Long-term all-cause mortality in people treated for tuberculosis: a systematic review and meta-analysis

Background Accurate estimates of long-term mortality following tuberculosis treatment are scarce. This systematic review and meta-analysis aimed to estimate the post-treatment mortality among tuberculosis survivors, and examine differences in mortality risk by demographic and clinical characteristics. Methods We systematically searched Embase, MEDLINE, and the Cochrane Database of Systematic Reviews for cohort studies published in English between Jan 1, 1997, and May 31, 2018. We included research papers that used a cohort study design, included bacteriological or clinical confirmation of tuberculosis disease for all participants, and reported, or provided enough data to calculate, mortality estimates for people with tuberculosis and a valid control group representative of the general population. We excluded studies that reported duplicate data, had a study population of fewer than 50 people overall, had a follow-up period shorter than 12 months after treatment completion, or had a loss to follow-up of more than 30%. From eligible studies, we extracted standardised mortality ratios (SMRs), or calculated them when the data were sufficient, by dividing the sum of the observed deaths by the sum of the expected deaths. For studies that did not report SMR as their mortality estimate, either mortality hazard ratios or mortality rate ratios were extracted and pooled with SMRs. Random-effects meta-analysis was used to obtain pooled SMRs. Between-study heterogeneity was estimated with I-2. This study was prospectively registered in PROSPERO (CRD42018092592). Findings Of the 7283 unique studies identified, data from ten studies, reporting on 40 781 individuals and 6922 deaths, were included. The pooled SMR for all-cause mortality among people with tuberculosis, compared with the control group, was 2.91 (95% CI 2.21-3.84; I-2 = 99%, p heterogeneity < 0.0001). When restricted to people with confirmed treatment completion or cure, the pooled SMR was 3.76 (95% CI 3.04-4.66; I-2 = 95%). Effect estimates were similar when stratified by tuberculosis type, sex, age, and country income category. Causes of mortality were extracted for 4226 deaths that occurred post-treatment, with most deaths attributable to cardiovascular disease (20% [95% CI 15-26]; I-2 = 92%). Interpretation People treated for tuberculosis have significantly increased mortality following treatment compared with the general population or matched controls. These findings support the need for further research to understand and address the biomedical and social factors that affect the long-term prognosis of this population.