1.1.1. Nomenclature

• Chem. Abstr. Serv. Reg. No.: 562-10-7
• Deleted CAS Reg. No.: 121367-03-1
• Chem. Abstr. Name: Butanedioic acid, compd. with N,N-dimethyl-2-[1-phenyl-1-(2-pyridinyl)-ethoxy]ethanamine (1:1); N,N-dimethyl-2-[1-phenyl-1-(2-pyridinyl)-ethoxy] ethanamine, butanedioate (1:1)
• IUPAC Systematic Name: Pyridine, 2-[α-[2-(dimethylamino)ethoxy]-α-methylbenzyl]-, succinate (1:1); succinic acid, compd. with 2-[α-[2-(dimethylamino)-ethoxy]-α-methylbenzyl]pyridine (1:1)
• Synonyms: Doxylamine succinate (1:1); histadoxylamine succinate

1.1.2. Structural and molecular formulae and relative molecular mass

1.1.3. Chemical and physical properties of the pure substance

• (a) Description: White or creamy-white powder (Gennaro, 1995)
• (b) Melting-point: 100–104 °C (Budavari, 2000)
• (c) Solubility: Very soluble in water (1000 g/L at 25 °C) and chloroform; soluble in ethanol; slightly soluble in benzene and diethyl ether (Gennaro, 1995; American Hospital Formulary Service, 2000; Budavari, 2000)
• (d) Spectroscopy data: Infrared [proton (9383), grating [45206P] spectral data have been reported (Sadtler Research Laboratories, 1980).
• (e) Ionization constants: pKa, 5.8 and 9.3 (American Hospital Formulary Service, 2000)

1.1.4. Technical products and impurities

Doxylamine is currently formulated exclusively as the succinate salt. Doxylamine succinate is commercially available as an over-the-counter 25-mg tablet and a 50-mg liquid-filled capsule; it is also available in combination with antitussives and decongestants. Inactive ingredients in commercial formulations may include dibasic calcium phosphate, FD&C Blue No. 1, aluminium lake, magnesium stearate, microcrystalline cellulose and sodium starch glycolate (Gennaro, 1995; American Hospital Formulary Service, 2000; Medical Economics Co., 2000).

Trade names for doxylamine succinate include Alsadorm, Decapryn, Donormyl, Dormacil, Dormidina, Doxised, Doxy-Sleep-Aid, Dozile, Duebien, Evigoa D, Gittalun, Hewedormir, Hoggar N, Lidène, Mereprine, Munleit, Nighttime Sleep Aid, Noctyl, Nytol, Restaid, Sanalepsi, SchlafTabs-ratiopharm, Sedaplus, Sleep Aid, Sleep Easy and Unisom (Royal Pharmaceutical Society of Great Britain, 2000; Swiss Pharmaceutical Society, 2000; Vidal, 2000).

Doxylamine succinate is also used in over 50 pharmaceutical preparations in combination with other drugs.

1.1.5. Analysis

The US Pharmacopeia specifies ultraviolet absorption spectrophotometry with comparison to standards as the method for identifying doxylamine succinate; titration with perchloric acid is used to assay its purity. In pharmaceutical preparations, doxylamine succinate is identified by infrared absorption spectrophotometry; ultraviolet absorption spectrophotometry and high-performance liquid chromatography are used to assay for content (US Pharmacopeial Convention, 1999).

4.1.1. Humans

The pharmacokinetics of doxylamine succinate was determined in 16 healthy male volunteers after administration of a single oral dose of 25 mg. The mean peak plasma concentration was 99 ng/mL, the peak time was 2.4 h after the dose, the elimination half-time was 10.1 h, and the apparent oral clearance was 217 mL/min (Friedman & Greenblatt, 1985). Clearance of a single oral dose of 25 mg of doxylamine succinate was slower (174 versus 240 mL/min) in older men (60–87 years) than in younger men (20–43 years), but no difference was seen in women (Friedman et al., 1989).

N-Desmethyldoxylamine, N,N-didesmethyldoxylamine and their respective N-acetyl conjugates were identified as urinary metabolites after oral administration of 50 mg of doxylamine succinate to a single volunteer (Ganes & Midha, 1987). Approximately 1% of an oral dose of 75–100 mg of doxylamine succinate was recovered in the urine of male volunteers as quaternary ammonium-linked glucuronides (Luo et al., 1991).

4.1.2. Experimental systems

Total urine and faecal recovery of radiolabel after oral administration of [¹⁴C]-doxylamine succinate at 13.3 or 133 mg/kg bw to male and female Fischer 344 rats was > 90%, regardless of sex and dose. The identified unconjugated urinary metabolites (representing 36–44% of the radiolabel) included doxylamine N-oxide, N-desmethyldoxylamine, N,N-didesmethyldoxylamine and ring-hydroxylated products of doxylamine and desmethyldoxylamine (Holder et al., 1987). The identified conjugated metabolites (representing 44–55% of the radiolabel) included doxylamine O-glucuronide, N-desmethyldoxylamine O-glucuronide and N,N-didesmethyldoxylamine O-glucuronide (Holder et al., 1990). N-Acetyl conjugates of N-desmethyl- and N,N-didesmethyldoxylamine were tentatively identified in rat urine (Ganes et al., 1986).

¹⁴C-Doxylamine succinate given as a single oral dose of 13 mg/kg bw to four female rhesus monkeys (Macaca mullata) was metabolized by at least four pathways: a major pathway to mono- and didesmethyldoxylamine via successive N-demethylation; a major pathway to side-chain cleavage products via direct side-chain oxidation and/or deamination; a minor pathway to the N-oxide; and a minor pathway to unknown polar metabolites (Slikker et al., 1986). In two male squirrel monkeys (Saimiri sciureus) given 20 mg of doxylamine succinate orally twice daily for 1 week, only the N-acetyl conjugate of N,N-didesmethyldoxylamine was detected in the urine (Ganes et al., 1986). Repeated oral administration of 7 mg/kg bw per day of Bendectin^® during days 22–50 of gestation to three cynomolgus monkeys (Macaca fascicularis), four rhesus monkeys (Macaca mulatta) and five baboons (Papio cynocephalus) did not alter the pharmacokinetics of doxylamine (Rowland et al., 1989).

Analysis of fetal plasma samples by high-performance liquid chromatography demonstrated the presence of doxylamine and metabolites after a single intravenous injection of 13.3 mg/kg bw doxylamine succinate to three rhesus monkeys in late-term pregnancy (Slikker et al., 1987, 1989).

4.1.3. Comparison of animals and humans

Similar demethylated metabolites and their N-acetylated derivatives have been determined in the urine of humans, monkeys and rats after oral doses of doxylamine succinate. In addition, rats metabolized the compound by N-oxidation, aromatic hydroxylation and ether cleavage.

4.2.1. Humans

In 109 cases of intoxication with doxylamine preparations, the most frequent symptoms included impaired consciousness, seizures, tachycardia, mydriasis and a psychosis-like state similar to catatonic stupor. A serious complication may be rhabdomyolysis with subsequent impairment of renal function and acute renal failure (Köppel et al., 1987).

4.2.2. Experimental systems

Administration of feed containing doxylamine succinate at a concentration of 0, 40, 375, 750 or 1500 mg/kg (expressed as free amine) to male and female B6C3F₁ mice for 7 or 15 days resulted in decreased serum thyroxine concentrations (approximately 80% of control) with compensatory increases (fourfold) in those of serum thyroid stimulating hormone, as also seen with phenobarbital at a concentration of 375 mg/kg of diet (expressed as free acid). No clear changes in serum triiodothyronine concentration were apparent (Bookstaff et al., 1996).

Groups of 12 male and female Fischer 344 rats were given diets containing doxylamine succinate at a concentration of 0, 162, 405, 1012, 2530 or 6325 ppm (expressed as free amine) for 90 days. The only histopathological changes seen were cytoplasmic vacuolization or fatty change in the liver and cytomegaly in the parotid salivary glands (Jackson & Blackwell, 1988a). In groups of 12 male and female B6C3F₁ mice given diets containing doxylamine succinate at a concentration of 0, 80, 162, 325, 750 or 1500 ppm (expressed as free amine) for 90 days, the liver was the only organ affected. The histological lesions consisted of mild to severe cytomegaly and/or karyomegaly and a high incidence of hepatocellular necrosis in animals of each sex at the highest concentration in feed, although necrosis was also observed at lower concentrations (Jackson & Blackwell, 1988b)

4.3.1. Humans

Doxylamine succinate has rarely been used on its own, but it has been used extensively in pregnancy in combination with pyridoxine as Bendectin^® for the treatment of nausea and vomiting. Reviews of the extensive literature on Bendectin^® use, covering many thousands of women exposed during pregnancy, have concluded that the drug is not a human teratogen (Shiono & Klebanoff, 1989; McKeigue et al., 1994; Brent, 1995).

4.3.2. Experimental systems

No satisfactory studies on doxylamine alone were available to the Working Group. Studies of Bendectin^® in rats and non-human primates were available.

Groups of 37–41 pregnant CD Sprague–Dawley rats were given 0, 200, 500 or 800 mg/kg bw Bendectin^® by gavage daily on days 6–15 of gestation (day 0 being considered that on which a positive vaginal smear was found) and were killed on day 20 for examination of their fetuses by standard teratological techniques. The highest dose was very toxic to the dams, killing 17%, and a marked reduction in body-weight gain was observed at the two higher doses. The lowest dose was minimally toxic. A small increase in the frequency of resorptions was seen at the highest dose and reduced fetal weight and reduced skeletal ossification at 500 and 800 mg/kg bw per day. There was no increase in the number of fetuses with malformations, but the number of litters with malformed fetuses was increased at the highest dose, due mainly to an increase in the number of fetuses with short 13th ribs. No increase in the incidence of external or visceral malformations was observed in any group (Tyl et al., 1988).

Groups of 21–24 pregnant cynomolgus monkeys (Macaca fascicularis) were given Bendectin^® at an approximate dose of 0, 1.3, 3.3 or 13.3 mg/kg bw (as 2/5, 1 and 4 tablets/day per animal, corresponding to two, five and 20 times the human dose) by nasogastric intubation daily on days 22–50 of gestation. The fetuses were removed surgically just prior to term for examination. No maternal toxicity and no embryofetal toxicity or teratogenicity was observed (Hendrickx et al., 1985).

4.4.1. Humans

A randomized, placebo-controlled study was conducted in 48 healthy male volunteers to compare the effects of doxylamine succinate on the pharmacokinetics of antipyrine metabolites and the urinary excretion of 6β-hydroxycortisol. Normal renal function, an age between 18 and 40 years and a normal diet were the criteria for inclusion. Groups of 16 men received 12.5 mg of doxylamine succinate, placebo or 30 mg of phenobarbital orally every 6 h for 17 days. No statistically significant differences indicative of enzyme induction were observed (Thompson et al., 1996).

4.4.2. Experimental systems

Doxylamine succinate caused a dose-dependent increase (up to 2.6-fold) in liver microsomal cytochrome P450 (CYP) enzyme activity in male and female B6C3F₁ mice (aged 45–52 days) after 7 or 15 days on a diet containing the compound at a concentration of 0, 40, 375, 750 or 1500 mg/kg (expressed as free amine). Doxylamine succinate caused marked induction of CYP2B enzymes, as demonstrated by a large increase in the O-dealkylation of 7-pentoxyresorufin (up to 38-fold) and in the 16β-hydroxylation of testosterone (up to 6.9-fold). In addition, treatment resulted in a modest induction of CYP3A and CYP2A and a 50% increase in thyroxine glucuronosyltransferase activity. No induction of CYP1A, CYP2E or CYP4A was found. These results suggest that doxylamine is a phenobarbital-type inducer of liver microsomal CYP enzymes in B6C3F₁ mice (Bookstaff et al., 1996).

Doxylamine inhibited aminopyrine N-demethylation by rat liver microsomes, with a median inhibitory concentration of 73 µmol/L (Brandes et al., 1994).

4.5.1. Humans

No data were available to the Working Group.

4.5.2. Experimental systems (see Table 1 for references)

Table 1

Genetic and related effects of doxylamine succinate.

Doxylamine succinate did not induce mutations in Salmonella typhimurium. It weakly induced unscheduled DNA synthesis in primary rat hepatocytes in vitro in the absence of an exogenous metabolic system. It inhibited gap-junctional intercellular communication, measured as metabolic cooperation, in Chinese hamster V79 cells. It did not induce sister chromatid exchange in human lymphocytes in vitro or micro-nucleus formation in hamster bone-marrow cells. In a transplacental system, a small dose-dependent induction of chromosomal aberrations, but no sister chromatid exchange or micronucleus formation, was found in the embryos of mice treated on day 11 of gestation[The Working Group was aware of unpublished studies of sister chromatid exchange and chromosomal aberrations in Chinese hamster ovary cells in vitro, which showed no effect of doxylamine succinate.].

Overall evaluation

Doxylamine succinate is not classifiable as to its carcinogenicity to humans (Group 3).

1.1.1. Nomenclature

1.1.2. Structural and molecular formulae and relative molecular masses

Phenobarbital

Sodium phenobarbital

1.1.3. Chemical and physical properties of the pure substances

1.1.4. Technical products and impurities

Phenobarbital is available as 8-, 16-, 32-, 65- and 100-mg tablets, as a 16-mg capsule and as a 15- or 20-mg/5 mL elixir. Sodium phenobarbital is available as 30-, 60-, 65- and 130-mg/mL injections and as a sterile powder in 120-mg ampules (Gennaro, 1995).

Trade names for phenobarbital include Adonal, Agrypnal, Amylofene, Barbenyl, Barbiphenyl, Barbipil, Barbita, Barbivis, Blu-phen, Cratecil, Dormiral, Doscalun, Duneryl, Eskabarb, Etilfen, Euneryl, Fenemal, Gardenal, Gardepanyl, Hysteps, Lepinal, Lepinaletten, Liquital, Lixophen, Lubergal, Lubrokal, Luminal, Neurobarb, Noptil, Nunol, Phenaemal, Phenemal, Phenobal, Phenoluric, Phenonyl, Phenyral, Phob, Sedonal, Sedophen, Sevenal, Somonal, Stental Extentabs, Teolaxin, Triphenatol and Versomnal. Trade names for sodium phenobarbital include Gardenal sodium, Linasen, Luminal sodium, PBS, Phenemalum, Phenobal sodium and Sodium luminal.

1.1.5. Analysis

Several international pharmacopoeias specify infrared absorption spectrophotometry with comparison to standards, thin-layer chromatography, high-performance liquid chromatography (HPLC) with ultraviolet detection and colorimetry as the methods for identifying phenobarbital; HPLC and titration with ethanolic potassium hydroxide are used to assay its purity. In pharmaceutical preparations, phenobarbital is identified by infrared absorption spectrophotometry, HPLC and colorimetry; HPLC and titration with ethanolic potassium hydroxide or silver nitrate are used to assay for phenobarbital content (British Pharmacopoeia Commission, 1993; Society of Japanese Pharmacopoeia, 1996; Council of Europe, 1997; US Pharmacopeial Convention, 1999).

Several international pharmacopoeias specify infrared absorption spectrophotometry with comparison to standards, thin-layer chromatography and HPLC as the methods for identifying sodium phenobarbital; HPLC with ultraviolet detection and potentiometric titration are used to assay its purity. In pharmaceutical preparations, sodium phenobarbital is identified by infrared absorption spectrophotometry and HPLC; HPLC with ultraviolet detection is used to assay for sodium phenobarbital content (British Pharmacopoeia Commission, 1993; Council of Europe, 1997; US Pharmacopeial Convention, 1999).

1.4.1. Occupational exposure

According to the 1981–83 National Occupational Exposure Survey (National Institute for Occupational Safety and Health, 2000), about 23 000 workers, including 20 000 nurses, nursing aides, health aides, pharmacists and laboratory workers in health services and 2900 chemical industry workers, were potentially exposed occupationally to phenobarbital in the USA.

1.4.2. Environmental occurrence

No data were available to the Working Group.

2.1.1. Studies of patients with seizures (see Table 1)

Table 1

Cohort studies of cancer incidence or mortality in epilepsy patients, generally treated with phenobarbital.

A cohort of (initially) 9136 patients hospitalized for epilepsy in the Filadelfia treatment community in Denmark between 1933 and 1962 was followed for cancer incidence from 1943 to 1967 (Clemmesen et al., 1974). Throughout the study period, phenobarbital was the basic therapeutic drug, given in daily doses of 100–300 mg. In the 1940s, phenytoin was introduced at the treatment centre, either alone or in combination with phenobarbital, and in the mid-1950s primidone was also used. Primidone is partly metabolized to phenobarbital. The files of the Danish Cancer Registry, established in 1942, were used to identify incident cases of cancer, including benign brain tumours and bladder papillomas (Olsen et al., 1989), in the patients. The expected numbers of cases were calculated from national cancer incidence rates, similarly based on the files of the Cancer Registry. In two updated reports with extended follow-up (Clemmesen & Hjalgrim-Jensen, 1977, 1981), the original roster was revised to include 8078 and 8077 subjects, respectively. The reduction of the cohort by 12% was due to exclusion of non-Danish citizens, patients who died before 1943 when the follow-up began, duplicate registrations and persons whose diagnosis of epilepsy was not sustained. In the most recent update (Clemmesen & Hjalgrim-Jensen, 1981), the period of follow-up was from 4 weeks after first admission to the treatment community or 1 January 1943, whichever came last, to the day of death or the end of 1976, whichever came first. A total of 467 cases of cancer were observed when 419.5 were expected [yielding an overall standardized incidence ratio (SIR) of 1.1; 95% confidence interval (CI), 1.0–1.2; or 1.25 (95% CI, 1.1–1.4) among men and 0.99 (95% CI, 0.87–1.13) among women]. The overall increase was due mainly to the observation of 71 tumours of the brain and nervous system when 13.4 were expected [yielding a significantly increased SIR of 5.3; 5.8 among men and 4.8 among women]. Excess rates of brain tumour were seen in particular during the first 15 years of follow-up [with SIRs of 12, 5.4, 1.8, 2.7, 1.4, 2.1 and 0.0 for latency periods of 0–9, 10–14, 15–19, 20–24, 25–29, 30–34 and ≥ 35 years, respectively]. The authors suggested that the seizures of some patients were early symptoms of their brain tumours. Thirteen cases of primary liver cancer were observed, when 3.44 cases were expected [SIR, 3.8; 95% CI, 2.0–6.5]. This excess of liver cancer, which was particularly evident in patients followed for 15 years or more, was ascribed to use of radioactive thorium dioxide (Thorotrast) for cerebral angiography in a subgroup of patients during diagnostic work-up. Only three of the 13 patients with primary liver cancer had had no documented exposure to Thorotrast, which was in agreement with the 3.4 cases expected among all cohort members. Finally, the increased SIR for lung cancer among male patients [SIR, 1.3; 95% CI, 0.94–1.7] was considered most likely to be due to a higher prevalence of smokers.

The Danish cohort was further evaluated by Olsen et al. (1989), who extended follow-up for cancer incidence from 1976 through 1984. Exclusion of additional duplicate admissions, non-Danish citizens and patients for whom the data were incomplete and the inclusion of patients hospitalized for fewer than 4 weeks reduced the cohort to 8004 patients. Of these, 140 had had documented exposure to Thorotrast. Linkage of the records of the remaining 7864 patients not known to have received this contrast medium with the files of the national Cancer Registry resulted in identification of 789 cancers, with 663.7 expected (SIR, 1.2; 95% CI, 1.1–1.3). A sixfold increased risk was seen for brain cancer on the basis of 118 observed cases; 43 were seen within 1 year of admission (SIR, 88). The risk for brain tumours in childhood was especially high. A significant excess also occurred for cancer of the lung, but with no clear trend over time. Non-significantly increased risks were seen for cancers of the liver (SIR, 1.9) and biliary tract (SIR, 1.7) on the basis of 9 and 11 observed cases, respectively. It was suggested by the authors that these non-significant increases in risk among patients not known to have received Thorotrast might be related to undocumented exposure to this hepatocarcinogen. Four cases of thyroid cancer were observed, with 3.2 cases expected (SIR, 1.2; 95% CI, 0.3–3.2). Significant deficits of bladder cancer (SIR, 0.6; 18 cases) and melanoma (SIR, 0.5; 7 cases) were observed. In the subgroup of 140 patients with documented exposure to Thorotrast, 17 cases of liver cancer (SIR, 202) and three cases of biliary-tract cancer (SIR, 28) were reported.

White et al. (1979) evaluated mortality among patients admitted to one treatment centre for epilepsy in England between 1931 and 1971. Patients were included in the study if their notes stated a diagnosis of epilepsy not due to trauma or progressive disease and if they had been prescribed long-term anti-convulsant drug therapy. Most of the patients severely affected by epilepsy had been treated with phenobarbital and phenytoin. After exclusion of patients who were known to have discontinued anticonvulsant drug therapy within 6 months of starting, 2099 subjects were linked to the files of the National Health Service Central Register of the United Kingdom for information on vital status and migration; for those who had died, a copy of the death certificate was provided by the Office of Population Censuses and Surveys. Personal data on 38 subjects (2%) could not be verified, and 81 were known to have died before the period of follow-up (1951–77), leaving 1980 patients for the analysis of mortality. A total of 78 deaths from cancer was recorded when 51.5 were expected on the basis of the age-, sex- and calendar year-specific rates for England and Wales (standardized mortality ratio [SMR],1.5; 95% CI, 1.2–1.9). Neoplasms of the brain and central nervous system accounted for six of the deaths, with 1.5 expected (SMR, 4.1; 95% CI, 1.5–8.9). When these neoplasms were excluded as possibly being associated with epilepsy, the SMR was 1.4, which was still significantly high (95% CI, 1.1–1.8), but there was no trend over time and no significant excess of deaths was associated with cancer at a particular site. A non-significantly increased risk of 1.4 was seen for lung cancer.

Through a medical records system in Rochester, Minnesota (USA), providing access to all medical contacts of citizens of the area covered by the system, Shirts et al. (1986) identified 959 patients in whom an unprovoked seizure disorder had been diagnosed between 1935 and 1979. On the basis of the same records system, patients were followed for new diagnoses of cancer from the date of the initial diagnosis of seizure to death, last contact or the end of 1982: on average, 13 years. A total of 65 primary cancers were diagnosed when 45.9 were expected from the age- and sex-specific incidence rates of the background population, yielding an overall SIR of 1.4 (95% CI, 1.1–1.8). The increased risk was largely attributable to 17 primary brain cancers, for which the expected number was < 1; an SIR of 47 (13 cases) during the first 5 years of follow-up decreased gradually to 12 (two cases) and 5.9 (two cases) in the subsequent 5–9 years and 10 or more years of follow-up, respectively. The risk for lung cancer was also significantly elevated, with an SIR of 2.7 (95% CI, 1.2–5.2); however, seven of the nine cases were observed during the first 5 years of follow-up, limiting the excess to this early period. Marginally significant increased risks were seen for breast cancer (SIR, 2.0; 95% CI, 0.98–3.8; 10 cases) and cancers of the lymphatic and haematopoietic system (SIR, 2.9; 1.0–5.0; seven cases), with no clear trends in risk over time. [The Working Group noted that no information was given on the completeness of follow-up.] The SIR for all cancers except primary brain cancers was 0.9 (95% CI, 0.6–1.4) for users of anticonvulsants and 1.3 (0.8–2.0) for non-users.

2.1.2. Studies of transplacental exposure

Annegers et al. (1979) reported on the occurrence of brain tumours in 177 individuals born in Rochester, Minnesota (USA), between 1939 and 1976 who had been exposed in utero to anti-convulsants during the first trimester of gestation. Barbiturates had been prescribed for the mothers of 135 of these patients. No case of brain tumour was observed during follow-up. [The Working Group noted that the expected number of brain tumours was not estimated, and that the number of person–years of follow-up was not provided]

Olsen et al. (1990) conducted a separate study of the incidence of cancer among 3727 offspring of 3758 women admitted for epilepsy to the Filadelfia treatment community in Denmark between 1933 and 1962. A survey of drug use by 130 of the patients indicated that 76% had been treated with phenobarbital and 30% with primidone. The records of the offspring, who were identified from hospital charts, population listings and parish registers, were linked with the files of the Danish Cancer Registry in order to follow-up for cancer through 1986. The expected age-, sex- and calendar time-specific cancer incidence rates for the general population were also derived from the Registry. Overall, 49 cancers were identified, with 53.8 expected, yielding an SIR of 0.9 (95% CI, 0.7–1.2). Among the 2579 children born after their mother’s first admission for epilepsy, and thus presumably exposed in utero to anticonvulsant drugs, 14 cases of cancer were identified (average follow-up period, 22.4 years; maximum, 50 years), with 13.8 expected (SIR, 1.0; 95% CI, 0.6–1.7). This sub-cohort of offspring showed no excess risk for any specific tumour type, including brain tumours (3 observed cases, 2.2 expected). Among the group of 1148 offspring born before the first admission of the mother to the treatment community, 35 developed cancer (average follow-up period, 37.5 years; maximum, 65 years), with 40.0 expected (SIR, 0.9; 95% CI, 0.6–1.2). No significant increase in the risk for any cancer was found.

2.1.3. Studies in the general population

Phenobarbital was included in a hypothesis-generating cohort study designed to screen a large number (215) of drugs for possible carcinogenicity, which covered more than 140 000 subscribers enrolled between July 1969 and August 1973 in a prepaid medical care programme in northern California (USA). Computer records of persons to whom at least one drug prescription has been dispensed were linked to the cancer records of hospitals covered by the medical care programme and the regional cancer registry. The observed numbers of cancers were compared with those expected, standardized for age and sex, for the entire cohort. Three publications summarized the findings for follow-up periods of up to 7 years (Friedman & Ury, 1980), up to 9 years (Friedman & Ury, 1983) and up to 15 years (Selby et al., 1989). Among 5834 persons who received phenobarbital, mostly as a sedative, associations were noted in the 7-year report for cancers of the lung (44 cases observed, 28.9 expected; p < 0.05), ovary (seven cases observed, 2.7 expected; p < 0.05) and gall-bladder (four cases observed, 1.0 expected; p < 0.05) and in the 9-year report for cancer of the lung [figures not given] and gallbladder and biliary tract (six cases observed, 1.8 expected; p < 0.05). In the 15-year report, associations were noted with cancers of the gall-bladder (eight cases observed, 3.2 expected; p < 0.05) and bone (three cases observed, 0.6 observed; p < 0.05), but not for cancer of the lung. [The Working Group noted, as did the authors, that, since some 12 000 comparisons were made in this hypothesis-generating study, the associations should be verified independently. Data on duration of use were generally not provided.]

In a post-hoc evaluation of the finding of lung cancer in the 7-year follow-up, Friedman (1981) merged this group with 2156 users of pentobarbital and 2884 users of secobarbital, two other commonly used barbiturates. In the combined group of 9816 users of one of these three barbiturates, the author observed 87 cases of lung cancer when 50.2 were expected (SIR, 1.7; p < 0.002). Data on smoking habits, collected at regular health check-ups, were available for 49% of the members of the combined cohort, and information on histological subtype of lung cancer was obtained from the medical charts of cancer patients. The resulting SIRs for lung cancer among barbiturate users were 1.5 [95% confidence interval (CI), 0.4–3.8] for non-smokers, 1.4 [0.5–3.0] for ex-smokers and 1.6 [1.1–2.3] for smokers. There was no change in the risk for lung cancer after the incorporation of a lag time of 1 or 2 years in the analysis. There was no particular association with any of the major sub-types of lung cancer.

The risk pattern for lung cancer was evaluated by Friedman and Habel (1999), who extended follow-up through to 1992 and further added users of mixtures of barbiturates, resulting in a group of 10 213 exposed individuals. An initially elevated SIR of 1.6 (95% CI, 1.3–1.9) for lung cancer for the combined group of barbiturate users with 3–7 years of follow-up, unadjusted for smoking habits, gradually decreased and stabilized at about 1.3 [1.2–1.4] after 11–15 years of follow-up. An initial, nonsignificant increase in the risk for lung cancer of 80% among people who had never smoked decreased to near unity in later periods of follow-up. A dose–response trend was observed, on the basis of the number of prescriptions dispensed, with an SIR of 3.4 (95% CI, 2.0–5.4) for individuals receiving the highest dose (20 or more prescriptions of barbiturates). Adjustment for smoking habits in a Cox model in the subgroup for which this information was available reduced, but did not eliminate, the dose–response trend. [The Working Group noted that the last two analytical studies were post-hoc evaluations of a finding of lung cancer in a large surveillance study with multiple testing. These studies did not include the results for phenobarbital users specifically but rather for the combined group of barbiturate users.]

In a companion study, based on the 10 368 users of barbiturates, Habel et al. (1998) observed 34 cases of bladder cancer in the 1992 follow-up, yielding an overall SIR of 0.71 (95% CI, 0.51–1.0). The SIRs for current smokers and former smokers were 0.56 (0.23–1.2) and 0.68 (0.27–1.4), respectively, whereas the SIR for people who had never smoked was 1.04 (0.48–2.0), indicating an inverse association between barbiturate treatment and bladder cancer risk only among current and former smokers.

3.3.1. Promotion in mouse liver

3.3.2. Promotion in rat liver

3.3.3. Promotion in hamster liver

Groups of male Syrian golden hamsters [initial numbers not specified], 5 weeks of age, received a single intraperitoneal injection of either 100 mg/kg bw NDEA or 20 mg/kg bw methylazoxymethanol acetate (MAMA) and 2 weeks later were given either tap-water or drinking-water containing 500 mg/L phenobarbital for 18, 45 or 62 weeks, and killed. Between 18 and 62 weeks, hepatocellular adenomas were induced by both NDEA (11/30, 37%) and MAMA (19/30, 63%), but carcinomas were seen only in MAMA-treated hamsters (2/18, 11%). Subsequent administration of phenobarbital for up to 45–62 weeks had no significant effect either on the incidence (adenoma: NDEA plus phenobarbital, 27%; MAMA plus phenobarbital, 67%; carcinoma: MAMA plus phenobarbital, 11%) or on the histological appearance of pre-neoplastic or neoplastic hepatocellular lesions in either NDEA- or MAMA-initiated hamsters. Phenobarbital treatment alone did not induce hepatocellular lesions (Diwan et al., 1986b).

Groups of male Syrian golden hamsters, 5–6 weeks of age (initial number, 163), were given subcutaneous injections of 500 mg/kg bw N-nitrosobis(2-hydroxypropyl)amine (NBHPA) once a week for 5 weeks, after which they were fed a basal diet alone or a diet containing 0.05% phenobarbital for 30 weeks. Control groups were given subcutaneous injections of saline once a week for 5 weeks followed by either a basal diet or diet containing 0.05% phenobarbital for 30 weeks. Phenobarbital treatment significantly decreased the final body weight of hamsters initiated with NBHPA (average, 135 g versus 145 g; p < 0.05). No significant effect was seen on the liver weights of animals exposed to phenobarbital or NBHPA plus phenobarbital. Hepatic hyperplastic nodules were found in 100% of animals exposed to the nitrosamine with or without phenobarbital. Hepatocellular carcinomas occurred in 13/24 (54%) hamsters given NBHPA plus phenobarbital but in only 2/15 (13%) given the nitrosamine alone (p < 0.02). Treatment with phenobarbital did not affect the development of bile-duct, gall-bladder or pancreatic lesions induced or initiated by NBHBA (Makino et al., 1986).

Groups of male Syrian golden hamsters, 6 weeks of age, received an intraperitoneal injection of 6 mg/kg bw NDMA and were then fed a basal diet (15 animals), the same dose of NDMA and 1 week later a diet containing 500 mg/kg diet phenobarbital for 30 weeks (11 animals), 1 week of basal diet and then the diet containing phenobarbital for 30 weeks (14 animals; controls), phenobarbital or basal diet continuously for 31 weeks (13 animals; untreated controls). All groups developed altered hepatic foci. The number (foci/cm²) and size (mm² × 10⁻²) of foci induced by NDMA did not significantly change after subsequent treatment with phenobarbital (NDMA: number, 6.3 ± 3.4; size, 10.8 ± 8.9; NDMA plus phenobarbital: number, 7.3 ± 3.4, size 11.5 ± 9.0). Animals treated with phenobarbital alone developed significantly more foci than the untreated group (1.8 ± 0.8 versus 0.2 ± 0.3 foci/cm², p < 0.01). One hamster exposed to NDMA alone developed three adenomas, while three given NDMA plus phenobarbital developed one adenoma each. Treatment with phenobarbital alone did not result in tumour formation, but enlarged hepatocytes with abundant cytoplasm were observed. Thus, phenobarbital failed to promote liver carcinogenesis initiated by NDMA in hamsters (Tanaka et al., 1987).

3.3.4. Promotion in monkey liver

In a first study, nine pregnant patas monkeys (Erythrocebus patas) were injected intravenously twice weekly with NDEA to provide a cumulative dose of 0.6–3.2 mmol/kg bw. None of the nine mothers or six offspring developed tumours during 4 years of subsequent observation. At that time, three mothers and three offspring were given drinking-water containing phenobarbital at a concentration providing a dose of 15 mg/kg bw per day for the remainder of their lives or up to 43 months. Within less than 2 years, multiple hepatocellular neoplasms had developed in both offspring (5.6 adenomas and 0.3 carcinomas per animal) and mothers (11.6 adenomas per animal) given NDEA followed by phenobarbital, but none were found in those given NDEA alone. A few hepatocellular foci (0.3 foci/cm²) were detected in monkeys treated with NDEA only. The authors concluded that, in patas monkeys, hepatocellular carcinogenesis could be initiated by prenatal exposure to NDEA and that the initiated cells could remain dormant for years. In a second study, young male and female patas monkeys [initial numbers not specified] received intraperitoneal injections of 0.1–0.4 mmol/kg bw NDEA in phosphate-buffered saline at 14-day intervals for a total of 20 injections. Fifteen days later, four of the seven survivors were given drinking-water containing phenobarbital at a concentration providing a dose of 15 mg/kg bw per day for 9 months. A significant increase in the multiplicity of hepatocellular neoplasms was found among animals given NDEA plus phenobarbital (adenomas, 17.25 per animal; carcinomas, 2.0 per animal) when compared with those given NDEA alone (adenomas, 1.6 per animal; carcinomas, 0.3 per animal), and the increase in the incidence of adenomas was statistically significantly greater than that with NDEA plus phenobarbital [p value not given]. Animals exposed to phenobarbital alone had no tumours or foci. Thus, phenobarbital is also an effective promoter of hepatocellular neoplasia in this non-rodent species (Rice et al., 1989).

3.3.5. Tumour promotion in the thyroid

Table 4 summarizes studies of the promotion of tumours in the thyroid with phenobarbital. Mouse: Pregnant B10.A mice were given a single intravenous injection of 0.5 mmol/kg bw ENU in citrate buffer on day 18 of gestation and were allowed to deliver normally. After sexing at 3 weeks of age, equal numbers of pups from each litter of dams treated with ENU were given ENU alone or ENU concomitantly with drinking-water containing 0.05% phenobarbital from 4 to 80 weeks of age. Postnatal administration of phenobarbital more than tripled the incidence of thyroid follicularcell adenomas in ENU-treated offspring (ENU: males, 1/14; females, 1/14; ENU plus phenobarbital: males, 5/14; females, 2/14; 7% versus 25% in males and females combined, p < 0.05). Postnatal administration of phenobarbital had no effect on the development of intestinal or renal tumours induced or initiated by ENU (Diwan et al., 1989d).

Rat: A total of 231 male Wistar rats, 6 weeks of age, were given a subcutaneous injection of 700 mg/kg bw NBHPA once a week for 4 or 6 weeks, and 1 week later were given either basal diet or a diet containing phenobarbital at 500 mg/kg during weeks 5–16 or weeks 7–18. Control groups received phenobarbital alone for similar durations. The experiment was terminated at 20 weeks. No significant effects were observed on the mean final body weights of rats given four injections of NBHPA, but those given six injections with or without phenobarbital weighed significantly less than the corresponding control groups (p < 0.05). The mean weights of the thyroid glands of rats exposed to phenobarbital alone or with NBHPA were higher than those of the corresponding control groups (p < 0.05). The incidence of thyroid tumours was 66% in the group exposed to phenobarbital after four injections of NBHPA, but no thyroid tumours were found in rats given four injections of NBHPA alone or phenobarbital alone. Thyroid tumours were found in 24% (5/21) of rats given six injections of NBHPA and in 100% (21/21) given six injections of the nitrosamine followed by phenobarbital. Two rats given four doses of NBHPA followed by phenobarbital and 5/21 (24%) rats given six doses of NBHPA followed by phenobarbital developed malignant thyroid tumours. All rats given four or six doses of NBHPA followed by phenobarbital had multiple follicular adenomas (106 and 239, respectively) (Hiasa et al., 1982).

Groups of male Wistar rats (total, 200), 6 weeks of age, received a single subcutaneous injection of 2800 mg/kg bw NBHPA, and 2 weeks later were given either a basal diet or a diet containing phenobarbital at 500 mg/kg for 6, 12 or 19 weeks. The study was terminated at 20 weeks. Of the rats given NBHPA, 21/24 (87%) developed thyroid follicular-cell adenomas when phenobarbital was given for 19 weeks, 19/24 (79%) when phenobarbital was given for 12 weeks and 10/24 (42%) when phenobarbital was given for 6 weeks. The incidences of thyroid adenomas in NBHPA-initiated rats exposed to phenobarbital for 19 and 12 weeks were significantly higher (p < 0.05) than that in rats given the nitrosamine alone (9/24, 37%). Ten, 5.2 and 2.6 thyroid tumours per rat were found after exposure to phenobarbital for 19, 12 and 6 weeks, respectively, and the total numbers of follicular-cell adenomas were 237, 124 and 62 in NBHPA-treated rats exposed to phenobarbital for 19, 12 and 6 weeks, respectively. Papillary adenomas occurred in three rats exposed to phenobarbital for 6 weeks and one each in those exposed for 19 and 12 weeks after initiation with NBHPA. Nine (37%) rats exposed to NBHPA only developed tumours, consisting of 23 follicular-cell adenomas. None of the rats exposed to phenobarbital only developed any thyroid tumours (Hiasa et al., 1983).

Groups of 10 male and 10 female Fischer 344 rats, 4 weeks of age, received either a single intravenous injection of 0.2 mmol/kg bw MNU or four weekly injections of 0.05 mmol/kg bw MNU followed 2 weeks later by or concurrently with drinking-water containing 0.05% phenobarbital, which was continued until 52 or 80 weeks of age. At 52 weeks, phenobarbital given subsequent to MNU or concurrently with divided doses of MNU significantly enhanced the incidence of thyroid follicular-cell tumours only in males (MNU, 10–20%; MNU plus phenobarbital, 50%; MNU followed by phenobarbital, 60–70%). Between 53 and 80 weeks of age, the thyroid tumour incidence was 70% in male rats exposed to MNU plus phenobarbital, 80–100% in male rats that received MNU followed by phenobarbital and about 30–40% in groups that received MNU only. The incidence of thyroid tumours in females exposed to MNU with or before phenobarbital (60–70%) was not significantly different from that in males, but a higher multiplicity of these tumours in males (average, 3.5 versus 2.1 per tumour-bearing rat) persisted in the groups given phenobarbital. Several animals developed both follicular-cell adenomas and carcinomas after treatment with MNU and phenobarbital (Diwan et al., 1985). [The Working Group noted the small number of animals per group.]

To investigate the effects of dose and sex on the development of thyroid tumours, groups of 24 male and 24 female Wistar rats, 6 weeks of age, received a single intraperitoneal injection of 2100 mg/kg bw (males) or 4200 mg/kg bw (females) NBHPA, and 1 week later were fed diets containing phenobarbital at a concentration of 20, 100, 500 or 2500 mg/kg for 19 weeks. Control groups received either NBHPA alone or various concentrations of phenobarbital in their diet. No significant differences were found in the body-weight gain of either male or female rats treated with NBHPA plus phenobarbital or phenobarbital alone, whereas rats exposed to NBHPA alone had lower body weights than controls. Thyroid follicular-cell tumours occurred only in groups that received NBHPA with or without phenobarbital. NBHPA treatment alone resulted in an 8% incidence of thyroid tumours in both male and female rats; with the addition of 500 mg/kg of diet phenobarbital, a threefold increase in tumour yield was found in male rats but no increase in females. At 2500 mg/kg of diet, a marked increase (about eightfold) in tumour yield was seen in male rats but a less than threefold increase in females. The incidences of follicular adenomas in rats given NBHPA followed by a concentration of 20, 100, 500 or 2500 mg/kg of diet phenobarbital were 8, 45, 70 and 66% for males and 12, 17, 50 and 58% for females, respectively. Papillary adenomas were seen in male rats only at 500 and 2500 mg/kg of diet phenobarbital (12% and 20%, respectively); only one female rat (4%) developed such a tumour when given 2500 mg/kg of diet. No thyroid tumours were found in control groups with or without phenobarbital treatment (Hiasa et al., 1985).

A total of 120 male Wistar rats, 6 weeks of age, were given a single intraperitoneal injection of 2100 mg/kg bw NBHPA at the end of the first experimental week. Groups of 20 of these rats were castrated either at the beginning of the experiment or at the end of the second week and received the basal diet containing phenobarbital at 500 mg/kg from week 3 to week 40 of the experiment. The other groups were given the basal diet without phenobarbital. Significant differences in mean body weight were found between groups that received NBHPA only or NBHPA plus phenobarbital and those of castrated rats. The mean weights of the thyroid of rats that received phenobarbital after NBHPA initiation with or without castration at the end of the second week were (not significantly) higher than that in the group exposed to NBHPA alone, while the thyroids of rats that were castrated at the beginning of the experiment weighed significantly less than those of rats that were similarly castrated and given phenobarbital (p < 0.05). The incidence of thyroid follicular-cell adenomas was 20% in animals that received NBHPA only, 75% in those that received the nitrosamine plus phenobarbital, 30% in those that were castrated at the end of second week and received NBHPA plus phenobarbital and 0% in those that were castrated at the end of second week and received NBHPA only. The incidences of these tumours were 20% and 0% in rats castrated at the beginning of the first week, treated with NBHPA and given phenobarbital or basal diet, respectively. The incidence of thyroid carcinomas was 10% with NBHPA only, 40% with NBHPA plus phenobarbital and 15% in animals that were castrated at the end of the second week and given phenobarbital. No thyroid carcinomas occurred in any other groups. The serum concentration of thyroid-stimulating hormone (TSH) in the group given NBHPA plus phenobarbital was significantly higher than in groups that received NBHPA only or NBHPA plus phenobarbital with castration. Thus, castration inhibited the development of thyroid tumours in rats treated with NBHPA, and the inhibition was greater when the rats were castrated before receiving the nitrosamine (Hiasa et al., 1987).

Groups of 20 male and 20 female CD(SD)BR rats (weighing 204 g and 143 g, respectively) were injected subcutaneously with 700 mg/kg bw NBHPA or saline once a week for 5 weeks and were then fed a control diet or a diet containing 500 mg/kg phenobarbital and/or l-thyroxine (T4) at a dose of 50 µg/kg bw per day from week 6 through week 20, when all animals were killed. The dose of T4 was required to maintain the normal thyroid gland weight in phenobarbital-treated rats and induced a slight increase in the serum concentration of T4 and a slight decrease in that of triiodothyronine (T3). The food consumption in all groups was comparable, but slightly decreased body weight was observed in all groups that received NBHPA. All rats treated with phenobarbital had increased liver weights, and NBHPA and T4 did not affect this increase. In male rats, phenobarbital increased the weight of the thyroid gland over that of controls (p ≤ 0.05). NBHPA alone did not alter thyroid weights (∼ 30 mg), but subsequent administration of phenobarbital resulted in a marked increase in thyroid weight (∼ 55 mg). T4 reduced this response in groups given phenobarbital with or without NBHPA (∼ 30 mg). No significant differences were seen in the weights of the thyroid gland in female rats. Six of 16 male rats exposed to NBHPA alone developed thyroid follicular-cell adenomas (20 tumour foci), and this incidence was significantly increased by treatment with phenobarbital (15/18, p ≤ 0.05) (107 tumour foci). T4 reduced the incidences of both adenomas (5/20) and tumour foci (11). Only one female rat given NBHPA plus phenobarbital developed an adenoma. None of the rats exposed to phenobarbital alone developed thyroid tumours (McClain et al., 1988). [The Working Group noted the short duration of exposure.]

Groups of 20 male Fischer 344 rats, 6 weeks of age, were given drinking-water containing 0.2% NBHPA for 1 week, and 1 week later were given either basal diet or a diet containing 0.05% phenobarbital, for 50 weeks. Controls received the diet containing 0.05% phenobarbital alone. The incidence of thyroid follicular-cell carcinomas in rats given NBHPA plus phenobarbital was significantly higher than that in rats given the nitrosamine alone (19/19 versus 14/20, p < 0.05). The mean number of tumours (adenomas and carcinomas) per rat given NBHPA plus phenobarbital was also significantly higher than in those given NBHPA only (2.63 ± 1.12 versus 1.56 ± 1.34, p < 0.02). One carcinoma in a rat given the combination metastasized to the lung. The difference in the incidence of thyroid adenomas in groups given NBHPA with and without phenobarbital (10/19 versus 7/20) was not statistically significant (Shirai et al., 1988).

Pregnant Fischer 344/NCr rats were given a single intravenous injection of 0.2 mmol/kg bw MNU in citrate buffer on day 18 of gestation. After sexing, equal numbers of pups from each litter of dams treated with MNU were given MNU alone with a normal diet and tap-water or with drinking-water containing 0.05% phenobarbital from 4 to 80 weeks of age. Postnatal administration of phenobarbital had no effect on the incidence or latency of either neurogenic or renal tumours induced or initiated by MNU. However, phenobarbital given postnatally promoted the development of thyroid tumours of follicular cell origin (MNU: males, 1/19; females, 0/17; MNU plus phenobarbital: males, 13/17; females, 3/14), especially in male offspring exposed prenatally to MNU (p < 0.01). The multiplicity of thyroid tumours in the MNU plus phenobarbital group was 2.6 per rat in males and 1.3 per rat in females, while that with MNU only was 1 and 0 per rat, respectively. Thyroid follicular-cell carcinomas were found only in male rats exposed transplacentally to MNU and postnatally to phenobarbital (6/17) (Diwan et al., 1989d).

4.1.1. Humans

Phenobarbital is readily absorbed from the gastrointestinal tract in humans (Lous, 1954) and is eliminated by hepatic extraction and renal excretion. In the liver, phenobarbital is para-hydroxylated and subsequently conjugated (Butler, 1956). After oral administration of [¹⁴C]phenobarbital (120 mg phenobarbital containing 31 µCi ¹⁴Clabel) to two healthy men, 78–87% of the dose was recovered in the urine over 16 days. Phenobarbital N-glucoside, para-hydroxyphenobarbital and unchanged phenobarbital accounted for approximately 27, 19 and 29% of the dose, respectively (average for the two subjects) (Tang et al., 1979). Small amounts of the O-methylcatechol metabolite of phenobarbital were identified in the urine of a single individual given 300 mg orally (Treston et al., 1987). In patients with epilepsy receiving long-term treatment, 57% of the daily dose was recovered in urine, 14% as the N-glucoside, 16% as para-hydroxyphenobarbital and 27% as unchanged phenobarbital (Bernus et al., 1994).

4.1.2. Experimental systems

After an intravenous injection of 50 mg/kg bw [¹⁴C]phenobarbital to rats, the total amount of radiolabel in the liver was higher than that in other organs. Urinary elimination of radiolabel associated with phenobarbital reached a peak after 6–8 h (Glasson et al., 1959). The major pathways for biotransformation of phenobarbital in rodents appear to involve para-hydroxylation, with excretion in either the free form or as a glucuronide conjugate. The N-glucoside of phenobarbital was formed at < 1% and excreted in urine only in mice; it was not detected in the urine of rats, guinea-pigs, rabbits, cats, dogs, pigs or monkeys (Soine et al., 1991). Less than 50% of a dose of phenobarbital administered to rodents was excreted in urine (Maynert, 1965). In addition to urinary excretion, 18% of a dose of 75 mg/kg bw phenobarbital was excreted in the bile of Sprague-Dawley rats within 6 h, the majority as conjugated metabolites (Klaassen, 1971). Very small amounts of meta-hydroxyphenobarbital, a 3,4-dihydrodiol and a 3,4-catechol derivative were also detected in the urine of rats and guinea-pigs given 110 and 56 mg/kg bw, respectively, by intraperitoneal injection (Harvey et al., 1972). When phenobarbital was administered at a dose of 15 mg/kg bw to beagle pups at 4, 10, 20, 40 and 60 days of age, no difference in the elimination half-time was observed with age, and there was no apparent formation of the para-hydroxylated metabolite (Ecobichon et al., 1988).

In male CFE rats and CF1 mice pretreated with phenobarbital, a total of 63% of phenobarbital and metabolites was excreted in 3 days in the urine and 30% with faeces in rats, and 72% and 11% in mice, respectively. After 2 weeks on a diet containing 0.1% (w/v) phenobarbital for rats and 0.05% (w/v) phenobarbital for mice, more phenobarbital and metabolites were excreted in the faeces of rats and less in those of mice than in animals that had not been pretreated. Most of the material excreted in urine was para-hydroxyphenobarbital and its conjugates, with lesser amounts of phenobarbital. The main effects of phenobarbital pretreatment on phenobarbital excretion in rats were increased urinary excretion of conjugated para-hydroxyphenobarbital, a similar decrease in the excretion of the free form of this metabolite and no change in total phenobarbital excretion. In mice, phenobarbital pretreatment caused a threefold increase in the urinary excretion of the free form of para-hydroxyphenobarbital, a small decrease in that of the conjugated form and about a twofold decrease in that of unchanged phenobarbital. The overall effect of phenobarbital pretreatment in mice appeared to be about a 50% increase in the amount of all forms of para-hydroxyphenobarbital, taking into account changes in urinary and faecal excretion. A minor urinary metabolite found in both species was not identified but was believed to be either the 3,4-dihydrodiol or the 3,4-catechol (Crayford & Hutson, 1980).

The half-time for clearance of phenobarbital was decreased to one-third by 70 weeks in C3H/He mice given phenobarbital at a dose of 85 mg/kg of diet for up to 90 weeks (Collins et al., 1984).

4.1.3. Comparison of animals and humans

The para-hydroxylated metabolite of phenobarbital and its conjugates as well as unchanged phenobarbital have been identified in the urine of humans and rodents. The formation and urinary excretion of the N-glucoside of phenobarbital appears to occur selectively in humans, with minor amounts excreted in mice but not in other species examined.

4.2.1. Humans

The most frequent side-effect of phenobarbital is sedation; however, tolerance develops after long-term treatment. Nystagmus and ataxia occur at excessive doses. Phenobarbital has produced irritability and hyperactivity in children and confusion in the elderly. Scarlatiniform or morbilliform rash, probably with other manifestations of drug allergy, occurred in 1–2% of patients (Hardman et al., 1996).

Oppenheimer et al. (1968) showed increased T4 turnover in humans treated with phenobarbital. Ohnhaus and Studer (1983) determined that an induction sufficient to increase antipyrine clearance by at least 60% was required to change steady-state thyroid hormone levels in phenobarbital-treated patients. Induced patients had decreased concentrations of T4, but the concentrations of T3, TSH and the TSH response to thyrotropin-releasing hormone usually remained within normal limits. The concentrations of circulating free T3 and free T4 were reported to be reduced in children with epilepsy maintained on phenobarbital, although clinical hypothyroidism was not seen (Yüksel et al., 1993).

4.2.2. Experimental systems

The intraperitoneal LD₅₀ of phenobarbital was reported to be 340 mg/kg bw in mice (Collins & Horlington, 1969), and the oral LD₅₀ was 162 mg/kg bw in rats (Goldenthal, 1971). The oral LD₅₀ in rats for sodium phenobarbital was 660 mg/kg bw (Stecher, 1968).

4.3.1. Humans

As reported in an abstract, 57 women receiving phenobarbital monotherapy during pregnancy were identified in a population of 61 090. When compared with a matched control group, the exposed infants showed significant increases in the frequency of major malformations or growth retardation (15.8% versus 6.7%), ‘anticonvulsant face’ (12.9% versus 2.5%) and fingernail hypoplasia (25.8% versus 4.5%) (Holmes et al., 1990).

As reported in an abstract, pregnancy outcome was evaluated prospectively in 84 women given phenobarbital monotherapy who were identified through the California Teratogen Information Service (USA). Forty-six of 63 ascertained liveborn infants (seven cases were lost to follow-up and there were 12 spontaneous and one therapeutic abortions) were evaluated by a dysmorphologist. Of these, seven (15%) had facial features characteristic of anti-epileptic therapy and 11 (24%) had hypoplastic fingernails (Jones et al., 1992a).

In a prospective cohort study of the pregnancy outcomes of women being treated for epilepsy with anti-convulsant therapy, 72 infants were born to mothers who had received phenobarbital monotherapy during the first trimester (Dravet et al., 1992). This group comprised 12 infants with microcephaly, 44 who were not microcephalic and 16 unrecorded outcomes [odds ratio apparently not significant].

In a study of the risk of intrauterine growth delay in the offspring of mothers with epilepsy, prospective data on 870 newborn infants in Canada, Italy and Japan were pooled and analysed. A total of 88 infants were born to mothers who had received phenobarbital monotherapy. By logistic regression, the risk for small head circumference was shown to be higher (relative risk, 3.6; 95% CI, 1.4–9.4) among the infants of phenobarbital-exposed mothers (Battino et al., 1999). Subsequent analysis showed statistically significant dose- and concentration-dependent effects of phenobarbital on small head circumference.

A study of the development of sexual identity was carried out among the offspring of mothers with epilepsy who had taken phenobarbital during the index pregnancy in the Amsterdam Academic Medical Centre between 1957 and 1972. Of 243 exposed subjects, 147 participated in the follow-up [age not indicated]. The controls were an equal number of persons from the original pool of 222, matched for birth date, sex and maternal age. Three tests of psychosexual development were used: the Gender Role Assessment Schedule, the Klein Sexual Orientation Grid and the Psychosexual Milestones in Puberty questionnaire. Exposed and control subjects did not differ with respect to ‘gender’ role behaviour, although greater numbers of persons exposed prenatally to anticonvulsants reported past or present cross-sexual behaviour and/or sexual dysphoria (Dessens et al., 1999).

The intelligence scores of adult men whose mothers had received phenobarbital during pregnancy and who had no history of a central nervous system disorder were measured. The population was drawn from the Danish Perinatal Cohort that was assembled in 1959–61 (Reinisch et al., 1995). A total of 114 exposed offspring and 153 controls were matched for a number of variables. Exposure to phenobarbital, especially during the last trimester, was associated with significantly lower verbal intelligence scores.

In a preliminary note on a prospective randomized double-blind trial on the efficacy of prenatal phenobarbital therapy in the prevention of neonatal hyperbilirubinaemia, 1522 exposed and 1553 control infants were studied (Yaffe & Dorn, 1991). The phenobarbital treatment consisted of administration of a 100-mg tablet daily during weeks 34–36 of gestation. The frequency of neonatal hyperbilirubinaemia was reduced in those exposed to phenobarbital. In a follow-up of 36% (719) of the 2003 children in one of the two geographical study areas at the age of 5.1–6.8 years in three intelligence tests, the 415 children who had been exposed to phenobarbital tended to score higher than the controls, but the difference was significant only on the Visuo-Motor Integration Test. The observations were continued on 341 adolescents. Pubertal development appeared to be affected, in that there was a trend for the pubertal stage to be delayed in the treated group. The boys showed significantly higher cognitive function as assessed by the Wechler Intelligence Scale. [The Working Group could not locate a subsequent publication on this cohort that contained a complete report.]

In a clinical intervention trial for infantile intracranial haemorrhage, 110 women in preterm labour were assigned to receive either placebo (n = 60) or phenobarbital (10 mg/kg bw intravenously followed by 100 mg/kg bw per day orally) until delivery (Shankaran et al., 1996a). There was a trend towards a decreased incidence of intracranial haemorrhage of any grade in the group given phenobarbital (22% versus 35%). The authors noted that the mode of action of this response is not clear but may be related to hypertensive peaks in the neonate. In a follow-up study, no adverse consequences on growth or in the McCarthy General Cognitive Index was seen in the phenobarbital-treated offspring up to 3 years of age (Shankaran et al., 1996b), but the study has been criticized for a low and potentially biased follow-up rate (Doyle, 1996).

The influence of prenatal exposure to phenobarbital on childhood IQ (measured by the Bayley Mental Development Index) at 2 years of age was studied in a double-blind, placebo-controlled trial in which the drug was given in utero for the prevention of intracranial haemorrhage to patients with an imminent risk of premature delivery (Thorp et al., 1999). The dose of phenobarbital was targeted to yield serum phenobarbital concentrations in the mother and infant of 15–17 µg/mL. [The Working Group noted that the study group also received 10 mg vitamin K every 4 days until delivery.] The incidence of antenatal fetal or postnatal death did not differ between the groups. In the 121 (32%) of 375 children who participated in the 2-year follow-up, a significantly lower Bayley mental developmental index was found in the treated group compared with the controls (104 versus 113). Backward regression analysis indicated the presence of five other covariates that were statistically significant, including maternal education and patent ductus arteriosis.

A prospective study of 983 infants born to women with epilepsy in Canada, Italy and Japan indicated that the incidence of malformations in the infants of women who had not taken an anti-epileptic drug (n = 98) was 3.1%, whereas that in the infants of women on phenobarbital monotherapy (n = 79) was 5.1%; the resulting odds ratio was 1.7 [no CI provided]. No specific pattern of malformations was identified after phenobarbital monotherapy. For any anti-epileptic therapy, the incidence of malformations was 9.0% (Kaneko et al., 1999).

In a review of exposure to anti-epileptic drugs and pregnancy outcome, it was noted that phenobarbital lowers folate concentrations in both humans and rats, which may contribute to its teratogenic potential (Lewis et al., 1998).

4.3.2. Experimental systems

The teratogenic effects of phenobarbital in experimental animals have been reviewed (Finnell & Dansky, 1991). A paucity of work on the effects of exposure to phenobarbital in utero on pregnancy outcome was noted, the available evidence suggesting that the potency was less than that of other anti-convulsants, such as phenytoin and valproic acid.

In a review of the literature, the influences of route of administration and dose on the nature of the adverse pregnancy outcome were emphasized (Middaugh, 1986). Dietary exposure resulted in high blood concentrations (40–200 µg/mL), reduced food consumption and maternal weight gain and low birth weights and anatomical and biochemical abnormalities in the offspring. Studies in which injections were given that tended to result in plasma concentrations similar to those for therapeutic uses (5–20 µg/mL) showed smaller effects on pregnancy outcome, although effects on behavioural and biochemical end-points were sometimes apparent.

Long-Evans rats were given phenobarbital by oral gavage at a dose of 40 mg/kg bw per day during the first 7 days of lactation to investigate whether neonatal exposure altered the sensitivity to carcinogens later in life. Twenty-four hours after injection of 37-week-old male offspring with [³H]aflatoxin B₁ at a dose of 1 mg/kg bw, more aflatoxin B₁–DNA adducts were found in the liver than in untreated controls. In other male offspring of the same age, neonatal exposure to phenobarbital caused a 2.6-fold increase in the activity of ethylmorphine-N-demethylase (Faris & Campbell, 1981).

4.4.1. Humans

In healthy volunteers, oral administration of phenobarbital at 2–3 mg/kg bw per day for 3 weeks increased antipyrine clearance but not the liver size (Roberts et al., 1976). Biopsy samples from the livers of patients treated for epilepsy with phenobarbital and phenytoin were found to have elevated total CYP activity, and the livers of these patients were enlarged (Pirttiaho et al., 1978).

4.4.2. Experimental systems

The ability of phenobarbital to induce the expression of CYP genes has been investigated and reviewed extensively (Conney, 1967; Okey, 1990; Waxman & Azaroff, 1992). The CYP types induced most effectively by phenobarbital in rat liver are CYP2B1 and CYP2B2. Other members of the superfamily induced by phenobarbital include CYP2A1, CYP2C6, CYP2C7, CYP2C11, CYP3A1 and CYP3A2. Relatively high concentrations of phenobarbital (typically 80 mg/kg bw per day given by intraperitoneal injection for 4 days) are required for this type of induction. Other enzymes induced by phenobarbital include aldehyde dehydrogenase, epoxide hydrolase, NADPH:cytochrome P450 reductase, UGT and several glutathione S-transferases. In addition, increased activities of γ-GT and catalase, but no change in glutathione peroxidase, have been reported (Furukawa et al., 1985). Phenobarbital at 0.05% (w/v) in drinking-water induced the activity of hepatic O⁶-methyltransferase in rats that had received a single dose of NDMA (O’Connor et al., 1988). The induction of detoxication enzymes may contribute to the observed inhibition of tumorigenesis by phenobarbital administered before or concurrently with DNA-reactive carcinogens.

The levels of 4-aminobiphenyl adducts in bladder DNA of rats given this carcinogen were decreased by phenobarbital at 0.1% (w/v) in drinking-water for 8 days (Olsen et al., 1993). This suggests that phenobarbital may induce drug-metabolizing enzymes that deactivate carcinogens.

The dose–response relationship for the effects of phenobarbital on total CYP, glucose-6-phosphatase and UGT activities was examined in male Sprague-Dawley rats (age not specified) given intraperitoneal injections of phenobarbital at doses ranging from 1 to 125 mg/kg bw per day on 6 consecutive days. The total CYP activity was determined spectrophotometrically in microsomal preparations. Dose-related increases in CYP activity were found: phenobarbital at 1 mg/kg bw per day had no effect, at 3 mg/kg bw per day it caused a significant increase, and a maximal increase (∼ two-fold) was observed at 75 mg/kg bw per day. For UGT, a similar pattern was seen: at 3 mg/kg bw per day, phenobarbital caused a significant increase, and a 2.8fold increase in activity was seen at 125 mg/kg bw per day. In contrast, the latter dose caused a 50% decrease in the activity of glucose-6-phosphatase in liver microsomes of these rats (Tavoloni et al., 1983).

CYP and related enzymes were induced for up to 90 weeks in C3H mice by exposing them to phenobarbital at 85 mg/kg bw per day. The concentration in the diet was adjusted between 0.048% and 0.083% in order to maintain the dose. After 4 weeks of exposure to phenobarbital, slight cellular and nuclear hypertrophy was seen in the immediate vicinity of hepatic veins. After 4 weeks of age, hypertrophy of centrilobular cells was evident. Between 25 and 40 weeks, focal proliferative lesions developed, followed by large eosinophilic nodules. The activities of microsomal ethylmorphine N-demethylase and 7-ethoxycoumarin O-deethylase were increased 8–11-fold and three- to sixfold, respectively, during the exposure period. The activities of CYP and NADPH-cytochrome c reductase were increased about twofold. Additionally, the activity of benzo[a]pyrene hydroxylase, which is associated with CYP1A1, was increased fivefold, and that of glutathione S-transferase was increased two-to-three fold (Collins et al., 1984).

C57BL, C3H, B6C3F₁ (C57BL × C3H) and C3B6F₁ (C3H × C57BL) mice were compared with respect to their susceptibility to induction of liver enzyme activities by phenobarbital, as they differ markedly in their susceptibility to phenobarbital-induced carcinogenesis. Four-week-old mice were given drinking-water containing 0.05% phenobarbital for 4 or 28 days. The ratio of liver:body weight was increased in all strains and at both times. Similarly, increased activities of CYP isozymes, aminopyrine-N-demethylase and 7-ethoxyresorufin-O-deethylase and an increased extent of testosterone oxidation were found in all strains. The authors concluded that the strain specificity of the cancer-promoting activity of phenobarbital was not due to any of these effects (Lin et al., 1989).

The phenobarbital dose–response relationships for CYP2B-mediated enzymatic activities were investigated in male and female Fischer 344/Ncr rats. Rats aged 8 weeks were given diets containing phenobarbital at concentrations ranging from 0 to 0.15%. After 14 days, the animals were killed, and the enzyme activities were determined in liver homogenates. The maximal increases over control were: benzyloxyresorufin O-dealkylation, 265-fold (female rats) and 54-fold (males); pentoxyresorufin O-dealkylation, 197-fold and 58-fold; and testosterone 16β-hydroxylation, 1320-fold (females) and 118-fold (males) (Nims et al., 1993).

The expression, inducibility and regulation of four CYP isozymes (PB₁, PB₂, MC₁ and MC₂) [CYP2C6, CYP2C11/12, CYP1A1, CYP1A2], glutathione transferases and microsomal epoxide hydrolase were studied in young female Wistar rats giving drinking-water containing 0.01% (w/v) NDEA for 10 days. One group of animals was not further exposed; a second group was given 3-methylcholanthrene by intraperitoneal injection at a dose of 40 mg/kg bw per day on 3 consecutive days before being killed, and a third group of animals was given NDEA and then a diet containing 0.05% (w/w) phenobarbital or an intraperitoneal injection at a dose of 80 mg/kg bw per day on 3 consecutive days before being killed. Frozen liver sections were analysed by immunohistochemistry for foci, nodules (adenomas) and carcinomas. Progressive loss of constitutive CYP expression was observed during tumorigenesis. Phenobarbital caused a heterogeneous pattern of response in preneoplastic and neoplastic lesions: some foci responded to the same degree as the surrounding normal liver, some were less, or not at all, inducible, and others were more inducible than the surrounding normal hepatocytes, particularly with regard to CYP2C11/12. The pattern of induction of this enzyme was closely linked to that of NADPH:cytochrome P450 reductase, suggesting that the regulation of these two enzymes is coordinated (Kunz et al., 1987).

Incubation with 1.5 mmol/L phenobarbital caused a 2.5-fold increase in the activity of 7-ethoxycoumarin O-deethylase in cultured human hepatocytes, but had no effect on the activity of aryl hydrocarbon hydroxylase. The median effective dose (ED₅₀) for this increase was about 0.5 mmol/L (Donato et al., 1990). In comparison, P450b, P450e and P450p [CYP2B1, CYP2B2 and CYP3A1] mRNA levels increased in primary cultures of adult rat hepatocytes exposed to various concentrations of phenobarbital. The concentrations that resulted in half-maximum increases in the activities of these three enzymes were 0.015, 0.0057 and 0.3 mmol/L, respectively, with maximal induction of 193-fold, 11-fold and 12.6-fold. Thus, there was only a threefold difference in ED₅₀ between CYP2B1 and CYP2B2, whereas there was a 20-fold difference between CYP2B1 and CYP3A1. The authors concluded from the different ED₅₀s that the classes of CYP gene were induced by different pathways (Kocarek et al., 1990).

Modification of gene expression is an important response to xenobiotic compounds, as changes in gene expression frequently modulate toxic response. In differential display approaches, the expression of over 7500 mRNAs was examined in the liver of chick embryos treated with phenobarbital in vivo. Twenty-nine cDNA fragments were significantly changed 48 h after treatment with phenobarbital in ovo. Of these, 18 were increased and 11 were decreased. The subcloning and sequencing of 20 of these fragments showed that CYP2H1, glutathione S-transferase, UGT, fibrinogen, glutamine synthetase and apolipoprotein B were up-regulated (Frueh et al., 1997).

The aryl hydrocarbon (Ah) receptor mediates the transcriptional response to a number of hydrocarbons, resulting in the induction of CYP1A enzymes. Phenobarbital did not appear to bind strongly to the Ah receptor but did induce CYP1A in liver cells. In wild-type and Ah-receptor-knockout C57BL/6J mice, phenobarbital induced CYP1A2, but not CYP1A1, even in the absence of the Ah receptor (Corcos et al., 1998).

A specific nuclear receptor (CAR) has now been identified that mediates induction by activation of the phenobarbital response element located in the 5′ flanking region of inducible CYP genes (Trottier et al., 1995; Zelko & Negishi, 2000). Activation of nuclear factor κB modifies the expression of many genes, and Mejdoubi et al. (1999) reported that it participates in the expression of phenobarbital-responsive genes.

Altered hepatocyte foci in the livers of male rats given phenobarbital overexpressed two forms of CYP, CYP2B1,2 and CYP2C7 (Decloître et al., 1990). In a study of the effects of phenobarbital on CYP2B1 and CYP2B2 in cultured rat hepatocytes, phenobarbital induced a concentration-dependent increase in benzyloxyresorufin O-deethylase activity up to 25-fold that of controls. Co-incubation with interleukin-6 or addition of interleukin-6 up to 12 h after phenobarbital inhibited this induction. The authors suggested that the inhibition is mediated by the early molecular events of the induction process (Clark et al., 1996). In a study in C57BL/6 mice and in mouse hepatocytes in culture exposed to phenobarbital, the overall constitutive expression of CYP2B10 was greater in male than in female mice. Phenobarbital induced the expression of both CYP2B9 and CYP2B10 (Jarukamjorn et al., 1999). The relationship between phenobarbital treatment, CYP gene expression and growth patterns in hepatic hyperplastic nodules induced by NDEA was studied in nodules from rats treated with phenobarbital. Phenobarbital increased the number of γ-GT-positive nodules, but it did not change the pattern of labelling indexes or the average labelling index. A slight but uniform increase in CYP1A2 expression was also seen. CYP2B1/2 was underexpressed in 53% of the γ-GT-positive nodules (Chen et al., 1992).

The induction of CYP2B, CYP2C and CYP3A by phenobarbital was examined in male and female patas (Erythrocebus patas) and cynomologus (Macaca fascicularis) monkeys. Hydroxylation of testosterone associated with CYP3A was increased two-to fivefold in phenobarbital-treated animals. Testosterone-16β hydroxylation activity was induced up to 15-fold, and benzyloxyresorufin O-dealkylation was induced 10fold by phenobarbital in both species of monkey. Induction of CYP2C was observed with phenobarbital and was more pronounced in the cynomologous monkeys than in the patas monkeys (Jones et al., 1992b).

4.5.1. Humans

The potential for phenobarbital to induce sister chromatid exchange in the peripheral lymphocytes of epilepsy patients on phenobarbital monotherapy was examined. Nine male patients (six smokers) of mean age 38.8 ± 1.25 (SE) years and nine male controls (six smokers) of mean age 38.4 ± 1.27 (SE) years were compared. The same exclusion criteria were applied to the two groups: age > 50 years, recent illness, use of medication other than phenobarbital, use of illicit drugs, alcoholism, surgery, anaesthesia, blood or blood product transfusions, chemotherapy, exposure to ionizing radiation and unusual exposure to ultraviolet radiation within 1 year of examination. The daily coffee consumption was 0–20 cups among the patients and 0–12 cups among the controls. The patients’ serum phenobarbital concentrations were consistently in the range 10–40 µg/mL, except for two patients on maintenance doses resulting in concentrations of 5 and 6 µg/mL. The mean numbers of sister chromatid exchanges/cell were 6.82 ± 0.54 (SE) in the exposed group and 6.14 ± 0.51 (SE) in the controls (Schaumann et al., 1989).

4.5.2. Experimental systems (see Table 6 for references)

Table 6

Genetic and related effects of phenobarbital.

In some studies, phenobarbital has been shown to be weakly mutagenic in Salmonella typhimurium TA1535 or TA100 in the absence, but not in the presence, of an exogenous metabolic activation system. The effect was observed in two of two studies with TA1535 and in two of five studies with TA100. Direct addition of phenobarbital enhanced the mutagenic effect in S. typhimurium TA1535 of sodium azide or 2-aminoanthracene, but not that of 4-nitroquinoline N-oxide or 2-nitrofluorene (Albertini & Gocke, 1992).

Phenobarbital induced aneuploidy, but not mutation or recombinational events in fungi; it did not induce sex-linked recessive lethal mutations, somatic cell mutations or mitotic recombination in Drosophila melanogaster.

In cultured, non-human mammalian cells, no DNA strand breakage or induction of unscheduled DNA synthesis was observed. Although mutations were induced in some studies, the inconsistent positive results were weak and not clearly associated with a particular locus; both positive and negative results were obtained in experiments at the Tk and Hprt loci and tests for ouabain resistance. Similarly, results for the induction of both sister chromatid exchange and chromosomal aberrations were equally divided between positive and negative in different laboratories. Single studies in cultured mammalian cells for the induction of micronuclei and aneuploidy did not show any effect of phenobarbital treatment. Variable results were also obtained in assays for cell transformation conducted in different laboratories, but the differences may have been due partly to the different transformation assays used. Assays for gap-junctional intercellular communication, however, also gave variable results in different laboratories even when the same Chinese hamster lung V79 cell system was used in five of six studies. In a single study with Djungarian hamster fibroblasts, cell-to-cell communication was actually enhanced. The negative result with NRK-52E cells, a rat kidney epithelial cell line, was interpreted by the authors as being consistent with the lack of effect of phenobarbital (which does not promote renal tumours) on the kidney, since barbital, a renal tumour promoter, did inhibit gap-junctional intercellular communication in these cells (Konishi et al., 1990).

In cultured human lymphoblastoid cells, no significant increase in the frequency of mutations was observed in one study, while, in cultured human lymphocytes, chromosomal aberrations were induced in two studies; sister chromatid exchange was not induced in another study.

In vivo, phenobarbital did not form adducts with DNA in mouse liver after either a single oral dose (200 mg/kg bw) or when given for 2 weeks in the diet (1000 mg/kg). A study of mutation induction in mice carrying the Escherichia coli lacI transgene as the target did not show any effect of phenobarbital. Studies of the bone-marrow cells of mice treated in vivo also did not show induction of sister chromatid exchange, micronuclei or chromosomal aberrations after treatment with phenobarbital alone. A significant response was observed in the single-cell gel electrophoresis assay with cells from the liver of mice given phenobarbital (140 mg/kg bw) by intraperitoneal injection 24 h before sacrifice, but not effect was seen with treatment 3 h before being killed; no effect was observed in a number of other organs. The frequency of morphologically abnormal sperm was not increased in mice treated with phenobarbital.

Phenobarbital was reported to induce chromosomal abnormalities, including translocations, in male mouse primary spermatocytes after oral administration for 5 or 60 days. [The Working Group noted that the level of the effects reported was little influenced by either the dose or the treatment regimen.]

A positive response was reported in a test for dominant lethal mutation in male Swiss mice (Nandan & Rao, 1983). [The Working Group noted that there was a significant response of all germ-cell stages, and a greater response in treated spermatogonial cells than in post-meiotic cells; this result is highly improbable and has not been reported previously with any chemical.]

In another study on mutation induction in the lacI transgene, transgenic mice were first given a single intraperitoneal injection of 50 mg/kg bw NDEA; 7 days later they were started on a diet containing 500 mg/kg sodium phenobarbital for 14 days, followed by a normal diet for 7 days. Increased liver weights were observed in the phenobarbital-treated mice, and the mutation frequencies in the lacZ gene recovered from liver were consistently higher than the values obtained from mice treated only with NDEA. Feeding a diet containing sodium phenobarbital at 500 mg/kg for 21 days did not affect the mutant frequency in the lacZ transgene in groups of mice given 100 mg/kg bw NDEA. The authors noted that no statistical analysis was performed, because of the small group sizes, which also limited interpretation of the data (Okada et al., 1997). In female Sprague-Dawley rats that had undergone a 70% partial hepatectomy and then received NDEA by gavage (10 mg/kg bw), subsequent treatment with diets containing 0.05% (w/w) sodium phenobarbital for 12 months produced a significant increase in the number of hepatocytes containing chromosomal aberrations. Most of the damaged cells were seen in γ-GT-positive foci. Karyotypic analysis indicated that the most frequent aberrations in these foci were a trisomy of chromosome 1 or of its long arm and a monosomy of chromosome 3 or its short arm (Sargent et al., 1992).

Klaunig and Ruch (1987a) found that phenobarbital inhibited gap-junctional intercellular communication most effectively in hepatocytes from B6C3F₁ mice, less so in primary hepatocytes from C3H and BALB/c mice and Fischer 344 rats and not at all in primary hepatocytes from C57BL mice. This finding was confirmed in part, and extended, in another study that showed that phenobarbital inhibited gap-junctional intercellular communication in primary hepatocytes from male Fischer 344 rats and B6C3F₁ mice, but did not do so in primary hepatocytes from male and female rhesus monkeys or from a human [sex unspecified] donor (Baker et al., 1995).

In studies of the mechanism of inhibition of gap-junctional intercellular communication, it was shown that phenobarbital (20–500 µg/mL) reduced gap-junctional intercellular communication between B6C3F₁ mouse hepatocytes in culture. Phenobarbital (250 µg/mL) also reduced cAMP levels 1 h after treatment, but no effect was observed 2, 4 or 8 h after treatment. The addition of dibutyryl cAMP (0.1 mM) increased the cAMP levels approximately 50-fold in these mouse hepatocytes and completely abolished the inhibition of gap-junctional intercellular communication by phenobarbital (Klaunig & Ruch, 1987b). Inhibition of protein synthesis by cycloheximide had no effect on the inhibition of gap-junctional intercellular communication by phenobarbital in cultured rat hepatocytes, but inhibition was enhanced by treatment with diethylmaleate, to deplete intracellular glutathione, or by the addition of the CYP inhibitors, SKF 525A or metyrapone (Guppy et al., 1994). It was later shown that phenobarbital-induced inhibition of gap-junctional intercellular communication is a complex phenomenon. Treatment with 2 mmol/L phenobarbital of primary rat hepatocytes co-cultured with WB-F344 rat liver epithelial cells for 1 h sharply reduced inter-hepatocyte dye-coupling from about 90% to 30%, but the cells fully recovered within 24 h, after which there was a more gradual reduction in dye-coupling to about 20% after 14 days. Dye-coupling between the WB-F344 cells was unaffected by phenobarbital over the same period and, in co-cultures, there was no dye transfer between the two cell types. The connexin (Cx32) steady-state mRNA levels were unaffected by treatment of the primary rat hepatocytes with phenobarbital for 14 days, and there was no change in Cx32 protein levels (Ren & Ruch, 1996). A similar transient decrease in gap-junctional intercellular communication after a 1-h exposure to phenobarbital was observed by Mesnil et al. (1993) in rat hepatocytes co-cultured with BALB/c 3T3 mouse-embryo cells, but the decrease was maintained at 50% of control values during a 3-week treatment period. Later studies from this group indicated that the inhibitory action of phenobarbital on gap-junctional intercellular communication is cell-specific rather than connexin-specific. Thus, phenobarbital inhibited gap-junctional intercellular communication in hepatocytes, which express primarily Cx32, but not in WBF344 rat epithelial cells, a highly communicating line that expresses Cx43, or in WBaB1 cells, a gap-junctional intercellular communication-incompetent line derived from WB-F344 that still expresses Cx43, or in WB-a/32-10 cells, which are derived from WB-aB1 by stable transduction with a Cx32 retroviral expression vector.

Cx32-deficient mice are resistant to liver tumour promotion by phenobarbital. Phenobarbital treatment led to an approximate fivefold increase in the volume fraction occupied by glucose-6-phosphatase-deficient liver lesions in Cx32^Y/+ mice, whereas there was no such increase in Cx32^Y/– mice. Even more pronounced differences were observed with respect to tumour response, phenobarbital clearly promoting the occurrence of large hepatomas in Cx32-proficient but not in Cx32-deficient mice. These results demonstrate that functional Cx32 protein is required for tumour promotion with phenobarbital (Moennikes et al., 2000).

After oral administration of phenobarbital (50 mg/kg bw per day for up to 6 weeks) to Sprague-Dawley rats, a direct microinjection dye-transfer assay was carried out on fresh liver slices (0.5–0.7 mm thick). The average area of dye spread decreased after 1 week and stayed at the same level up to week 6. The area and number of Cx32 spots per hepatocyte in the centrilobular zones of the liver also decreased between week 1 and week 6, whereas there was no change in the spots in the perilobular areas. No changes were observed in Cx26 (Ito et al., 1998). Using a higher dose (400 mg/kg bw per day for 5 weeks), Krutovskikh et al. (1995) observed nearly total disappearance from the plasma membrane of cells in the centrilobular region of both the principal hepatocyte connexin Cx32 and Cx26, while Cx43 protein and the expression of its mRNA were stimulated.

DNA sequence analysis of the H-ras gene in liver tissue from male B6C3F₁ mice treated with phenobarbital (0.05% in the drinking-water for 1 year) revealed a point mutation (AAA) in codon 61 (normal sequence CAA) in one of nine hepatocellular adenomas and in none of five hepatocellular carcinomas. When 50 liver tumours found in the control group were analysed, an activated H-ras gene was found in 15/18 hepatocellular adenomas and 10/14 hepatocellular carcinomas. The most frequent mutation in these tumours was a CG → AT transversion (59%), which is probably the result of a polymerase error (Fox et al., 1990). During DNA synthesis on a non-instructional template, the most frequent polymerase error involves the preferential insertion of an adenine, the so-called ‘A rule’ (Strauss et al., 1982).

In C3H/He mice, Ha-ras codon 61 mutations occurred in 6/21 (29%) liver tumours from untreated mice, but in 0/15 (0%) tumours from mice given a diet containing phenobarbital at a concentration of 0.04–0.07% providing a dose of 85 mg/kg bw per day. The absence of mutations in codon 61 (or in codon 12) in the phenobarbital-treated mice suggests a tumorigenic mechanism different from that of spontaneous tumours. In contrast, codon 61 mutations were found in 19/46 tumours from mice treated with NDEA (Rumsby et al., 1991). In a similar study, two of eight CF1 mouse liver carcinomas that occurred after prolonged exposure to diets containing 1000 mg/kg phenobarbital were shown to carry a CG → AT transversion in the Ha-ras gene. One of eight tumours from control mice also carried a codon-61 mutation (Bauer-Hofmann et al., 1990).

The same group analysed the pattern of codon 61 mutations in the Ha-ras gene of glucose-6-phosphatase-deficient hepatic lesions of male C3H/He mice given a diet containing 500 mg/kg sodium phenobarbital for 52 weeks. Ha-ras mutations were found in 12/21 lesions (57%) from untreated mice and in 4/16 (25%) in the phenobarbital-treated group (p < 0.01, Fisher’s exact test). The commonest mutation was a CG → AT transversion (eight in untreated mice, three in treated mice). The wild-type sequence, CAA, was present in all the enzyme-deficient lesions, and, when a mutation was present, the signals for the wild-type and the mutant sequences were very similar. This result suggested that both normal and mutant alleles were present in each cell of the lesions (Bauer-Hofmann et al., 1992).

Tumours from C3H/He mice were also screened for p53 mutations in exons 5, 7 and 8, which contain nearly all the mutations so far described. No p53 mutations were found in any of eight tumours recovered from mice treated with phenobarbital (Rumsby et al., 1994).

4.6.1. Liver tumours

The available evidence indicates that phenobarbital is generally not genotoxic, and genotoxicity does not appear to play a role in its hepatocarcinogenicity. In particular, DNA adducts have not been detected with ³²P-postlabelling methods in animals given doses that produce liver tumours.

Phenobarbital is a microsomal enzyme inducer and has been studied extensively for its ability to promote hepatic tumours. There is evidence that the microsomal enzyme induction is correlated with hepatic tumour promotion by phenobarbital. Rats and mice that develop tumours after treatment with phenobarbital show expression of CYP enzymes, CYP2B1 and CYP2B2 being the most important. The potency for CYP induction in mice and rats also correlates with the degree of tumour promotion. Phenobarbital does not induce these enzymes in hamsters and does not produce liver tumours in this species.

The enhancement of hepatic tumorigenesis by phenobarbital was shown to be due to tumour promotion rather than a syncarcinogenic effect. In an initiation–promotion model, phenobarbital administered before or with a carcinogen produced no tumours, whereas repeated exposure to phenobarbital after the carcinogen produced tumours.

Although the mechanisms of tumour promotion are not completely known, effects on the control of cell proliferation appear to play a role. Long-term exposure of rodents to phenobarbital produces hepatomegaly and hepatocellular hypertrophy and hyperplasia. Several studies have demonstrated a transient increase in DNA synthesis in normal hepatocytes. In initiation–promotion models, phenobarbital selectively increased the labelling index in foci as compared with normal surrounding liver. The foci progress to the stage of hepatocellular adenoma, in which cellular proliferation no longer depends on the presence of phenobarbital.

The mitogenic and tumour-promoting effects of phenobarbital appear to involve changes in growth factors, intracellular communication, gene expression and cell cycle signal transduction. These can mediate the effects of phenobarbital, including the transient increase in DNA synthesis, the selective mitogenic effects in foci and inhibition of apoptosis.

In addition, there are marked species and strain differences in susceptibility to hepatic tumour promotion by phenobarbital, which is genetically determined and heritable.

Overall, the experimental evidence supports the conclusion that the mode of action of phenobarbital in the production of hepatic tumours is non-genotoxic and involves tumour promotion.

4.6.2. Thyroidal effects

Although phenobarbital has not been shown to produce thyroid gland tumours in a bioassay for carcinogenicity, it has been shown to promote these tumours after administration of a carcinogen (NBHPA).

Phenobarbital is a microsomal enzyme inducer that increases the hepatic activity of thyroxin-UGT and alters thyroid function by enhancing the peripheral disposition of thyroid hormones. Phenobarbital-treated rats show decreased serum concentrations of thyroid hormones, increased concentrations of TSH, increased thyroid gland weights and follicular-cell hypertrophy and/or hyperplasia.

The tumour-promoting effects of phenobarbital have been shown to be mediated by increased pituitary secretion of TSH as a compensatory response to increased hepatic disposition of thyroid hormone, as opposed to a direct tumour-promoting or carcinogenic effect in the thyroid follicular epithelium.

Overall evaluation

Phenobarbital is possibly carcinogenic to humans (Group 2B).