Characteristics, predictors and consequences of tuberculosis treatment interruption: A multicentre retrospective...

Treatment interruption is associated with poor tuberculosis (TB) treatment outcomes and increased drug resistance. To address the issue, we aimed to investigate the characteristics, predictors and consequences of treatment interruption.

We conducted a retrospective cohort study by retrieving 4 years (2018-2021) of TB patients' records at 10 public health clinics in Sarawak, Malaysia. Adult patients (≥18 years) with drug-susceptible TB were selected. Treatment interruption was defined as ≥2 weeks of cumulative interruption during treatment. The Chi-square test, Mann-Whitney U test, Kaplan-Meier and Cox proportional hazards regression were used to analyse the data, with p < 0.05 being considered statistically significant.

Out of 2953 eligible patients, 475 (16.1%) experienced TB treatment interruption. Interruptions were most frequent during the intensive phase (46.9%, n = 223), with the greatest risk within the first 4 weeks of treatment. The median time to interruption was 2 weeks in the intensive phase and the cumulative interruption probability at the end of the intensive phase was 12.9%. Notably, treatment interruption occurred during both intensive and continuation phases for 144 patients (30.3%), while the remaining 108 (22.7%) experienced interruptions only during the continuation phase with a median time to interruption of 16 weeks. Three predictors were identified to increase the risk of treatment interruption: adverse drug reaction (aHR = 8.53, 95% Cl: 6.73-10.82), smoking (aHR = 2.67, 95% Cl: 2.03-3.53) and illicit drug use (aHR = 1.88, 95% Cl: 1.03-3.45). Conversely, underlying diabetes was associated with a reduced likelihood of treatment interruption (aHR = 0.72, 95% Cl: 0.58-0.90). Treatment interruption led to significant differences in treatment restarts (62.3% vs. 0.7%), changes in medications (47.8% vs. 4.9%), prolonged treatment duration (247 days [IQR = 105] vs. 194 days [IQR = 44.3]) and lower successful outcomes (86.5% vs. 99.9%).

Understanding the temporal characteristics, predictors and negative consequences of treatment interruption can guide the development of time-relevant approaches to mitigate the problem.