The IMPI Africa registry is a simple pragmatic multi-centre prospective observational study of patients admitted to hospital with suspected tuberculous pericarditis which was designed to assess the impact of HIV infection on clinical presentation, diagnostic evaluation, initial treatment, and outcome of patients with tuberculous pericarditis in Africa. We set out to enrol a minimum of 100 consecutive suspected tuberculous pericarditis patients presenting to collaborating physicians over 6 months, and follow them prospectively for 6–12 months. The study was approved by the research ethics committee of the University of Cape Town, South Africa, and all participants gave written informed consent.
Twenty seven hospital-based physicians from eight African countries (Cameroon, 2; Ghana, 1; Kenya, 1; Lesotho, 1; Nigeria, 1; South Africa, 19; Uganda, 1; Zimbabwe, 1) were invited by electronic mail in January 2004 to participate in the study. These physicians had expressed an interest in the project to one of the investigators (BMM). Fifteen physicians from three countries (Cameroon, 1; Nigeria, 1; and South Africa, 13) contributed patients to the IMPI Africa Registry (56% response rate). Twelve of the 15 physicians were affiliated to medical schools, and three (all in South Africa) were based in district general hospitals.
Suspected tuberculous pericarditis was defined as a patient presenting to hospital with a clinical syndrome of pericardial disease which was suspected to be caused by tuberculosis on the basis of clinical and/or laboratory findings leading to the commencement of anti-tuberculosis chemotherapy according to the national tuberculosis control programme  Consecutive incident cases of suspected tuberculous pericarditis were enrolled. Adult patients were eligible for inclusion in the study if the collaborating physician felt sufficiently confident with the diagnosis of tuberculous pericarditis to commence anti-tuberculosis treatment. The management of each patient was at the discretion of the collaborating physician in keeping with the observational nature of the study.
Pericardial disease was classified as acute non-effusive pericarditis, pericardial effusion, effusive-constrictive pericarditis, or constrictive pericarditis on the basis of the collaborating physician's assessment of clinical and imaging information.
We classified tuberculous pericarditis as 'definite' or 'probable' depending on the amount of clinical and bacteriological information supporting the diagnosis of tuberculosis, or 'not tuberculous pericarditis' when an alternative (non-tuberculous) cause of pericarditis was found. A diagnosis of 'definite' tuberculous pericarditis was based on the demonstration of tubercle bacilli in pericardial fluid (smear and/or culture) or on histologic section of the pericardium; 'probable' tuberculous pericarditis on the proof of tuberculosis elsewhere in a patient with otherwise unexplained pericarditis, on the basis of indirect tests (e.g., adenosine deaminase level ≥ 40 IU/l in pericardial fluid) and/or an appropriate response to a trial of anti-tuberculosis chemotherapy.
The HIV status of the patients was based on the results of serological testing for HIV. However, serological testing for HIV is not always available or offered in some medical institutions in Africa. Therefore, the physicians were requested to state whether they suspected HIV infection on clinical grounds and to classify 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. This assessment was left to the discretion of the collaborating physician, and no criteria were specified.
The effect of pericardial disease on the functional capacity of each patient was defined using New York Heart Association (NYHA) criteria as class I (no limitation in physical activity), class II (ordinary physical activity results in dyspnoea), class III (minimal physical activity results in dyspnoea), or class IV (dyspnoea at rest) . Haemodynamic instability was defined by the presence of at least two of these signs: pulse rate > 100 beats/minute, systolic blood pressure < 100 mmHg, and cardiac tamponade requiring pericardiocentesis. The radiological, electrocardiographic, echocardiographic, and laboratory results were based on the report provided by the collaborating physician without central verification.
Demographic, clinical, diagnostic and therapeutic information was captured by means of a standardized data collection form (available on request) and transmitted (by fax or e-mail) to the IMPI Africa Coordinating Centre at the Cardiac Clinic, Groote Schuur Hospital, Cape Town, South Africa. Enrolment into the registry started on 1 March 2004 in 11 hospitals and up to two months later in the other four, and ended on 31 October 2004. Patients were reviewed for the study outcomes at three, six and 12 months following entry into the registry, with a minimum follow-up period of 6 months for all participants. Patient follow-up was carried out by personal interview, failing which telephonic enquiry and postal services were used.
On enrolment, each patient provided a contact person (next of kin or neighbour) who could be contacted if the patient's whereabouts were not known at the time of follow-up. In the event of death or illness the contact person would be able to provide information. In the event of death, every effort was made to obtain a copy of the death certificate. We also used the services of the South African Department of Home Affairs (for South African centres) and a private detective company to obtain the vital status of patients lost to follow-up. The follow-up phase of the study ended on 30 April 2005, and (follow-up) data will be reported separately.
Data were analysed using Epi Info 3.3 (CDC, Atlanta, GA, USA). We used the chi-square test (or Fisher's exact test for variables with small number of expected frequencies) and analysis of variance to asses differences between categorical and continuous variables respectively. Significance tests were two-tailed and statistical significance was defined at the 5% alpha level. All patients were stratified by clinical HIV disease status and entered in the analyses regardless of the final diagnosis.