Antiepileptic drugs and prostate cancer risk in the Finnish Randomized Study of Screening for Prostate Cancer

Antiepileptic drugs (AEDs) with histone deacetylase (HDAC) inhibitor properties decrease prostate cancer (PCa) cell proliferation in vitro. A population‐based cohort of 78 615 men was used to evaluate the risk of PCa among users of AEDs. Study population was linked to the Finnish national prescription database to obtain information on individual medication reimbursements in 1996 to 2015. Cox regression with antiepileptic medication use as a time‐dependent variable was used to analyze PCa risk overall, and low, medium and high‐risk PCa separately. The analysis was adjusted for age, screening trial arm, and other drugs in use, including statins, antidiabetic drugs, antihypertensive drugs, aspirin, and nonsteroidal anti‐inflammatory drugs. Compared to the nonusers of AEDs, overall PCa risk was decreased among AED users (hazard ratio [HR] = 0.86, 95% confidence interval [CI] = 0.76‐0.96). A similar PCa risk decrease was observed among users of HDACi AEDs (HR = 0.87, 95% CI = 0.76‐1.01), but no risk difference was found when comparing HDACi AED users to users of other AEDs (HR = 0.98, 95% CI = 0.76‐1.27). Our study showed a decrease in overall PCa risk among men using AEDs compared to nonusers. The risk associations were similar for HDAC inhibitors as for AEDs in general.

Previous epidemiological studies on association of AEDs on PCa risk have given conflicting results. In a small UK cohort study, VPA users showed a nonsignificant trend of increased PCa risk. 11 Another study found that VPA, carbamazepine, oxcarbazepine, lamotrigine, and levetiracetam may decrease serum levels of prostate-specific antigen (PSA). 12 Two studies have reported no association between PCa risk and VPA usage. 13,14 In our previous case-control study, we found a decreased risk of PCa associated with usage of VPA, carbamazepine, and phenobarbital. 15 We did not analyze HDACi AEDs as a group or compared to users of other types of AEDs.
People with epilepsy are known to have increased risk of cancer and increased cancer mortality overall. 16,17 The mechanism behind the association is unclear, but is believed to be due to conditions associated with epilepsy. There are, however, no data available of PCa risk in people with epilepsy that we know of.
We examined the association between AEDs and PCa risk among men in the Finnish Randomized Study of Prostate Cancer Screening (FinRSPC) during 1996 to 2015. To our knowledge, this is the first study to evaluate PCa risk comprehensively among AED users.

| The study cohort
The Finnish Randomized Study of Prostate Cancer Screening includes 80 458 men residing in the metropolitan areas of Helsinki and Tampere. In 1996 to 1999, all men aged 55, 59, 63 and 67 in the target population were identified annually from the Population Register in Finland. After exclusion of prevalent PCa cases, they were randomly assigned 1:1.5 into two groups: the screening arm (32 000 men) and the control arm (48 458 men). Family history of PCa was obtained at baseline participating in the screening arm. Men in the screening arm were invited to screening with PSA test every 4 years until the age of 71 years, excluding men who were diagnosed with PCa, had moved abroad or died. They were considered screening positive if the PSA was above 4 ng/mL, or if they had PSA at 3.0 to 3.9 ng/mL and percentage of free PSA less or equal to 15%. No intervention was offered to the men in the control arm.
Incident cases in the control arm and in the screening arm were identified from the Finnish Cancer Registry, which covers practically all cancer cases in Finland. 18   medication, antihypertensive medication, nonsteroidal antiinflammatory drugs (NSAIDs) and aspirin were obtained as potential confounders from the database.

| Statistical analysis
Cox regression was used to estimate hazard ratios (HRs) and 95% confidence intervals (95% CIs) for PCa by medication usage. The time metric was years and months (as decimals of year) since the screening trial randomization (the baseline). Cox regression was adjusted for age at randomization and in the multivariable analyses also for simultaneous usage of other drugs. Analyses of the full trial cohort were also adjusted for the screening trial arm. Validity of proportional hazards assumption was not formally tested as exposure of interest is ana-

| Population characteristics
Prevalence of AED use was 11.4%. The median age at randomization and BMI were comparable between AED users and nonusers (

| AEDs and PCa risk
The age-adjusted incidence of PCa was 78.5/10 000 person years for users of any AED, 74.1/10 000 person years for those using HDACi AED and 78.9/10 000 person years for men not using AED.
When compared to the nonusers, overall PCa risk was lower among AED users (multivariable-adjusted HR = 0.86, 95% CI = 0.76-0.96), with a decreasing trend with intensity of AED use (

| Lag-time analyses
Users of AEDs had lower PCa risk compared to nonusers of AEDs also

| Sensitivity analyses
When limiting users of AEDs to those men not using AEDs before randomization, risk of PCa was lower compared to nonusers AEDs (HR = 0.89, 95% CI = 0.78-1.01 in multivariable adjusted analysis).

When comparing users of HDACi AEDs and other types of AEDs
including only men with no AED use before baseline, no difference in PCa risk was seen (HR = 0.89, 95% CI = 0.76-1.05 in multivariable adjusted analysis).
We performed a sensitivity analysis were only men with at least one PSA test during the follow-up were included to the analysis.

| DISCUSSION
Our study showed a decreased risk of PCa among users of AEDs compared to nonusers. This finding was significant only for low-risk PCa; nonsignificant risk reduction was found also for high-risk cancer. Similar risk reductions were found in users of HDACi AEDs and users of other types of AEDs. Results of previous epidemiological studies on the subject have been inconsistent. In our previous study, we found a significant PCa risk reduction among users of VPA, carbamazepine and phenobarbital. 15 However, only users of carbamazepine had a significant risk reduction for advanced (lymph-node positive or metastatic) disease. A cohort study of 26 911 US veterans found no association between VPA use and PCa risk overall. 14 A Danish population-based casecontrol study found no decrease in PCa risk among VPA users, but their follow-up time was only 5 years, with no more than six exposed cases. 13 A British cohort study of 3000 patients with epilepsy found increased PCa risk among men using VPA, but it was based on only eight exposed cases. 11 In lag-time analyses, PCa risk was lower in AED users in both 1-year and 3-year analysis. These findings indicate that PCa protective association with AED use is most likely associated with long-term influence of drug use on development of PCa or related to properties of men using AEDs in long-term, that is, people with epilepsy rather than protopathic bias due to cancer-related condition, such as brain-metastases.
Our study has several strengths. A large population-based cohort was used to evaluate PCa risk among AED users. We were able to obtain accurate information on PCa cases and AED use through comprehensive nationwide registers. Since the detailed exposure information was obtained objectively from a prescription database, which comprehensively records drug reimbursements including the number of packages, amount and dose for the drug purchased, no recall bias affected the estimation of exposure. Our study setting also allowed analyzing PCa risk for both AED users with and without systematic screening for PCa. Exceptional accuracy in our information on medication purchases allowed analyzing usage in time-dependent manner to minimize immortal time bias.
Our study also has some limitations. We did not know the indication for the drug usage except for a small subgroup of people with confirmed epilepsy, although we had information on background comorbidities and of diagnoses within the study population.
Some AEDs are also used in management of nonepileptic conditions, such as neuropathic pain and migraine. No significant PCa risk difference was found between users of HDACi AEDs and other types of AEDs when analysis was limited only to people with confirmed epilepsy, which is in line with the findings in the whole study population. We had only information on special reimbursement for years 1995 to 2009, so the statistical power of this subgroup analysis is limited. Drugs administered for hospital inpatients are not recorded in the prescription database causing underestimation and misclassification of the exposure. Also, we did not have information of the actual intake of medication, which might lead to overestimation of the exposure. However, continued unused medication purchases should be uncommon. Alcohol usage, smoking and physical activity are lifestyle factors that could cause confounding in our results and we did not have information on these, but their role as a risk factor for PCa is not well defined.
Our results show that AED use is associated with a reduced risk of PCa. However, the risk reduction does not depend on HDAC inhibitory properties. Further studies are required to evaluate whether this risk reduction leads to lower PCa mortality.

DATA AVAILABILITY STATEMENT
The data sets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

ETHICS STATEMENT AND CONSENT TO PARTICIPATE
The study was performed in accordance with the Declaration of Helsinki and the protocol was approved by the ethics committee of Pirkanmaa Hospital District. The informed consent was waived as the study is retrospective.