To the Editor:
We thank the authors for their careful review of our evidence-based assessment of new antiepileptic drugs (AEDs), published in Epilepsia and Neurology in 2004 (1–4). The authors have raised some concerns. We attempt to address our methods for this evidence-based report.
1. The authors have made several points regarding the nomenclature for the types of epilepsy that are addressed in our guideline. This is quite understandable, considering that classification is an extremely important aspect of epilepsy care. This is not, however, a guideline on diagnosis and classification. When assessing the evidence to make evidence-based conclusions, one can essentially only “go where the money is.” That is, one can address only the studies that have been performed in a controlled fashion, describe them completely, and reach conclusions. One cannot reperform the studies in the way that one might have designed them oneself. This is a frustrating aspect of evidence-based medicine, but not one that is novel to this particular assessment.
It is certainly the duty of the committee writing the report to try to address methodology, where possible. More important, one must state the conclusions based on the evidence, not based on what one would like the evidence to be. We have made every effort to do this in our parameter, but understandably, this leads to some unusual nomenclature, mimicking the categorizations that were used in the research studies themselves.
2. A major concern of the authors is the inability to distinguish studies performed in idiopathic generalized epilepsy from those performed in partial epilepsy, or in newly diagnosed epilepsy. Of course, we realize that ideal method would consist of studies performed in a single epilepsy syndrome. Our method for this evidence-based review was to assess the data that were available to us.
Very few studies have been performed that were powered to assess syndromes in isolation. Differences were not readily discernible within the studies themselves. For example, in the study by Brodie et al. (5), comparing carbamazepine (CBZ) with lamotrigine (LTG) in newly diagnosed epilepsy, no differences in outcome were found between patients with primary generalized, tonic–clonic convulsions and patients with a focal onset. However, because of small numbers, we chose not to form any conclusions based on this. In our recommendations for future research, we very clearly make a plea for studies to be performed in specific syndromes.
We agree that “mixed seizure disorders” is a term that could be misunderstood. We tried to clarify the meaning to the best of our ability. We also believe that physicians often find themselves in the circumstance of selecting an initial AED before a definitive syndromic diagnosis can be made. In the article by King et al. (6), cited by Panayiotopoulos, epilepsy syndrome was diagnosable in only 47% of patients on clinical grounds, and in 77% with addition of the EEG, in the hands of experienced epileptologists, at first seizure presentation. Thus understanding the outcome in patients who are unclassified is a clinically important concept.
3. The authors make the curious claim that the term new-onsetepilepsy is not of importance, and that “efficacy and safety of AEDs are not determined by how long the diagnosis has been established.” A great deal of recent investigation has regarded the prognosis in patients with newly diagnosed epilepsy versus those for whom AEDs already have failed. The studies seem to provide evidence that time since diagnosis may be as important as epilepsy syndrome in determining outcome of AED therapy (7). We had very little choice in selection of terms, because newlydiagnosed versus refractory was the way patients were defined for the purposes of the studies that were performed.
4. The authors are concerned about a class I designation for a study that in their minds presents “violations of evidence-based clinical management.” The authors, many of whom are from outside the United States, and perhaps are not familiar with the classification scheme of the American Academy of Neurology, may not be aware that the issues that they discussed have no bearing on designation as class I, class II, etc. These classification schemes have the sole purpose of eliminating bias in studies. They do not address whether the study results are clinically valid. Because the particular study that they questioned was appropriately randomized and blinded, no question exists that the assignment of classification was correct. In this blinded study, patients were randomized to topiramate versus either valproate (VPA) or CBZ, based on the choice of the treating physician. The authors correctly point out that 27% of the patients randomized to treatment with CBZ had idiopathic generalized epilepsy.
Randomization to treatment based on physician preference may be considered to be as valid a method as any other, because it may mimic “real life.” The guideline does not specifically say that topiramate (TPM) is effective in either partial or generalized epilepsy.
5. Another concern is that seizure freedom was not used as an outcome measure, and this “might bias toward less effective drugs.” As noted in the guideline, seizure freedom rarely is reported in the literature and therefore could not consistently be used. Even if this number is included, typically insufficient information is provided to determine whether this number is based on completers or includes dropouts. In several examples, high seizure-free outcomes have been reported, but subsequently it has been discovered that many of these “seizure-free patients” had dropped out within days of starting the study.
6. Again, the authors are concerned about imprecise and ambiguous classification. As noted earlier, the recommendations are based on the populations that were included in the trial, whether they represent a precise seizure classification or not. It is unclear where the authors' concerns rest. They claim that “[generalized tonic–clonic seizures] GTCSs are equated with any type of generalized seizures.” This is not the case. Under the category of generalized epilepsy, data are presented for any study that would be included in that category. The only study that was identified was one performed with TPM in GTCSs. It is very clearly stated both in the practice parameter and in the table that accompanies it, that the evidence specifically addresses refractory GTCSs. In addition, a discussion mentions that this may not be translatable to patients with similar syndromes, but not refractory disease. Another example relates to the abstract. The authors have extracted a portion of the sentence, which in its entirety reads, “There is evidence either from comparative or dose-controlled trials that gabapentin (GBP), LTG, TPM, and oxcarbazepine (OXC) have efficacy as monotherapy in newly diagnosed adolescents and adults with either partial or mixed seizure disorders.”
This is consistent with the body of the document, because the statement in no way implies that each drug is effective in both partial and mixed seizure disorders.
7. The authors claim that the QSS and TTA do not provide any recommendations regarding established documentation of harmful effects of certain AEDs. They provide examples such as the proabsence effects of tiagabine (TGB) and OXC. Once again, these suggestions are based on “common knowledge,” rather than on controlled trial data. We had a great deal of difficulty handling adverse events in an evidence-based fashion. Of course, most adverse-event data of this type derives from case reports rather than randomized controlled trials. We did indicate that TGB was associated with spike–wave stupor as an adverse event. Going beyond the statement would have been going beyond the available evidence.
8. The final, and perhaps most damning, claim is that “the QSS and TTA reports make invalid presumptions rather than impartial evidence-based recommendations.” We truly hope that this is not the case. Two examples are provided for this claim. The first is that, when talking about GBP for refractory GTCSs, we suggested that the ineffectiveness of GBP may be a dose effect. This is based on a statement that “in retrospect, it is possible that the dose was too low.” The dose used in this study was 1,200 mg GBP per day, which in refractory epilepsy would seem to be a low dose. However, this did not affect our recommendations whatsoever.
The recommendation simply stated, “There is insufficient evidence to recommend gabapentin for the treatment of refractory generalized tonic–clonic seizures.” The second example was directed at the statement indicating, “The possibility that once an AED has demonstrated efficacy as adjunctive therapy in refractory partial seizures in adults, the AED will demonstrate the same efficacy as adjunctive therapy in children older than 2 years.” The authors do not provide the sentence after this one, which read, “However, trials in pediatric populations remain critically important to establish efficacy in this as well as other pediatric specific epilepsy syndromes.” Again, the statement did not in any way affect the recommendations for use of drugs in children, which were completely impartial and evidence based.
In summary, we fully support the authors' conclusion that clinical trials should insist on precise definitions of populations studied. We cannot agree that our practice parameter was in any way misleading, incorrect, or biased. The authors would like to see recommendations that “reinforce best clinical practice.” Unfortunately, evidence-based guidelines are not based on best clinical practice. Rather they are based on the available evidence.