A HIERARCHICAL CLUSTER ANALYSIS OF CONCURRENT INITIAL ADVERSE EVENTS IN PATIENTS RECEIVING BPAL/BPALM REGIMENS FOR DRUG-RESISTANT TUBERCULOSIS IN LOWER MYANMAR

Abstract

Myanmar has recently scaled up all-oral BPaL (bedaquiline, pretomanid, and linezolid) and BPaLM (BPaL with moxifloxacin) regimens nationwide for multidrug-resistant or rifampicin-resistant tuberculosis (MDR/RR-TB). While clinically effective, the safety profile of these regimens is characterized by frequent, often concurrent adverse events (AEs). To date, real-world evidence regarding the clustering patterns of these toxicities within the Myanmar context remains limited. We conducted a retrospective analysis of electronic health records from 729 patients with MDR/RR-TB in lower Myanmar, utilizing data from the National Tuberculosis Programme’s Open Medical Record System (Open MRS). To characterize early-onset toxicity, analysis was restricted to first-occurrence AEs. We calculated pairwise mean differences (MDs) in time-to-first AE across all event types. These MDs were normalized between 0 and 1 and visualized via heatmaps with 0.1 cut-off intervals. Hierarchical cluster analysis using Euclidean distance and Ward’s linkage was applied to identify AE phenotypes based on temporal similarity. Seven distinct AE clusters were identified, with cluster sizes ranging from three to eight events. The most substantial cluster demonstrated multi-system involvement, encompassing palpitations, lower gastrointestinal symptoms, arthralgia, elevated liver enzymes, myalgia, headache, electrolyte disturbances, and QT prolongation. In conclusion, AEs associated with BPaL/BPaLM regimens exhibit distinct multi-systemic clustering patterns, suggestive of a treatment-related syndrome rather than sporadic side effects. These data support a transition toward cluster-based clinical monitoring and integrated safety surveillance to optimize patient management and treatment adherence for patients with MDR/RR-TB.