Background: International guidelines recommend monitoring weight as an indicator of therapeutic response in childhood tuberculosis (TB) disease. This recommendation is based on observations in adults. In the current study, we evaluated the association between weight change and treatment outcome, the accuracy of using weight change to predict regimen efficacy, and whether successfully treated children achieve catch-up weight gain. Methods: We enrolled children treated for drug-susceptible TB disease (group 1) and multidrug-resistant TB disease (group 2) in Peru. We calculated the change in weight-for-age z score (ΔWAZ) between baseline and the end of treatment months 2-5 for group 1, and between baseline and months 2-8 for group 2. We used logistic regression and generalized estimating equation models to evaluate the relationship between ΔWAZ and outcome. We plotted receiver operating characteristic curves to determine the accuracy of ΔWAZ for predicting treatment failure or death. Results: Groups 1 and 2 included 100 and 94 children, respectively. In logistic regression, lower ΔWAZ in months 3-5 and month 7 was associated with treatment failure or death in groups 1 and 2, respectively. In generalized estimating equation models, children in both groups who experienced treatment failure or death had lower ΔWAZ than successfully treated children. The ΔWAZ predicted treatment failure or death with 60%-90% sensitivity and 60%-86% specificity in months 2-5 for group 1 and months 7-8 for group 2. All successfully treated children-except group 2 participants with unknown microbiologic confirmation status-achieved catch-up weight gain. Conclusions: Weight change early in therapy can predict the outcome of childhood TB treatment.