Background: Inter-individual variability in exposure to anti-tuberculous drugs is well-described during therapy. However, relationships between pharmacokinetic parameters and clinical outcomes are poorly understood, particularly in HIV-endemic regions of southern Africa. Clinical pharmacokineticpharmacodynamic (PK-PD) studies are required to evaluate these associations and inform the development of shorter treatment regimens. Methodology: Adults with a first presentation of smear ‘++’ or ‘+++’ pulmonary TB were treated with standard 6 month therapy. HIV status, CD4 count and prior ART were documented at baseline. Plasma was sampled at 0, 2 and 6 hours post-dose on day 14 or 21 to assay concentrations of all four firstline anti-TB drugs. Mixed effects modelling of liquid culture Time to Positivity (TTP) and Serial Sputum Colony Counting (SSCC) data was used to generate individual patient estimates of the bacillary elimination rate during the first 8 weeks. Patients were followed until 1 year post-treatment and final outcomes were defined as favourable (sustained cure) or unfavourable (failure/relapse). Relationships between PK parameters and PD measurements of treatment response were assessed by logistic regression. Results: 75/133 (56%) patients were HIV-infected (median CD4 count: 166 [range 6-783] cells/μl) and 24 (18%) were on ART. Although 95% patients reported excellent adherence, 98/113 (87%) had a low rifampicin Cmax, 59/118 (50%) had a low isoniazid Cmax, 46/118 (39%) had a low pyrazinamide Cmax and 24/104 (23%) had a low ethambutol Cmax compared to published reference ranges. HIV associated factors did not explain inter-individual variability in the PK parameters of any first-line drug, although there was a trend towards a lower Cmax of rifampicin amongst those who were HIV-infected (p=0.09). A low AUC0-6h of isoniazid was associated with slower bacillary clearance in TTP and SSCC models (p=0.038 and 0.040 respectively), a lower likelihood of 2 month sputum culture conversion (p=0.045) and a higher likelihood of unfavourable final outcomes (p=0.035). Conclusions: The frequent finding of a low rifampicin Cmax advocates for acceleration in the conduct of rifamycin dose escalation studies in HIV-endemic populations. The association between low isoniazid exposure and all PD measures of treatment response has not been previously described and highlights the importance of clinical PK-PD studies in unravelling the determinants of early and late TB outcomes.