Tuberculosis incidence among household contacts of individuals with symptomatic and asymptomatic tuberculosis: A historical cohort study in the Western Cape province of South Africa

Longitudinal, individual-level data describing long-term tuberculosis (TB) risk in high-burden communities are limited. We linked participants from the 2010 Zambia, South Africa Tuberculosis and AIDS Reduction (ZAMSTAR) prevalence survey to routine health records in the Western Cape Provincial Health Data Centre (PHDC) to estimate TB incidence of diagnosis over 11 years (to December 2020) among individuals without prevalent TB at enrolment. Cox proportional hazards models assessed predictors of TB diagnosis, including sex, age, HIV status, and household exposure to symptomatic or asymptomatic TB. Of 30,017 participants enrolled in the ZAMSTAR prevalence survey, 15,887 (52.9%) were linked to the PHDC, including 15,471 without prevalent TB at enrolment. Of the 15,471 individuals, 1,819 (11.8%) developed incident TB over the subsequent 11 years. TB incidence declined from 1,614 per 100,000 person-years in 2011-1,302 per 100,000 person-years in 2019. Over follow-up, individuals who had been exposed to symptomatic TB within their household had the highest cumulative probability of TB diagnosis (13%; 95% CI: 7.7-18%). In a multivariable Cox model adjusted for sex and HIV status, individuals living with a household contact with symptomatic TB had a higher hazard of TB incidence than those from unexposed households (aHR 1.53, 95% CI: 1.09-2.13); the hazard of incident TB among those exposed to asymptomatic TB was not significantly elevated (aHR 1.12, 95% CI: 0.67-1.87). In sensitivity analyses restricting follow-up to two and three years, hazard ratios were higher among individuals living with a symptomatic household contact (two years: 1.82 (95% CI: 0.97-3.41) and three years: 1.73 (95% CI: 0.98-3.10)). This long-term linked cohort demonstrates sustained TB incidence in this high-burden community. Incomplete linkage of individuals to the PHDC highlights the need for improving data collection in research studies and national patient identification systems to support long-term cohort analyses.