Historically, an association between tuberculosis and diabetes was recognised clinically, and the recent global rise in diabetes prevalence has reignited interest. We therefore quantified the tuberculosis–diabetes association using US survey data. A case-control analysis was performed using cross-sectional data from the second National Health and Nutrition Examination Survey (1976–1980; civilian non-institutionalised US population aged 20–74). Cases were respondents ever diagnosed with tuberculosis, and controls were respondents who reported never receiving a tuberculosis diagnosis. Exposure to diabetes and intermediate hyperglycaemia was defined using a self-reported measure, an oral glucose tolerance test, or both. We used logistic regression to estimate an adjusted odds ratio, controlling for potential major confounders. In relation to the main exposure measure, the adjusted odds ratio for the association between tuberculosis and diabetes varied between 2.31 (95% confidence interval 1.36–3.93) and 2.36 (95% confidence interval 1.40–3.97), depending on the model. No association was found for intermediate hyperglycaemia, with adjusted odds ratio varying between 1.33 (95% confidence interval 0.49–3.64) and 1.34 (95% confidence interval 0.50–3.62), depending on model. Irrespective of the exposure measure and the confounders controlled for, diabetes was associated with an increased tuberculosis risk. This study may underestimate the true association due to exposure misclassification.