Socio-ecologic Factors Impacting Medication Adherence Trajectories in Patients With Drug-Resistant Tuberculosis and Human Immunodeficiency Virus

Abstract Rationale: Poor treatment adherence is associated with poor outcomes in patients with multidrug-resistant TB (MDR-TB) and HIV. Despite therapeutic advances, adherence challenges and care gaps persist, partly due to socio-demographic factors. While psychosocial support and digital adherence tools show promise, more research is needed to identify vulnerable subpopulations to optimize resource allocation for interventions. Methods: This is a prospective study of South African patients with MDR-TB and HIV (NCT03162107) treated with bedaquiline (BDQ) based regimens and ART from 2016 to 2020. We included individuals with adherence data measured via a cellular-enabled electronic dose monitor (Wisepill RT2000), allowing separate BDQ and ART adherence.We conducted a 1 to 5-class Latent Class Analysis (LCA) based on 21 socio-demographic variables. LCA was run as a mixture model with maximum likelihood estimates for convergence. Model selection was based on Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), entropy, and clinical relevance. Class-based trajectories were used to evaluate dynamic changes in adherence over time. Results: A 3-class solution was chosen based on optimal BIC (9829) and entropy (0.863), improving upon the 2-class model (BIC 9850, entropy 0.773). Significant socio-demographic differences emerged between classes. Class 1 (high risk) mainly consisted of younger, predominantly male (63%) individuals who were single, unemployed, had unstable housing, histories of imprisonment, and substance use, and were associated with poor TB outcomes (60% positive). Class 2 (intermediate risk) mainly consisted of younger females (65%), single, unemployed, with more stable housing and less substance use than Class 1. This group was more likely to receive government grants and live in larger households. Their treatment outcomes were moderate (75% positive), suggesting possible protective factors despite socioeconomic challenges. Class 3 (low risk) was predominantly older males (73%), employed, with higher BMIs, better functional status, and stable marital status, and had the best treatment outcomes (87% positive). Although there were no significant differences in 6-month cumulative ART or BDQ adherence between classes, BDQ adherence trajectories showed that Class 3 started and maintained higher adherence over 24 weeks, while Class 1 had persistently lower adherence which declined over time, highlighting the impact of socio-demographic combinations on adherence patterns and treatment outcomes. Conclusion: Our results identified 3 MDR-TB and HIV subclasses with distinct socio-ecological characteristics and adherence trajectories. Although cumulative adherence was similar across groups, treatment outcome differences suggest varying levels of risk that may be partially explained by distinct adherence trajectories rather than cumulative adherence alone.