Statistics from Altmetric.com
This letter concerns the recently published study by Thompsonet al,1 reporting the authors' findings on cognitive effects with topiramate. Firstly, I want to correct the authors' mischaracterisation of a review paper of mine. The authors state that “the literature on antiepileptic drugs ... emphasised positive psychotropic effects,” referencing only a 1998 review in which I discussed cognitive and psychotropic effects of antiepileptic drugs. Although I mention some positive effects, I also discussed negative effects, including studies from my own centre that have shown significant negative psychotropic and cognitive drug effects.
Secondly, I provide some perspective on the report of Thompsonet al of clinically significant cognitive declines in 18 patients treated with topiramate as adjunctive therapy. The authors correctly conclude that “caution is warranted in the interpretation of the findings due to methodological limitations of the study design.” Because their study was retrospective and observational, it is susceptible to considerable subject selection bias. For example, five of the 18 patients were specifically included in the topiramate sample because they reported cognitive effects.
The only way to minimise effects that may bias study conclusions is to conduct a prospective randomised controlled study. Two such studies have recently compared topiramate and valproate as add on therapy to carbamazepine, using comprehensive neuropsychological batteries to objectively measure drug effects.
At the end of 3 months of maintenance therapy, only one of 17 (6%) variables in one study2 and only two of 30 variables (7%) in the other3 were significantly worse for topiramate compared with valproate. For the three variables with statistically significant differences, the mean differences in change scores were modest. Analysis of individual data showed that scores were unchanged or even improved in most patients receiving topiramate and valproate. Statistically significant differences could be accounted for by a minority of patients receiving topiramate in whom scores deteriorated>1 SD from baseline. I suggest that the patients reported by Thompson et al likely represent a similar subgroup of patients.
Physicians should be aware that a subgroup of patients treated with topiramate may experience clinically significant cognitive effects. When these effects occur, they are generally apparent to the patient or family members and can therefore be monitored with routine clinical evaluations. Alternatively, a brief cognitive test (for example, a verbal fluency test or symbol digit modalities test) should easily detect changes of the magnitude reported. In a subgroup of patients, topiramate may need to be discontinued if cognitive effects do not resolve over time with slowed titration or dosage reduction.
The authors reply:
We write in response to the letter of Meador. We concur that his review article discusses both positive and negative psychotropic effects of antiepileptic medication.
Meador comments on the biased nature of our patient sample. However, we think that we adequately emphasised that our study was not randomised or controlled but carried out in a clinical environment. We pointed out that some of the patients had been referred due to cognitive complaints. Furthermore we acknowledged studies which reported no cognitive effects of the drug including the valproate and topiramate study and drew attention to the conflicting findings. We agree that prospective randomised controlled trials are the way to minimise selection bias. However, they are not the be all and end all. Bone marrow aplasia with felbamate and visual field constriction with vigabatrin treatment were not found in randomised controlled trials but by the careful clinical study of patients. This is an analogous situation. Randomised controlled trials are not without their own biases as most will be sponsored by the pharmaceutical industry and it would be naïve to conclude that this does not influence the presentation of the results.
We agree with Meador that the adverse effects of topiramate reported are likely to occur in a minority of patients treated with the drug. However, we think that this may represent a clinically significant number of patients, particularly in those attending tertiary referral centres. Negative effects, however small the numbers, are worthy of reporting and of further exploration. The question our findings raise is not does topiramate have adverse effects but rather why does it have adverse effects in some people?
We agree with Meador's final point and indeed this was one of the intended take home messages of our paper. This is why we chose to submit to a journal with a broad readership who would have much less experience with topiramate. We hoped that our paper would draw attention to a group of patients who should be prioritised for neuropsychological monitoring and highlight the type of measures that could be employed showing that an extensive assessment is not necessary.
We, however, do not think that the cognitive changes experienced would be obviously attributed to topiramate treatment. Most patients in the study were not referred because of cognitive complaints. Six were being seen as part of their presurgical assessment and indeed were not considered good candidates due to their neuropsychological test profiles. For some the cognitive complaints did not occur in association with the introduction of topiramate or with any change in dosage and did not seem to develop until they had been on the drug for several months. For such patients, particularly those with left hemispheric pathology, increases in word finding problems and other verbal difficulties are more likely to be attributed to the underlying pathology and ongoing seizures than to a drug effect. Topiramate is a useful antiepileptic drug but it may lead to adverse cognitive changes and we need to be alert to this.