Early detection of HIV in infants: Aspects of design and analysis for diagnostic test studies

Objective: To review investigations on the early detection of HIV infection in infants to determine adherence to traditional methods of study design and analysis for evaluating new laboratory tests. Data Sources: A National Library of Medicine (MEDLINE) search was conducted to identify such investigations through 1993. Cited references in identified manuscripts were also considered. The search was restricted to investigations of human subjects and those published in the English language. Study Selection: Final inclusion criteria included (1) report of the age and human immunodeficiency virus (HIV) infection status of the subjects at the time of the diagnostic testing, and (2) presentation of data allowing confirmation of presented analyses and additional analyses. Data Extraction: Criteria for judging the investigations included (1) whether the criteria used to determine the positive and negative test results were defined; (2) whether the necessary sample size for the study was calculated; (3) whether the patients studied were representative of the patients to whom the test would be applied; (4) whether a gold standard evaluation was performed; (5) whether the outcomes included in the analyses were independent; (6) whether the test characteristics were properly analyzed; and, (7) whether confidence intervals were presented. Data Synthesis: An informative presentation of a diagnostic test should include as a minimum the seven criteria listed above. Only 21 of 36 (58%) of the studies incorporated at least three of the criteria. Conclusions: There is a wide variation in the manner in which investigations of diagnostic tests are conducted and the results reported. Increased awareness and use of standard study designs and analyses will allow the application of metanalyses. Such analyses will help guide the direction taken for finding and establishing early diagnostic procedures for HIV infection at birth or during infancy.