Secondary Logo

Share this article on:

A review of epidemiological data on epilepsy, phenobarbital, and risk of liver cancer

La Vecchia, Carloa,b; Negri, Evaa

European Journal of Cancer Prevention: January 2014 - Volume 23 - Issue 1 - p 1–7
doi: 10.1097/CEJ.0b013e32836014c8
Review Article: Gastrointestinal Cancer

Phenobarbital is not genotoxic, but has been related to promotion of liver cancer (as well as inhibition) in rodents. In October 2012, we carried out a systematic literature search in the Medline database and searched reference lists of retrieved publications. We identified 15 relevant papers. Epidemiological data on epileptics/anticonvulsant use and liver cancer were retrieved from eight reports from seven cohort (record linkage) studies of epileptics, and data on phenobarbital use from a pharmacy-based record linkage investigation of patients treated with phenobarbital (three reports), plus a case–control study nested in one of the cohort studies and including information on phenobarbital use. Of the studies of cancer in epileptics, two showed no excess risk of liver cancer. A long-term (1933–1984) Danish cohort study of epileptics found relative risks (RRs) of 4.7 [95% confidence interval (CI) 3.2–6.8] of liver cancer and of 2.2 (95% CI 1.2–3.5) of biliary tract cancers. Such apparent excess risks could, however, be largely or completely attributed to thorotrast, a contrast medium used in the past in epileptic patients for cerebral angiography. A Finnish cohort study of epileptics obtained an RR of 1.7 (95% CI 1.2–2.4). Such an apparent excess risk, however, was not related to phenobarbital or to any specific anticonvulsant drug. The long-term follow-up of two UK cohorts found some excess risk of liver cancer among severe, but not among mild, epileptics. Some excess risk of liver cancer was also found in cohort studies of patients hospitalized for epilepsy in Sweden and Taiwan, in the absence, however, of association with any specific drugs. A UK General Practice database, comparing epileptics treated with valproate with unexposed ones, found a very low incidence of liver cancer. Of the studies of cancer in patients treated with phenobarbital, a large US pharmacy-based cohort investigation showed no excess risk of liver cancer. In a case–control study, nested in the Danish cohort of epileptics, no association was observed between phenobarbital and liver cancer among patients who had not received thorotrast (RR=1.0 for liver and 0.8 for biliary tract cancers). Thus, some, although not all, studies reported excess risk of all cancers and liver cancer in severe, but not in milder epileptics. There is no evidence of a specific role of phenobarbital in human liver cancer risk, but data on the topic are limited.

aDepartment of Epidemiology, Mario Negri Institute for Pharmacological Research

bDepartment of Clinical Sciences and Community Health, University of Milan, Milan, Italy

Correspondence to Carlo La Vecchia, MD, Department of Epidemiology, Mario Negri Institute for Pharmacological Research, Via Giuseppe La Masa, 1920156 Milan, Italy Tel: +39 02 39014 527; fax: +39 02 33200 231; e-mail:

Received January 21, 2013

Accepted January 21, 2013

Back to Top | Article Outline


Primary liver cancer is the sixth most common cancer in the world, and the third most common cause of cancer mortality. The estimated number of new cases in 2008 was about 750 000 worldwide, with about 700 000 deaths (Jemal et al., 2011). Of these, over 80% are from low-income and middle-income countries, and about 50% from China alone. Hepatocellular carcinoma (HCC) is the main type of liver cancer worldwide, accounting for over 90% of cases.

Chronic infections with hepatitis B and C viruses are the main causes of HCC worldwide. Other relevant risk factors are alcohol and alcohol-related cirrhosis, tobacco, being overweight, and diabetes and, in selected low-income countries, contamination of cereal foodstuff with aflatoxin. Oral contraceptives and selected other drugs have also been related to the risk of HCC (Franceschi et al., 2006; Chuang et al., 2009; La Vecchia and Bosetti, 2009; Trichopoulos et al., 2011).

Among drugs related to liver cancer is phenobarbital, as studies on rodents had shown that phenobarbital has a promoting effect on liver carcinogenesis (Peraino et al., 1971, 1980; Diwan et al., 1995). The experimental evidence indicates that the mechanism of phenobarbital in the production of hepatic tumors is likely non genotoxic and involves tumor promotion (IARC, 2001). In other studies of rodents, however, phenobarbital appeared to inhibit the development of liver cancer (Pereira et al., 1986).

The topic of phenobarbital and the risk of cancer in humans has been addressed in a number of epidemiological studies, following the reports in the mid 1970s and early 1980s from a Danish cohort of epileptics (Clemmesen et al., 1974; Clemmesen and Hjalgrim-Jensen, 1981; Olsen et al., 1995), which showed an over three-fold excess risk of liver cancer. These epidemiological investigations include both a number of studies considering the risk of cancer in patients treated (or hospitalized) for epilepsy in general, and more specifically, two studies addressing the issue of phenobarbital and the risk of (liver) cancer. Their main findings are reviewed below. Most studies consider liver cancer, including mainly HCC, but some also include intrahepatic bile duct and other rarer or unspecified types of liver cancer. The scanty evidence on bile duct cancers is also considered separately, whenever available. We did not consider case series on alterations of liver enzymes and morphology (Aiges et al., 1980), or single case reports of liver tumors (Ferko et al., 2003; Cerminara et al., 2012).

Back to Top | Article Outline

Materials and methods

In October 2012, we carried out a systematic literature search in the Medline database, using PubMed, and the following search string: (‘Epilepsy’[Mesh] OR ‘Phenobarbital’[Mesh] OR ‘Anticonvulsants’[Mesh]) AND (‘Risk’[MESH] OR ‘Epidemiologic Studies’[Mesh]) AND {‘Neoplasms/epidemiology’[Mesh] OR ‘Neoplasms/etiology’[Mesh] OR ‘Neoplasms/prevention and control’[Mesh] OR (‘mortality’[Subheading] OR ‘mortality’[All Fields] OR ‘mortality’[MeSH Terms])}. This identified 2077 papers.

To identify further articles, we searched the reference lists of retrieved publications and of earlier reviews, including the IARC Monograph 79 (IARC, 2001). We also searched for articles citing seminal articles using Google Scholar. This led to the identification of four additional papers.

Each publication identified in this process was reviewed and included in the analysis if the following criteria were fulfilled:

  • Cohort or case–control design.
  • Studying the risk of cancer in epileptics or users of phenobarbital or of anticonvulsants.
  • Reporting the number of observed and expected cases and/or the estimates of the standardized incidence ratio (SIR) or the standardized mortality ratio (SMR), of the odds ratio (OR) or relative risk (RR), and the corresponding confidence intervals (CIs) – or information sufficient to calculate them – for the risk of liver cancer/HCC/bile duct cancer/gallbladder cancer.

We identified 15 papers from seven cohort studies (Clemmesen et al., 1974; White et al., 1979; Friedman and Ury, 1980; Clemmesen and Hjalgrim-Jensen, 1981; Shirts et al., 1986; Olsen et al., 1989, 1995; Selby et al., 1989; Klenerman et al., 1993; Nilsson et al., 1997; Lamminpaa et al., 2002; Singh et al., 2009, 2012; Chang et al., 2012) and two nested case–control studies in two of the above cohorts (Olsen et al., 1995; Friedman et al., 2009). Various reports were considered for each study, when providing additional information.

We reviewed all the studies and abstracted the following information in a standard format: study design; country; period of enrollment and/or follow-up; number of participants (cases, controls, or noncases or cohort size); covariates adjusted for in the analysis; risk estimates SIRs, SMRs, (ORs or RRs, collectively referred to as RRs) for categories of exposure and the corresponding 95% CIs; and, when available, the number of cases and noncases for each exposure of interest.

Back to Top | Article Outline

Studies of cancer in epileptics

Table 1 presents the summary results of studies of cancer in epileptics.

Table 1

Table 1

Back to Top | Article Outline

Clemmesen et al. (1974), Clemmesen and Hjalgrim-Jensen (1981), Olsen et al. (1989, 1995)

The issue of a possible relationship between antiepileptic drugs and the risk of cancer emerged in a Danish cohort study on 4627 male and 3810 female patients admitted to a Danish epilepsy center in the period 1939–1962 and followed up over the period 1943–1976 using population registers of deaths and cancer registries (i.e. a record linkage approach; Clemmesen et al., 1974). Initially, that study found a gross excess risk of brain cancer, which, however, leveled off over time (Clemmesen and Hjalgrim-Jensen, 1981; Olsen et al., 1989), and an over three-fold excess risk of liver cancer.

The most recent follow-up of that cohort study (Olsen et al., 1989, 1995) included 8004 patients hospitalized for epilepsy between 1933 and 1962, and followed up through 1984. A total of 26 cases of liver cancer were observed in the cohort study, corresponding to an RR of 4.7 (95% CI 3.2–6.8). Corresponding figures for biliary tract cancers were 14 cases and an RR of 2.2 (95% CI 1.2–3.5). The excess liver cancer could, however, be largely or completely attributed to thorotrast, a contrast medium used in the past during cerebral angiography in a subgroup of patients, and a known liver carcinogen. Excluding patients known to have received thorotrast, there were nine deaths from liver cancer versus 4.3 expected, and the association was no longer significant (RR=1.9, 95% CI 0.9–3.6) (Olsen et al., 1989).

Back to Top | Article Outline

White et al. (1979)

A cohort of 2000 epileptic patients admitted between 1931 and 1971 to a referral center (Chalfont Centre for Epilepsy) in England were followed up for mortality between 1951 and 1977 through record linkage with the National Health Service Central Register and with national mortality data (White et al., 1979). Two subsequent, partly overlapping, reports are available for the same Chalfont Centre (Klenerman et al., 1993; Singh et al., 2009). Overall, 78 cancer deaths were observed versus 51.5 expected (RR=1.5, 95% CI 1.2–1.9). The only significant excess was for brain cancer (six observed versus 1.5 expected, RR=4.1, 95% CI 1.5–9.9). Some excess, of borderline significance, was also observed for lung (23 observed vs. 16.3 expected, RR=1.4, 95% CI 0.9–2.1) and breast cancers (eight observed vs. 3.6 expected, RR=2.2, 95% CI 1.0–1.4). There was no death from liver cancer and one death from gallbladder cancer. The expected number of liver and gallbladder cancers combined was 0.6, corresponding to a nonsignificant RR of 1.5 (95% CI 0.0–8.5) for both sites combined.

Thus, this study, based on a long-term follow-up of a group of severe epileptic patients, allowed the exclusion of an appreciable excess risk of liver cancer, although the number of expected cases and observed events was limited. No information was available on the type of drugs used. Nevertheless, the issue of phenobarbital was mentioned in the introduction and discussion, implying that at least a proportion of patients had used it.

Back to Top | Article Outline

Shirts et al. (1986)

In a population-based cohort of 959 patients with seizure disorders from Rochester, Minnesota, followed up between 1935 and 1979, 10 cases of all digestive tract cancers combined were observed versus 15.4 expected (RR=0.8, 95% CI 0.4–1.4). No cases of liver cancer occurred. Besides brain cancer (17 observed vs. 0.7 expected, RR=24.3), respiratory cancers (9 observed vs. 3.3 expected, RR=2.7) were significantly elevated (Shirts et al., 1986). Overall, there were 65 observed cancers versus 45.7 expected, corresponding to an RR of 1.6 (95% CI 1.1–1.8). Most of the excess risk for brain and for all cancers combined was restricted to the 5–10 years since the diagnosis of seizure: the RR at least 10 years after the diagnosis of seizure was 5.9 (95% CI 0.7–21.2) for brain tumors and 1.0 for all cancers, none of these being significant. Similarly, there was no duration–risk relationship for brain cancer in relation to medication use. This study, however, had no specific information on phenobarbital use. Nevertheless, on the basis of a uniquely large follow-up in a period (1935–1979) when phenobarbital was a largely used treatment for epilepsy, it allows the exclusion of an appreciable excess risk of (liver) cancer.

Back to Top | Article Outline

Klenerman et al. (1993)

An additional report is available from the Chalfont Centre for Epilepsy in England, including deaths that occurred between 1 January 1980 and 31 December 1990 among patients in the long-term residential care unit (Klenerman et al., 1993). The total number of patients is not reported and the total number of person-years was 3392. The mean age was 52 years and the male : female ratio was 2 : 1.

Overall, 113 deaths were observed versus 58.3 expected on the basis of national mortality rates, yielding an SMR of 1.9 (95% CI 1.6–2.3). A total of 29 deaths from all cancers were observed versus 14.8 expected (SMR 2.0, 95% CI 1.3–2.9). There were three deaths from hepatobiliary cancers versus 0.2 expected, corresponding to an SMR of 17.6 (95% CI 3.6–51.5). None of these was a primary liver cancer, two were gallbladder cancers, and one was a cholangiocarcinoma. Significant excesses were also observed for lung (SMR 3.3, 11 cases) and pancreatic cancer (SMR 6.4, four cases). No correlation with drug treatment was attempted.

It is not clear, however, whether some of the patients (and which ones) in the study by White et al. (1979) were included in this study. Information is also lacking on the total number of patients and on other characteristics. Further, this study is based on a highly selected population with severe chronic epilepsy, which is by no means representative of the generality of patients with epilepsy.

Back to Top | Article Outline

Nilsson et al. (1997)

The Stockholm County In-Patient Register was used to identify all individuals aged 15 years and above admitted to a hospital at least once during the period 1980–1989 who were discharged with a diagnosis of epilepsy (Nilsson et al., 1997). The cohort included 9061 individuals (5375 men and 3686 women) with a mean age of 54.3 years (range 15–97 years).

For 40% of patients, epilepsy was the only diagnosis at discharge of the index hospital admission, and it was the main diagnosis for an additional 20%. For the remaining patients, the main diagnoses were alcoholism (7%), cerebrovascular disease (3%), diabetes, heart disease, and dementia (1% each), and for the other 27% a wide range of diagnoses each representing less than 1% of patients. The reliability of the diagnosis of epilepsy was reviewed in a subsample of 765 patients using medical records. The diagnosis was erroneous in 21% of the subsample.

The end of follow-up was 31 December 1992. The person-years were 53 520, and during the follow-up, 4001 deaths were recorded (2503 men and 1498 women).

Expected numbers of cause-specific deaths were computed using mortality rates in Stockholm County and the person-years in the cohort by sex, quinquennia of age, and calendar period.

The SMR for total mortality was 3.6 (95% CI 3.5–3.7); it was 3.7 (95% CI 3.6–3.9) in men and 3.4 (95% CI 3.3–3.6) in women. There were 647 deaths from malignant neoplasms versus 253.3 expected, corresponding to an SMR of 2.6 (95% CI 2.4–2.8). When brain tumors were excluded, the observed number became 506 and the SMR 2.0 (95% CI 1.9–2.2). There were 23 observed deaths from cancers of the liver or intrahepatic bile ducts, as compared with 8.2 expected, and the SMR was 2.8 (95% CI 1.8–4.2). SMRs were significantly increased for cancers of the brain, oral cavity, esophagus, respiratory organs, melanoma, female breast, lymphatic, and hematopoietic tissue. Mortality was also significantly increased for several other non-neoplastic causes, including diseases of the circulatory, respiratory, and digestive systems, and injury and poisoning.

This is a population-based study where selection bias should be reduced. However, there was a high percentage of erroneous diagnoses (21%) in the validation study. The main issue that renders the study difficult to interpret in terms of excess liver cancer is that mortality from several neoplasms, as well as non-neoplastic causes, is significantly increased. There is therefore a lack of specificity in the observed association between epilepsy and (liver) cancer.

Back to Top | Article Outline

Lamminpaa et al. (2002)

A Finnish cohort study included 14 487 men and 13 932 women who had received reimbursement for antiepileptic drugs in 1979–1981, and were followed up to 1997 for cancer incidence through the Finnish Cancer Register, using a standard record linkage approach (Lamminpaa et al., 2002). Overall, 2242 cancer cases were observed versus 1743 expected, corresponding to an RR for all cancers combined of 1.29 (95% CI 1.23–1.34). Excluding nervous system cancers, the corresponding figures were 1968 cancers observed versus 1679 expected (RR=1.19, 95% CI 1.12–1.22). There were 37 cases of liver cancer versus 22 expected, corresponding to an RR of 1.71 (95% CI 1.20–2.35), and the RR was similar for those younger than 60 years of age (RR=2.12, nine cases) and 60 years of age and older (RR=1.60, 28 cases). Thorotrast, used in one hospital in Finland in the early 1950s, did not appear to explain the excess risk.

No information was provided on potential risks related to any specific antiepileptic drug, including phenobarbital. In the discussion, the frequency of use is given of the most common antiepileptic drugs, that is carbamazepine [2.8 defined daily doses (DDD) in Finland per 1000 inhabitants], valproate (1.31 DDD), and phenytoin (1.28 DDD), but no figure is given for phenobarbital.

The role of alcohol is also discussed, as 20% of newly diagnosed epilepsies in adults in Finland are alcoholic epilepsies, thus explaining – at least in part – the apparent excess of liver cancer incidence, as liver cancer is strongly related to alcohol drinking (Franceschi et al., 2006; Pelucchi et al., 2008; Chuang et al., 2009). This study is incidence-registration based, and epileptic patients are more medicalized than the general population. Thus, a role of increased diagnostic ascertainment in liver cancer incidence is also possible – although difficult to quantify. The overall excess of all cancers except brain cancer in this study is also at variance with most other investigations of epileptic patients, and points to a greater incidence of cancer in this cohort of epileptics because of smoking or other causes besides alcohol, as the RR for lung cancer was 1.29 (95% CI 1.15–1.43), and significant excess risks were observed from larynx, pancreas, and other major tobacco-related cancers. As liver cancer is a tobacco-related cancer (Pelucchi et al., 2008; Trichopoulos et al., 2011), smoking may also account for part of the apparent excess risk. Most likely, such a generalized excess risk of cancer can be attributed to differences with the reference population or biases, which may be less subject to cancer diagnosis and ascertainment.

Thus, this study shows an apparent excess risk of (liver) cancer in a relatively short-term (<10 years) follow-up of a large cohort of Finnish epileptic patients. There is no indication, however, that such an excess can be attributed to phenobarbital use.

Back to Top | Article Outline

Singh et al. (2009)

This is an additional report from the Chalfont Centre for Epilepsy, including a follow-up to 31 December 2003 of the severe epilepsy cohort of patients admitted from 1931 to 1971 originally reported by White et al. (1979) (n=1358 of the original 2000 patients). In addition, it includes a milder epilepsy cohort that included 4494 patients aged 35 years or older, notified for epilepsy to the same center from 1974 to 1979. The median age at entry was 44 years in the severe epilepsy cohort and 46 years in the milder epilepsy one. The median age at death or censoring was 65 and 70 years, respectively. Among severe epileptics, 68% were men, and among milder ones, 61% were men.

Among severe epileptics, there were 602 deaths versus 257.2 expected, corresponding to an SMR of 2.3 (95% CI 2.2–2.5). Of these, 125 were deaths from cancer (SMR 1.4, 95% CI 1.2–1.7). Among milder epileptics, there were 1114 deaths from all causes versus 1183.3 expected, corresponding to an SMR of 0.9 (95% CI 0.9–1.0). Overall, observed cancer deaths were 396 versus 425.2 expected (SMR 0.9, 95% CI 0.8–1.0). Among severe epileptics, there were four deaths from liver cancer versus 0.97 expected (SMR 4.1, 95% CI 1.1–10.6) and three deaths from gallbladder cancer versus 0.41 expected (SMR 7.3, 95% CI 1.5–21.4). Among milder epileptics, there were three deaths from liver cancer versus 4.79 expected, corresponding to an SMR of 0.6 (95% CI 0.1–1.8), and two deaths from gallbladder cancer (SMR 1.0, 95% CI 0.1–3.7). No information is provided on the type of treatment for both cohorts.

The severe epileptic cohort overlapped partly with the reports of White et al. (1979) and Klenerman et al. (1993); thus, the results are not independent. The excess risk of cancer in severe, but not in milder, epileptics can be attributed to deprivation and lifestyle factors associated with institutionalization. No information is available on treatment.

Back to Top | Article Outline

Chang et al. (2012)

A total of 2180 epileptic patients treated in a tertiary care hospital from Taiwan between 1989 and 2008 were compared for mortality with the general population of Taiwan. Follow-up was up to 31 December 2008. Overall, 256 deaths were observed versus 106.6 expected, corresponding to an SMR of 2.5 (95% CI 2.2–2.8). There were 48 cancer deaths versus 28.1 expected, corresponding to an SMR of 1.7 (no correct CI provided), 40 deaths from all cancers except brain (SMR 1.4, of borderline significance, no correct CI provided) and 15 deaths from liver cancer versus 5.8 expected (SMR 2.6, no correct CI provided). No significant excess was found for lung cancer (SMR=1.1). The high liver cancer rates were related to the elevated prevalence of hepatitis B in this population, but no reference was made on any specific antiepileptic drugs, including phenobarbital (Chang et al., 2012).

Back to Top | Article Outline

Singh et al. (2012)

This study includes information relevant only to valproate. Using the UK General Practice Research Database, 2997 epileptic patients who were prescribed valproate between 1987 and 1992 were compared with 11 988 unexposed epileptics (Singh et al., 2012). Follow-up ended in December 2005. During 30 327 person-years of follow-up in the exposed group, no case of liver cancer was observed versus 1 case among 120 866 person-years in the unexposed group. No information on phenobarbital was available. This study shows a low incidence of liver cancer among epileptic patients treated in general practice, in contrast with some reports of severe, hospitalized epileptics (Selby et al., 1989; Nilsson et al., 1997; Singh et al., 2009; Chang et al., 2012).

Back to Top | Article Outline

Studies of cancer in patients treated with phenobarbital

Table 2 presents the summary findings of studies including data on phenobarbital users and risk of liver cancer.

Table 2

Table 2

Back to Top | Article Outline

Friedman and Ury (1980, 1983), Selby et al. (1989)

A pharmacy-based record linkage study of 143 584 outpatients at the Kaiser-Permanente Medical Care Program (KPMCP) in San Francisco considered the incidence of cancer in relation to a large number (43) of medications used (Friedman and Ury, 1980, 1983), starting in July 1969.

In the 1976 follow-up of this database (Friedman and Ury, 1980), there were 235 cancers of all sites among phenobarbital users versus 211.6 expected, on the basis of age-specific and sex-specific incidence data for all pharmacy users, corresponding to an RR of 1.1, nonsignificant. No association was reported with liver cancer. The limitations of this study include multiple testing, because of the large numbers of medications used and type of cancers considered, and the relatively short follow-up, with the consequent limited number of expected cases for most cancers considered. In a subsequent follow-up of the same dataset from 1969 to 1978 (Friedman and Ury, 1983), six cases of gallbladder cancer were observed versus 1.8 expected (significant). No data on liver cancer were reported, indicating that there was no significant excess risk of liver cancer, as only the significant associations were given in the results.

A subsequent follow-up through 1984 is available from the same study (Selby et al., 1989). The only sites showing significant associations with phenobarbital use were inversely related to risk, that is large intestine (RR=0.72, 36 cases) and bladder (RR=0.64, eight cases) cancers. There was no longer an association with gallbladder cancer – nor was any association observed with liver cancer even at this longer follow-up – and the RR for all cancers considered was 1.01, nonsignificant.

Thus, in three subsequent follow-up reports from this large record linkage study, there was no indication that phenobarbital is related to the risk of liver cancer in this Californian population from a large health maintenance organization.

Back to Top | Article Outline

Olsen et al. (1995)

A nested case–control study was carried out on the 26 cases of liver cancer in the above-described Danish cohort (and 49 controls matched by sex, age, and survival time) and on 13 of the 14 cases of biliary tract cancer (and 24 controls) (Olsen et al., 1995) to specifically consider and disentangle exposure to phenobarbital (and phenytoin), and to thorotrast, among cases and controls. Information on phenobarbital (and other anticonvulsant) use was obtained from medical records of the epilepsy center. Overall, phenobarbital use was associated with a nonsignificant excess risk of cancers of the liver (RR=2.0, 95% CI 0.5–7.2) and of the biliary tract (RR=1.5, 95% CI 0.4–6.7). There was, however, no association among patients who did not receive thorotrast, with RRs of 1.0 (95% CI 0.1–8.0, on the basis of six cases and 11 controls) for liver cancer, and of 0.8 (95% CI 0.1–4.2) for biliary tract cancers (seven cases, 14 controls). Even at the highest dose of phenobarbital use (≥750 g), the association with liver cancer was nonsignificant among non thorotrast users (RR=2.8, 95% CI 0.2–39). Thus, the apparent excess liver and biliary tract cancers in this Danish cohort was attributed to factors other than anticonvulsant (and specifically other than phenobarbital) use, including essentially thorotrast used in the past as a contrast medium for cerebral angiography, and factors associated with related cirrhosis of the liver.

Back to Top | Article Outline

Friedman et al. (2009)

The KPMCP of northern California database described above (Friedman and Ury, 1980, 1983; Selby et al., 1989) was subsequently utilized in a nested case–control study including 131 743 patients with at least one incident cancer between August 1994 and June 2006 and 10 controls per case randomly selected and free of the cancer of interest on the date of the case’s diagnosis. SIRs were derived using the method of external adjustment by Schneeweiss (2006). There was no association with gallbladder (SIR=0.6), lung (SIR=1.0), ovary (SIR=1.1), or brain (SIR=1.2) cancers. No data were given for liver cancer, indicating an absence of association. The only cancer site showing a significant excess for phenobarbital users was small intestine (10 cases, SIR=5.1, 95% CI 2.4–11.0). This, however, may reflect the role of chance in multiple testing.

Back to Top | Article Outline


Anticonvulsant drugs and, specifically, phenobarbital, are known to increase liver enzymes related to the cytochrome 450 system and consequently, under selected circumstances, may both activate and detoxify carcinogens (Olsen et al., 1989; Singh et al., 2005). In rodents exposed to hepatocarcinogens, high doses of phenobarbital have been show to increase liver cancer promotion but also to inhibit liver cancer development (Peraino et al., 1971, 1980, 1986; IARC, 2001).

Epidemiologic data on phenobarbital and liver cancer are limited, but indicate an absence of any specific association and, in any case, allow the exclusion of an appreciable excess risk.

We reviewed eight reports from seven cohort studies of epileptic patients providing some data on liver cancer. Some, although not all, showed some excess risk of (liver) cancer among severe, but not among mild, epileptics (Selby et al., 1989; Nilsson et al., 1997; Singh et al., 2009, 2012; Chang et al., 2012). This is attributed to deprivation and consequent unfavorable lifestyle factors among severe (hospitalized) epileptics (Forsgren et al., 2005; Neligan et al., 2010), although data on lifestyle factors are scarce and inconsistent, particularly for outpatients (Cockerell et al., 1996). A Danish cohort study of epileptics originally reported an excess risk of liver cancer (Clemmesen et al., 1974; Olsen et al., 1989, 1995), but showed, in subsequent analyses, that such an apparent excess risk was no longer significant among patients not receiving thorotrast as a contrast medium for brain angiography. An English (White et al., 1979) and a US cohort study (Shirts et al., 1986) showed no association with liver cancer, whereas a Finnish study (Lamminpaa et al., 2002), a subsequent report from the English cohort (Singh et al., 2009), a Swedish (Nilsson et al., 1997), and a Taiwanese (Chang et al., 2012) cohort showed some excess (liver) cancer among (severe) epileptics, which, however, was not related to the use of phenobarbital.

Information on the risk of liver cancer among phenobarbital users – the issue specifically relevant to the present report – was available from a large US pharmacy-based record linkage investigation (Friedman and Ury, 1980, 1983; Selby et al., 1989), showing no association between phenobarbital and liver cancer. A case–control study nested in the Danish cohort (Olsen et al., 1995), and including detailed information on exposure to phenobarbital and thorotrast, showed no excess liver or biliary tract neoplasms in relation to phenobarbital use among patients who had not received thorotrast. Any modest excess of liver and biliary tract cancers among patients with epilepsy who were not known to have received thorotrast may also be attributed to undocumented exposure to that potent liver carcinogen (Olsen et al., 1989).

These data indicate therefore that factors other than anticonvulsant treatment, and specifically thorotrast administration and factors associated with cirrhosis, were responsible of the earlier apparent excess risk of liver cancer in selected groups of epileptic patients, and that phenobarbital is of little importance in the etiology of liver cancer in humans (Olsen et al., 1995). That Danish study remains the largest dataset of liver cancer in phenobarbital users with adequate allowance for covariates, and its null results have been confirmed subsequently, as no appreciable excess liver cancer has been related to phenobarbital.

Consequently, despite the limitations of published data and the evidence of some excess risk of (liver) cancer among severe, but not milder, epileptics, there is no convincing evidence of a specific and relevant role of phenobarbital in human liver cancer risk (Singh et al., 2005).

Back to Top | Article Outline


This work was supported by an unconditional consulting agreement with BASF SE. This work of C.L.V. and E.N. is also supported by a grant from the Italian Association for Cancer Research (AIRC contract 10068).

Back to Top | Article Outline

Conflicts of interest

There are no conflicts of interest.

Back to Top | Article Outline


Aiges HW, Daum F, Olson M, Kahn E, Teichberg S.The effects of phenobarbital and diphenylhydantoin on liver function and morphology.J Pediatr1980;97:22–26.
Cerminara C, Bagnolo V, De Leonardis F, Coniglio A, Roberto D, Compagnone E, et al..Hepatocellular adenoma associated with long-term exposure to phenobarbital: a paediatric case report.Childs Nerv Syst2012;28:939–941.
Chang YH, Ho WC, Tsai JJ, Li CY, Lu TH.Risk of mortality among patients with epilepsy in southern Taiwan.Seizure2012;21:254–259.
Chuang SC, La Vecchia C, Boffetta P.Liver cancer: descriptive epidemiology and risk factors other than HBV and HCV infection.Cancer Lett2009;286:9–14.
Clemmesen J, Hjalgrim-Jensen S.Does phenobarbital cause intracranial tumors? A follow-up through 35 years.Ecotoxicol Environ Saf1981;5:255–260.
Clemmesen J, Fuglsang-Frederiksen V, Plum CM.Are anticonvulsants oncogenic?Lancet1974;1:705–707.
Cockerell O, Gupta S, Sander JWAS, Shorvon SD.Risk factors for cancer and vascular deaths in patients with epilepsy in a community and a residential population: a case-controlled study.J Epilepsy1996;9:23–26.
Diwan BA, Henneman JR, Rice JM.Further evidence for promoter-dependent development of hepatoblastoma in the mouse.Cancer Lett1995;89:29–35.
Ferko A, Bedrna J, Nozicka J.Pigmented hepatocellular adenoma of the liver caused by long-term use of phenobarbital.Rozhl Chir2003;82:192–195.
Forsgren L, Hauser WA, Olafsson E, Sander JW, Sillanpaa M, Tomson T.Mortality of epilepsy in developed countries: a review.Epilepsia2005;46Suppl 1118–27.
Franceschi S, Montella M, Polesel J, La Vecchia C, Crispo A, Dal Maso L, et al..Hepatitis viruses, alcohol, and tobacco in the etiology of hepatocellular carcinoma in Italy.Cancer Epidemiol Biomarkers Prev2006;15:683–689.
Friedman GD, Ury HK.Initial screening for carcinogenicity of commonly used drugs.J Natl Cancer Inst1980;65:723–733.
Friedman GD, Ury HK.Screening for possible drug carcinogenicity: second report of findings.J Natl Cancer Inst1983;71:1165–1175.
Friedman GD, Jiang SF, Udaltsova N, Quesenberry CP Jr, Chan J, Habel LA.Epidemiologic evaluation of pharmaceuticals with limited evidence of carcinogenicity.Int J Cancer2009;125:2173–2178.
.Phenobarbital and its sodium salt. In: IARC monographs on the evaluation of carcinogenic risks to humans, vol. 79, Some thyrotropic agents2001.Lyon:IARC Press;161–288.
Jemal A, Bray F, Center MM, Ferlay J, Ward E, Forman D.Global cancer statistics.CA Cancer J Clin2011;61:69–90.
Klenerman P, Sander JW, Shorvon SD.Mortality in patients with epilepsy: a study of patients in long term residential care.J Neurol Neurosurg Psychiatry1993;56:149–152.
Lamminpaa A, Pukkala E, Teppo L, Neuvonen PJ.Cancer incidence among patients using antiepileptic drugs: a long-term follow-up of 28,000 patients.Eur J Clin Pharmacol2002;58:137–141.
La Vecchia C, Bosetti C.Oral contraceptives and neoplasms other than breast and female genital tract.Eur J Cancer Prev2009;18:407–411.
Neligan A, Bell GS, Shorvon SD, Sander JW.Temporal trends in the mortality of people with epilepsy: a review.Epilepsia2010;51:2241–2246.
Nilsson L, Tomson T, Farahmand BY, Diwan V, Persson PG.Cause-specific mortality in epilepsy: a cohort study of more than 9,000 patients once hospitalized for epilepsy.Epilepsia1997;38:1062–1068.
Olsen JH, Boice JD Jr., Jensen JP, Fraumeni JF Jr.Cancer among epileptic patients exposed to anticonvulsant drugs.J Natl Cancer Inst1989;81:803–808.
Olsen JH, Schulgen G, Boice JD Jr, Whysner J, Travis LB, Williams GM, et al..Antiepileptic treatment and risk for hepatobiliary cancer and malignant lymphoma.Cancer Res1995;55:294–297.
Pelucchi C, Gallus S, Garavello W, Bosetti C, La Vecchia C.Alcohol and tobacco use, and cancer risk for upper aerodigestive tract and liver.Eur J Cancer Prev2008;17:340–344.
Peraino C, Fry RJ, Staffeldt E.Reduction and enhancement by phenobarbital of hepatocarcinogenesis induced in the rat by 2-acetylaminofluorene.Cancer Res1971;31:1506–1512.
Peraino C, Staffeldt EF, Haugen DA, Lombard LS, Stevens FJ, Fry RJ.Effects of varying the dietary concentration of phenobarbital on its enhancement of 2-acetylaminofluorene-induced hepatic tumorigenesis.Cancer Res1980;40:3268–3273.
Pereira MA, Klaunig JE, Herren-Freund SL, Ruch RJ.Effect of phenobarbital on the development of liver tumors in juvenile and adult mice.J Natl Cancer Inst1986;77:449–452.
Schneeweiss S.Sensitivity analysis and external adjustment for unmeasured confounders in epidemiologic database studies of therapeutics.Pharmacoepidemiol Drug Saf2006;15:291–303.
Selby JV, Friedman GD, Fireman BH.Screening prescription drugs for possible carcinogenicity: eleven to fifteen years of follow-up.Cancer Res1989;49:5736–5747.
Shirts SB, Annegers JF, Hauser WA, Kurland LT.Cancer incidence in a cohort of patients with seizure disorders.J Natl Cancer Inst1986;77:83–87.
Singh G, Driever PH, Sander JW.Cancer risk in people with epilepsy: the role of antiepileptic drugs.Brain2005;128:7–17.
Singh G, Fletcher O, Bell GS, McLean AE, Sander JW.Cancer mortality amongst people with epilepsy: a study of two cohorts with severe and presumed milder epilepsy.Epilepsy Res2009;83:190–197.
Singh G, Bell GS, Driever PH, Sander JW.Cancer risk in people with epilepsy using valproate-sodium.Acta Neurol Scand2012;125:234–240.
Trichopoulos D, Bamia C, Lagiou P, Fedirko V, Trepo E, Jenab M, et al..Hepatocellular carcinoma risk factors and disease burden in a European cohort: a nested case–control study.J Natl Cancer Inst2011;103:1686–1695.
White SJ, McLean AE, Howland C.Anticonvulsant drugs and cancer. A cohort study in patients with severe epilepsy.Lancet1979;2:458–461.

anticonvulsant; humans; liver cancer; phenobarbital; risk

© 2014 Lippincott Williams & Wilkins, Inc.