The new antiepileptic drugs: a systematic review of their efficacy and tolerability

PURPOSE: Gabapentin (GBP), lamotrigine (LTG), tiagabine (TGB), topiramate (TPM), vigabatrin (VGB), and zonisamide (ZNS) are all in use as "add-on" treatment for patients with refractory epilepsy. There have been no comparative randomized controlled trials allowing an evidence-based choice between these drugs. We report a series of meta-analyses of randomized placebo-controlled add-on trials in which these drugs have been tested in patients with partial epilepsy. This work provides an estimate of each drug's efficacy and tolerability compared with placebo. These estimates are compared across drugs to give broad estimates of comparative efficacy and tolerability. METHODS: Trial reports were found by searching Medline, by searching through journals by hand, and by contacting the pharmaceutical industry. The outcomes chosen were the proportion of patients who (a) have a > or = 50% reduction in seizure frequency (50% responders); (b) withdrew from the study (any reason); or (c) reported the following side effects: ataxia, dizziness, fatigue, nausea, or somnolence. Overall odds ratio (OR) with 95% confidence intervals (CIs; 50% responders) or 99% CIs; side effects) were calculated. RESULTS: Twenty-nine trials were included, representing 4,091 randomized patients. The ORs for 50% response (95% CI) were GBP, 2.29 (1.53-3.43); LTG, 2.32 (1.47-3.68); TGB, 3.03 (2.01-4.58); TPM, 4.07 (2.87-5.78); VGB, 3.67 (2.44-5.51); and ZNS, 2.7 (1.36-4.47). ORs for discontinuation were GBP, 1.36 (0.75-2.49); LTG, 1.19 (0.79-1.79); TGB, 1.81 (1.21-2.70); TPM, 2.56 (1.64-4.00); VGB, 2.58 (126-5.27); and ZNS, 4.23 (1.71-10.49). CONCLUSIONS: We have clear evidence that each of these drugs is better than placebo at preventing seizures. When results are compared across drugs, the confidence intervals overlap, and we have no conclusive evidence of differences in efficacy or tolerability. Despite this, the agent that appears most effective may be twice as effective as the agent that appears least effective, and the agent that appears most likely to cause discontinuation may be 4 times more likely to do so than the treatment that appears least likely to do so. Comparative randomized studies are needed further to evaluate these drugs.