Phenobarbital for the Treatment of Epilepsy in the 21st Century: A Critical Review


Address correspondence and reprint requests to Dr. M.J. Brodie at Epilepsy Unit, Western Infirmary, Glasgow G11 6NT, Scotland. E-mail:


Summary:  Phenobarbital (PB) is the most widely used antiepileptic drug (AED) in the developing world and remains a popular choice in many industrialized countries. Meta-analyses of randomized controlled trials suggest that few differences in efficacy exist between PB and other established AEDs, but its possible deleterious cognitive and behavioral side effects remain a concern in the developed world. In contrast, high degrees of efficacy and tolerability in everyday clinical use have been demonstrated consistently in observational studies in developing countries. We propose that a pragmatic, comprehensive outcomes program be carried out, perhaps under the aegis of the Global Campaign Against Epilepsy, to optimize the conditions of the use of PB, so that more people around the world can benefit from this cost-effective medication and live more fulfilling lives.

Phenobarbital (phenobarbitone; PB) was synthesized by Emil Fischer, the doyen of Germany's organic chemists, in 1911. Its anticonvulsant properties were discovered serendipitously by Alfred Hauptmann, who originally used it as a hypnotic for his epilepsy patients (1). Despite the development of successive generations of antiepileptic drugs (AEDs) (2), PB has retained a unique position in the therapeutic armamentarium and is still the most widely prescribed treatment for epilepsy worldwide. It is recommended by the World Health Organization as first-line for partial and generalized tonic–clonic seizures in developing countries (3). Although its purported propensity to cause sedation and other cognitive and behavioral side effects has relegated it to second- or third-line use in many parts of the industrialized world, it remains a popular choice in many developed countries (4). Indeed, in the Italian FIRST study comparing the effect of treating the first or second generalized seizure on long-term prognosis, more physicians chose PB (47% after the first seizure and 39% after the second seizure) than carbamazepine (CBZ), sodium valproate (VPA), or phenytoin (PHT) as the initiating AED for their patients (5).

Epilepsy is among the most common serious brain disorders worldwide. The disorder can be successfully controlled with a single well-tolerated AED in the majority of cases (6), and yet 85% of the 50 million epilepsy patients living in the developing world do not receive any treatment (7). In support of the efforts of the Global Campaign Against Epilepsy, a joint initiative of the International League Against Epilepsy, International Bureau for Epilepsy, and World Health Organization to bring epilepsy “out of the shadows” and improve patient care around the world (8), we review the clinical pharmacology and therapeutics of PB and reappraise its global role as an AED in the 21st century. Its pros and cons are discussed, and clinical evidence from developing and developed countries, critically reviewed. Use of PB for status epilepticus and in neonates is not be considered in this review.

Mechanisms of action and pharmacokinetics

Mechanisms of action

PB and other barbiturates exert their antiepileptic effect primarily by facilitating γ-aminobutyric acid (GABA)-mediated inhibition via allosteric interaction with the neuronal postsynaptic GABAA receptors (9,10). It enhances the activation of GABAA receptors by increasing the mean channel open duration without affecting open frequency or conductance (11). The resultant increase in Cl flux hyperpolarizes the postsynaptic neuronal cell membrane, thus impeding the transmission of epileptic activity. Barbiturates also activate the GABAA receptor directly, in the absence of GABA, an effect that may underlie their sedative properties (12).


PB (5-ethyl-5-phenylbarbituric acid) is a substituted barbituric acid (Fig. 1) with a pKa approximating physiologic pH (13). It is well absorbed with >95% bioavailability, reaching peak plasma concentration after 0.5 to 4 h (14,15). Concentrations of PB in the CSF correlate with unbound serum levels (13). PB has a low clearance with a long elimination half-life (3 to 5 days in adults and 1.5 days in children), affording once daily dosing (13) (Table 1).

Figure 1.

Structure of phenobarbital.

Table 1. Clinical pharmacology of phenobarbital
  1. aWide interindividual variation. See text.

  2. From Brodie MJ, Dichter MA. Antiepileptic drugs. N Engl J Med 1996;334:168–75.

IndicationPartial and generalized
  tonic–clonic seizures,
  neonatal seizures, or status epilepticus
Protein binding48–54%
Elimination half-life72–144 h
Route of eliminationHepatic metabolism
  25% renally excreted unchanged
Standard maintenance dosea
 Children4–8 mg/kg/day
 Adults60–240 mg/day
Daily doses1–2
Target plasma concentration10–40 μg/ml

The long half-life of PB may protect against the occurrence of withdrawal seizures, should intake be ceased abruptly, which can occur in developing countries because of interrupted supply. Little relation exists between the rate of PB withdrawal or initial serum level and seizure control (16), and PB can be discontinued safely in patients maintaining on other AEDs without increasing seizure frequency (17). In the United Kingdom Medical Research Council Antiepileptic Drug Withdrawal Study, no evidence indicated that discontinuation of PB, compared with other established agents, was associated with exacerbation of seizures (18).

Approximately 25% of a PB dose is excreted unchanged in urine (19). The majority of absorbed PB is metabolized in the liver to form two inactive metabolites, p-hydroxyphenobarbital by aromatic hydroxylation, and 9-d-glucopyranosylphenobarbital by glucosidation, but large interindividual variability is found in the proportions of contribution from these routes (13). Hydroxylation of PB is attributed to the cytochrome P450 (CYP) enzyme system, primarily CYP2C9, with minor contributions from CYP2C19 and CYP2E1 (20). Genetic polymorphism of the CYP enzymes affects their expression and is an important determinant of individual susceptibility to drug toxicity (21,22). Poor CYP2C19 metabolizers are found in 5% of white people, but in up to 20% of Asians (23). In contrast, variants of CYP2C9 are more prevalent among whites (∼35%) compared with African-American and Asian populations (<10%) (24). Polymorphism of CYP2C19 has been shown to influence clearance of PB in Japanese patients (25), although this effect was not confirmed in a subsequent study (26). Despite its supposed greater contribution, whether genetic variation of CYP2C9 affects the metabolism of PB has not been reported. Given the diversity of the elimination pathways and the variable contribution from the different metabolizing enzymes, the significance of genetic polymorphisms on the overall biotransformation of PB remains to be determined.

Brain disposition

Overexpression of drug transporters at cerebral capillary endothelium is recognized to play a potentially important role in limiting access of AEDs across the blood–brain barrier, resulting in pharmacologic intractability (27). PB has been suggested to be a substrate of P-glycoprotein, the prototype of such drug transporters, in some animal models (28) but not in others (29), whereas another drug transporter, multidrug resistance protein (MRP2), does not seem to influence the cerebral distribution of PB (30). Expression of P-glycoprotein is partly determined by genetic polymorphism of the encoding gene, MDR1 or ABCB1 (31), and the frequency of polymorphism exhibits wide ethnic variation (32). Recent in vitro studies found no influence of PB on expression of the drug transporter genes MDR1 or MRP1 (33,34).

Drug interactions

PB can be the target and effector in hepatic enzyme induction and inhibition interactions involving AEDs and other lipid-soluble drugs (35,36). VPA inhibits both the hydroxylation (37) and glucosidation (38) of PB, (39) resulting in reduced clearance and prolonged half-life (37). Non-AEDs that induce or inhibit CYP enzymes also may modulate the pharmacokinetics of PB. Examples include the reduction of PB clearance by chloramphenicol (35) and, to a lesser extent, by dextropropoxyphene (40). PB has been described as the archetypal enzyme inducer (41). It increases the expression of a range of CYP and phase II metabolizing enzymes (36,42), although disagreement exists on the exact subfamilies under influence (32,36). The induction of CYP enzymes by PB is a result of altered transcriptional regulation mediated through orphan nuclear receptors, including pregnane X receptor and constitutive androstane receptors (43,44). Enzyme induction with PB results in increased clearance and reduced circulating concentrations of many xenobiotics that undergo hepatic biotransformation, including other AEDs [e.g., PHT, CBZ, VPA, lamotrigine (LTG), ethosuximide (ESM), felbamate (FBM), topiramate (TPM), zonisamide (ZNS), and tiagabine (TGB)], many commonly used lipid-soluble drugs (e.g., oral contraceptives, warfarin, corticosteriods) and some endogenous hormones (e.g., vitamin D, sex hormones) (35,36).

Clinical efficacy and tolerability

Sources of evidence

PB is effective against partial and generalized tonic–clonic seizures (45). Although it has been used as an AED for >90 years, its risk/benefit ratio remains unclear, because objective analysis of its effectiveness is relatively deficient. This is partly due to a lack of commercial interest in PB. The perception of PB as a highly neurotoxic compound may also have hampered modern investigators' enthusiasm to evaluate its efficacy and tolerability in a systematic fashion. Indeed, the last published randomized controlled trial involving PB in the developed world, although reported in 1996, ceased patient recruitment in 1987 (46). Many of the “first generation” comparative AED studies with PB may not be regarded as conforming to current recommendations for clinical trial methods (47,48). Although randomized controlled trials are the recognized gold standard for evaluating the efficacy of a new therapy, whether their results can be readily extrapolated to clinical practice has been questioned (49,50). Pragmatic, observational studies may assist the translation of trial data into everyday use (51,52). In the case of PB, the vast experience from developing countries where the drug is widely used remains an important source of clinical evidence.

Randomized controlled trials: meta-analyses

PB has been compared with the established AEDs primidone (PRM), PHT, and CBZ in several major head-to-head randomized trials (46,53–55) and a number of smaller studies. Two of the major trials recruited adults (53,54), and the other two included just children (46,55). Meta-analyses of these trials have been performed by using individual patient data. However, important differences exist between the trial protocols (Table 2), as demonstrated by evidence of statistical heterogeneity in some analyses (56). The Veterans Administration study (53), performed in the United States, used a double-blind, “double-dummy” design and enrolled adult patients with partial-onset seizures with or without secondarily generalized tonic–clonic seizures. Untreated and “undertreated” patients were included. The latter, accounting for 21% of subjects, were defined as having “subtherapeutic” blood levels of AEDs. Because the patients were recruited from Veterans Administration medical centers, the great majority (87%) were men. The United Kingdom and the Indian studies were open-label and enrolled patients with both partial-onset seizures and primary generalized tonic–clonic seizures (46,54,55). Only patients not receiving treatment at randomization were included. These design features resulted in demographic differences compared with the Veteran Administration study, such as a more equal sex distribution and exclusion of previously treated patients, who may respond differently to AED therapy (6).

Table 2. Characteristics and results of major randomized controlled trials of phenobarbital
 Mattson et al. (53)Feksi et al. (59)Heller et al. (54)De Silva et al. (46)Pal et al.(55)
  1. GTCSs, generalized tonic–clonic seizures.

Year published19851991199519961998
Number of centers101251
BlindingDouble-blind, double-dummyOpen labelOpen labelOpen labelOpen label
SettingVeterans Administration Medical Centers in U.S.Hospital clinic in KenyaTeaching hospitals in U.K.Teaching hospitals and district clinics in U.K.Rural district in India
Age (yr)18–706–65≥163–162–18
Male sex (%)87Not reported485152
SeizuresPartial onsetGTCS ± other seizure typesPartial-onset or primary GTCSPartial-onset or primary GTCSPartial-onset or primary GTCS
Previous antiepileptic drugs (%)42260046
Antiepileptic drugs randomized (patient number)Phenobarbital (155)
Phenytoin (165)
Carbamazepine (155)
Primidone (145)
Phenobarbital (150)
Carbamazepine (152)
Phenobarbital (58)
Phenytoin (63)
Carbamazepine (61)
Valproate (61)
Phenobarbital (10)
Phenytoin (54)
Carbamazepine (54)
Valproate (49)
Phenobarbital (47)
Phenytoin (47)
Median follow-up (range, mo)20 (0 – 66)1230 (1–91)44 (3–88)9 (0–12)
Serum monitoringYesNoYesYesNo
Phenobarbital dosageFlexible. Serum level maintained at “mid to high therapeutic range”Flexible. Starting daily dose, 30–60 mg, depending on age.Flexible. Starting daily dose, 60 mg Maximum when serum level, 30–40 mg/LFlexible. Starting daily dose. 3 mg/kg Maximum when serum level 30–40 mg/LFlexible. Starting daily dose, 1.5 mg/kg. Maximum, 5 mg/kg
Primary outcomeRetention timePercentage of completers seizure free in 6- to 12-month follow-upTime to first seizure, time to 1-year remissionTime to first seizure, time to 1-year remissionFrequency of side effects
Primary outcome resultsCarbamazepine, phenobarbital, phenytoin better than primidone No difference between phenobarbital and carbamazepineNo difference among all four groupsNo difference among all four groupsNo difference between phenobarbital and phenytoin
Secondary outcome resultsCarbamazepine, phenobarbital, phenytoin better than primidone for tonic–clonic seizures; carbamazepine, phenytoin better than phenobarbital, primidone for partial seizuresNot applicable likely to bePhenobarbital more likely to be discontinued than othersPhenobarbital more efficacy outcome discontinued than othersNo difference in

Phenobarbital versus phenytoin

This meta-analysis (56) included the four major trials accounting for between 499 and 592 patients for various outcome measures (46,53–55). For the individual trials, no difference was found between PB and PHT in the primary outcome measures (see Table 2). Meta-analysis found that PB was more likely to be discontinued than PHT (hazard ratio, 1.62; 95% confidence interval, 1.22–2.14), but no difference was found between patients randomized to the two drugs in time to 12-month remission (hazard ratio, 0.93; 95% confidence interval, 0.70–1.23) or time to first seizure (hazard ratio, 0.84; 95% confidence interval, 0.68–1.05), implying that the difference in retention time was primarily because of a greater propensity for PB to produce side effects. However, significant quantitative heterogeneity was found among the trials (p = 0.009), with the highest discontinuation rates for PB in the two unblinded United Kingdom studies (46,54). Indeed, de Silva et al. (46) discontinued the PB arm in their trial after randomizing only 10 children to the drug because of perceived poor tolerability. In a study conducted in rural India, Pal et al. (55) found no difference in behavior rating scores between 94 children randomized to PHT and PB. Unfortunately the recruited sample size was only 67% of the original target, undermining the statistical power of the study to draw firm conclusions. Nevertheless, the question remains whether the higher discontinuation rate of PB compared with PHT in the UK trials could be due to local physician/patient bias in an open-label design.

Phenobarbital versus carbamazepine

Individual patient data were available from four trials (46,54,55,58) to the reviewers (57), accounting for 684 patients and only 59% from all eligible trials. For patients with generalized seizures, no significant difference was found between the two drugs in time to discontinuation (hazard ratio, 1.79; 95% confidence interval, 0.87–3.62). However, CBZ was less likely to be discontinued for patients with partial-onset seizures (hazard ratio, 1.6; 95% confidence interval, 1.18–2.17). No difference was seen in time to 12-month remission between the two groups for all seizure types. Contradictory to clinical experience and the reviewers' expectation, analysis for time to first seizure suggested a trend in favor of CBZ for generalized-onset seizures (hazard ratio, 1.50; 95% confidence interval, 0.95–2.35), whereas PB was better for partial-onset seizures (hazard ratio, 0.71; 95% confidence interval, 0.55–0.91).

Similar to the comparison between PB and PHT, the apparent higher discontinuation rate of PB was likely to be confounded by the open labeling of treatment in all but the Veterans Administration study (see Table 2). Another potentially important caveat in the meta-analysis was the exclusion of a large trial performed in rural Kenya (59) because of failure to obtain the final dataset by the reviewers. In this open-label study, 302 untreated pediatric and adult patients were randomized to receive PB or CBZ for 12 months. In the published report, the investigators found no difference in the proportion of patients remaining seizure free in the 6- to 12-month follow-up period between the two groups in both per-protocol and intention-to-treat analyses (see Table 2). Although more reports of adverse events occurred in the PB group, no significant difference was noted in the number of patients experiencing side effects between the two groups. PB was discontinued in only five (3.3%) patients because of adverse effects compared with eight (5.3%) in the CBZ group. Omission of this large trial could have had a significant impact on the validity of the meta-analysis.

Observational studies

The use of PB for the treatment of epilepsy has been investigated in a number of observational studies in developing countries. Their conduct was driven by the practical need to establish a feasible and sustainable clinical service for this prevalent condition, and tended to include largely unselected, untreated patients with a range of seizure types across all ages.

In a retrospective study conducted in rural Tanzania, among 164 patients newly started on AED therapy between 1959 and 1963 (all but six received PB), 52.4% were reported to have no further seizures while maintained on treatment (60). Sixty-nine percent of patients were treated for 6 to 10 years, but unfortunately AED supply ended in 1971. The mortality was high in the patient cohort (67% at 30 years), and 21 (13%) patients were known to die of epilepsy-related events, even during the period of reliable drug supply. The seizure types were not well documented but appeared to be mixture of partial and generalized seizures.

In a hospital clinic in urban Nigeria where 90% of the 344 children with epilepsy were treated with PB, complete seizure control was achieved in 50.6%, and in only two patients was PB discontinued because of intolerable side effects (61). However, 94 (27%) patients were lost to follow-up after one or two visits for unknown reasons, rendering it difficult to draw valid conclusions from the data.

To overcome these limitations, the Yelandur study, conducted in rural south India, employed a team of paramedical workers and specialists in neurology and pediatrics (62). Patients with focal and tonic–clonic seizures were enrolled, whereas those with other generalized seizures, such as absence, myoclonic, or atonic seizures were excluded. Only 15 (11%) among the 135 recruited patients were lost to follow-up during the 5-year study period. Seventy-five (55%) patients took PB, 68 (50%) as monotherapy, whereas the rest were treated with PHT. Bearing in mind that drug assignment was not randomized, only three (4%) patients receiving PB reported adverse events, compared with 43% of those taking PHT (mostly in the form of gingival hyperplasia). Direct comparison in seizure control between the two groups was not made.

Nimaga et al. (63) enlisted support from the local village authorities in an open study of PB in the extremely poor and inaccessible rural areas of Mali. PB was given to 96 children and adults with partial and generalized seizures for an average of 12 months. A high level of drug adherence was achieved by regular domiciliary monitoring of each patient by trained rural physicians. Seizure control for ≥5 months was noted in 81% of patients, with associated life-changing improvement in social functioning, such as employment and education, in many. Initial minor treatment side effects (e.g., dizziness and drowsiness) did not persist with prolonged therapy. After a year of treatment, only three (3%) patients experienced adverse events, none of which was severe enough to lead to drug discontinuation.

In summary, in contrast to the high discontinuation rates reported in some randomized trials performed in industrialized countries, PB appears to be efficacious and well tolerated when used in resource-poor parts of the world. In addition, the effectiveness of PB has often been demonstrated in spite of a long history of untreated seizures and a general lack of sophisticated drug-level monitoring or even basic clinical investigation. In the study in Nigeria, only 11 of 344 patients could afford an EEG examination (61). Therefore seizure classification in these studies was largely semiologic and tended to be poorly defined in the published reports. In addition to this shortcoming, many of the observational studies in developing countries have low follow-up rate because of dropouts or deaths, variable follow-up duration, and lack of systematic data collection for adverse events and formal testing for cognitive and behavioral side effects. It also may be argued that the lack of alternative AED options in many cases had deterred the patients or their carers from withdrawing from treatment, even in the presence of significant side effects.


Although PB demonstrates overall tolerability similar to that of other established AEDs, and serious systemic side effects are uncommon (Table 3)(64), its potential neurotoxicity remains a topic of major concern, particularly in the developed world, where a range of alternative AEDs is readily available (65). These neurotoxic effects include sedation, behavioral problems (in particular, hyperactivity), impaired cognition, and depressed mood and affect. A number of studies have evaluated the neurotoxicity of PB in a range of clinical settings, but only a few had a randomized controlled design.

Table 3. Potential side effects of phenobarbital
Relatively commonUncommon
  1. From Baulac M, Cramer JA, Mattson RH. Phenobarbital and other barbiturates: adverse effects. In: Levy RH, Mattson RH, Meldrum BS, et al., eds. Antiepilepticdrugs. 5th ed. Philadelphia: Lippincott Williams & Wilkins, 2002:528–40.

NeurotoxicityMegaloblastic anemia
 CognitionAggravation of porphyria
 Mood and affectHypersensitivity
Connective tissue disordersTeratogenicity
 Dupuytren's contracture 
 Frozen shoulder 
 Folate deficiency 

Reinisch et al. (66) conducted two case–control studies on 114 men, in Copenhagen, Denmark, who were exposed to PB during gestation via maternal medical treatment, mostly for hypertension. When tested in early adulthood, they had significantly lower verbal (but not performance) intelligence score compared with controls matched for a spectrum of maternal variables.

When phenobarbital was given to children to prevent recurrent febrile convulsions in a double-blind placebo-controlled study, a significant negative correlation was found between serum PB level and five Binet subscores but no difference in hyperactivity between the two groups (67). Farwell et al. (68,69) performed prospective serial analysis of the cognitive effects of early treatment with PB for febrile seizures. Children aged 8–36 months with at least one febrile seizure were randomized to receive either PB (4 to 5 mg/kg/day) or placebo for 2 years. A trend of higher dropout rate in the PB group (23%) was noted, compared with the placebo group (14%), but life-table analysis showed that the probability of staying on the assigned treatment in fact tended to be higher in the former (0.82 at 1 year and 0.66 at 2 years) than the latter (0.65 at 1 year and 0.46 at 2 years). When tested at 2 years, children in the PB group scored significantly lower than the controls in the Stanford-Binet Scales of Intelligence. The difference narrowed to a nonsignificant level 6 months after drug discontinuation, when the children were tested at 2½ years. These investigators reassessed the IQ and academic achievement of 139 children (64% of the original cohort) at the median age of 7 years 8 months (range, 6 years 8 months to 10 years 1 month) when they completed at least the first grade in school (69). The difference in the Stanford-Binet IQ score had decreased further. However, in the Wide Range Achievement Test, treated children achieved significantly lower reading scores and nonsignificantly lower spelling and arithmetic scores compared with placebo controls.

It may be argued that such placebo-controlled studies do not have direct relevance to patients with epilepsy for whom AED therapy is indicated. When PB was given for the treatment of epilepsy, reports of its cognitive and behavioral adverse effects in comparison with those of other established AEDs appeared to be more conflicting. Most of the data are derived from pediatric populations. Vining et al. (70) performed a double-blind crossover study of 28 children with epilepsy treated with PB or VPA for 6 months. Seven children dropped out, four because of severe behavioral problems while receiving PB. Among the 21 completers, children scored significantly less (uncorrected p < 0.01) while taking PB in four of 35 compared items. In the parental questionnaires with 48 items, a general trend to better behavior was seen with VPA compared with the PB phase, although the difference was significant in three items only. Mitchell and Chavez (71) randomized 33 children with newly diagnosed partial-onset seizures to either PB or CBZ. No difference was noted in psychometric and behavioral evaluations at 6- and 12-month follow-up by both parents and the investigators. Meador et al. (72) compared the neuropsychological effects of PB, CBZ, and PHT after 3-month treatment in 15 patients with complex partial seizures by using a randomized, double-blind, triple crossover design. Among a range of psychometric testing, the only significant difference was for digit symbol, with patients scoring worse while receiving PB than with the other two drugs.

Seventy-six Chinese children with new-onset epilepsy were randomized to receive PB, CBZ, or VPA in a Taiwan study (73). No significant difference was found in psychometric testing scores among the three groups after 6 and 12 months of treatment, but P300 event-related potential latencies were prolonged in the PB group compared with the other two groups. Pal et al. (55) applied validated behavior rating scales in their randomized trial in India and found no significant overall difference in scores between the PB and PHT groups. Indeed, of the 20 children rated as having behavioral problems, 13 belonged to the PHT group.

Effectiveness of phenobarbital: developing versus developed countries

Evidence from developing and developed countries, derived from either randomized trials or observational studies, suggests that PB has an efficacy similar to that of other established AEDs. High discontinuation rates due to neurotoxicity tended to be observed in studies conducted in developed countries, whereas the drug did not appear to show excess neuropsychological toxicity when used in the developing world. Although methodologic deficiencies of the observational studies conducted in resource-poor countries are clearly recognized, a possible explanation for this discrepancy in the tolerability of PB might be the dose, which tended to be higher in developed than in developing countries (see Table 2). Recent long-term outcome studies of newly diagnosed epilepsy suggest that the majority (>90%) of patients who enter remission on first-ever monotherapy require no more than a modest or moderate drug dose (74). This could account for the comparable degree of efficacy of the lower doses of PB used in the open-label studies in developing countries, while avoiding excessive toxicity associated with the higher doses used in trials in developed countries.

Ingrained preconceptions against PB by treating clinicians, patients, and their carers may have introduced bias in some of the open-label studies. The threshold at which cognitive and behavioral adverse effects becomes intolerable may be influenced by the medicosocial context in which PB is evaluated. With limited treatment options, patients (and carers) in developing countries with a long history of untreated recurrent seizures may be willing and able to tolerate a higher level of neurotoxic side effects while benefiting from the success in seizure control. Last, the possibility of genetic influence on the tolerability of PB has received little attention. Pharmacogenetic factors could contribute to differential response to PB, in terms of both efficacy and neurocognitive side effects, across different ethnic groups.

Economic factors

Perhaps the greatest advantage of PB is its unparalleled low net cost. According to 2003 figures, a box of 1,000 generic PB tablets (100 mg each) costs US$7.12 (75). In the Mali study (63), each patient required in average 1.1 tablets per day, translating to 401.5 tablets, or US$2.56 per patient per year, which is far less than the transport costs for physician visits and delivery of supplies to the patients in the villages (∼US$915 for 100 patients per year). In India, as of 2002, if the per-diem cost of PB is defined as one unit, the corresponding figure for PHT is 1.6; CBZ, 4.7; and VPA, 5.3 (62). The differences in price are even more staggering and up to several hundredfold when comparing PB with the newer agents. For developing countries with low purchasing power, cost-effectiveness is of top priority in choosing treatment. As Nimaga et al. (63) pointed out, in poor regions, “the choice is not between phenobarbital and a new medicament but between phenobarbital and no treatment at all.”

Proposal for a pragmatic outcomes program

In view of its affordable cost, ease of use with once-daily dosing, reliable supply, broad spectrum of action, and comparative efficacy to other established AEDs, it is not surprising that PB is the most prescribed AED worldwide. However, its purported behavioral and cognitive side effects, identified in randomized controlled trials in industrialized countries in particular, have raised genuine ethical concerns. Yet observational studies performed in developing countries reported a high degree of effectiveness, but these studies had serious methodologic deficiencies. A pragmatic response to this dilemma would be the conduction of a modern, comprehensive, prospective evaluation program with the primary goal of optimizing the use of PB as monotherapy for untreated epilepsy in the developing world. This project could help solve the pressing need to devise cost-effective and sustainable treatment programs for epilepsy in this setting, while providing long-term outcome data and evaluating factors that influence treatment response. To fulfill these aims, studies should have the following ambitious objectives:

  • 1To document the tolerability of PB (particularly its cognitive and behavioral side effects)
  • 2To measure the efficacy of PB
  • 3To correlate these two outcome parameters with dosage
  • 4To determine whether and to what degree outcome is influenced by genetic factors
  • 5To assess the impact of PB treatment on socioeconomic functioning

This approach could be incorporated into the Global Campaign Against Epilepsy's ongoing “demonstration projects” (76,77). In view of the wide ethnic variations in frequency of genetic polymorphisms, separate studies in Africa, Asia, and Latin America should be considered, and their results compared. Particular effort should be made to ensure reliable drug supplies and meticulous follow-up (e.g., by domiciliary visits) to minimize drop-out rates. Because one of the prime objectives would be the assessment of the tolerability of PB, a low starting dose and slow titration schedule should be used. Pragmatic outcome measurements including educational, employment, and family questionnaires would allow evaluation of the real impact of PB treatment on socioeconomic functioning in patients with epilepsy in the developing world.

Pharmacogenetics also should form a component of the program. A population-genetic study design with appropriate stratification (78) could be used to test the hypothesis that the differential response to PB, in particular relating to the occurrence of neurocognitive side effects, is in part due to genetic variability. Materials for genotyping should be collected from all patients at treatment initiation and stored for later analysis, so that treating clinicians remain blinded to the patient's genotype. The study should include genes already reported to be associated with epilepsy outcome [e.g., ABCB1 encoding P-glycoprotein (79), PRNP encoding cellular prion protein (80)], as well as other candidate genes (e.g., CYP enzyme and GABA receptor genes), and any associating genes reported in the future.


On a global scale, PB remains one of the most important AEDs because of its continuing widespread use in developing (and many developed) countries. It has many advantages, including reliability of supply, affordable cost, broad spectrum of action, and ease of use. Reports of cognitive and behavioral side effects are conflicting. As Eadie (81) pointed out, PB “may be allowed to fade from use, at least in affluent societies, not so much because of its limitations but because its virtues are no longer promoted.” Pragmatic, comprehensive studies should be carried out to optimize the condition of its use, so that more people around the world can benefit from effective AED treatment and live more fulfilling lives.