Analysis of three lamotrigine extended-release clinical trials: comparison of pragmatic ITT and LOCF methodologies

Abstract

Early withdrawal of patients from a clinical trial can compromise the robustness of the data by introducing bias into the analysis. This is most commonly addressed by using the "intent to treat" (ITT) population and "last observation carried forward" (LOCF) methodology, where a patient's last assessment is carried forward. This can lead to overstatement of treatment efficacy especially if events indicative of treatment failure are infrequent. An alternative methodology, labeled "pragmatic ITT" (P-ITT), requires patients to have a positive outcome and to complete the trial in order to be considered a treatment success by that outcome measure. Data from 3 randomized multicenter lamotrigine extended-release (LTG XR) trials were analyzed and response (proportions seizure-free and with 50% response) were compared using LOCF and P-ITT methodologies. In 2 of the 3 trials, a lower response for both seizure freedom and 50% response was seen during the Maintenance phase using the P-ITT methodology. In the trial that did not show a difference, only a small number of patients withdrew early, thus negating the benefit brought by the P-ITT method. Differences between methodologies were not noted when evaluation was applied to the entire treatment period, most likely a reflection of the fact that a therapeutic dose of lamotrigine is not rapidly achieved. We propose that the P-ITT may be a simpler, more informative method for evaluating the effectiveness of a drug, especially in comparison to another active drug(s).