Association of diabetes, smoking, and alcohol use with subclinical-to-symptomatic spectrum of tuberculosis in 16 countries: an individual participant data meta-analysis of national tuberculosis prevalence surveys

Abstract

BACKGROUND Non-communicable diseases (NCDs) and NCD risk factors, such as smoking, increase the risk for tuberculosis (TB). Data are scarce on the risk of prevalent TB associated with these factors in the context of population-wide systematic screening and on the association between NCDs and NCD risk factors with different manifestations of TB, where ∼50% being asymptomatic but bacteriologically positive (subclinical). We did an individual participant data (IPD) meta-analysis of national and sub-national TB prevalence surveys to synthesise the evidence on the risk of symptomatic and subclinical TB in people with NCDs or risk factors, which could help countries to plan screening activities. METHODS In this systematic review and IPD meta-analysis, we identified eligible prevalence surveys in low-income and middle-income countries that reported at least one NCD (e.g., diabetes) or NCD risk factor (e.g., smoking, alcohol use) through the archive maintained by the World Health Organization and by searching in Medline and Embase from January 1, 2000 to August 10, 2021. The search was updated on March 23, 2023. We performed a one-stage meta-analysis using multivariable multinomial models. We estimated the proportion of and the odds ratio for subclinical and symptomatic TB compared to people without TB for current smoking, alcohol use, and selfreported diabetes, adjusted for age and gender. Subclinical TB was defined as microbiologically confirmed TB without symptoms of current cough, fever, night sweats, or weight loss and symptomatic TB with at least one of these symptoms. We assessed heterogeneity using forest plots and I2 statistic. Missing variables were imputed through multi-level multiple imputation. This study is registered with PROSPERO (CRD42021272679). FINDINGS We obtained IPD from 16 national surveys out of 21 national and five sub-national surveys identified (five in Asia and 11 in Africa, N = 740,815). Across surveys, 15.1%–56.7% of TB were subclinical (median: 38.1%). In the multivariable model, current smoking was associated with both subclinical (OR 1.67, 95% CI 1.27–2.40) and symptomatic TB (OR 1.49, 95% CI 1.34–1.66). Self-reported diabetes was associated with symptomatic TB (OR 1.67, 95% CI 1.17–2.40) but not with subclinical TB (OR 0.92, 95% CI 0.55–1.55). For alcohol drinking ≥ twice per week vs no alcohol drinking, the estimates were imprecise (OR 1.59, 95% CI 0.70–3.62) for subclinical TB and OR 1.43, 95% CI 0.59–3.46 for symptomatic TB). For the association between current smoking and symptomatic TB, I2 was high (76.5% (95% CI 62.0–85.4), while the direction of the point estimates was consistent except for three surveys with wide CIs. INTERPRETATION Our findings suggest that current smokers are more likely to have both symptomatic and subclinical TB. These individuals can, therefore, be prioritised for intensified screening, such as the use of chest X-ray in the context of community-based screening. People with self-reported diabetes are also more likely to have symptomatic TB, but the association is unclear for subclinical TB.