Contact tracing versus facility-based screening for active TB case finding in rural South Africa: A pragmatic cluster-randomized trial (Kharitode TB)

Background There is a dearth of comparative effectiveness research examining the implementation of different strategies for active tuberculosis (TB) case finding, particularly in rural settings, which represent 60% of the population of sub-Saharan Africa. Methods and findings We conducted a pragmatic, cluster-randomized comparative effectiveness trial of two TB case finding strategies (facility-based screening and contact tracing) in 56 public primary care clinics in two largely rural districts of Limpopo Province, South Africa. In the facility-based screening arm, sputum Xpert MTB/RIF was performed on all patients presenting (for any reason) with TB symptoms to 28 study clinics, and no contact tracing was performed. In the contact-tracing arm, contacts of patients with active TB were identified (via household tracing in 14 clinics and using small monetary incentives in the other 14 clinics), screened for TB symptoms, and offered Xpert MTB/RIF testing. The primary outcome was the number of newly identified patients with TB started on treatment. The analysis used multivariable Poisson regression adjusted for historical clinic-level TB case volumes and district. The trial was registered with ClinicalTrials.gov (NCT02808507). From July 18, 2017, to January 17, 2019, a total of 3,755 individuals started TB treatment across 56 study clinics in the 18-month period. Clinic characteristics and clinic-level averages of patient characteristics were similar across the two arms: 40/56 (71%) clinics were in a rural location, 2,136/3,655 (58%) patients were male, and 2,243 (61%) were HIV positive. The treatment initiation ratio comparing the yield of TB patients started on treatment in the facility-based arm compared to that from the contact-tracing arm was 1.04 (95% confidence interval [CI] 0.83-1.30, p = 0. 73). In the contact-tracing arm, 1,677 contacts of 788 new TB index patients were screened, yielding 12 new patients with TB. Prespecified subgroup analyses resulted in similar results, with estimated treatment initiation ratios of 0.96 (95% CI 0.64-1.27; p = 0.78) and 1.23 (95% CI 0.87-1.59; p = 0.29) among historically smaller and historically larger clinics, respectively. This ratio was 1.02 (95% CI 0.66-1.37; p = 0.93) and 1.08 (95% CI 0.74-1.42; p = 0.68) in the Vhembe and Waterberg districts, respectively. The estimated treatment initiation ratio was unchanged in sensitivity analyses excluding 24 records whose TB registration numbers could not be verified (1.03, 95% CI 0.82-1.29; p = 0.78) and excluding transfers-in (1.02, 95% CI 0.80-1.29; p = 0.71). Study limitations include the possibility of imbalance on cluster size owing to changes in catchment population over time and the inability to distinguish the independent effects of the two contact investigation strategies. Conclusions Contact tracing based on symptom screening and Xpert MTB/RIF testing did not increase the rate of treatment initiation for TB relative to the less resource-intensive approach of facility-based screening in this rural sub-Saharan setting.