Network meta-analysis and the comparison of efficacy and tolerability of anti-epileptic drugs for treatment of refractory focal epilepsy

We read with interest the recent paper on network meta-analysis [1] in which efficacy and tolerability of selected anti-epileptic drugs in people with refractory focal epilepsy were compared. Whilst network meta-analysis may be a powerful tool, the validity of the results must depend on the selection of studies included. In this respect, we are curious about several methodological points. The authors include some studies but not others that are of the same nature. The authors deem as appropriate doses those that most experienced epilepsy clinicians would consider unusual and the results that emerge would be considered misleading by most experienced epilepsy clinicians. Only one of the four drugs the authors suggest have efficacy and tolerability in focal epilepsies would be accepted as such by experienced epilepsy clinicians. Their recommendation of vigabatrin in this category is most surprising, given the significant adverse effects associated with its use. The paper illustrates the importance of ensuring that the results of any complex statistical process should always be checked against actual clinical experience. The authors selected for comparison only people receiving doses of the target drugs which they deemed appropriate. Despite the authors’ claim that they chose the selected doses based on their clinical experience, the choice seems arbitrary, is out of keeping with clinical experience and is not based on any stated evidence. For instance, only someone with little clinical experience of oxcarbazepine would consider doses over 1800 mg day−1 as appropriate. The dose of 2400 mg is poorly tolerated by most people. It is also unlikely that many people would tolerate doses of topiramate over 300 mg day−1. Several relevant studies were not included, without clear explanation. For instance, a key lacosamide study [2] was not included. The same apply for studies with topiramate [3] and levetiracetam [4]. Two relevant lamotrigine studies [5, 6] were also excluded. One could suggest that these were excluded as they were crossover studies, but then other crossover studies were included, such as a levetiracetam study [7]. It is imperative that consistent criteria are applied for study selection if this type of analysis is to be meaningful. The authors conclude that in their mixed-treatment network meta-analysis, levetiracetam, vigabatrin, sodium valproate and gabapentin emerge as the anti-epileptic drugs with the best combination of short term efficacy and tolerability. As experienced epilepsy clinicians having seen many thousands of patients, we do not consider that valproate and gabapentin should be considered as first line anti-epileptic drugs for people with focal epilepsy, whilst any epilepsy clinician would know that the risk of visual field loss with vigabatrin is far too high to contemplate its regular use. The authors’ recommendation on valproate appears to emanate from one old study comparing vigabatrin with valproate, in which a small number of people with epilepsy not controlled by carbamazepine were given either valproate or vigabatrin [8]. It is not clear why better quality data [9] were omitted from the analysis. For these reasons, the conclusions cannot be considered particularly helpful. The need for answers to the questions the authors set out to address remains. One important thing to remember is that epilepsy, a symptom-complex, is not a single disease but a collection of many different conditions for which the current clinical trials paradigm is completely inappropriate [10, 11].