Statins and Finasteride Use Differentially Modify the Impact of Metformin on Prostate Cancer Incidence in Men with Type 2 Diabetes

Abstract

Background: Metformin has shown promise for cancer prevention. Prior studies suggested that metformin might interact potential prostate cancer (PCa) prevention agents: finasteride and statins. This study assessed if concurrent use of statins or finasteride modified the long-term impact of metformin on PCa risk in men with type 2 diabetes (T2DM).

Materials and Methods: The study cohort consisted of 71,999 men with T2DM seen in the Veteran Administration Health Care System, without prior cancer or liver diseases, nor prescription of thiazolidinediones or insulin between FY2003-FY2013. Cox proportional hazard analyses (adjusting for covariates and propensity scores of metformin use) were conducted to compare the hazard ratio (HR) of PCa associated with metformin use between statins or finasteride users and none users.

Results: Mean follow-up was 6.4±2.8 years; 5.2% (N= 3,756) of the cohort subsequently received a PCa diagnosis. Both statins and finasteride significantly modified the impact of metformin on PCa incidence (p–value<0.001): HR’s of PCa associated with metformin use were 0.89 (p–value=0.02) among non-statin /non-finasteride users, 0.73 (p–value<0.001) among statin users, and 1.42 (p–value<0.001) among finasteride users.

Conclusion: Metformin was associated with reduced PCa risk in men with T2DM. This impact was enhanced by statins but reversed by finasteride. Metformin, statins, and finasteride are potential PCa prevention agents. The interaction of these drugs on PCa risk needs further confirmation in other cohorts. Our finding of differential impacts of metformin, statins, and finasteride (alone or in combination) on PCa risk is informative for treatment management in men at risk for PCa and T2DM.

Keywords: Prostate cancer; Metformin; Statins; Finasteride; Type 2 diabetes

Introduction

Prostate Cancer (PCa) prevention is an attractive strategy in men with type 2 diabetes (T2DM) as this population appears to be at a higher risk for high-grade, potentially-lethal PCa compared to men without diabetes [1,2]. Metformin, a biguanide drug, is commonly prescribed as first-line glucose-lowering therapy for patients with T2DM and endorsed by a recent consensus statement of the American Diabetes Association and the European Association for the Study of Diabetes [3]. Metformin has also showed promise for cancer prevention [4-10]. The potential mechanisms of this antineoplastic effect include a reduction of hyperinsulinemia through activation of the AMP-Activated Protein Kinase (AMPK) pathway, inhibition of the mammalian target of rapamycin pathway, blockade of cell cycle progression, and alteration of anti-inflammatory properties [4,11-16]. The effect of metformin on PCa risk is unclear. While two observational human studies found metformin use to be associated with a decreased PCa risk in the general population [17,18], two studies among diabetic subjects did not find a significant reduction in PCa risk with metformin use [19,20]; a dose-dependent increase was observed in a more recent study [21]. The discrepancy of these findings could be due to the difference in study design, some of which could be subject to time period bias [22].

Our group previously examined 5,042 veterans with T2DM treated with either sulfonylureas or metformin as their sole glucose-lowering agent [21]. In that study that carefully avoided confounding and the time period bias by design [22], we found that the incidence of PCa among concurrent metformin and statin users was significantly lower as compared to the incidence among those with sulfonylureas use but not statins. Paradoxically, the incidence of PCa was significantly higher among metformin users without concurrent statin use [21]. Statins are 3-Hydroxy-3-Methylglutaryl-Coenzyme A (HMG-CoA) reductase inhibitors, which are recommended for patients with T2DM primarily for preventing cardiovascular diseases. Statins may also affect PCa tumorigenesis by blocking the mevalonate pathway and reducing cholesterol and/or through multiple pleiotropic effects [23]. The favorable joint effect of metformin and statins on PCa may be due to their joint effects via lipid-lowering or a combination of pleiotropic effects (e.g., via improving inflammatory conditions) [21,24,25]. The interactive effect of metformin and statins on PCa found in our previous analysis is yet to be confirmed in a broader population with T2DM, especially including a comparison between a broader range of metformin users (e.g., not limited to sole metformin users) and non-metformin users (e.g., not limited to sulfonylurea users).

Finasteride, a 5α-reductase inhibitor used for treating benign prostatic hypertrophy (BPH), was initially found to reduce PCa risk by 24.8% in the Prostate Cancer Prevention Trial (PCPT); and longterm follow-up found the risk reduction to be 30% [26,27]. Recently, studies have shown that finasteride attenuated the insulin-like growth factor binding protein (IGFBP2) or C-peptide induced PCa risk in the PCPT [28,29], which suggests that the beneficial effect of finasteride on reduced PCa could be attenuated among individuals who are less hyperinsulinemic, such as those treated with an insulin sensitizer like metformin. To date, no studies have examined whether finasteride modulates the impact of metformin on PCa.

With these questions in mind, we examined an 11-year longitudinal historical population of 71,999 men with T2DM who were insulin naïve in the Veteran Administration Health Care System (VAHCS) to determine if the impact of metformin on PCa risk was modified by concurrent use of statins or finasteride.

Materials and Methods

Study cohort

We drew our study sample from the 268,136 VAHCS enrollees who were ≥40 years of age in 2003, and had at least one primary care visit (defined as any visit to the general medicine, geriatric, or diabetes clinics) as well as a diagnosis of T2DM (diagnosis using ICD-9 CM codes of 250.00 or 250.02) each year during FY2001-FY2002. Our criteria of identifying patients with T2DM could miss at most 3% T2DM who only had diagnoses of diabetes complications [30]. We further narrowed the study cohort to 71,999 men who also met the following criteria: (i) no prescription of any glucoselowering medication in FY2002; (ii) having had prescription(s) of metformin as a glucose-lowering medication for ≥180 days or none during the study period, (iii) having had prescription(s) of statin as a lipid-lowering medication for ≥180 days or none during the study period, (iv) having had prescription(s) of finasteride for ≥180 days or none during the study period, (v) no prescription for insulin or any thiazolidinedione (TZD) during the study period; (vi) no liver or renal diseases during the study period; (vii) no cancer diagnosis nor metformin exposure prior to FY2003; and (viii) no missing data on any baseline covariates (age, race, hemoglobin A1c (HbA1c), body mass index (BMI), and age-adjusted Charlson co-morbidity score [31]). Regarding criterion (v), we excluded patients who were on a TZD and/or insulin to better assess the association between metformin and PCa risk to eliminate potential confounding effects of other glucose-lowering medications which could themselves have an effect on PCa incidence or progression as reported in the literature [32,33]. All study procedures were approved by the Institutional Review Board of the UT Health Science Center San Antonioand Research and Development of the South Texas Audie L. Murphy Veterans Hospital.

Data sources

We used four VAHCS datasets for this study. VAHCS Inpatient and Outpatient Medical SAS Datasets were used to identify the cohort of men with T2DM and their associated characteristics, including demographic variables and comorbidities (based on diagnosis codes). Additional clinical variables were extracted from the VA Decision Support System (medication prescription records, HbA1c, LDL, and prostate specific antigen (PSA) lab results and dates of measurements) and VAHCS Corporate Data Warehouse (height and weight values for deriving BMI) [34].

Outcomes of interest

The outcome of interest in this study is the incidence (rate) of the initial PCa diagnosis during the study period. The dependent variable used in our analyses is the time interval between the starting date (October 1 2002, the starting date of FY2003) for those without use of any oral glucose-lowering medication, the initiation of nonmetformin glucose-lowering medication for those non-metformin glucose-lowering medication users, and the initiation of metformin for metformin users) to the initial PCa diagnosis observed during the study period. A PCa event was defined as having an ICD-9 diagnosis of 185 from either primary or secondary diagnosis code. Those who died prior to the first PCa event observed during the study period, or those who remained alive but never had a PCa event during the study period were treated as censored data. The study termination date for each patient corresponded to either the date of the initial PCa diagnosis, the date of death, or September 30, 2012 (the end of followup), whichever came first.

Predictors and measures

Medication exposure: For this study, the metformin group consisted of patients who had metformin prescription of any dose for 180 days or more but never had any TZD or insulin during the study period. We regarded sufficient metformin exposure as a minimum of 180 days since most clinical trials on metformin were 24 weeks or longer and other studies have used a similar exposure cut-point [20]. Non-metformin users were patients who had no prescription for metformin, and never had any TZD or insulin during the study period. Statin users consisted of patients who had any type of statin prescription at any dose for 180 days or more during the study period. Non-statin users consisted of patients who never had any type of statin prescription during the study period. Finasteride users were patients who had any finasteride prescription for 180 days or more during the study period.

Covariates: Covariates adjusted for in the analyses included demographic and clinical characteristics of the patient. Demographic characteristics included age, and race/ethnicity (Caucasian, African- American, Hispanic, others). Clinical characteristics included ageadjusted Charlson co-morbidity score, and the mean change of low density lipoprotein (LDL) and HbA1c levels from baseline, and the maximum PSA during the study period.

Statistical analyses

The Cox proportional hazard model adjusting for covariates and the propensity scores of metformin use was conducted to assess the hazard ratio of PCa associated with metformin use, where the propensity scores of metformin use were incorporated as the inverse probability weights in the analyses. Each of these propensity scores was the likelihood of being treated with metformin conditioned on each subject’s pretreatment characteristics. Using these scores, individuals were weighted differently to achieve balance in covariates at baseline between the metformin users and non-metformin users, and therefore potential confounding due to pretreatment characteristics was minimized [35]. The propensity scores were calculated using logistic regression analysis, where the dependent variable was the indicator of receiving metformin treatment and the independent variables were patients’ pretreatment characteristics. We included all significant predictors (i.e., the pretreatment/baseline covariates) in the propensity score models to ensure a credible estimation of the main outcome (PCa incidence) model [35]. The propensity score model of metformin use included HbA1c, age, and race/ethnicity, and Charlson comorbidity score.

To assess the interactive effect between metformin and statin and that between metformin and finasteride, we used the model with metformin by statin interaction and metformin by finasteride interaction. Each of these interaction terms corresponded to the product of the associated indicators of medication use, and its coefficient was assessed by the Wald test with p–value <0.05 being considered significant. In the Cox regression model setup, the interactive effect between metformin and statins compared the hazard ratio associated with metformin use between statin users and nonstatin users, and this effect was assessed using the contrast statement. Similarly, the interactive effect between metformin and finasteride compared the hazard ratio associated with metformin use between finasteride users and non-finasteride users. All statistical analyses were conducted using SAS 9.1.

Results

In our final cohort of 71,999 men with T2DM, 28,308 (39.32%) had prescription(s) for metformin. Among 28,308 men who had prescription(s) for metformin, 24,384 (86.14%) were statin users and 3,744 (13.23%) were finasteride users. Among 43,691 men (60.68%) who did not receive prescription(s) for metformin, 31,900 (73.01%) were statin users and 6,107 (13.98%) were finasteride users. Study subject characteristics are shown in Table 1. Mean age at cohort entry was 66.91±10.22 years, and the mean study period was 6.41±2.76 years. The mean HbA1c at baseline was 6.51±0.96%. A total of 3,756 patients (5.2%) were diagnosed with PCa during the study period.