BACKGROUND: It is very difficult to observe tuberculosis (TB) transmission chains and thus, identify superspreaders. We investigate cough duration as a proxy measure of transmission to assess the presence of potential TB superspreaders. DESIGN: We analyzed six studies from China, Peru, The Gambia and Uganda, and determined the distribution of cough duration and compared it with several theoretical distributions. To determine factors associated with cough duration, we used linear regression and boosted regression trees to examine the predictive power of patient, clinical and environmental characteristics. RESULTS: We found within-study heterogeneity in cough duration and strong similarities across studies. Approximately 20% of patients contributed 50% of total cough days, and around 50% of patients contributed 80% of total cough days. The cough duration distribution suggested an initially increasing, and subsequently, decreasing hazard of diagnosis. While some of the exposure variables showed statistically significant associations with cough duration, none of them had a strong effect. Multivariate analyses of different model types did not produce a model that had good predictive power. CONCLUSION: We found consistent evidence for the presence of supercoughers, but no characteristics predictive of such individuals.