This correspondence critiques the analysis by Templin et al. of tuberculosis (TB) incidence after TB preventive therapy (TPT) among persons living with HIV who recently initiated antiretroviral therapy (ART) in Mozambique. The author acknowledges the clinical importance of the question but identifies several design and reporting choices that complicate causal interpretation of the reported findings.
A central methodological concern is how exposure to TPT was defined. The study classified exposure based on eventual treatment completion—specifically, more than 170 days of isoniazid or more than 80 days of 3HP (the 12-dose isoniazid/rifapentine regimen)—while starting follow-up at the time of ART initiation. The letter points out that classifying participants at baseline using information that becomes known only after they survive long enough to complete therapy introduces potential immortal time bias unless completion is modeled as a time-varying exposure.
Immortal time bias occurs when a period during which the outcome cannot occur (because the exposure definition requires survival to a later date) is incorrectly attributed to the exposed group. The critique notes that unless analyses explicitly treat completion as time-varying or otherwise account for this period, estimated associations between completion and TB risk may be biased.
The authors of the original study reported a longer median time to TB diagnosis among patients who completed TPT. The letter argues that this observed delay could reflect the analytic structure (immortal time) rather than a true biological delay in TB onset produced by TPT. The magnitude of any resulting bias is not estimable from the published tables, according to the critique.
To evaluate whether immortal time or related biases influenced the results, the letter recommends several analytic strategies:
These approaches would help determine whether the longer time to TB diagnosis among completers is robust to bias-reducing methods.
The letter also addresses the use of routine programmatic data. Although such data can be valuable, missing values and timing relative to outcomes matter for causal inference. Specifically, HIV viral load and CD4 results were reported but the critique highlights the need to consider the timing of those laboratory measures relative to TB diagnosis and to handle missing CD4 data cautiously. The author suggests that these factors should be explicitly discussed when constructing causal models.
A reproducibility issue is raised regarding discrepant exclusion windows for ART pickup: the Methods section reportedly described a 3-month exclusion window, while the Results section used a 30-day window. The letter requests clarification of this inconsistency to improve transparency and reproducibility of the analysis.
The author is careful to state that these methodological critiques do not overturn the established evidence that TPT administered with ART provides benefit. Rather, the issues introduce uncertainty about the magnitude and timing of benefit reported in this specific analysis. The letter emphasizes that analytic choices susceptible to immortal time bias could distort estimates and should be addressed before drawing strong conclusions about timing or repeat courses of TPT.
The correspondence further notes that randomized trial evidence did not demonstrate added benefit from annual repeat 3HP in trials conducted in Ethiopia, Mozambique, and South Africa. On that basis, the author recommends caution about endorsing a second course of TPT after one year until analyses that are less susceptible to immortal time bias are available.
In summary, the critique identifies potentially important analytic and reporting limitations in the observational study of TB after TPT among people with HIV initiating ART. The principal concern is possible immortal time bias from defining exposure by eventual TPT completion while starting follow-up at ART initiation. Additional concerns relate to timing and missingness of programmatic laboratory data and a reported discrepancy in ART-pickup exclusion windows. The letter calls for alternative analyses and clearer reporting to support causal interpretation, while acknowledging established benefits of TPT with ART and urging caution regarding repeat annual TPT recommendations until more robust analyses are presented.