Time-Dependent Confounding in Tuberculosis Treatment Outcome Analyses: A Review of a Source of Bias.


In 2018, 484,000 people became sick with tuberculosis (TB) resistant to at least rifampin, the most effective drug to treat TB.(1) Patients with rifampin-resistant or multidrug-resistant TB (MDR-TB) must be treated for up to 24 months with≥ 4 drugs, many of which are toxic.(2) Research priorities for MDR-TB treatment are evolving rapidly, compromising the relevance of randomized controlled trials (RCTs) that take 5-10 years from planning to reporting.(3) RCT evidence remains the gold standard in assessing causal effects. Observational studies may also be used to infer causality when they generate high-quality data, specify potential sources of bias, and employ appropriate analytic methods to account for such threats, like confounding.

MIDAS Network Members