¹Claude Bernard University, Debrousse Hospital, Rene Descartes University, St Vincent de Paul and Necker-Enfants Malades Hospitals, Biocodex, Louis Pradel Hospital, Lyon, France

Abstract

Background: Because meta-analysis of clinical trials allows evaluating more precisely the treatment effect and the heterogeneity between original studies, it has been recommended for small clinical trials. We performed a meta-analysis on individual patient data of randomized clinical trials (RCTs) comparing STP and placebo in SMEI.

Methods: We searched Medline from 1966 to 2003, Cochrane data base and contacted Biocodex for unpublished RCTs. We used a fixed effect model to summarize the relative risk of seizures and a logistic model to evaluate the influence of age, sex, and study center on treatment effect.

Results: Among 542 abstracts screened, two studies1 published, 1 unpublished, met the inclusion criteria. Individual patient data were available for 64 children aged between 3 and 20y. Responders were defined as children having more than 50% reduction in seizure frequency on STP added to clobazam and valproate for 2 months. STP multiplied by 10 the proportion of responders (CI: 2.6, 39), by 7.3 (1.1, 50) in study 1 and 14 (2.1, 98) in study 2. No heterogeneity was detected between included studies (P Heterogeneity= 0.63). Moreover, it was not clear whether the treatment effect was influenced by sex and age.

Conclusions: METASTIRI shows that meta-analysis on individual patient data appears feasible in the context of small RCTs; it increases the precision of the effect size and the generalizability of results.