PURPOSE: Because of the lack of head-to-head adjunctive-therapy trials of antiepileptic drugs (AEDs) in refractory partial epilepsy, meta-analyses of placebo-controlled randomized controlled trials (RCTs) represent a potentially important source of evidence to guide treatment decisions. However, such indirect comparisons raise various methodologic issues that may hamper their relevance. METHODS: All RCTs in adult refractory partial epilepsy were analyzed to assess whether efficacy outcomes are influenced by: characteristics of patients and trials ; use of last observation carried forward (LOCF) analysis; evaluation period (entire period versus maintenance period); and year of publication. A meta-analysis of these AEDs was then performed taking these factors into consideration. KEY FINDINGS: Sixty-three RCTs evaluating 20 AEDs were included. The following variables influenced efficacy estimates: (1) responder rates correlated positively with duration of the entire treatment period (p = 0.038); (2) response to placebo was significantly greater in the maintenance period than in the entire treatment period (p = 0.005); (3) responder rates increased over the years both for AEDs (p < 0.001) and for placebo (p = 0.001); (4) LOCF analysis overestimated responder rates for AEDs (p < 0.001) and for placebo (p = 0.001) compared with completer-based analysis, and the overestimation correlated positively with withdrawal rates (p < 0.001). A meta-analysis of available data showed large differences in efficacy ranking in relation to dose selection and type of analysis, but these were mostly nonsignificant due to statistical power limitations. SIGNIFICANCE: Several methodologic issues hamper the relevance of indirect comparisons of AEDs in the adjunctive-therapy of refractory partial epilepsy. Some of these issues could be overcome by improved standardization in the reporting of efficacy outcomes.