Association between antiepileptic drugs and hepatocellular carcinoma in patients with epilepsy: a population‐based case–control study

Abstract Background This study explored whether antiepileptic drugs (AEDs) use increases the risk of hepatocellular carcinoma (HCC). Methods We conducted a case–control study using data from the National Health Insurance system of Taiwan. The case group comprised 1,454 epilepsy patients with newly diagnosed HCC, and the control group comprised 1,448 epilepsy patients without HCC. Both groups had similar distributions of sex and age, and follow‐up duration. Possible associations with the AEDs in Taiwan were examined. Results After adjusted for AEDs (phenobarbital and primidone, clonazepam, clorazepate and diazepam, and other AEDs), and for the comorbidities of diabetes, chronic liver disease and cirrhosis, hepatitis B and C virus infection, and alcoholism, the odds ratio (OR) of HCC was 1.22 (95% confidence interval [CI]: 1.01–1.47) for the group of phenytoin users compared with nonphenytoin users. An annual means of 61–120, 121–180, and >180 of defined daily doses (DDDs) of phenytoin (OR: 4.07, 95% CI: 2.03–8.18; OR: 7.51, 95% CI: 3.03–18.7, and OR: 14.6, 95% CI: 7.88–26.9, respectively) were significantly correlated with the risk of HCC but not with a DDD of ≤60. Compared with nonphenytoin users, HCC patients who had used phenytoin within 1 year of HCC diagnosis were at a greatest risk of HCC (adjusted OR: 2.29, 95% CI: 1.71–3.08), followed by who had used phenytoin within 2 years of diagnosis (adjusted OR: 1.92, 95% CI: 1.44–2.56). Conclusion The results indicate that high dose of phenytoin was associated with a statistically significant increased OR for HCC, which was not demonstrated for low‐dose phenytoin.

Cancer remains a leading cause of morbidity and mortality worldwide. In 2012, nearly 14 million new cases arose and 8.2 million cancerrelated deaths occurred (Stewart & Wild, 2014). Cancer can complicate numerous chronic diseases, including epilepsy (Singh, Driever, & Sander, 2005). AEDs are often used over long periods of time, which may increase the tendency for some adverse effects, including carcinogenesis. The purpose of this study was to determine whether there was an association between any specific AED and hepatocellular carcinoma (HCC) in patients with epilepsy in Taiwan. In 2014, HCC was the second leading cause of cancer deaths (after lung cancer) in Taiwan, with a mortality rate of 34.9 per 100,000 people (18.6% of the total cases) (Department of Health 2012). Because of the importance of epilepsy treatment, the potential of AEDs to cause hepatotoxicity must be considered.
To the best of our knowledge, no large-scale case-control study to date has addressed the association between HCC and the use of AEDs in Taiwan. Thus, we conducted a population-based case-control study using data from the National Health Insurance (NHI) program of Taiwan to determine the association between AEDs and the risk of HCC.

| Data sources
Taiwan's universal NHI program has been in operation since 1995, and nearly 100% of all residents are enrolled. The National Health Research Institutes (NHRI) was commissioned to construct and maintain the National Health Insurance Research Database (NHIRD) for researchers. The NHIRD provides detailed information on the healthcare services used by each patient, as well as the demographic characteristics, complete outpatient visits, hospital admissions, International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnostic codes, prescriptions, and clinical orders (such as surgery) of participants and utilities. The NHRI encrypts patients' personal information for privacy protection and provides researchers with anonymous identification numbers associated with the relevant claims information. This study was approved by the Institutional Review Board of China Medical University Hospital (CMUH104-REC2-115).

| Study patients
Our research was based on case-control study patients with epilepsy (ICD-9-CM code 345) who were indentified using two subdatasets of the NHIRD, medical claims data from the Catastrophic Illness Patients Therefore, the diagnoses of HCC, and the ICD-9 codes, in this study were highly reliable.

| AEDs exposure and comorbidities
We conducted a comprehensive review and classified several categories of commercially available AEDs in Taiwan

| Statistical analysis
The distributions of sex, age (20-39, 40-64, 65-74, and ≥75 years), AED medications, and baseline comorbidities were compared between the HCC group and the non-HCC group, and were examined using the chi-square test for categorical variables and the Student's t test for continuous variables. Univariate and multivariate logistic regression models were used to calculate the odds ratio (OR) and 95% confidence intervals (CIs) for the association between HCC and AEDs use. The multivariate analysis was performed to adjust for the AEDs of phenobarbital and primidone, clonazepam, clorazepate and diazepam, and other AEDs, and the comorbidities of diabetes, chronic liver disease and cirrhosis, HBV infection, HCV infection, and alcoholism. We also estimated the risk of HCC by cumulative dose for phenytoin use.
Furthermore, we analyzed the association between HCC and the time difference between the last phenytoin in use and the index date. All analyses were performed using SAS statistical software for Windows (Version 9.4; SAS Institute Inc., Cary, NC, USA), and the significance level was set at .05.

| RESULTS
The case group comprised 1,454 epilepsy patients with newly diagnosed HCC, and the control group comprised 1,448 epilepsy patients without HCC. Both groups had similar distributions of sex and age, and were predominantly male (73.5%). Approximately, 40% were older than 75 years of age. The mean ages of the HCC patients and non-HCC controls were 64.9 ± 13.6 and 64.7 ± 13.7 years, respectively. The HCC group exhibited a higher prevalence of all baseline comorbidities than did the non-HCC control patients (Table 1). The risk of HCC increased significantly with the comorbidities of diabetes, chronic liver disease and cirrhosis, HBV infection, HCV infection, and alcoholism.
In this study, the association between HCC risk and the annual mean DDD of phenytoin were analyzed for the period that patients were undergoing monotherapy. The association between HCC risk and the annual mean DDD of phenytoin use is shown in Table 3.
Compared with nonphenytoin users, the HCC risk was highest in patients who were administered >180 annual mean DDDs of phenyt-

| DISCUSSION
We conducted a comprehensive population-based case-control study, using the NHIRD to investigate the association between and Allison (2014) also indicated a higher risk of cancer in the first year after the start of AEDs treatment.
Use of specific AEDs is based on several different factors, including patient age, sex, epileptic syndromes, and comorbid diseases.
We used the Taiwan NHIRD, which follows the diagnostic coding system recommended by the ICD-9-CM; according to the criteria of ICD-9-CM, there are separate codes indicating epilepsy and actively treated epilepsy (defined as use of an AED more than once). In Taiwan, 71% of patients were prescribed monotherapy (Hsieh & Huang, 2009), and it is reported that, assuming appropriate choice of monotherapy for newly diagnosed epilepsy, 75% of cases will be free from seizures (Mattson et al., 1985). Monotherapy is reported to be the gold standard for the treatment of epilepsy as it is effective in nearly 80% of patients (Deckers, 2002). Carbamazepine, phenytoin, valproic acid are the most common choices for initial monotherapy in Taiwan (Hsieh & Huang, 2009 liver cancer. In addition, some authors observed a strong association between epilepsy and the rate of cancer of the CNS and of the mouth and throat (IRRs 2.00-3.91), and a small to moderate association between epilepsy and the prevalence of digestive organs and respiratory tract cancers (IRRs 1.17-1.35), independent of AEDs use.
There was no association between epilepsy and the other specified forms of cancer (including liver) in patients with no record of AEDs use. The risk of CNS cancers almost doubled (IRRs 3.91 vs. 2.00) in epilepsy patients treated with AEDs in comparison with that in epilepsy patients not treated with AEDs. They also found that, an increased risk of liver cancer is associated with epilepsy in people using AEDs (Kaae et al., 2014). A follow-up study from Olsen et al., found that epilepsy patients were at significantly increased risks of brain and CNS (RR = 5.7) as well as lung (RR = 1.4) cancer. However, the incidence of bladder cancer and malignant melanoma significantly decreased (RR = 0.6 vs. 0.5). The risk of non-Hodgkin's lymphoma increased but the change was not significant (RR = 1.4), this may have been associated with the use of phenytoin (Olsen, Boice, Jensen, & Fraumeni, 1989). Another nested case-control study revealed a significantly high risk of lung cancer but a low risk of bladder cancer among 8,004 epileptic patients in Denmark (Olsen et al., 1993). However, lung cancer was not associated with any AEDs, and the use of phenobarbital was inversely correlated with bladder cancer (Olsen et al., 1993).
Previous studies using phenobarbital and phenytoin in a rodent model revealed that phenobarbital caused liver tumors, whereas phenytoin promoted not only liver tumors but also lymphoma (Braeuning et al., 2014;Diwan, Henneman, Nims, & Rice, 1993;Murray, Hill, Hegemier, & Hurwitz, 1996). Although phenobarbital represented a potent nongenotoxic liver tumor promoter in rodents, the role of phenobarbital in the cancer risk of epilepsy patients remains controversial (La Vecchia & Negri, 2014). Phenytoin has been implicated in the causation of human lymphoma, myeloma, and neuroblastoma (Tittle & Schaumann, 1992). Evidence of carcinogenicity from long-term pharmacoepidemiological data on these two types of AEDs is inconsistent.
Our results did not show a correlation between phenobarbital and HCC; however, we observed that phenytoin was associated with the risk of HCC, particularly in the patients with the DDD of >60.
Valproate has been reported to have an antiproliferative effect in some cancers, both in vitro and in vivo (Abaza, Bahman, & Al-Attiyah, 2014;Wang et al., 2013). A multicenter, randomized Phase II study was conducted to combine weekly paclitaxel medication with valproate as second-line chemotherapy to improve survival in gastric cancer patients (Fushida et al., 2015). Recently, new AEDs appear to be safe; however, there is no regulatory testing to monitor carcinogenicity after listing.
Consistent with our findings, higher doses of phenytoin (>60 DDDs) were highly correlated with the risk of HCC; however, this association was not found in DDD of ≤60.
Our study has two limitations. First, the NHIRD does not provide patient details that might indicate confounding risk factors for HCC, such as body mass index, lifestyle factors (including smoking, alcohol consumption, and physical activity), environmental exposure, and family history of malignancy, which may play a critical role in the final outcome. Second, we were unable to contact the patients directly to obtain additional information because of the anonymized nature of the database. The claims data in the NHIRD were used primarily for administrative billing purposes and were not analyzed for scientific reliability for clinical studies.

| CONCLUSION
We have found an association between higher doses of phenytoin and HCC, after having adjusted for phenobarbital and primidone, clonazepam, clorazepate and diazepam, and other AEDs, and for the comorbidities. This association was not confirmed for low-dose phenytoin.

CONFLICTS OF INTERESTS
All authors declare that they have no conflict of interest.