Search

Objectives To identify the pyrazinamide MIC above which standard combination therapy fails. Methods MICs of pyrazinamide were determined for Mycobacterium tuberculosis isolates, cultured from 58 patients in a previous randomized clinical trial in Cape Town, South Africa. The MICs were determined using BACTEC MGIT 960 for isolates that were collected before standard treatment with isoniazid, rifampicin, pyrazinamide and ethambutol commenced. Weekly sputum collections were subsequently made for 8 weeks in order to culture M. tuberculosis in Middlebrook broth medium. Classification and regression tree (CART) analysis was utilized to identify the pyrazinamide MIC predictive of sputum culture results at the end of pyrazinamide therapy. The machine learning-derived susceptibility breakpoints were then confirmed using standard association statistics that took into account confounders of 2 month sputum conversion. Results The pyrazinamide MIC range was 12.5 to >100 mg/L for the isolates prior to therapy. The epidemiological 95% cut-off value was >100 mg/L. The 2 month sputum conversion rate in liquid cultures was 26% by stringent criteria and 48% by less stringent criteria. CART analysis identified an MIC breakpoint of 50 mg/L, above which patients had poor sputum conversion rates. The relative risk of poor sputum conversion was 1.5 (95% CI: 1.2–1.8) for an MIC >50 mg/L compared with an MIC ≤50 mg/L. Conclusions We propose a pyrazinamide susceptibility breakpoint of 50 mg/L for clinical decision making and for development of rapid susceptibility assays. This breakpoint is identical to that identified using computer-aided simulations of hollow fibre system output.