Background:Novel regimens have revolutionized multidrug-resistant tuberculosis (MDR-TB) treatment; however, medication adherence remains challenging and poorly characterized. We hypothesized that bedaquiline adherence, measured using electronic dose monitoring, would predict MDR-TB treatment outcomes.Setting:This is a prospective cohort study conducted in KwaZulu-Natal, South Africa.Methods:Adults with MDR-TB and HIV initiating bedaquiline and on antiretroviral therapy (ART) were eligible. Separate electronic dose monitoring devices measured bedaquiline and ART adherence through 6 months, calculated as observed versus expected doses. Whole-genome sequencing was performed to identify bedaquiline resistance-Associated variants.Results:From November 2016 through February 2018, 199 participants with MDR-TB and HIV were enrolled and followed up through treatment completion (median 17.2 months interquartile range 12.2-19.6). The median bedaquiline adherence was higher than ART adherence (97 vs. 89%, P < 0.001) but correlated (r2= 0.68, P < 0.001). High bedaquiline adherence (≥90%) compared with lower adherence was associated with improved end of treatment successful outcome (83.4% vs. 46.3%, P < 0.001), decreased mortality (11.0% vs. 29.6% P = 0.004), and improved retention in care through end of treatment (94.5% vs. 79.6% P = 0.002). Modeling identified a highly significant but linear association between bedaquiline adherence and outcome. On multivariable analysis, bedaquiline adherence was independently associated with mortality and outcome. Bedaquiline resistance-Associated variants were seen in 12% (7/57) of sequenced isolates (7% baseline, 5% emergent) with only 28.6% experiencing successful treatment outcome.Conclusions:Bedaquiline adherence through 6 months independently predicted end of MDR-TB treatment outcome, but a specific bedaquiline adherence threshold was not identified. Interventions to optimize bedaquiline adherence are urgently needed to improve MDR-TB HIV treatment outcomes.