Background: Many molecular epidemiology studies have been conducted to identify risk factors for clustering of tuberculosis (TB) cases in the population.
Objective: To estimate the impact of commonly investigated risk factors on TB clustering.
Methods: Ten electronic databases were searched up to January 2006 along with a hand search of the International Journal of Tuberculosis and Lung Disease and bibliographies of review articles. Meta-analyses of odds ratios (ORs) for various risk factors were conducted using random effect models, stratified by TB incidence. Meta-regressions were employed to account for the heterogeneity in clustering proportions and the magnitudes of risk.
Findings: The TB clustering proportion varied greatly (7.0-72.3%) among 36 studies in 17 countries. In multiple meta-regression analyses, high TB incidence, mean cluster size and conventional contact tracing were significantly associated with higher clustering. The pooled ORs (95%CIs) for low and high/intermediate TB incidence studies, using a cut off of 25/100000 per year, were 3.4 (2.7- 4.2) and 1.6 (1.3-2.1) for local-born status, 1.6 (1.5-1.7) and 1.7 (1.3-2.2) for pulmonary TB and 1.2 (1.1-1.3) and 1.3 (1.1-1.7) for smear-positive cases, respectively. Male sex, local birth, alcohol abuse and injection drug use were significantly higher risks in low TB incidence studies than in the high/intermediate ones.
Interpretation: Meta-analyses yielded significant estimates of ORs for several risk factors across both levels of TB incidence. Alcohol abuse, injection drug use and homelessness--all characteristics of marginalized populations--were found to be consistently significant in populations of low TB incidence. More research is needed to better understand TB transmission dynamics in high-burden countries.