Antidiabetic Medication and Prostate Cancer Risk: A Population-based Case-Control Study

Abstract

Decreased risk of prostate cancer in diabetic men has been reported. The authors evaluated the association between antidiabetic medication use and prostate cancer at the population level. All incident prostate cancer cases in Finland during 1995–2002 were identified from the Finnish Cancer Registry. Matched controls were provided by the Population Register Center (24,723 case-control pairs). Information on medication use was obtained from a comprehensive prescription database. Multivariable-adjusted odds ratios were computed by using conditional logistic regression. The authors found that prostate cancer risk was decreased for antidiabetic medication users (odds ratio = 0.87, 95% confidence interval: 0.82, 0.92). The decrease was observed for most drug groups. The odds ratio decreased in a dose-dependent fashion by quantity of use. Duration of antidiabetic treatment was inversely associated with overall prostate cancer risk and risk of advanced cancer. Similar risk reduction for users of different antidiabetic drugs suggests that diabetes, instead of the medication itself, is behind the association. This finding is unlikely to be secondary because of differential uptake of the prostate-specific antigen test or different prostate-specific antigen levels between medication users and nonusers; prevalence of testing in Finland is low. Dose and time dependency of the relation probably indicates that duration of diabetes is negatively associated with risk.

case-control studies, diabetes mellitus, drug therapy, Finland, incidence, population, prostatic neoplasms

Type 2 (adult-type) diabetes is a condition currently affecting a substantial proportion of the Western population. Its development is often linked with obesity and resulting insensitivity to endogenous insulin, leading to impaired glucose balance. Type 1 (juvenile) diabetes is characterized by complete absence of endogenous insulin production and is not associated with obesity (1).

Medical therapy for adult-type diabetes is often started with oral drugs that improve glucose tolerance or increase insulin production, later possibly combined with injectable insulin treatment. Therapy for juvenile diabetes involves insulin when treatment begins.

Recent studies have reported a decreased prostate cancer risk for diabetic men, although the evidence is controversial (2). An inverse association of prostate cancer with metabolic syndrome has also been described, of which adult-type diabetes is an integral part (3, 4).

It is currently unclear whether use of antidiabetic medication affects the association between diabetes and prostate cancer. One study has reported that adjustment for antidiabetic medication did not affect the negative association between diabetes and prostate cancer (5); another study suggested that the risk reduction for diabetic men could be restricted to users of sulfonylureas and insulin only (6).

Biologic effects of oral antidiabetic drugs in prostate cancer cells are not well known, whereas the effect of insulin metabolism on prostate cancer growth has been studied more extensively (7). To our knowledge, only 1 study has reported growth reduction via cell cycle arrest in prostate cancer cells and xenografts after metformin treatment (8). However, all types of antidiabetic drugs affect insulin metabolism, providing a possible indirect mechanism for the effect on prostate cancer.

In the current study, we evaluated prostate cancer risk for users of antidiabetic medication in a population-based setting.

MATERIALS AND METHODS

Study design

The Finnish Cancer Registry identified all newly diagnosed prostate cancer cases in Finland during 1995–2002, a total of 25,029 men. The registry collects data through mandatory notifications of all cancer diagnoses made by the Finnish health care units. Thus, it is a nationwide register covering more than 99% of all cancer patients in Finland (9). The register information includes the primary site of cancer, histology, date, and method of diagnosis, but the register does not record differentiation, such as Gleason score, or serum prostate-specific antigen (PSA) values.

Practically all cases were histologically confirmed (99.3%). In addition, cases whose diagnosis was based solely on clinical (0.4%), radiologic (0.3%), or specific laboratory (0.02%) findings were included. A total of 185 cases (0.7%) for whom method of diagnosis was unknown were excluded.

Information on the stage of prostate cancer was available for 55% of the cases (n = 13,616). Of these, 73% of the cancers were localized. Median age did not substantially differ between cases with or without information on stage (68 years vs. 69 years). The yearly number of new cases and the proportion of cases without information on stage of disease tended to rise during the study period—from 2,328 new cases (34% without stage information) in 1995 to 3,840 new cases (64%) in 2002.

Controls were individually matched to cases by age and residential area at the time (month and year) of the corresponding case's prostate cancer diagnosis. The Population Register Center of Finland randomly selected 24,723 male controls. A total of 963 men were considered twice in the analysis, first as a control and later as a case in another case-control pair after being subsequently diagnosed with prostate cancer later in the study period. Matched controls could not be found from the same municipality for 121 cases in the oldest age group, resulting in their exclusion. A total of 24,723 case-control pairs were included in the analyses.

Approval was received from the ethics committee of the Pirkanmaa health care district, Finland (Pirkanmaa University Hospital ethics committee license code ETL R03290). However, obtaining informed consent from the study population was not required because of the large size of the population and the fact that some of the subjects could not be contacted (because of death or emigration) by the time the study began.

Information on antidiabetic medication purchased by the study population and for which the cost was reimbursed by the Social Insurance Institution of Finland during 1995–2002 was obtained from the comprehensive nationwide prescription database of this institution. The prescription database (10) and the reimbursement system (11) have been described in detail previously. In short, the Social Insurance Institution of Finland reimburses the cost of medication for each physician-prescribed drug approved as reimbursable. Reimbursement is available for all Finnish residents. The amount and dose of the drug, as well as the date of each reimbursed purchase, are recorded in the prescription database. All antidiabetic drugs in clinical use in Finland during the study period were reimbursable and were available through a physician's prescription only, thus comprehensively documented by the registry.

The database provided detailed information on the quantity and time of medication purchases for each person in the study population for a maximum of 8 years. The drugs available for the entire study period were human insulin, metformin, guar gum, and the sulfonylureas glibenclamide and glipizide. Glimepiride was available beginning in 1997, rosiglitazone in 2002, insulin aspart in 2001, and insulin lispro in 1996.

Medication use was followed until the diagnosis date (cases) or index date (corresponding controls), ensuring identical exposure time within each case-control pair. The defined daily doses (DDDs) recommended by the World Health Organization (12) were used to quantify the amount of use of antidiabetic drugs. For each year during the study period, cumulative usage (in milligrams or international units for insulin) for each drug was calculated based on all purchases reimbursed that year. Yearly usage was divided by the quantity corresponding to 1 DDD. The total number of DDDs used for each drug during the study period was obtained as the sum of yearly DDDs. Total DDDs for all separate drugs in the sulfonylurea or insulin categories were combined to obtain the overall amount of sulfonylurea or insulin use during the study period. For the subjects with multiple prescriptions, for example, metformin initially and insulin subsequently, the cumulative quantity was calculated for each drug and the subject was considered both a metformin and an insulin user.

Statistical analysis

All medication reimbursements between January 1, 1995, and the month of diagnosis were included in the analyses, regardless of length of use. For controls, the month of diagnosis of their matched case was used as the reference month for medication use, serving as the end of the exposure period.

Conditional logistic regression was used to estimate the odds ratios and likelihood-based 95% confidence intervals for the odds ratios of prostate cancer related to medication use in Stata 8.2 software (Stata Corporation, College Station, Texas). All reported P values are 2 sided.

Age-adjusted and multivariable-adjusted odds ratios were calculated. The multivariable-adjusted model included age, place of residence (municipality), and use of drugs commonly combined with antidiabetic drugs (aspirin, cholesterol-lowering drugs, and antihypertensive drugs) as covariates.

To estimate dose dependence between antidiabetic medication use and prostate cancer risk, users were stratified into quartiles by amount of DDDs, and the risk was analyzed separately in each stratum. Similarly, time dependence was analyzed by stratifying users by the length of antidiabetic medication use. Nonusers were considered the reference group in all analyses.

RESULTS

Because of matching, the age distribution was identical between the cases and controls. Median age was 68 years (range, 20–96) for both groups.

The prevalence of oral antidiabetic drug use was 7.5% for cases and 8.4% for controls (Table 1). Similarly, the prevalences of insulin use were 2.5% and 3.0%, respectively. The most commonly used drugs were glibenclamide (5.1% of cases and 5.8% of controls), metformin (3.9% and 4.6%), and human insulin (2.4% and 3.0%). Neither oral antidiabetic drug use (7.6% vs. 8.8%) nor insulin use (2.2% vs. 2.8%) differed substantially between cases for whom information on stage was or was not available.

Table 1.

Distribution of Medication Use in a Study Population of 24,723 Finnish Prostate Cancer Cases and Matched Controls During 1995–2002

	Cases		Controls
	No.	%	No.	%
Total	24,723	50	24,723	50
Antidiabetic medication use
Oral antidiabetic medication	1,852	7.5	2,082	8.4
Insulin	607	2.5	731	3.0
Both oral medication and insulin	422	1.7	521	2.1
Other medication use
Aspirin	1,943	7.8	2,025	8.2
Cholesterol-lowering drugsa	2,960	11.9	2,791	11.3
Antihypertensive drugsb	12,765	51.5	11,749	47.5

a

Includes statins, fibric acid derivatives, bile-acid binding resins, and acipimox.

b

Includes diuretics, beta-blockers, calcium-channel blockers, angiotensin-converting enzyme inhibitors, and angiotensin receptor blockers.

Overall, ever use of any antidiabetic drugs was associated with a decreased prostate cancer risk (Table 2). The risk decrease was observed for users of metformin, glibenclamide, glipizide, and human insulin. The risk was borderline decreased for users of glimepiride and guar gum (Table 2). Adjustment for multiple covariates compared with only age strengthened the association. We were not able to analyze the risk for users of rosiglitazone or insulin aspart because of the small number of users of these drugs (4 men for each drug).

Table 2.

Age-adjusted and Multivariable-adjusted Odds Ratios for Prostate Cancer (With 95% Confidence Intervals) in Users of Antidiabetic Drugs Compared With Nonusers, by Type of Medication Used, Among 24,723 Finnish Prostate Cancer Cases and Matched Controls During 1995–2002

Type of Medication (Ever vs. Never Use)	No. of Discordant Pairsa	ORage adjusted	95% CI	ORmultivariable adjustedb	95% CI
Any antidiabetic drug	1,953/2,194	0.89	0.84, 0.94	0.84	0.79, 0.90
Oral drugs	1,812/2,036	0.89	0.84, 0.95	0.85	0.79, 0.91
Metformin	904/1,064	0.85	0.78, 0.93	0.80	0.73, 0.88
Sulfonylureas	1,532/1,781	0.86	0.80, 0.92	0.82	0.77, 0.88
Glibenclamide	1,185/1,362	0.87	0.80, 0.94	0.83	0.77, 0.90
Glimepiride	282/307	0.92	0.78, 1.08	0.88	0.75, 1.04
Glipizide	322/388	0.83	0.72, 0.97	0.80	0.69, 0.93
Guar gum	322/339	0.95	0.82, 1.11	0.89	0.75, 1.04
Any insulin	588/717	0.82	0.74, 0.92	0.78	0.70, 0.87
Human insulin	588/717	0.82	0.74, 0.92	0.78	0.70, 0.87
Insulin lispro	31/29	1.07	0.64, 1.77	1.00	0.60, 1.67

Abbreviations: CI, confidence interval; OR, odds ratio.

a

Number of pairs including an exposed case and unexposed control/number of pairs including an unexposed case and exposed control.

b

Adjusted for age, place of residence, and simultaneous use of other medications (aspirin, cholesterol-lowering drugs, or antihypertensive drugs).

The overall prostate cancer risk decreased with amount of oral drugs and insulin used (P for trend < 0.001 and P for trend = 0.009, respectively) (Table 3). The risk of advanced prostate cancer showed no dose dependence with oral drugs or insulin. However, the risk estimates for advanced cancer were below 1 for all strata, except for the men using the smallest amount (≤228 DDD) of oral antidiabetic drugs (Table 3).

Table 3.

Odds Ratios for Risk of Any Prostate Cancer and Advanced Prostate Cancer (With 95% Confidence Intervals), by Amount of Antidiabetic Medication Use, Among 24,723 Finnish Prostate Cancer Cases and Matched Controls During 1995–2002

Total Amount of Medication for Which Subjects Were Reimbursed, DDDa	No. of Discordant Pairsb	Overall Cancer		Advanced Cancerc
		ORd	95% CI	OR	95% CI
Oral drugs
≤228	598/604	0.99	0.88, 1.11	1.06	0.79, 1.42
229–700	575/632	0.91	0.81, 1.02	0.73	0.53, 0.99
701–1,650	559/621	0.90	0.80, 1.01	0.81	0.59, 1.11
≥1,651	509/688	0.74	0.66, 0.84	0.80	0.56, 1.14
		P for trend < 0.001e		P for trend = 0.294
Insulin
≤288	148/172	0.86	0.69, 1.08	0.49	0.27, 0.87
289–744	142/290	0.70	0.56, 0.86	0.65	0.37, 1.14
745–1,707	142/192	0.74	0.60, 0.93	0.74	0.40, 1.35
≥1,708	156/184	0.85	0.68, 1.05	0.63	0.35, 1.16
		P for trend = 0.009		P for trend = 0.129

Abbreviations: CI, confidence interval; DDD, defined daily dose; OR, odds ratio.

a

Medication users were stratified by quartiles of reimbursements.

b

Number of pairs including an exposed case and unexposed control/number of pairs including an unexposed case and exposed control.

c

Locally or regionally invasive and metastatic prostate cancer.

d

Adjusted for age, place of residence, and simultaneous use of other medications (aspirin, cholesterol-lowering drugs, or antihypertensive drugs).

e

P values for trend were computed by including the cumulative total quantity of medication purchases in the multivariable-adjusted logistic regression model as a continuous covariate.

When use of antidiabetic medication was stratified by time since the first drug purchase during the study period, both the overall prostate cancer risk and the risk of advanced cancer showed inverse relations with duration of medical treatment (Table 4). Compared with that for men not using any antidiabetic medications, the overall risk was decreased by 34% for men with 7 years of antidiabetic drug treatment before the diagnosis/reference date, whereas the risk of advanced cancer was decreased by 39% in the same group (P for trend < 0.001 and P for trend = 0.003, respectively).

Table 4.

Odds Ratios for Prostate Cancer Overall and for Advanced Prostate Cancer (With 95% Confidence Intervals), by Time Since Onset of Antidiabetic Medication Use, in a Study Population of 24,723 Finnish Prostate Cancer Cases and Matched Controls During 1995–2002

Time Since Start of Treatment, years	No. of Discordant Pairsa	Overall Cancer		Advanced Cancerb
		ORc	95% CI	OR	95% CI
≤1	606/631	0.96	0.85, 1.07	0.96	0.74, 1.25
2	312/359	0.87	0.74, 1.01	0.61	0.42, 0.90
3	268/353	0.76	0.65, 0.89	1.12	0.70, 1.79
4	250/281	0.89	0.75, 1.06	0.79	0.49, 1.27
5	209/268	0.78	0.65, 0.94	0.64	0.37, 1.11
6	171/225	0.76	0.62, 0.92	0.62	0.34, 1.13
7	137/208	0.66	0.53, 0.83	0.61	0.28, 1.34
		P for trend < 0.001d		P for trend = 0.003

Abbreviations: CI, confidence interval; OR, odds ratio.

a

Number of pairs including an exposed case and unexposed control/number of pairs including an unexposed case and exposed control.

b

Locally or regionally invasive and metastatic prostate cancer.

c

Adjusted for age, place of residence, and simultaneous use of other medications (aspirin, cholesterol-lowering drugs, or antihypertensive drugs).

d

P values for trend were computed by including the cumulative total quantity of medication purchases in the multivariable-adjusted logistic regression model as a continuous covariate.

DISCUSSION

Our study showed a 16% decrease in the odds ratio for prostate cancer among users of antidiabetic drugs. The decrease was observed for users of multiple oral antidiabetic drugs and also for insulin users. This finding suggests that the effect is likely not associated with antidiabetic drug therapy per se but more likely with diabetes, the indication for the medication use. This finding, as well as the magnitude of the risk decrease, is in line with previous studies on this topic (2, 5).

Previous studies have suggested an inverse correlation between prostate cancer risk and time since diagnosis of diabetes, that is, a lower risk for men who have had diabetes for a longer period (13, 14). Our results concur with this finding because length of medical treatment was inversely associated with risk. The risk reduction was also dependent on the amount of medication used, also supporting the importance of treatment duration because use of large quantities of antidiabetic drugs requires a longer time of use. The risk of advanced prostate cancer was inversely associated with duration but not with cumulative amount of medication. However, note that, in most cases, treatment of adult-type diabetes is started with nutritional and lifestyle counseling before initiation of medical therapy. Thus, time since onset of diabetes is longer than it appears to be in the analysis for most men in our study population.

To our knowledge, this study is the first to evaluate the risk for diabetic men by adjusting for other medications commonly prescribed with antidiabetic drugs, namely, aspirin, cholesterol-lowering drugs, and antihypertensive drugs. Each of these drug groups possibly affects prostate cancer risk, thus potentially confounding the effect of diabetes and its treatment (10, 15, 16). Additionally, hypercholesterolemia and hypertension are components of the metabolic syndrome, another condition possibly affecting prostate cancer risk (3, 4). Therefore, adjusting for cholesterol-lowering drugs and antihypertensive drugs enabled us to control for the effect of metabolic syndrome to some degree. Controlling for several covariates strengthened the observed association of lowered prostate cancer risk compared with controlling for age only, confirming the confounding.

This study is thus far the largest single one known to estimate prostate cancer risk for diabetic men. Because of the comprehensive national health care registers in Finland, we were able to carry out a large, population-based case-control study with minimal influence of chance or selection bias.

Detailed exposure information was obtained objectively from a prescription database unaffected by disease status. Thus, recall bias, a common problem in case-control studies, did not affect our results. The representativeness of the study is demonstrated by the comparative medication use between our study population and the overall Finnish population. In 2003, overall use of metformin, sulfonylureas, and insulin by Finnish men was 14.07 DDD, 23.44 DDD, and 20.16 DDD per 1,000 persons per day (17), respectively. In our study population, the respective observed use in 2002 was 12.42 DDD, 22.61 DDD, and 20.36 DDD per 1,000 persons per day, which is highly consistent with the population estimates. All antidiabetic drugs in Finland during the study period were available through a physician's prescription only. Therefore, medication purchases were comprehensively documented by the prescription database. However, we did not have information on actual intake of medication because men used their own discretion. The main limitation of the Social Insurance Institution of Finland database is lack of information on medication for institutionalized patients.

Some exposure misclassification could have been caused by the fact that information on medication purchases was available since 1995 only, although metformin, glibenclamide, glipizide, guar gum, and human insulin were licensed in Finland earlier. Thus, some information on actual medication use was probably missing, likely to result in underestimation of exposure. For instance, some subjects may have had a longer history of use than found in our study. Because antidiabetic medication is not curative, it is rarely discontinued, decreasing the probability of misclassifying past users as nonusers. Information on medication use was obtained in a similar fashion for cases and controls; therefore, nondifferential misclassification is likely to result, which may dilute the observed association. However, the distortion is likely to be small because the risk estimates were not systematically different for cases diagnosed during the early period (with less complete coverage of recent use) versus later.

A limitation of our study is the missing information on serum PSA testing within the study population. The prevalence of latent prostate cancer, already high among men in their forties, increases with age (18, 19). Introduction of serum PSA testing in prostate cancer diagnostics and as a screening tool has led to detection of many prostate tumors in their latent, clinically nondetectable phase and an increase in incidence rates of prostate cancer (20). The benefits of prostate cancer screening with the PSA test are yet to be proven, and systematic screening is not officially recommended in most countries (21). Nevertheless, opportunistic screening occurs, although the prevalence is low in Finland—less than 20% annually (22).

Clinical management of diabetes includes frequent control of blood glucose balance and serum cholesterol level. It is plausible that PSA testing is more frequent among diabetic men than among nondiabetic men, who probably use health services less often. Thus, there could be more opportunistic prostate cancer screening of diabetic men than nondiabetic men, causing a positive detection bias, that is, increasing the probability of prostate cancer diagnosis. However, the lack of mass screening for prostate cancer and the low prevalence of opportunistic screening in Finland (22) is another strength of our study, decreasing the likelihood of such bias. Furthermore, such bias would not jeopardize our conclusions because we observed a lower risk for diabetic men.

On the other hand, serum PSA level has been reported to be 21.6% lower in diabetic men compared with nondiabetic men, not depending on antidiabetic medication use (23). This difference could cause a negative detection bias, that is, fewer prostate cancers being diagnosed in diabetic men because fewer prostate biopsies are being performed based on elevated PSA levels. Lower PSA levels could explain some of the observed risk decrease in our study and in previous studies on this subject. An inverse detection bias is also supported by the fact that prostate cancer risk for diabetic men has been consistently lower in studies performed after the introduction of the PSA test compared with the studies conducted in the pre-PSA era (2). However, the pre-PSA studies have also reported lower prostate cancer risk for diabetic men (2).

A 25% decrease in average PSA (e.g., from 4 ng/mL to 3 ng/mL) can decrease the prostate cancer detection rate by 36% (24). Thus, our results need to be confirmed in further studies with adjustment for PSA level.

Information on prostate cancer stage was available for only slightly more than half of the cases, which impeded our analyses of advanced cancer. There were no substantial differences in age between the cases for whom information on stage was or was not available. However, the proportion of cases without stage information steadily increased throughout the study period, from 7.0% in 1995 to 21.8% in 2002. The largest increase occurred between 1999 and 2000, from 11.6% to 17.1%, which could reflect an increasing prevalence of opportunistic PSA screening in Finland, with a larger proportion of early cancers. For these cancers, complete staging is not routine. However, the most prominent short-term effect of screening is an increase in early cancer. Thus, more common screening in men with diabetes would most likely lead to a decrease or reversal in protective effect regarding early prostate cancer. When the results were analyzed separately by using cases diagnosed before and after 2000, there was no difference in overall prostate cancer risk but some indication of a decrease in risk of advanced cancer. Therefore, it is unlikely that detection bias by opportunistic PSA testing would substantially explain the observed lower odds ratio of prostate cancer among antidiabetic medication users.

Overall prevalence of antidiabetic drug use was slightly lower in cases for whom stage information was available (8.3% vs. 9.4% for cases without such information), which, on the other hand, could have diminished the observed association with advanced cancer.

Age and ethnicity are established risk factors for prostate cancer (25). We controlled the confounding effect of age by individual matching of cases and controls. No significant effect modification by age was observed. We did not have information on the race of our study subjects. However, more than 98% of the Finnish population is Caucasian, minimizing potential for confounding by ethnicity in our study population (26). We could not control for study subjects’ family history of prostate cancer. An inherited predisposition is a strong prostate cancer risk factor. However, hereditary factors have been estimated to account for only a minor proportion of prostate tumors (27), 5%–10% of all Finnish prostate cancers (28). To generate confounding, prostate cancer family history would need to be associated with antidiabetic medication, for which there is little indication.

Furthermore, we did not have data on obesity and a Western-style high-fat diet. These are potential confounding factors because they are frequently associated with adult-type diabetes and possibly affect prostate cancer risk (29, 30) and serum PSA level (31). However, their role as prostate cancer risk factors is debatable.

We could not distinguish between adult-type and juvenile diabetics in our analyses. Thus, we could have missed possible subgroup effects within these 2 groups of diabetics. One might assume that virtually all men using oral drugs have type 2 diabetes mellitus. With more advanced type 2 diabetes, an estimate is that 30%–40% of the patients are using insulin (1). Also given the substantially higher prevalence of type 2 than type 1 diabetes, most patients using insulin are likely to have type 2 diabetes.

The potential mechanism behind decreased prostate cancer risk for diabetic men is currently unclear. Most likely, the changes in endogenous hormone metabolism occurring in diabetes have an important role.

Diabetic men have a lower serum testosterone concentration compared with nondiabetic men (32). Conversely, low serum testosterone is linked with increased risk of developing diabetes (33–35). Because growth of prostate cancer is androgen dependent, changes in serum testosterone in diabetic men could provide a possible explanation for their lowered prostate cancer risk.

Fasting insulin levels (3) or insulin resistance (36) have not been reported to affect prostate cancer risk. However, insulin-like growth factors, overexpression of which is reported to increase prostate cancer risk, are dependent on serum insulin levels (7). Therefore, altered insulin metabolism is another possible explanation for decreased prostate cancer risk for diabetic men.

Our large, population-based study showed decreased prostate cancer risk for users of antidiabetic drugs. A similar decrease was observed for users of metformin, sulfonylureas, and insulin, suggesting an overall lowered risk for diabetic men. The decrease was dose dependent with quantity and length of medication use, suggesting that duration of diabetes is also a determinant of risk. These findings are consistent with recent results in this field. We also showed that the risk of advanced prostate cancer decreases for diabetic men, depending on the length of antidiabetic treatment. Ours is the first study known to demonstrate that the association could not be explained by the medications commonly prescribed along with antidiabetic medication. However, because varying PSA testing activity within the study populations and lower PSA levels in diabetic men can introduce detection bias, future studies that evaluate prostate cancer risk while effectively controlling for serum PSA are needed.

Abbreviations

Abbreviations
• DDD

  defined daily dose

• PSA

  prostate-specific antigen

Author affiliations: School of Public Health, University of Tampere, Tampere, Finland (Teemu J. Murtola, Anssi Auvinen); Department of Urology, Tampere University Hospital and Medical School, University of Tampere, Tampere, Finland (Teuvo L. J. Tammela); Department of Internal Medicine, Tampere University Hospital, Tampere, Finland (Jorma Lahtela); and The Finnish Cancer Institute, Helsinki, Finland (Anssi Auvinen).

Funding was received from the Academy of Finland (grant 205 862); the Sigrid Juselius Foundation; the Finnish Cancer Society; the Pirkanmaa Regional Fund of the Finnish Cultural Foundation; the Medical Research Fund of Tampere University Hospital; the Irja Karvonen cancer trust; and the Astellas, Lilly Foundation, Schering Foundation, and research foundation of Orion Pharma (nonrestricted grants to T. J. M.).

Conflict of interest: none declared.