Longitudinal-Survival Models for Case-Based Tuberculosis Progression

Area of Science:

• Epidemiology
• Biostatistics

Background:

• Tuberculosis (TB) remains a major global health burden, causing millions of illnesses and deaths annually.
• Existing analyses often overlook the dynamic nature of sputum smear results during TB treatment.
• Repeated sputum smear measures and survival times in TB patients necessitate advanced statistical approaches like joint modeling.

Purpose of the Study:

• To investigate the association between sputum smear results and unfavorable outcomes in TB patients.
• To dynamically predict survival probabilities using a joint modeling framework.
• To leverage routine TB data for improved patient outcome prediction.

Main Methods:

• A joint model for longitudinal and time-to-event data was employed.
• The longitudinal submodel used a generalized linear mixed-effects model for smear results.
• The time-to-event submodel utilized a Cox proportional hazards model, linked via a current value association structure.
• Bayesian inference with Markov Chain Monte Carlo (MCMC) approximation was used for parameter estimation.

Main Results:

• A positive and significant association (α = 1.033) was found between sputum smear positivity and the hazard of unfavorable outcomes.
• Factors like HIV status, previous TB treatment history, sex, and BMI significantly influenced outcomes.
• Previously treated patients (HR=2.52) and TB/HIV co-infected patients (HR=47.87%) faced substantially higher risks.

Conclusions:

• Sputum smear results are crucial for estimating unfavorable outcomes in TB patients.
• Joint modeling provides a dynamic approach to survival prediction based on longitudinal smear data.
• Specific patient groups, including men, previously treated individuals, TB/HIV co-infected, and severely malnourished patients, require targeted interventions due to higher risk.