Not all that glitters is gold: A guide to the critical interpretation of drug trials in epilepsy

Abstract

Clinical trials represent the best source of evidence on which to base treatment decisions. For such evidence to be utilized meaningfully, however, it is essential that results are interpreted correctly. This requires a good understanding of strengths and weaknesses of the adopted design, the clinical relevance of the outcome measures, and the many factors that could affect such outcomes. As a general rule, uncontrolled studies tend to provide misleading evidence as a result of the impact of confounders such as regression to the mean, patient-related bias, and observer bias. On the other hand, although randomized controlled trials (RCTs) are qualitatively superior, aspects of their execution may still decrease their validity. Bias and decreased validity in RCTs may occur by chance alone (for example, treatment groups may not necessarily be balanced for important variables despite randomization) or because of specific features of the trial design. In the case of industry-driven studies, bias often influences the outcome in favor of the sponsor's product. Factors that need to be carefully scrutinized include (1) the purpose for which the trial is conducted; (2) potential bias due to unblinding or lack of blinding; (3) the appropriateness of the control group; (4) the power of the study in detecting clinically relevant differences; (5) the extent to which eligibility criteria could affect outcomes and be representative of routine clinical practice; (6) whether the treatments being compared are used optimally in terms of dosing, duration of treatment, and other variables; (7) the appropriateness of the statistical comparisons; (8) the clinical relevance of the outcome measures and whether all key outcome information is reported (for example, responder rates in completers); and (9) potential bias in the way results are presented and discussed. This article discusses each of these aspects and illustrates the discussion with examples taken from published antiepileptic drug trials.

Keywords: Antiepileptic drugs; Clinical trials; Epilepsy; Interpretation; Randomized trials.

Figure 1

Responder rates (proportions of patients with at least 50% reduction in primary generalized tonic‐clonic seizure frequency compared with baseline) in two placebo‐controlled, adjunctive‐therapy RCTs of topiramate and lamotrigine in patients with primarily generalized tonic‐clonic seizures.28, 29 Despite use of a very similar design in both trials, there was a prominent difference in responder rates in the groups assigned to placebo treatment. Epilepsia Open © ILAE

Figure 2

Retention in the trial (a combined measure of efficacy and tolerability) in two double‐blind RCTs comparing the outcome of treatment with lamotrigine and carbamazepine (CBZ) monotherapy in patients aged 65 years and over with newly diagnosed epilepsy with onset in old age.55, 56 Outcome with lamotrigine was similar in the two trials, whereas outcome on carbamazepine was better in the trial that used the controlled‐release formulation. Both trials enrolled very similar populations and used identical dosing schemes. Duration of follow‐up was longer for the trial that used controlled‐release carbamazepine (40 vs. 24 weeks). Reproduced from Perucca53 with permission. Epilepsia Open © ILAE

Figure 3

Relationship between percent seizure reduction (vs. baseline) during the last 12 weeks of a 28‐week placebo‐controlled adjunctive‐therapy trial and mean change in health‐related quality of life (QOLIE‐89 score).67 A significant improvement in quality of life was found only for patients who achieved complete freedom from seizures. Epilepsia Open © ILAE

Figure 4

Responder rates (proportion of patients with >50% decrease in seizure frequency compared with baseline) and rates of premature discontinuation from the trial in a placebo‐controlled adjunctive‐therapy trial of oxcarbazepine (OXC) in a total of 694 patients with focal seizures.70 For oxcarbazepine‐treated groups, most premature discontinuations were due to adverse events. Because of the application of last‐observation‐carried‐forward analysis, the proportion of responders at the 2,400‐mg dose was lower than the proportion of patients who discontinued prematurely owing to adverse events. Epilepsia Open © ILAE