BACKGROUND: Eight decades ago, the ratio of monocytes to lymphocytes (hereafter, the "ML ratio") was noted to affect outcomes of mycobacterial infection in rabbits. Recent transcriptomic studies support a role for relative proportions of myeloid and lymphoid transcripts in tuberculosis outcomes. The ML ratio in peripheral blood is known to be governed by hematopoietic stem cells with distinct biases. METHODS: The predictive value of the baseline ML ratio was modeled in 2 prospective cohorts of HIV-infected adults starting cART in South Africa (primary cohort, 1862 participants; replication cohort, 345 participants). Incident tuberculosis was diagnosed with clinical, radiographic, and microbiologic methods per contemporary guidelines. Kaplan-Meier survival analyses and Cox proportional hazards modeling were conducted. RESULTS: The incidence rate of tuberculosis differed significantly by baseline ML ratio: 32.61 (95% confidence interval [CI], 15.38-61.54), 16.36 (95% CI, 12.39-21.23), and 51.80 (95% CI, 23.10-101.71) per 1000 patient-years for ML ratios of less than the 5th percentile, between the 5th and 95th percentiles, and greater than the 95th percentile, respectively (P = .007). Neither monocyte counts nor lymphocyte counts alone were associated with tuberculosis. After adjustment for sex, World Health Organization human immunodeficiency virus disease stage, CD4(+) T-cell counts, and previous history of tuberculosis, hazards of disease were significantly higher for patients with ML ratios of less than the 5th percentile or greater than the 95th percentile (adjusted hazard ratio, 2.47; 95% CI, 1.39-4.40; P = .002). CONCLUSIONS: The ML ratio may be a useful, readily available tool to stratify the risk of tuberculosis and suggests involvement of hematopoietic stem cell bias in tuberculosis pathogenesis.