Impact of a multicomponent strategy including decentralized molecular testing for tuberculosis on mortality: planned analysis of a cluster-randomized trial in Uganda.

Background: Rapid diagnosis of tuberculosis (TB) is important for improving outcomes and reducing transmission. Previous studies assessing the impact of Xpert MTB/RIF (Xpert), a molecular assay that provides results within 2 h, on mortality have been inconclusive. In this planned analysis of a pragmatic cluster-randomized trial in Uganda, we assessed whether a multicomponent strategy, including decentralized Xpert testing, decreased mortality among adults evaluated for TB.

Methods: Ten community health centers were randomized, using a computer-generated randomization sequence, to the XPEL-TB intervention (on-site Xpert testing plus implementation supports) and ten to routine TB care without any modifications (on-site smear microscopy and referral-based Xpert testing for selected patients). The trial included all adults ( $\ge$ 18 years of age) undergoing evaluation for presumptive TB at each trial health center. All-cause mortality was a secondary outcome of the trial. For this analysis, the primary outcome was the mortality rate (censored at 18 months), and the secondary outcome was the six-month mortality risk. We compared the outcomes between trial arms using cluster-level analyses to account for stratified randomization and patient-level covariates. The trial was registered with the US National Institutes of Health (identifier: NCT03044158) and the Pan African Clinical Trials Registry (identifier: PACTR201610001763265).

Findings: Vital status was ascertained for 8413 of 9563 (88%) XPEL-TB trial participants who presented at the health centers from October 22, 2018 through February 29, 2020. The adjusted rate ratio (aRR) was 0.77 (95% CI: 0.47-1.28), comparing the intervention (145 deaths/3655 person-years) to routine care (154 deaths/3015 person-years). In sub-group analyses, point estimates for mortality were lower in the intervention arm among people without HIV (aRR = 0.50, 95% CI: 0.26-0.96) and among females (aRR = 0.64, 95% CI: 0.33-1.23). The mortality risk analysis yielded similar results.

Interpretation: Consistent point estimates favoring the intervention in our trial and previous ones suggest that Xpert testing may have an impact on mortality at community health centers. However, the magnitude of effect is small, and statistically significant results are unlikely to be attained within a single trial. Future trials of novel TB diagnostics at community health centers should focus on more proximal outcomes including TB detection and treatment initiation.

Funding: This work was supported by the National Heart, Lung, and Blood Institute of the US National Institutes of Health under award number R01HL130192.