Sputum induction (SI) has been proposed as the optimal sample collection method for patients with paucibacillary tuberculosis (TB). Studies reporting the culture of Mycobacterium tuberculosis from SI were reviewed. A random-effects meta-analysis of diagnostic yield (numerator M. tuberculosis SI culture-positive cases; denominator all culture-positive cases) was conducted. Diagnostic yields (95% confidence intervals, CIs) were displayed as Forest plots. Heterogeneity was evaluated using Chi-squared and I-squared tests and meta-regression analysis. Ninety publications were screened, 28 full-text papers reviewed, and 17 analyzed. Collectively, n=627 SI culture-positive cases among n=975 culture-confirmed TB cases were reported. The diagnostic yield of SI ranged from 35 to 95%. The pooled diagnostic yield was 74% (CI 65-81%), with significant heterogeneity (p<0.0001, I2=86%). There were no statistically significant differences in the yield between sub-groups defined by human immunodeficiency virus (HIV) prevalence or age. Univariate analysis demonstrated that the use of fiberoptic bronchoscopy (FOB) as the comparator method was associated with a 22% reduction (CI 2-42%) in the diagnostic yield of SI. However, after adjustment for confounding, the meta-regression analysis showed that FOB usage (p=0.21) and saline concentration (p=0.31) were not independently associated with the diagnostic yield. SI will detect approximately three-quarters of M. tuberculosis culture-positive cases under study conditions. Significant heterogeneity in the diagnostic yield was not explained by HIV prevalence, age, or the use of FOB as the comparator method. The use of a particular nebulized saline concentration for SI cannot be recommended on the basis of this meta-regression analysis.