Debate: Substitution of generic drugs in epilepsy: Is there cause for concern?


Address correspondence to Barry E. Gidal, School of Pharmacy & Dept. of Neurology, University of Wisconsin, 777 Highland Ave, Madison, WI 53705, U.S.A. E-mail:


The role of generic drugs in both the US and global marketplace has been steadily increasing over the last few years. Although generic drugs clearly represent an important economic alternative for many patients, there are reasons for concern in certain disease states. Recently, the substitution of antiepileptic drugs in patients with epilepsy has gained increased attention. Concerns over potential therapeutic inequivalence has prompted many clinicians to question current regulatory requirements for both establishing bioequivalence, as well as product substitution. The objective of this article is to present arguments both for and against the use of generic drugs and practice of generic substitution in patients with epilepsy. Regulatory requirements, pharmacokinetic methodology, and biopharmaceutical considerations are discussed.

Both in the U.S.A. and Europe, the increasing use of generic antiepileptic drugs (AEDs) has resulted in considerable controversy. The arguments for an increased use of generics are mainly economic but the regulatory criteria for bioequivalence that are applied in the USA and elsewhere have been questioned, and concerns have been raised about the safety of generic substitution in epilepsy treatment. The authors of this debate article have been assigned opposite positions and asked to present arguments against generic substitution of AEDs (Barry Gidal) or for an increased use of generic AEDs (Torbjörn Tomson).

Con: Generic Substitution of AEDs Barry E. Gidal


Several surveys have suggested that although our regulatory bodies have assured us that approved generic formulations of AEDs are therapeutically equivalent to their branded counterparts, a growing number of neurologists and patients remain unconvinced, with many feeling that substitution had resulted in either breakthrough seizures or new adverse effects in some patients (Guberman & Corman, 2000; Wilner, 2002, 2004; Haskins et al., 2005; Berg et al., 2006). These concerns have prompted several professional societies to issue position papers ( Liow et al., 2007) expressing concern for widespread and indiscriminate substitution. Given that it is generally believed that optimization of AED pharmacotherapy requires individualization, perhaps even small deviations in bioavailability have the potential to result in loss of seizure control in some patients (Kramer et al., 2007).

At the heart of this controversy is the notion of therapeutic equivalence. In order to fully understand both the scientific merits and potential limitations of this concept, the clinician must first understand the methodologic process used by the US Food and Drug Administration (FDA).

Approval of generic drugs in the U.S.A.

For the purposes of interchangeability, a generic drug must meet a number of fairly strict criteria. First, the generic product must be pharmaceutically equivalent, meaning that it must contain the same amount of active drug as the innovator product, and must meet United States Pharmacopeia (USP) standards for purity, strength, and quality. Inactive ingredients must be recognized as safe, but do not need to be identical to those used in the innovator, or branded product. The generic product must also meet FDA requirements for adequate labeling, and the manufacturer of the product must be able to demonstrate to the FDA that the production facility is in compliance with Good Manufacture Practices (GMP) {FD&C act 505 Sections 505 (j) and 505 (b) (2) (21 CFR 314.54)}.

For an orally administered drug such as an AED, therefore, once these criteria have been met, the key determinant of therapeutic equivalence is that of bioequivalence. The formal, FDA definition of bioequivalence is as follows: “The absence of a significant difference in the rate and extent to which the active ingredient or active moiety in pharmaceutical equivalents or pharmaceutical alternatives becomes available at the site of drug action when administered at the same molar dose under similar conditions in an appropriately designed study”{Federal Register (21 CFR 320.1)}. In essence, this means that to be considered therapeutically equivalent, a generic formulation must display essentially similar pharmacokinetic properties as the innovator product. It is this definition of equivalence that has resulted in both confusion and controversy.

To be accepted as being bioequivalent, the FDA requires that new oral drug formulations be compared to the branded product in relatively small (typically 24–36 healthy adult volunteers) single-dose crossover studies. Bioequivalence is evaluated by comparison of the area under the concentration–time curve (AUC), a measure of systemic exposure, and peak plasma concentration (Cmax), a measure of absorption rate. For a generic product to be considered equivalent, the 90% confidence interval (CI) of the log-transformed ratios of AUC and Cmax between brand and generic product must fall between goal-posts of 80 and 125%. This has been commonly misinterpreted to mean that the FDA will accept a difference in AUC of 20–25% for a generic product. This is an incorrect interpretation. Indeed, although early regulatory definitions of bioequivalence did allow for a ±20% variability in pharmacokinetic parameters, this was changed in the 1980s and early 1990s to the current, more stringent requirement that the 90% CI fall within the 80–125% limits (Sommerville, 2006). Under the newer rules, this does not mean that there can be a 20–25% difference between the mean pharmacokinetic parameters of the two products. If in fact, ratios are close to 80% or 125%, it is more likely that upper or lower confidence limit will fall outside accepted limits. Importantly, it must be recognized that the intent of the regulations is to ensure interchangeability between innovator product and an individual generic formulation. There is no requirement that generic products be tested to prove bioequivalence with another generic formulation; they are assumed to be equivalent.

Given these pharmacokinetically and statistically sound testing approaches, how do we reconcile the apparent discordance between regulatory pharmacokinetic trials and widespread clinical experience and perception? In other words, although statistically valid, do these statistics really answer the important clinical question? What is the essential issue? It is reasonable to argue that there are several issues including the notion of “prescribability versus switchability,” as well as the validity of generalizing single-dose pharmacokinetic studies in young healthy individuals to other patient populations.

Bioequivalent drug formulations: Biopharmaceutical considerations

There are numerous factors that can influence the bioavailability of a drug product. Chief among these is aqueous solubility. For a drug to reach the systemic circulation where it is available to exert a pharmacologic effect, a number of steps or processes must occur, including drug disintegration and dissolution, diffusion through gastrointestinal fluids, and, finally, mucosal membrane permeation and uptake into blood or lymph. Important factors that can affect any of these steps include alterations in aqueous solubility, membrane permeability, and presystemic metabolism. For a generic drug product, given that the new formulation contains the same active ingredient, pharmaceutical characteristics such as solubility and permeability of the compound are likely to be the most important predictors of bioequivalence (Amidon et al., 1995).

Drugs that exhibit high solubility and high permeability are generally considered to be very well absorbed. Factors (either pathophysiologic, or drug related) that affect either gut surface area or gut transit time, may be expected to alter bioavailability. Drugs that display low aqueous solubility, but high membrane permeability tend to display dissolution rate-limited oral absorption. Anything that increases the rate of in vivo dissolution will tend to increase bioavailability of these compounds. Drugs with low water solubility may be more likely to display variable oral absorption patterns, owing to variable dissolution rates. Changes in product formulation that alter (either improving or impairing) dissolution may be clinically significant. In addition, because drugs are more soluble when ionized, increased gastric pH can decrease drug dissolution of weak bases and increase the solubility of weakly acidic drugs (and hence fraction dissolved and subsequently absorbed) such as phenytoin. Drugs such as zonisamide (pKa of 9.66) also display solubility that is pH dependent, becoming more soluble as pH increases.

Although there are requirements that the active ingredient be identical to the branded product, there are no such requirements regarding other ingredients such as excipients, binders, fillers, and lubricants. Variances in these factors may be of only modest importance for many drugs, for drugs such as phenytoin, carbamazepine, lamotrigine, and oxcarbazepine that have relatively poor aqueous solubility (Bialer, 2002; Dickens & Chen, 2002; Spina, 2002); however, formulation changes that affect dissolution may ultimately impact oral absorption for some agents (Meyer & Straugh, 1993).

In Australia, Tyrer et al. (1970) reported an outbreak in cases of apparent phenytoin toxicity following a change in excipient content, resulting in improved phenytoin absorption and increased phenytoin serum concentrations. Changes in dissolution rate, and wide variances in absorption rate, may also explain apparent therapeutic failure of a generic carbamazepine formulation, another drug where in vitro dissolution rate appears to correlate with bioavailability. In 1988, the FDA investigated reports of unexpected seizure breakthrough in patients receiving a previously FDA-approved generic carbamazepine formulation (Meyer & Straugh, 1993). Both in vitro and in vivo studies demonstrated substantial variability in both dissolution rates and oral bioavailability in three lots of this formulation as compared to brand product. This particular formulation was eventually withdrawn from the market (Meyer et al., 1992, 1998). Similarly, Revankar et al. (1999) noted differences in dissolution rates in vitro that corresponded to significant differences in Cmax and Tmax between brand and a generic immediate-release carbamazepine formulation, leading these authors to conclude that the marketed generic was in fact not bioequivalent to the brand product. These reports highlight a potential limitation to the current approval process: that being, once a product formulation has been approved, unless a formulation change occurs, there is no requirement for additional in vivo pharmacokinetic studies.

Are current testing methodologies appropriate for all AEDs?

As described previously, current testing to establish bioequivalence is based upon single-dose crossover experiments in young, healthy volunteers. Although this approach is certainly pharmacokinetically valid, one must question whether this is always appropriate, given the unique biopharmaceutical properties of several of our commonly used AEDs. In other words, can we always extrapolate data generated from young, healthy adults to other patient populations (e.g., elderly and intellectually developmentally disabled children) that may have multiple comorbid medical conditions?

For example, phenytoin, still one of the most commonly prescribed AEDs in the USA, displays nonlinear pharmacokinetics, owing to saturation of the cytochrome P450 isozymes responsible for its metabolism. Practically speaking, this implies that modest changes in dose, or amount absorbed, may result in disproportionate changes in serum concentration. Although this phenomenon is well recognized, it is important to appreciate that the elderly patient may be more susceptible, presumably because of age-related declines in hepatic metabolic capacity. In addition, as discussed previously, phenytoin absorption may display variability because of physiologic factors.

Several well-documented reports have suggested that there may be problems with generic formulations of this agent. Rosenbaum et al. (1994) reported that mean phenytoin serum concentrations declined by about 30% when branded drug was replaced by an approved generic phenytoin preparation in 10 patients. More recently, in a retrospective review, Burkhardt et al. (2004) reported increased seizures in a group of 11 patients over a 5-month period. Ten of these patients had undergone substitution of a generic formulation of phenytoin. Following substitution with the generic, phenytoin serum concentrations were noted to decline by approximately 30%. Of note, once the agent was switched back to the brand product, serum concentrations returned to pre-switch values (Burkhardt et al., 2004).

How can these observations be explained? Well-conducted regulatory studies confirm that these products are essentially the same from a pharmacokinetic perspective. One possible explanation is that given the nonlinear, saturable elimination of phenytoin, extrapolating results from single-dose studies may not necessarily predict outcomes when the medication is given chronically. To illustrate this, in one study, a group of healthy subjects were given single doses of both an approved generic phenytoin formulation, and branded Dilantin, in a cross-over pharmacokinetic study (Wilder et al., 2001). In this study, bioequivalence was assessed using accepted FDA testing methodology, except that subjects were given both the generic formulation and branded product with a high-fat meal. Under these testing conditions, AUC was only about 13% lower for the generic product versus brand. Interestingly, however, the investigators used pharmacokinetic parameters derived from this experiment, and conducted simulations to assess the impact this otherwise modest reduction in bioavailability might have when given chronically. These simulations suggested that a 13% decrease in bioavailability could result in a 37% reduction in mean serum concentration if this formulation was given with food. Taken together, data from this study, as well as the report of Burkhardt and colleagues (2004) suggest that extrapolation of data from single-dose studies to chronic, multiple-dosing situations may be problematic for a nonlinear medication such as phenytoin. Although clearly the generic formulation is prescribable, one must question whether it is truly switchable under all clinical circumstances.

Clinical and economic consequence of generic substitution

Although the previous discussion has focused strictly on biopharmaceutical considerations, ultimately, one must question whether these theoretical concerns are of any “real-world” clinical importance. Several recent reports would suggest that they are.

In a retrospective claims database (Ingenix LabRx) analysis, Zachry et al. (2007) sought to determine if patients who received urgent, seizure-related care in an inpatient hospital or emergency/urgent care setting, were at greater risk of having had a substitution of their branded AED to an A-rated generic product during the previous 6 months when compared with epilepsy patients with no evidence of receiving seizure-related care in similar settings. In this analysis, cases (patients requiring urgent care) had an 81% greater odds of having a generic AED formulation switch in the previous 6 months as compared to controls (11.3% vs. 6.2%).

Other potential indicators of undesired clinical outcomes associated with generic substitution may be found in switchback rates from generic alternatives back to branded products. In one retrospective analysis from Canada, Andermann et al. (2007) evaluated switchback rates of several classes of drugs including AEDs (lamotrigine, clobazam, and valproic acid), as well as several antidepressants and cholesterol-lowering drugs. Strict Ontario rules favoring generic formulations require a physician letter of medical necessity before switchback from generic to the original brand product can be allowed. In this analysis, a high rate of switchback to branded AEDs (12.9% to 20.9%) compared with non-AED drug classes such as selective serotonin reuptake inhibitor (SSRI) antidepressants (1.5% to 2.9%) was found.

Similar findings were noted by LeLorier et al. (2008a), who found markedly higher switchback rates for AEDs as compared to antihypertensive or lipid-lowering drugs. In addition, for patients switching from generic to branded lamotrigine, significantly higher rates of medical service utilization and longer hospital stays were noted during the time when patients were receiving the generic product. Daily doses of generic lamotrigine were also about 5% higher as compared to the brand product.

Although the reasons for this apparently high switchback rate for AEDs as compared to non-AEDs is still uncertain, these data do clearly suggest that patients, physicians, or perhaps both, are feeling compelled to return to using higher-cost branded medications for epilepsy, as compared to other medical conditions. It must be acknowledged that retrospective analyses clearly have limitations, especially with respect to establishing a causal relationship with product substitution; however, collectively, these reports do raise concerns.

Finally, it is important that any discussion of generic drug substitution includes an economic perspective. In several recent analyses (Duh et al., 2007; Lelorier et al., 2008b), Canadian investigators have evaluated the potential cost implications of generic substitution. Using health claims data from Quebec’s provincial health plan, Lelorier and colleagues (2008b) identified 671 patients with epilepsy who were being treated with branded lamotrigine (Lamictal, GlaxoSmithKline, Philadelphia, PA, U.S.A.). Periods of branded and generic use of lamotrigine were compared for total health care costs, stratified into prescription drugs, and inpatient services and outpatient services. Interestingly, despite the lower cost of the generic product, using two different cost-estimation methods, these authors calculated significant increase in total projected health care costs during the time patients were receiving the generic product as compared to the brand (US$17502500 per person-year, p < 0.01). By way of these pharmacoeconomic analyses, it is reasonable to suggest that the potential savings associated with a switch to a generic product (at least for lamotrigine) may not be as great as one might anticipate, owing to higher costs likely associated with increased physician visits, hospitalizations, and utilization of medical/pharmacy services. Although these observations clearly require confirmation, they do suggest that whether because of anticipated or actual untoward clinical consequences, generic substitution may actually be increasing overall cost of patient care.


Substitution of generic AEDs is clearly a controversial subject. Generic products do offer an important alternative for most patients. Although FDA guidelines are sound for most medications, it may be argued, however, that current pharmacokinetic testing methodologies are not always appropriate for drugs that either display variable absorption patterns or nonlinear pharmacokinetics. Given the potential consequences of nontherapeutic equivalence, increased clinical vigilance when switching between formulations would seem prudent. In addition, patients should be discouraged from frequent switching between generic products, and increased communication among patient, pharmacist, and physician should be encouraged.

Advantages of increased use of generic drugs

Generic AEDs should be used more often because they save money, because their contribution to the overall variability in clinical response is negligible when the FDA criteria for bioequivalence are applied, and because the use of approved generics is safe.

The main argument for an increased use of generic AEDs is economic. There are FDA-approved generic products for most AEDs, and these are generally considerably less costly than the brand AEDs. A switch in prescribing from brand to generic AEDs can thus be assumed to reduce health-care costs and save money that could be used for other needs. For this reason, some countries have in recent years introduced mechanisms for mandatory substitution of brand products with less expensive generics. As an example, there is since 2002 in Sweden a compulsory generic substitution (with some exceptions) at the pharmacy level to the cheapest available approved generic. The introduction of this law was immediately followed by a significant fall in the price of generics, as well as brand products of drugs for which the patent had expired. It has been estimated that the total costs for prescribed drugs in the country decreased by 15% as a result of this law, and that this saving resulted partly from a decline in prices because of market competition and partly because of increased use of generic drugs (Lundin et al., 2007). If an increased use of generics reduced the costs by only 10%, this would save 14 billion US dollars (USD) annually only in the USA, conservatively assuming a cost for prescription drugs of 141 billion USD per year. The potential economic gains through generic prescribing are thus huge.

There have been concerns that an increased use of generic AEDs would make brand pharmaceutical companies reluctant to invest in the development of new AEDs, as their return on the brand AEDs will be reduced after the patents have expired. However, the opposite effect is at least as likely: the fact that their old brand AED is under pressure from cheaper generic products can serve as an incentive for the development of the much needed new and more innovative compounds.

Pro: The Argument for Increased Use of Generic AEDs Torbjörn Tomson

Bioequivalence of generics and variability in drug concentrations with unchanged brand

The arguments for an increased use of generic AEDs rest on the assumption that these are essentially similar to the brand AEDs in clinical effects. This assumption is based on bioequivalence criteria that have been defined by FDA and the European drug agency and that have to be met for a generic drug to be approved. Some have questioned whether these criteria, described in detail above, are sufficiently strict for drugs used to treat epilepsy (Heaney & Sander, 2007). The FDA criteria require that the 90% confidence intervals of bioavailability have to fall within the 80–125% limits of the brand. The difference between mean plasma concentrations is considerable smaller, and typically differs between the generic and the brand by no more than 5–7% (Perucca et al., 2006). In fact, on examination of approved generics, the FDA found a mean bioavailability difference between the generic and the brand product of only 3.5% (Bialer, 2007).

These differences should be seen in the context of the overall variability in AED plasma concentrations in patients on continuous treatment with unchanged brand medication, and this can be substantial. The individual coefficient of variation (CV) of consecutive plasma concentrations was calculated from three or more visits without dose (or brand) change in outpatients from the Comprehensive Epilepsy Program of Minnesota (Leppik, 1988). The mode CV for carbamazepine was 23.3% (n = 206), for phenytoin 25.2% (n = 192), and for valproate 27.1% (n = 181). In the elderly, phenytoin plasma concentrations have been reported to vary over time as much as two- to threefold despite unchanged doses among nursing home residents with a secured drug intake (Birnbaum et al., 2003). There are many factors that contribute to this variability in patients with unchanged treatment. However, it is probably less well known that the same brand AED may change pharmacokinetic properties over time as a result of alterations in excipients and the manufacturing process. The thus modified brand has to pass the same bioequivalence studies and criteria as a generic, but unlike generic substitution, the prescribers and the patients are rarely aware of the possible change in the composition of the brand product.

Hence, for various reasons, there is a substantial intra-patient variation in plasma concentrations of AEDs over time with unchanged treatment, and with the current criteria for bioequivalence the contribution of generic substitution to the overall variation in therapeutic response is negligible (Richens, 1997).

Clinical consequences of generic substitution

There is a widespread concern about the risks with generic substitution among physicians as well as patients, and that the cost savings may be outweighed by the cost of adverse consequences (Haskins et al., 2005; Crawford et al., 2006; Berg, 2007). These reports are, however, surveys of opinions rather than of facts, and the results could reflect that the regulatory bodies have been less successful in explaining their position than have the marketing activities of stakeholders in brand AEDs. A recent study from Canada analyzed switchback rates from generics to brand AEDs in comparison with antihyperlipidemics and antidepressants (Andermann et al., 2007). Switchback was defined as switching a patient from the branded drug to generic, and then back to the generic. The switchback rates were substantially higher for AEDs than for non-AEDs. Although the authors interpretation is that the high rates for AEDs may be associated with adverse clinical consequences resulting from switching from branded to generic AEDs, the reasons for switching back was not investigated. The results may thus just as well reflect expectations and attitudes and differences between epilepsy patients and the other groups in this respect.

Although there are many uncontrolled case reports and studies reporting increase in seizure frequency or adverse events after switching to a generic AED, the causal relationship is unclear. The FDA was unable to document a single example of therapeutic failure when an FDA-approved generic product substituted the corresponding brand drug (Perucca et al., 2006). The so far only randomized study comparing a brand and a generic AED was a small open-label crossover study of Depakene versus generic valproic acid (Vadney & Kraushaar, 1997). This study found no significant differences in seizures or blood levels between the treatment arms.

Hence, although there is clearly a widespread concern about the risks associated with generic substitution, the evidence for adverse clinical consequences are lacking. In this context it is interesting to consider the attitudes toward moderate alterations in AED plasma levels by other mechanisms such as pharmacokinetic interactions. The following representative citation from an summary of product characteristics (SPC) of a brand AED illustrates a different attitude to changes in AED plasma concentrations caused by drug–drug interactions, although exceeding those that might be a result of generic substitution: “15 mg olanzapine reduced the AUC and Cmax of lamotrigine by an average of 24% and 20%, respectively. Therefore an effect of this magnitude is generally not expected to be clinically relevant” (Lamictal SPC; GlaxoSmithKline, Updated Sept. 17, 2007). Furthermore, as another example, there is no general advice in the SPC against combining sequential contraceptive pills and lamotrigine, or any suggestions for dosage adjustments for the pill-free week, although lamotrigine plasma concentrations may increase by up to 100% at the end of that week (Christensen et al., 2007).


In summary, following the strict criteria applied by the FDA, the contribution of generic substitution to the overall variability in AED plasma concentrations is minimal, and evidence for adverse clinical effects is lacking. Potential economic savings, on the other hand, are considerable. A conservative recommendation for the use of generic AEDs, respecting the concerns of seizure-free patients, could be:

  • 1Prescribing a generic AED is rational in a newly diagnosed patient.
  • 2Generic substitution could be justified and acceptable in patients that are not fully controlled.
  • 3Generic substitution is best avoided in seizure-free patients.


We confirm that we have read the Journal’s position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.

Conflicts of interest: TT received speaker’s fees and/or research grants from GSK, Sanofi-Aventis, Novartis, Janssen-Cilag, Pfizer, UCB, and Eisai. BG has received honoraria and/or research grants from GSK, UCB, and Abbott Labs.