Inclusion of newly licensed or repurposed drugs in regimens to treat children for multidrug-resistant-tuberculosis may lead to therapy that is shorter than traditional regimens and composed only of oral medications. As an all-oral regimen may be more acceptable and have a better safety profile than current regimens, demonstrating non-inferiority may be satisfactory. To demonstrate non-inferior efficacy it is necessary to set a non-inferiority margin, and the study must have assay sensitivity. Multi-arm multi-stage designs may currently not be appropriate in pediatric trials because of the lack of sensitive and specific intermediate outcomes. However, including an arm with an agent added to ameliorate toxicity would be efficient. Covariates can be used to stratify randomization, define subgroups, and improve efficiency of analysis. Powering or enriching the confirmed-TB subgroup may be important. Primary outcomes using a fixed time point from randomization for all study arms may result in variation in duration post-treatment completion but may be the best choice. While blinding of site personnel and patients may not be possible when regimens differ substantially in drugs and modes of administration, blinding should be maintained for independent endpoint review groups and other personnel. Type I error and family-wise error rates should be tightly controlled.