Objectives: To describe patterns of clinical and public health research into tuberculosis in South Africa from 1994-1998, and to assess the quality of studies of diagnostic accuracy and treatment.
Design: MEDLINE and the African Health Anthology were searched for journal articles published from 1994-1998 reporting clinical and public health research into tuberculosis in South Africa. Information extracted from the reports included research setting, study design, aspects of the quality of study design and reporting, and evidence of attention to statistical power.
Results: One hundred and thirty-five eligible studies were identified; 39 (28.9%) of the studies were case reports or case series. Important methodological flaws were identified in studies of diagnosis and treatment that compromised the validity of the study results. Of the 28 studies assessing interventions, 10 (35.7%) were randomised controlled trials with only one reporting adequate allocation. Of the 34 studies assessing diagnosis, 16 (47.0%) were cross-sectional assessments of diagnostic accuracy, and none reported that comparison of the reference standard was independent or blind. Fifty-four (79.4%) of the analytical studies with statistically non-significant results showed no evidence of consideration of sample size.
Conclusion: The usefulness of tuberculosis research in South Africa is compromised by the questionable validity of many studies and a lack of attention to sample size.