Temporal Factors and Missed Doses of Tuberculosis Treatment: A Causal Associations Approach to Analyses of Digital Adherence Data

Helen R Stagg, James J Lewis, Xiaoqiu Liu, Shitong Huan, Shiwen Jiang, Daniel P Chin, Katherine L Fielding

Research output: Contribution to journalArticlepeer-review


RATIONALE: Tuberculosis treatment lasts for six months or more. Treatment adherence is critical; regimen length, among other factors, makes this challenging. Globally, analyses mapping common types of non-adherence are lacking. For example, is there a greater challenge from early treatment cessation (discontinuation) or intermittent missed doses (suboptimal dosing implementation)? This is essential knowledge for the development of effective interventions, more 'forgiving' regimens, and to direct National Tuberculosis Programs.
OBJECTIVE: Granularly describe how patients take their tuberculosis medication and the temporal factors associated with missed doses.
METHODS: Pulmonary tuberculosis patients enrolled in the control arm of a pragmatic cluster-randomized trial in China of electronic reminders to improve treatment adherence were included. Treatment was the standard six-month course (180 days), dosed every other day (90 doses). Medication monitor boxes recorded adherence (box opening) without prompting reminders.Patterns of adherence were visualized and described. Mixed-effects logistic regression models examined the temporal factors associated with per-dose suboptimal dosing implementation, adjusting for clustering by participant. Cox regression models examined the association between early suboptimal dosing implementation and permanent discontinuation.
RESULTS: Across 780 patients, 16,794 of 70,200 doses were missed (23.9%), 9,487 from suboptimal dosing implementation (56.5%). By 60 days, 5.1% of participants had discontinued, 14.4% by 120 days. Most participants (95.9%) missed at least one dose. The majority of gaps were of a single dose (71.4%), although 22.6% of participants had at least one gap of two weeks' or more. In adjusted models, the initiation-continuation phase transition (odds ratio 3.07 [95% confidence interval 2.68-3.51]) and national holidays (1.52 [1.39-1.65]) were associated with increasing odds of suboptimal dosing implementation. Early-stage suboptimal dosing implementation was associated with increased discontinuation rates.
CONCLUSIONS: Digital tools provide an unprecedented step-change in describing and addressing non-adherence. In our setting, non-adherence was common; patients displayed a complex range of patterns. Dividing non-adherence into suboptimal dosing implementation and discontinuation, both were found to increase over time. Discontinuation was associated with early suboptimal dosing implementation. These apparent causal associations between temporal factors and non-adherence present opportunities for targeted interventions. Clinical trial registration ISRCTN46846388 Primary source of funding Bill & Melinda Gates Foundation (51914).
Original languageEnglish
JournalAnnals of the American Thoracic Society
Early online date20 Dec 2019
Publication statusPublished - 1 Apr 2020


Dive into the research topics of 'Temporal Factors and Missed Doses of Tuberculosis Treatment: A Causal Associations Approach to Analyses of Digital Adherence Data'. Together they form a unique fingerprint.

Cite this