A study analyzed adherence patterns and their relationship with treatment outcomes among patients with multidrug-resistant or rifampicin-resistant tuberculosis (MDR/RR-TB) enrolled in the endTB Observational Study, a prospective multicountry cohort conducted between April 2015 and December 2019. The study included patients treated with regimens containing at least bedaquiline and/or delamanid across 12 countries. The objective was to identify distinct adherence trajectories during treatment and assess how these trajectories were associated with unsuccessful treatment outcomes, defined as treatment failure, death, or loss to follow-up.
A total of 1,787 patients were included from an original cohort of 2,803 consenting participants. Eligible patients had started an endTB regimen after enrollment, had at least one month of adherence data, a recorded final treatment outcome, and complete covariate information. Monthly adherence was calculated as the proportion of prescribed treatment days on which all medications were taken as prescribed. Adherence data were collected through directly observed therapy (DOT), self-report, or pill counts, depending on treatment delivery. The investigators applied a joint latent class mixed model consisting of a multinomial logistic model for class membership, a class-specific linear mixed model for adherence trajectories, and a class-specific survival model for time to unsuccessful treatment outcome. The survival model adjusted for age, sex, previous TB treatment, HIV/antiretroviral therapy status, hepatitis C virus (HCV) infection, diabetes, extensive disease, low BMI, fluoroquinolone resistance, baseline regimen drugs, and study site.
The median age was 35 years (IQR 26-45), 36.9% were female, 65.0% had fluoroquinolone resistance, and 65.7% had extensive disease. Median treatment duration was 20 months, and median monthly adherence was 95.9% (IQR 88.8%-100%). Overall, 19.0% of patients experienced an unsuccessful outcome, including 7.6% deaths, 3.3% treatment failures, and 8.1% loss to follow-up. Four adherence trajectory classes were identified: "consistently high" (72.5%), "high to low" (14.3%), "low to high" (7.3%), and "consistently low" (5.9%). Median adherence ranged from 98.0% in the consistently high group to 42.1% in the consistently low group. Unsuccessful outcomes occurred in 74.3% of the consistently low group, 1.5% of the low-to-high group, and 6.8% of the consistently high group. Compared with the consistently high group, the adjusted hazard ratio (aHR) for unsuccessful outcomes was 23.2 (95% CI 15.7-24.3) in the high-to-low group and 43.2 (95% CI 26.2-71.5) in the consistently low group. The low-to-high group did not have a significantly different risk (aHR 0.7, 95% CI 0.1-3.8). Adherence trajectory classification predicted unsuccessful outcomes substantially better than conventional adherence measures, with an AUROC of 0.84 (95% CI 0.82-0.86) versus approximately 0.65 for classifications based on overall adherence rates.
Distinct longitudinal adherence trajectories were strongly associated with MDR/RR-TB treatment outcomes, and trajectory-based classification predicted unsuccessful outcomes more accurately than conventional summary adherence measures. These findings suggest that patterns of adherence over time may be more clinically informative than overall adherence percentages alone. Limitations include exclusion of several study sites because of adherence data quality concerns, reliance on adherence measures that partly used self-report or pill counts, and inclusion only of patients with complete data. As an observational cohort study, residual confounding cannot be excluded.
Source: Law S, Fulcher I, Ashraf S, Bastard M, Docteur W, Franke MF, Guerra D, Hewison C, Huerga H, Khan M, Khan P. Characterizing Treatment Adherence Trajectories in the endTB Multisite Cohort of Drug-Resistant Tuberculosis Patients: An Application of Group-Based Trajectory Modeling. Clinical Infectious Diseases. 2026 Mar 15;82(3):e571-9.
