BACKGROUND: In low-incidence countries, clinical experience of tuberculosis is becoming more limited, with potential consequences for patient outcomes. In 2007, the Department of Health released a guidance 'toolkit' recommending that tuberculosis patients in England should not be solely managed by clinicians who see fewer than 10 cases per year. This caseload threshold was established to try to improve treatment outcomes and reduce transmission, but was not evidence based. We aimed to assess the association between clinician or hospital caseload and treatment outcomes, as well as the relative suitability of making recommendations using each caseload parameter. METHODS: Demographic and clinical data for tuberculosis cases in England notified to Public Health England's Enhanced Tuberculosis Surveillance system between 2003 and 2012 were extracted. Mean clinician and hospital caseload over the past 3 years were calculated and treatment outcomes grouped into good/neutral and unfavourable. Caseloads over time and their relationship with outcomes were described and analysed using random effects logistic regression, adjusted for clustering. RESULTS: In a fully adjusted multivariable model (34,707 cases)there was very strong evidence that management of tuberculosis by clinicians with fewer than 10 cases per year was associated with greater odds of an unfavourable outcome compared to clinicians who managed greater numbers of cases (cluster-specific odds ratio, 1.14; 95 % confidence interval, 1.05-1.25; P = 0.002). The relationship between hospital caseload and treatment outcomes was more complex and modified by a patient's place of birth and ethnicity. The clinician caseload association held after adjustment for hospital caseload and when the clinician caseload threshold was reduced down to one. CONCLUSIONS: Despite the relative ease of making recommendations at the hospital level and the greater reliability of recorded hospital versus named clinician, our results suggest that clinician caseload thresholds are more suitable for clinical guidance. The current recommended clinician caseload threshold is functional. Sensitivity analyses reducing the threshold indicated that clinical experience is pertinent even at very low average caseloads, which is encouraging for low burden settings.