Predicting Treatment Outcomes in MDR-TB: Comparing Electronic and Self-Reported Adherence Measures for Bedaquiline

Abstract Rationale Emergent drug-resistance to bedaquiline, novel antimycobacterial for treatment of multi-drug resistant tuberculosis, has highlighted the need to improve adherence support measures. Self-report is a commonly used method of medication adherence monitoring in both the research and clinical settings. Electronic dose monitoring (EDM) devices have been increasingly used to monitor adherence in real-time. EDM methods may be less subject to recall and social-desirability bias, but increase costs. The purpose of this analysis is to evaluate the accuracy and utility of self-report measures compared to an EDM device (Wisepill RT 2000) for measuring medication adherence to bedaquiline and the association of adherence with clinical outcomes. Methods Participants were enrolled from 2016-2018 into a prospective observational cohort study of patients co-infected with drug-resistant tuberculosis (DR-TB) and HIV on ART (the PRAXIS study, NCT03162107, NCT04032730) in Durban, KwaZulu Natal, South Africa and were initiating a bedaquiline-containing medication regimen. Pill-to-pill medication adherence was measured continuously using the Wisepill EDM device and by self-report at clinical visits, months 1-6. Monthly self-report data was calculated using the validated 3-item Wilson scoring tool of 30-day recall. Both EDM and self-report data is reported as a cumulative 6-month score. Negative clinical outcomes were defined as either treatment failure, death, or loss-to-retention-in-care. Results Of 283 enrolled patients, 282 provided EDM bedaquiline adherence data. Cumulative 6-month adherence was linearly associated with the proportion of negative end-of-treatment outcomes by both EDM monitoring (p<0.0001) and self-report (p=0.0002) (Figure 1, panels A and B). When controlling for age and gender, the odds of a negative outcome decreased by 8% for each percentage increase in self-reported Wilson score (p=0.0004), and the odds of a negative outcome decreased by 5.5% for each percentage increase in EDM monitored adherence (p<0.0001). The performance of each adherence measurement method for predicting outcomes improved across the study period; the AUC increased from 0.542 to 0.741 for EDM-measured adherence and 0.565 to 0.688 for self-report (Figure 1, panels C and D) from month 1 to 6, cumulatively. Conclusions The linear association between bedaquiline adherence and treatment outcomes suggests that each additional dose taken as prescribed significantly improves the probability of a positive outcome, which underscores the importance for accurate adherence monitoring. Both EDM monitoring and self-report as a cumulative 6-month adherence score significantly predicted outcomes, but further exploration of both tools is required to optimize sensitivity and specificity to correctly identify patients at risk for negative treatment outcomes.