Methods for selecting regimen duration to prevent relapse in drug-susceptible and drug-resistant TB

Predicting the required duration of treatment necessary to yield an acceptable risk of recurrence is a key question facing Phase III trials in both drug-susceptible (DS) and multidrug-resistant tuberculosis (MDR-TB). Data on treatment duration from animal models are increasingly a focus of such studies, but they have not been calibrated against human clinical trials and are lacking in MDR-TB. Empirical meta-regression models based on clinical trials in DS-TB suggest that early bacteriological results and treatment duration may have value in predicting relapse, and have been prospectively validated against the results of three large randomised controlled trials in DS-TB. While few trials have been conducted in MDR-TB to date, and observational cohort data should be interpreted carefully due to bias and confounding, these models also appeared to perform well in two recent cohort studies of MDR-TB. Applying these insights in practice may require innovations in clinical trial design, such as more extensive selection, adaptation and use of multiple durations during Phases II and III. While several studies have identified important individual level prognostic variables that could improve the accuracy of relapse prediction, attempts to stratify treatment duration for individual patients based on these factors have so far met with limited success.