1report the results of an observational study on the mortality rate and its predictors in patients with a presumptive diagnosis of tuberculous pericarditis in sub-Saharan Africa. By enrolling patients with presumed tuberculous pericarditis from 15 referral hospitals in Cameroon, Nigeria and South Africa, and not insisting on gold-standard diagnostic criteria, the investigators have managed to enroll 185 patients within a 6month period between March and October 2004. Patients were eligible for enrolment if the attending physician felt sufficiently confident with the diagnosis of tuberculous pericarditis to commence antituberculosis treatment, and the management of each patient was at the discretion of the attending physician, in keeping with the observational nature of the study. In the majority of patients, HIV status was based on serological testing for HIV, but due to the fact that, at the time of the study, voluntary and confidential testing for HIV was not always offered at the participating medical institutions, the investigators included the category of suspected HIV infection based on clinical grounds, and classified each patient as either having evidence of ‘clinical HIV disease’ or ‘no clinical HIV disease’, without regard to the HIV serological status of the patient. The HIV assessment was left to the discretion of the collaborating physician, and no criteria were specified. This approach may lack some scientific rigour but it allowed rapid enrolment of all suspected cases and reflects the real-life experience of clinicians across Africa who do not always have access to diagnostic facilities and who do not find it difficult to diagnose tuberculous pericarditis on clinical grounds, however difficult it may be to confirm the diagnosis in the laboratory. 2,3 Using Cox regression, they assessed the effect of baseline clinical and therapeutic characteristics on mortality during a 6month period of follow-up. In a group of 174 patients (median age 33; range 14 - 87 years), the overall mortality rate was 26%, and it was significantly higher in patients who had clinical features of HIV infection than in those who did not (40% v. 17%, p=0.001). Independent predictors of death during followup were: (i) a proven non-tuberculosis final diagnosis (hazard ratio (HR) 5.35, 95% confidence interval (CI) 1.76 - 16.25), (ii) the presence of clinical signs of HIV infection (HR 2.28, CI 1.14 - 4.56), (iii) coexistent pulmonary tuberculosis (HR 2.33, 1.20 - 4.54), and (iv) older age (HR 1.02, CI 1.01 - 1.05). There was also a trend towards an increase in death rate in patients with haemodynamic instability (HR 1.80, CI 0.90 - 3.58) and a decrease in those who underwent pericardiocentesis (HR 0.34, CI 0.10 - 1.19). The strength of the study is that it reflects what clinicians experience in the field and what they perceive to be HIVand tuberculosis-related deaths. This experience may differ substantially from statistics captured by national tuberculosis control programmes. This study by Mayosi and colleagues 1