Antiepileptic Drugs, 5th Edition

General Principles


The Development of Antiepileptic Drugs: Regulatory Perspective

Russell Katz MD*

* Division of Neuropharmacological Drug Products Center for Drug Evaluation and Research, United States Food and Drug Administration, Washington, DC

Dr. Katz is with the U.S. Food and Drug Administration (FDA). The views expressed in this chapter are his own and do not reflect an official statement from the FDA.


In the United States, the legal bases for the use in people of drugs not approved for marketing (investigational drugs) and the standards for drug approval for marketing are contained in The Federal Food, Drug, and Cosmetic Act (the Act), a statute passed by Congress in 1938 and amended in 1962 and several times thereafter. To enact the provisions of the Act, the U.S. Food and Drug Administration (FDA), the governmental agency charged with regulating human research with investigational drugs and making decisions about drug approvability, has the authority to promulgate regulations that describe the various requirements of drug development and approval. These regulations, although not enacted by duly elected representatives (Congress), are the product of a detailed process of public notification and rule making, and have the force of law. Although the Act and regulations (and various documents written by the FDA and other organizations that provide nonbinding guidance to drug sponsors) together describe in detail all aspects of drug development and approval, this chapter focuses primarily on issues related to the requirements for assessing the effectiveness of proposed drug products, with particular emphasis on issues related to the demonstration of effectiveness for antiepileptic drugs (AEDs). A discussion of the many other issues involved in drug development in general can be found elsewhere (1).

The sine qua non for drug approval in the United States is a demonstration that the proposed drug product is effective; regardless of how safe a treatment may be, if it is not effective, it may not be approved. The standard applied by the FDA in determining effectiveness is defined in the Act as “substantial evidence” of effectiveness for the conditions described in approved product labeling. The Act defines substantial evidence as:

…evidence consisting of adequate and well-controlled investigations, including clinical investigations, by experts qualified by scientific training and experience to evaluate the effectiveness of the drug involved, on the basis of which it could be fairly and responsibly concluded by such experts that the drug will have the effect it purports or is represented to have under the conditions of use prescribed, recommended, or suggested in the labeling or proposed labeling thereof.

It is critical to recognize that the approval of a product is inherently linked to the labeling approved with it. That is, the labeling must accurately reflect what is known about the drug at the time of approval, and must include, among other things, a description of the patient population in whom the drug is considered to be effective, a description of the symptoms/signs benefited by the treatment, and a description of the effective dose(s) and of the risks associated with its use.

It also is critical to recognize that traditionally, the plural investigations has been interpreted to mean at least two such trials, thereby incorporating the well accepted scientific principle of independent replication or corroboration of experimental findings before they can be generally accepted. Typically, the FDA is expected to take an initial action on a new drug application within 10 months of its submission; for drugs that represent an advance over existing therapies, the initial action is supposed to occur within 6 months of the submission of the application.

In 1997, the Act was amended to include a new definition of substantial evidence, which permitted the FDA to make a finding of substantial evidence on the basis of a single adequate and well controlled clinical investigation and “confirmatory evidence.” The relevant language in the amended Act is as follows:

If the Secretary determines, based on relevant science, that data from one adequate and well-controlled clinical investigation and confirmatory evidence (obtained prior to or after such investigation) are sufficient to establish effectiveness, the Secretary may consider such data and evidence to constitute substantial evidence….


Although Congress gave no indication when this new standard should be used, or what it considered “confirmatory evidence,” the FDA has produced a document that gives guidance on both of these points (2). In general, the FDA may rely on a single study to establish substantial evidence of effectiveness when that trial has demonstrated an important effect on mortality or irreversible morbidity or prevention of a disease with a potentially serious outcome, and where independent confirmation of the result would be difficult or impossible, for ethical or practical reasons. It is understood, therefore, that this new standard applies only in unusual circumstances; typically, drug approval still requires that substantial evidence consist of at least two, independent clinical trials demonstrating that the drug has the effect claimed for it in labeling.

Although the Act itself does not further describe the specific attributes of clinical investigations that the FDA may rely on as sources of evidence of effectiveness, these are described in detail in the regulations found at Title 21 of the Code of Federal Regulations, section 314.126 (21 CFR 314.126, “Adequate and well-controlled studies”). This section describes five specific trial designs (trials may incorporate aspects of several of these designs):

  1. Placebo concurrent control. In this design, some patients are treated with investigational drug, whereas others are treated with placebo; treatment assignment invariably is random, and usually both patients and investigators are unaware of the treatment assignments (double-blind).
  2. Dose-comparison concurrent control. In this design, patients are assigned to receive one of several doses of investigational drug; as in the placebo concurrent control, this design usually is random and double-blind.
  3. No treatment concurrent control. In this design, some patients receive the investigational drug, whereas others receive no treatment; as in the other designs, treatment usually is randomly assigned.
  4. Active treatment concurrent control. In this design, some patients are assigned (again, usually randomly) to the investigational drug, whereas others are assigned to treatment with a standard drug.
  5. Historical control. In this design, all patients are treated with the investigational drug, and their responses are compared with “experience historically derived from the adequately documented natural history of the disease….”

Although there are clinical circumstances in which each of these study designs is appropriate, because the natural history of most seizures in most patients with epilepsy is highly variable and unpredictable, well controlled studies of putative AEDs always must include a concurrent control, usually placebo or an “ersatz” placebo (see later).

The 1997 amended Act also introduced several additional provisions that have the potential to affect the development and approval of new treatments for epilepsy.

The Act, under its so-called Fast Track provisions, now provides for the approval of treatments for serious or life-threatening illnesses that demonstrate the “potential to address unmet medical needs” on the basis of substantial evidence of effectiveness on a surrogate marker that is reasonably likely to predict the clinical benefit of interest. Although this provision was first introduced into the Act in 1997, a similar provision has existed in the regulations since 1992. In those regulations, referred to as the Accelerated Approval regulations, the FDA for the first time codified its position that treatments for serious or life-threatening illnesses that provided meaningful benefits over existing treatments could be approved on the basis of an effect on a surrogate end point that “is reasonably likely, based on epidemiologic, therapeutic, pathophysiologic, or other evidence, to predict clinical benefit….”

validated surrogate marker is a laboratory test or other measure that is not immediately linked to the patient's clinical symptoms or signs, but that is correlated with the patient's condition and its response to the applied treatment predicts the patient's clinical response on an appropriate outcome. For example, blood pressure is a validated surrogate marker for stroke, heart attack, and other serious outcomes because, in general, a hypertensive patient's blood pressure does not, in the short term, correlate with clinical symptoms (outside the far extremes of blood pressure, the patient is asymptomatic), but drug-induced decreases in elevated blood pressure have been shown to correlate with and predict decreasing rates of the long-term outcomes listed previously. An antihypertensive treatment therefore can be approved on the basis of its effect on blood pressure, not patient symptoms, because this effect has been shown to confer a clinical benefit in the future.

The law now, however, permits approval on the basis of a drug's effect on an unvalidated surrogate marker, one that can reasonably be expected (but not yet demonstrated) to predict the clinical outcome of interest (again, the regulations have permitted such approval since 1992). This new provision does require that an attempt be made to validate this surrogate after the drug is approved; if the surrogate is shown not to predict the clinical benefit of interest, it may be withdrawn from the market under expedited procedures.

The advantage of relying on an effect on a surrogate outcome as a basis for drug approval is that if the drug's desired effect is on a clinical outcome likely to occur far in the future after treatment initiation, the trials may be done in a reasonable length of time. In the example of the antihypertensive given previously, the clinical events of interest are expected to occur many years after treatment initiation. The reliance on a drug's effect on blood pressure to support approval permits the studies to be practically done. However, there are many potential pitfalls in relying on a drug's effect on an unvalidated surrogate marker as the basis for drug approval (3). Most critically, the fact that the drug has the desired effect on the surrogate may not mean that the drug will have the


desired effect on the clinical outcome; indeed, the drug may have the expected effect on the surrogate, but an unexpected, deleterious effect on the clinical outcome of interest. Nonetheless, there are situations in which reliance on an unvalidated surrogate marker may be acceptable. To date, no AED has been approved on the basis of its effect on a surrogate marker; it is expected that an AED should produce a detectable beneficial effect on the clinical measure of interest (e.g., seizures) before it may be approved. Such effects are demonstrable in studies of reasonable duration, and therefore there is no need to rely on an indirect (surrogate) measure. However, there may be circumstances in which an effect on a surrogate marker may play an important part in the approval of AEDs in the future (see later).

The amended Act also offers an incentive to pharmaceutical sponsors to undertake studies in the pediatric population (defined as patients 16 years of age and younger), a segment of the population traditionally not included in studies performed to gain drug approval. Under this provision, sponsors can gain 6 months of additional marketing exclusivity for a drug (assuming that there is still a period of exclusivity attached to the product) if they perform studies in pediatric patients as requested by the FDA. This program is voluntary, but the prospect of 6 months of additional marketing has prompted many sponsors to conduct adequate studies in pediatric patients with seizures that presumably otherwise would not have been conducted.

However, because this provision is voluntary, and because the FDA had determined that studies in pediatric patients were vital for many drug products known to be used off-label in these patients, shortly after this amendment was instituted in 1997, the FDA adopted, in December 1998, regulations requiring that sponsors perform adequate trials in pediatric patients for those indications for which the drug is being developed in adults (21 CFR 314.55). Although the rule technically applies to all drugs, including drugs already approved, it has been applied primarily to drugs under development or for which new drug applications have been submitted since the rule went into effect in April 1999. Sponsors may submit an application with adult data only, but they must commit to performing the appropriate studies in pediatric patients for the same indications for which the drug is approved in adults in a specific time frame after initial approval; various penalties may be imposed if the studies are not completed in the appropriate time. Pediatric studies are required when gaining the appropriate claim in that population will provide a meaningful clinical benefit (defined as a significant improvement in the treatment, diagnosis, or prevention of a disease compared with other treatments approved for that indication in the relevant pediatric population, and the drug is in a drug class or indication for which there is a need for additional therapeutic options), or there would be substantial pediatric use for the indication and the absence of adequate pediatric labeling would pose a health risk. This requirement for pediatric studies may be waived if the treatment does not meet the criteria for meaningful benefit and substantial use, the sponsor demonstrates that the studies would be impossible or impractical to perform, the use would be unsafe in pediatric patients, or reasonable efforts to produce an appropriate dosage form (if necessary) have failed. Taken together, the 1997 provisions of the Act and the pediatric rule have had a profound effect on the number of adequate studies performed in pediatric patients, and in particular in pediatric patients with epilepsy.


Until the early 1990s, the last new chemical entity approved in the United States for the treatment of epilepsy was valproic acid, approved in 1978 for the treatment of absence seizures and multiple seizure types that include absence seizures. In 1993, however, two new chemical entities were approved, felbamate and gabapentin, followed in 1994 by lamotrigine. Including these three, as of this writing, nine new chemical entities have been approved for the treatment of various seizure types, as well as various new dosage forms and routes of administration for several older AEDs (e.g., controlled-release forms of carbamazepine, injectable forms of valproic acid, rectally administered diazepam). In addition, many of the newer drugs initially approved for a specific indication (most new AEDs are initially approved for the treatment of partial seizures), as well as some of the older drugs, have subsequently been approved for additional seizure types and populations (e.g., primary generalized seizures, Lennox-Gastaut syndrome).

Starting with the newly approved drugs in the early 1990s, several issues have been considered important in the approval process.


As noted earlier, the approval of a drug product in the United States is inextricably linked to the product labeling. Specifically, the labeling must accurately describe the population for whom the drug is useful as well as the effect seen for the drug in the trials that supported approval. Of course, in addition, the label must accurately reflect the known toxicities of the drug. Because of the requirement that the labeling accurately reflect the effects of the drug and the population in whom it has shown to be effective, the labeled indications for newly approved AEDs reflect a number of the specific conditions under which the drugs are studied. Specifically, drugs are approved as treatments for the specific seizure types found to be benefited by the treatment, for the specific population studied (adults, or, if studied in the pediatric population, the earliest age shown to be benefited is given), and the conditions of study (adjunctive therapy or monotherapy). Typically, AEDs are studied first in adults with partial seizures whose epilepsy is not adequately controlled and who already are being treated


with several other AEDs. For this reason, a typical indication in product labeling for a newly approved AEDs might state that the drug is indicated as adjunctive therapy in adults with partial seizures.

This approach to labeling a newly approved AED is based on several considerations. First, when the new drug applications for the first three newly approved AEDs were presented to the FDA's Peripheral and Central Nervous Systems Drugs Advisory Committee (a group of non-FDA experts in various areas of neurology impaneled by the FDA and to whom the FDA presents selected issues and from whom the FDA solicits nonbinding advice), the Committee recommended this approach. Although the FDA was, and is, not obliged to take the Committee's advice, it was thought prudent to do so in this case. First, the FDA agreed that studying a drug solely under conditions in which the drug is given against a background of other AEDs could not support a conclusion that the drug would be effective when given alone. Although it may be generally considered that any AED that works when given in conjunction with other AEDs must be effective when given alone, there is little to no evidence from adequate controlled trials that establishes this for most AEDs. Even if one were to conclude (without evidence) that this must be true, the effective dose when the drug is given as monotherapy would not be known; this could be reliably known only if it was studied in an adequate trial. Even if a therapeutic plasma range were known for the drug under adjunctive conditions (although this is not known for any of the recently approved AEDs), establishing a dose of the drug that results in these plasma levels when the drug is given as monotherapy (e.g., from a pharmacokinetic study) still would not be adequate to establish that this plasma range confers seizure control when the drug is given alone. For these reasons, drugs studied only under adjunctive conditions are specifically and explicitly indicated as adjunctive treatment.

Similarly, drugs studied only in adults are not indicated as being effective in pediatric patients. The FDA has not been willing to extrapolate findings from adults to pediatric patients. At the moment, if a drug sponsor wishes to obtain approval for an indication in pediatric patients, they must perform an adequate study in the relevant subsegments of the pediatric population or submit other evidence that would convince the FDA that the drug's effect in adults implies that it also is effective in pediatric patients. Such an approach would need to establish that the disease (in this case, a particular seizure type) is the same in both populations, that the two populations are known to respond the same way to the applied treatment, and that there exists sufficient pharmacokinetic information in the relevant pediatric groups to support dosing recommendations. Although it may appear logical to assume, for example, that partial seizures in the adult are “the same” as partial seizures in pediatric patients, it is not immediately obvious that the pathophysiologic events leading to a partial seizure in the developed brain are identical to those causing a partial seizure in the developing brain, even though the two events may appear phenomenologically similar. Further, given our lack of knowledge about (a) these pathophysiologic events, and (b) the complete mechanism of action of any AED, it is not obvious, in the absence of controlled trial data, how one could reasonably conclude that a drug will work the same way in the two populations. Finally, even if one were to conclude that these first two points could be assumed, one could not know what an effective dose would be in the pediatric patients without data from a controlled trial in pediatric patients. For these reasons, the FDA requires studies in pediatric patients to support a claim in this group.

Finally, the effects of a drug on a given seizure type are not ordinarily considered evidence of effectiveness for other seizure types. As with the aforementioned case with pediatric patients, given our lack of complete understanding about the biologic events underlying the cause of any specific seizure type and occurrence and our lack of detailed understanding of the mechanism of action of any of the available AEDs, one could not be certain of the effects of a drug on a seizure type not studied based on its effects on a different seizure type. Typically, therefore, for a sponsor to gain approval for a drug for a specific seizure type, the drug's effect on that seizure type must be documented in at least one adequate clinical trial. Specifically, if a drug is approved initially to treat, for example, partial seizures (on the basis of at least two clinical trials), it may be approved for another seizure type (e.g., primary generalized seizures) on the basis of a single additional, well controlled trial. This requirement for only a single trial in the latter seizure type is based on the view that adequate evidence already established the drug's effect as an AED; the FDA ordinarily considers this prior demonstration as “lending strength” to the finding from the single trial in the new seizure type, and concludes that substantial evidence of effectiveness has been presented for the new indication. There may, however, be instances in which the new seizure type is considered so fundamentally different from the previously approved seizure type that two trials may be required for approval of the former indication.

Clinical Trial Issues

Although theoretically any of the previously described study designs could be used to demonstrate an effect of a proposed AED, in practice only a relatively few are considered capable of yielding the valid evidence on which to base a finding of substantial evidence.

The overarching principle applied by the FDA in interpreting clinical trials with AEDs to date has been the principle that a (statistically significant) difference between the proposed treatment and the control treatment (which need not be placebo) must be demonstrated for the trial to be unambiguously interpretable. Such an outcome, in which a difference is seen between the proposed and control treatments, is the only outcome that can be interpreted without


reliance on data external to the trial itself, making it the most powerful and reliable sort of evidence of effectiveness.

That this is so can be seen from an examination of the results of a trial in which no difference is seen between a proposed AED and an active control drug. This outcome is subject to two interpretive problems, only one of which can be potentially addressed without resort to information outside of the trial itself.

The problem with this outcome that is potentially addressable is that the new drug may, in fact, be worse than the old drug, but the study was not adequately powered (did not have enough patients) to detect this difference. This problem might be able to be addressed if sufficient numbers of patients were enrolled.

The more fundamental interpretive problem with such an outcome is that to support the conclusion that the new drug is effective, one must assume that the old (control) drug was effective, in this specific trial (4,5). Although this may seem to be a reasonable assumption if the old drug is approved, and therefore known to be effective, for the seizure type under study, it is not always true that drugs known to be effective are effective at all times in all patients, and, specifically, it is not known if it was effective in the particular trial performed. There are many examples of cases in which drugs known to be effective have been shown not to be distinguished from a placebo in a given clinical trial (4).

As noted, then, for an active controlled trial that does not demonstrate a difference between the treatment groups to be interpreted as demonstrating the effectiveness of the new drug, it must be concluded that the old drug was effective in that particular trial. However, as we have just seen, that may not always be true. The only way in which the effectiveness of the old (active control) drug in a particular trial can reasonably be assumed to be true is if it can be known with great certainty that the patients assigned to the active control would not have responded as they did had they been assigned to placebo. This, in turn, can be known reliably only if the natural history of the untreated condition is known with great precision, or if there is a robust database in which the active control has been shown repeatedly to be superior to placebo by a given amount (this is why active controlled trials that do not detect a difference between treatments often are referred to as a subset of historical controlled trials, the weakest source of evidence). Unfortunately, for most types of seizures, and for most, if not all, AEDs used as active controls, both types of information (natural history and response rates) are unavailable. Because this information is not available, active controlled trials of AEDs that do not demonstrate a difference between treatments are not interpretable, and are not relied on by the FDA as being capable of providing evidence of effectiveness.

On the other hand, an active controlled trial in which the new treatment is seen to be superior to the active control is easily interpreted as supporting effectiveness, assuming that the active control did not make the control patients worse than they would have been without treatment. Although this assumption also may be untestable in a given trial, in most cases such an assumption is reasonable because it is quite rare that a drug known to be effective will make a separate cohort of patients worse than they would have been had they not been treated; the point is that it usually is unknown if an effective treatment will be effective at all times in all patient samples, or it is known that not infrequently many effective drugs are not effective in a given patient sample in a controlled trial. It is critical to point out, however, that such an outcome (in which patients assigned to the new drug perform better than those assigned to the old drug) ordinarily cannot support the conclusion that the new drug is superior to the old drug. For such a conclusion to be reached, one would need to be sure that the new drug was compared with the most effective dose of the old drug. Such a trial ordinarily would compare several doses of the new drug with the full range of available doses of the old drug, to ensure that the comparison was “fair”; such trials are difficult to do, and have rarely been done successfully.

Given, then, that a difference between treatments is necessary for a study to be accepted as contributing to a finding of substantial evidence of effectiveness, a number of clinical trial designs are considered acceptable by the FDA.

The most common study performed is the so-called placebo add-on design. In this study, patients on one to three concomitant AEDs are randomized to receive new drug or placebo added on to their current AED regimen. Such studies usually maintain patients on a dose of the experimental treatment thought to be therapeutic, or placebo, for 3 months, and the frequency of seizures is compared between the two groups. Again, this design is not capable of determining the superiority of the new drug to the concomitant AEDs being taken by the patients; it simply is capable of supporting the conclusion that the new drug, when added on to other AEDs, is superior to placebo added on to other AEDs; this is sufficient to establish that the drug has an antiseizure effect (when given as adjunctive therapy), and, if replicated, can support a finding of substantial evidence of effectiveness. In this study design, it is critical to ensure that the plasma levels of the concomitant AEDs are not systematically elevated in the drug-treated group compared with those in the placebo group (as might happen secondary to a pharmacokinetic interaction). If the plasma levels were significantly elevated, any effect seen might be due to the higher levels of the concomitant AEDs, and not to any intrinsic activity of the new drug.

Many other designs have contributed to a finding of substantial evidence of effectiveness, depending on the clinical condition under study. All are useful if, as noted, they can detect a favorable difference between the patients assigned to the investigational drug and those assigned to the control treatment.

A number of designs have been used to determine the effectiveness of an AED when used as monotherapy. Monotherapy studies pose a number of unique problems, mainly because some authors consider it unethical to treat


patients with seizures with placebo alone (6). In some cases, however, patients have been randomized to receive different fixed doses of the experimental drug, with the goal of showing superiority of the high dose to the low dose. Such a study can be considered ethically acceptable if there is uncertainty about which dose (if any) of the drug is effective. For example, intravenous lorazepam was studied in patients with status epilepticus; in this study, patients were randomized to receive one of two doses. In this trial, the high dose was superior to the low dose, permitting a conclusion that the drug (at the high dose) was effective. Similarly, patients have been randomized to investigational drug or low-dose active control as monotherapy, and in these studies a finding of superiority of the investigational drug has supported a conclusion that the investigational drug is effective (but again, it does not support the conclusion that the investigational drug is superior to the active control). Some authors believe that this design is unethical because patients are randomized to a dose of the active control that is known not to be optimum, and that therefore the control patients are not being adequately treated; others feel that the design is ethical because they believe that the low dose of the active control is sufficient to prevent any serious seizure event (e.g., status), and it may, in any event, be effective (although such an outcome probably would make the study uninterpretable; see earlier).

A monotherapy design in which patients are treated with placebo alone has been used in patients with severe refractory epilepsy who are being evaluated for surgical treatment. In these patients, an attempt is made to remove all AEDs to evaluate further their potential for surgery. Given that these patients are having all of their medications removed for a period as part of their presurgical evaluation, investigators have randomized these patients to drug or placebo for brief periods (usually no more than 10 days) and assessed their response. Although this design is capable of detecting an antiseizure effect of some drugs (a therapeutic dose of drugs used in this paradigm must be able to be achieved quite rapidly), the study period is so brief that the FDA does not currently consider such a study as providing substantial evidence of effectiveness in monotherapy.

A placebo control-only design also has served as the basis for a finding of effectiveness in patients with a history of only a few seizures or in patients newly diagnosed with epilepsy. These patients would be expected to have few seizures, and would not be expected to suffer serious harm if not treated with active drug for a defined period. In addition, studies in patients not well controlled on several AEDs in which patients are randomly assigned to monotherapy with an investigational drug or placebo (or low-dose AED as a control) have been performed. In these studies, the usual primary measure of effectiveness is the time to meeting various exit criteria (several measures of increasing seizure activity). Some find this design acceptable because it has the advantage of being interpretable, while protecting the patients' safety because they are withdrawn from the study when their seizures become worse (i.e., they meet exit criteria).

Another proposed monotherapy design involves treating patients with an investigational drug in open-label, uncontrolled conditions for a given duration (may be many months), and then randomly assigning them to continue on drug or placebo (randomized withdrawal design). This design has the additional advantage of permitting an assessment of long-term seizure control; if the drug is superior to placebo in the randomized phase, this implies that it had been effective during the open-label, uncontrolled, prerandomization phase.

However, as noted earlier, there is increasing concern among many (although certainly not all) investigators that the treatment of patients with seizures with placebo alone is unethical, and it is becoming increasingly difficult for sponsors to perform adequately designed monotherapy studies. The design of ethically acceptable but scientifically sound and interpretable monotherapy studies represents one of the great challenges for the future.

Finally, many investigators have expressed interest in developing trial designs that are capable of determining whether a proposed treatment can prevent or cure epilepsy. There are many unanswered questions about what features should be incorporated into such studies, including the population to be studied, the duration of such a study, and the appropriate outcome measures (e.g., it may be necessary for a sponsor to show not only that patients have no seizures, but that their electroencephalogram normalizes), including perhaps surrogate markers (e.g., imaging, structural, or functional). Although these are difficult questions, drugs developed to prevent or cure epilepsy, and the designs that would adequately demonstrate these effects, present perhaps the greatest and most exciting challenges for the future of the treatment of the patient with seizures.


  1. Katz R. The introduction of new drugs. In: Gennaro AR, ed. Remington: the science and practice of pharmacy.Easton, PA: Mack, 1995.
  2. U.S. Department of Health and Human Services. Guidance for industry: providing clinical evidence of effectiveness for human drug and biological products.Washington, DC: U.S. Department of Health and Human Services, 1998.
  3. Fleming TR, DeMets DL. Surrogate end points in clinical trials: are we being misled? Ann Intern Med1996;125:605-613.
  4. Temple RJ, Ellenberg SS. Placebo-controlled trials and active-control trials in the evaluation of new treatments. Ann Intern Med2000;133:455-463.
  5. Leber P. Hazards of inference: the active control investigation. Epilepsia1989;30[Suppl 1]: S57-S63.
  6. Chadwick D, Privitera M. Placebo-controlled studies in neurology: where do they stop? Neurology1999;52:682-685.