Statistical Analysis

We defined the intention-to-treat (ITT) population as all participants who received at least one dose of study medication. The modified intention-to-treat (mITT) population (prespecified as primary for efficacy) included those in the ITT population with more than one nonmissing, detectable, and plausible quantitative culture result. The per-protocol (PP) population included those in the mITT population whose experimental rifampin dose was not affected by study halts. Sensitivity analyses were performed on all participants with at least one log₁₀ cfu/ml count. We analyzed log₁₀ cfu/ml using linear and nonlinear mixed effects models in NONMEM 7.2 (ICON plc), Perl-speaks-NONMEM 4.4.0 (Uppsala University), and R version 3.3.2 (R Foundation for Statistical Computing). Monophasic and biphasic structural models were fitted, by treatment arm and pharmacokinetic exposure, using the M3 partial likelihood method and Laplacian estimation with interaction.

We used the one-sided Cochran-Armitage test for trend with 5% significance to compare proportions of participants who experienced at least one grade 2 or higher rifampin-related AE. We used the log-rank test to compare the time to AEs. We performed similar tests for secondary analyses of rifampin-related hepatic AEs and SAEs. A sensitivity analysis excluded AEs that occurred during study halts in the 15- and 20-mg/kg arms. We performed all safety analyses in Stata/SE 14.2 (StataCorp LLC).