prostate cancer stages III and IV and those with a Gleason score of 8 or higher. nosis and lower risk with long duration of the disease was observed for both aggressive and nonaggressive prostate cancer (table 3). No significant interactions were observed between diabetes and any of the other potential prostate cancer risk factors included in this analysis. Having a history of prostate-specific antigen testing did not modify the association between diabetes and risk of prostate cancer (data not shown). Because information on prostate-specific antigen testing was not collected until the 1997 questionnaire, we conducted a sensitivity analysis starting with follow-up in During this follow-up period, diabetes was associated with a 30 percent decreased risk of incident prostate cancer (RR = 0.69, 95 percent CI: 0.58, 0.82). Nearly all men, Person-years regardless of diabetes status, were receiving prostatespecific antigen testing during this period. DISCUSSION Rate ratio* 95% confidence interval* Nonaggressive prostate cancer Diabetes No 3, , Yes , , 0.80 Time since diabetes diagnosis Not diabetic 3, , <4 years 34 5, , years , , 0.76 p homogeneity <7 years 63 11, , <10 years 62 12, , <13 years 67 11, , years 43 7, , 0.87 p trend 0.76 Aggressive prostate cancer Diabetes No 1, , Yes 74 48, , 0.72 Time since diabetes diagnosis Not diabetic 1, , <4 years 12 5, , years 62 42, , 0.66 p homogeneity <7 years 16 11, , <10 years 20 12, , <13 years 18 11, , years 8 7, , 0.84 p trend 0.62 In this large prospective study, self-reported diabetes was associated with slightly increased risk of incident prostate cancer during the first few years after diagnosis of diabetes and reduced risk among long-term diabetic men. Our results are consistent with the hypothesis that risk of prostate cancer differs by time since diagnosis of diabetes. Our results with respect to time since diabetes diagnosis and prostate cancer risk are similar to those observed in the Health Professionals Follow-up Study (7), a cohort of similar design, but they differed from those of the US Physi-

5 Diabetes and Prostate Cancer 151 cians Health Study (8) and the CPS-I cohort (9). In the Health Professionals Follow-up Study, risk of prostate cancer was 24 percent higher in the first 5 years after diagnosis of diabetes (RR = 1.24, 95 percent CI: 0.87, 1.77) and was significantly decreased after 10 years (RR = 0.54, 95 percent CI: 0.37, 0.78). It is not clear why our results differed from those for the CPS-I cohort (9), in which men with diabetes for more than 5 years had significantly increased risk of incident prostate cancer. The CPS-I was conducted from 1959 to 1972, a period of time with no early prostate cancer detection practices other than the digital rectal examination. It is possible that, during this time period, prostate cancers would have been more likely to be detected among diabetic men than among nondiabetic men. Diabetic men in the CPS-II Nutrition Cohort were somewhat less likely to receive prostate-specific antigen testing for prostate cancer than were men with no diabetes. Therefore, we cannot rule out detection bias among nondiabetic men as a possible explanation for the lower prostate cancer rates among diabetic men. However, detection bias seems unlikely to fully account for our results, given that the association between diabetes and prostate cancer was seen for both nonaggressive and aggressive prostate cancer at diagnosis. In addition, we observed reduced risk among diabetics during the follow-up period, a time when prostate-specific antigen testing was very prevalent among both diabetic and nondiabetic men in our study cohort. The inverse association between diabetes and prostate cancer observed in this study has also been reported in epidemiologic studies conducted before the advent of widespread prostate-specific antigen testing, including a very large Swedish study (5). It is plausible that prostate-specific antigen testing could sometimes be conducted as part of diagnostic work-up for diabetic symptoms. In theory, this could have resulted in detection bias, contributing to the increased risk of prostate cancer observed among men with a recent diabetes diagnosis. Such detection bias is unlikely to have strongly influenced our results, because men were not classified as newly diabetic in the proportional hazards model until the date they reported diabetes on a questionnaire, usually many months after the date of diabetes diagnosis. Metabolic and hormonal changes associated with diabetes could explain the temporal relation observed in this and one previous study (7). First, insulin levels vary dramatically with increasing duration of diabetes; insulin levels are elevated as a result of insulin resistance during the years leading up to a diabetes diagnosis, but they later decrease to levels lower than those of nondiabetic men as a result of damage to pancreatic B cells (22). Insulin has been reported to stimulate the growth of a rat prostate cancer cell line in vitro (23) and has been associated with higher risk of prostate cancer (2) and higher recurrence of the disease (24). Second, insulin downregulates insulin-like growth factor binding protein 1 that controls the free fraction of insulinlike growth factor I (25). Plasma insulin-like growth factor I levels, therefore, are elevated at the beginning of the disease and decrease as the disease progresses (26, 27). Insulin-like growth factor I has been associated with increased prostate cancer risk in several prospective studies (28, 29). Finally, serum levels of insulin correlate negatively with circulating levels of testosterone, dihydrotestosterone, and sex hormone-binding globulin and correlate positively with the testosterone/sex hormone-binding globulin ratio (2, 30). Therefore, serum circulating levels of testosterone are lower in diabetic than nondiabetic men (31, 32), and testosterone has been associated with elevated risk of prostate cancer (33), although not consistently (34). The reliance on self-report of diabetes is a limitation of this study. We cannot differentiate between type I and type II diabetes, although we excluded men reporting insulin use before the age of 30 years, and the great majority of diabetes diagnoses after this age would be expected to be type II. Second, it is possible that some men reported impaired glucose intolerance as diabetes, which could lead to misclassification of diabetes status. However, this type of misclassification should bias the estimate toward the null. It is noteworthy that the percentage (13.8 percent) of men reporting diabetes in our cohort is in agreement with the prevalence (12.3 percent) of diabetes in the National Health and Nutrition Survey for people years of age (35). Results of this study are consistent with the hypothesis that diabetes reduces the risk of prostate cancer starting several years postdiagnosis, and they provide limited support for a role of insulin and testosterone in prostate carcinogenesis. REFERENCES 1. Hsing AW, Gao YT, Chua S Jr, et al. Insulin resistance and prostate cancer risk. J Natl Cancer Inst 2003;95: Hsing AW, Chua S Jr, Gao YT, et al. Prostate cancer risk and serum levels of insulin and leptin: a population-based study. J Natl Cancer Inst 2001;93: Barnard RJ, Aronson WJ, Tymchuk CN, et al. Prostate cancer: another aspect of the insulin-resistance syndrome? Obes Rev 2002;3: Giovannucci E. Medical history and etiology of prostate cancer. Epidemiol Rev 2001;23: Weiderpass E, Ye W, Vainio H, et al. Reduced risk of prostate cancer among patients with diabetes mellitus. Int J Cancer 2002;102: Wideroff L, Gridley G, Mellemkjaer L, et al. Cancer incidence in a population-based cohort of patients hospitalized with diabetes mellitus in Denmark. J Natl Cancer Inst 1997;89: Giovannucci E, Rimm EB, Stampfer MJ, et al. Diabetes mellitus and risk of prostate cancer (United States). Cancer Causes Control 1998;9: Zhu K, Lee IM, Sesso HD, et al. History of diabetes mellitus and risk of prostate cancer in physicians. Am J Epidemiol 2004; 159: Will JC, Vinicor F, Calle EE. Is diabetes mellitus associated with prostate cancer incidence and survival? Epidemiology 1999;10: Thompson MM, Garland C, Barrett-Connor E, et al. Heart disease risk factors, diabetes, and prostatic cancer in an adult community. Am J Epidemiol 1989;129: Henderson BE, Bogdanoff E, Gerkins VR, et al. Evaluation of cancer risk factors in a retirement community. Cancer Res 1974;34: Rosenberg DJ, Neugut A, Ahsan H, et al. Diabetes mellitus and the risk of prostate cancer. Cancer Invest 2002;20:

6 152 Rodriguez et al. 13. Ragozzino M, Melton LJ 3rd, Chu CP, et al. Subsequent cancer risk in the incidence cohort of Rochester, Minnesota, residents with diabetes mellitus. J Chronic Dis 1982;35: Steenland K, Nowlin S, Palu S. Cancer incidence in the National Health Nutrition Survey follow-up data: diabetes, cholesterol, pulse, and physical activity. Cancer Epidemiol Biomarkers Prev 1995;4: La Vecchia C, Negri E, Franceschi S, et al. A case-control study of diabetes mellitus and cancer risk. Br J Cancer 1994;70: Coughlin SS, Neaton JD, Segupta A. Cigarrette smoking as a predictor of death from prostate cancer in 348,874 men screened for the Multiple Risk Factor Intervention Trial. Am J Epidemiol 1996;143: Tavani A, Gallus S, Bosetti C, et al. Diabetes and the risk of prostate cancer. Eur J Cancer Prev 2002;11: Garfinkel L. Selection, follow-up, and analysis in the American Cancer Society prospective studies. Natl Cancer Inst Monogr 1985;67: Calle EE, Rodriguez C, Jacobs EJ, et al. The American Cancer Society Cancer Prevention Study II Nutrition Cohort: rationale, study design, and baseline characteristics. Cancer 2002;94: Calle EE, Terrell DD. Utility of the National Death Index for ascertainment of mortality among Cancer Prevention Study II participants. Am J Epidemiol 1993;137: Cox DR. Regression models and life tables (with discussion). J R Stat Soc (B) 1972;34: Polonsky KS. Evolution of beta-cell dysfunction in impaired glucose tolerance and diabetes. Exp Clin Endocrinol Diabetes 1999;107(suppl 4):S Polychronakos C, Janthly U, Lehoux JG, et al. Mitogenic effects of insulin and insulin-like growth factors on PA-III rat prostate adenocarcinoma cells; characterization of the receptors involved. Prostate 1991;19: Lehrer S, Diamond EJ, Stagger S, et al. Increased serum insulin associated with increased risk of prostate cancer recurrence. Prostate 2002;50: Bach LA, Rechler MM. Insulin-like growth factors and diabetes. Diabetes Metab Rev 1992;8: Clauson PG, Brismar K, Hall K, et al. Insulin-like growth factor-i and insulin-like growth factor binding protein 1 in a representative population of type 2 diabetic patients in Sweden. Scand J Clin Lab Invest 1998;58: Haffner SM, Katz MS, Dunn JF. The relationship of insulin sensitivity and metabolic clearance of insulin to adiposity and sex hormone binding globulin. Endocrinol Res 1990;16: Stattin P, Bylund A, Rinaldi S, et al. Plasma insulin-like growth factor-1, insulin-like growth factor-binding proteins, and prostate cancer risk: a prospective study. J Natl Cancer Inst 2000; 92: Chan JM, Stampfer MJ, Ma J, et al. Insulin-like growth factor- I (IGF-I) and IGF binding protein-3 as predictors of advancedstage prostate cancer. J Natl Cancer Inst 2002;94: Haffner SM. Sex hormones, obesity, fat distribution, type 2 diabetes and insulin resistance: epidemiological and clinical correlation. Int J Obes Relat Metab Disord 2000;24(suppl 2):S Andersson B, Marin P, Lissner L, et al. Testosterone concentrations in women and men with NIDDM. Diabetes Care 1994;17: Haffner SM, Shaten J, Stern MP, et al. Low levels of sex hormone-binding globulin and testosterone predict the development of non-insulin-dependent diabetes mellitus in men. MRFIT Research Group. Multiple Risk Factor Intervention Trial. Am J Epidemiol 1996;143: Gann PH, Hennekens CH, Longcope C, et al. Prospective study of sex hormone levels and risk of prostate cancer. J Natl Cancer Inst 1996;88: Stattin P, Lumme S, Tenkanen L, et al. High levels of circulating testosterone are not associated with increased prostate cancer risk: a pooled prospective study. Int J Cancer 2004;108: Harris MI, Flegal KM, Cowie CC, et al. Prevalence of diabetes, impaired fasting glucose, and impaired glucose tolerance in U.S. adults. Diabetes Care 1998;21: