Examining nonadherence in the treatment of tuberculosis: The patterns that lead to failure.

AIMS: Adherence has been shown to be a major predictor of tuberculosis treatment failure and relapse. The current adherence metrics can be improved to provide higher resolution of adherence patterns and identify patients in need of alternative treatment interventions. We investigated how adherence patterns affect treatment outcomes, when adherence is likely to decrease during treatment and which patients are at risk of being nonadherent. METHODS: Individual-level data were pooled from 3 clinical trials (n = 3724) for treatment of drug susceptible tuberculosis where monthly or weekly adherence patterns were collected and adherence patterns were quantified to assess the impact of clustered missed doses vs. randomly missed doses on tuberculosis treatment outcomes. Significance was determined through univariate and multivariate cox regression models. RESULTS: Patients who miss doses in clusters have an increased hazard risk for unfavourable outcomes and missing as little as 4 treatment days in 1 month resulted in 61% higher risk of unfavourable outcomes compared to patients who missed no treatment days (P < .01). Patients older than 50 years, and patients who experienced an adverse event were associated with lower adherence. CONCLUSION: Our results show that the pattern in which patients miss their drugs is important to their overall outcomes and missing treatment days in clusters rather than randomly increases the risk of poor outcomes. In the future more intensive and longitudinal adherence measurements will be valuable for clinical trials and regimen design and interpretation.