This review concluded that serological assays had reasonable sensitivity for recent infection with human immunodeficiency virus (HIV), but might misclassify established infections as recent. The conclusion was limited by the absence of a standard approach to assay evaluation. This caution was appropriate, given the methodological limitations of the review and the highlighted weaknesses in the dataset.

To assess the accuracy of serological tests for recent human immunodeficiency virus (HIV) infections.

PubMed was searched to the end of June 2009, for English-language articles and the search terms were reported. The bibliographies of selected studies were screened for other potentially relevant studies. Conference presentations were included if the corresponding full report was not available.

Primary studies that reported either test performance characteristics or a validation of assay-derived incidence estimates, for serological tests for recent HIV infection, were eligible for inclusion. Studies of tests based on the detection of viral nucleic acid or p24 antigen and studies that used another assay for recent infection to establish how recent the HIV infection in the specimens was, were excluded.

Test performance characteristics had to be assessed by collection of serum specimens from people infected with HIV for a known duration, the application of the test to these specimens, and the calculation of the sensitivity of the test to detect infections of short duration (recent) and the specificity of the test to detect infections of longer duration (not recent). A recent infection was defined as one documented to be shorter than the length of the window period of the assay under investigation; an established infection was defined as one of longer duration than the assay window period; and a longstanding infection was defined as one documented for a year or more. Validation of the assay-derived incidence estimates had to be by the collection of serum specimens from a population that had an established estimate of the incidence of HIV, the application of the tests to these specimens, and the comparison of the assay estimate with the established estimate.

The included studies described the performance characteristics or validation of 13 tests, which included 3A11-LS, Avidity-AxSym GU, BED, DE-V3, Serodia, Uni-Gold, and Vironostika-LS. The assay window period, where reported, ranged from 133 to 190 days. Most of the included studies were conducted in the USA or Europe. Specific risk categories assessed included heterosexuals from countries with high HIV prevalence, illegal intravenous drug users, and men who have sex with men.

The authors did not state how many reviewers assessed studies for inclusion.

The authors did not state that they assessed methodological quality.

For test performance, for each population in each study, the test sensitivity (proportion of recent infections detected as recent by the test) and specificity (proportion of non-recent infections detected as non-recent), with 95% confidence intervals, were extracted.

For incidence validation, the assay-derived incidence estimates and the source of the reference incidence estimate were extracted, as well as any correction factors applied to the assay-derived estimate of incidence.

The authors did not state how many reviewers extracted the data.

Studies were combined in a narrative synthesis.

Forty-five studies (number of participants not reported) were included in the review.

Test performance: Twenty-six studies, with 81 datasets, assessed the performance characteristics of 13 tests. Just over half of the datasets (48 of 81) reported the number of people tested to assess the performance characteristics, while 58 datasets reported the number of specimens tested. The median number of specimens tested per dataset was 100 (range seven to 2,749).

For all seven tests, 30 sensitivity estimates were reported and the median was 88·8% (range 42·3 to 100). For tests with more than three estimates available, the median sensitivity was 92·5% for Avidity-AxSym GU (range 74·0 to 100), 81·2% for BED (range 76·8 to 100), and 92·3% for Vironostika-LS (range 73·9 to 100).

The median, from 25 estimates of specificity to detect established infections (excluding longstanding infections and individuals receiving treatment with antiretroviral therapy), was 86·8% (range 49·5 to 100·0). For assays with more than three estimates available, the median specificity was 89·4% for Avidity-AxSym GU (range 49·5 to 99·5), 79·9% for BED (range 67·8 to 94·4), and 89·5% for Vironostika-LS (range 79·8 to 100). For longstanding infections, the median specificity, from 11 estimates was 98% (range 31·5 to 100·0).

Incidence validation: Twenty-three studies reported comparisons between assay-derived estimates of incidence, and corresponding established estimates. These comparisons were for four different tests: BED, Vironostika-LS, 3A11-LS, and IDE-V3. The median percentage difference between the assay-derived incidence and established incidence was 26·0% (range 0·0 to 483%).

Serological assays had reasonable sensitivity for the detection of recent infection with HIV, but were vulnerable to misclassifying established infections as recent, which could lead to bias in incidence estimates. There was no standard approach to assay evaluation.

The inclusion criteria were poorly defined and the search was limited to a single bibliographic database plus reference screening. The restriction to English-language studies introduced the possibility of language bias and relevant studies might have been missed. Reporting of the review process was poor; no measures to minimise error and bias were recorded and no assessment of the methodological quality of the included studies was reported. The narrative synthesis was appropriate for the data presented.

The authors' conclusions were appropriately cautious, given the methodological limitations of the review and the weaknesses in the dataset, which they highlighted.

One of the authors declared a financial interest in the BED assay.

Practice: The authors made no recommendations for practice.

Research: The authors stated that there was an urgent need for an internationally agreed framework for evaluating and comparing these tests.

Not stated.

This is a critical abstract of a systematic review that meets the criteria for inclusion on DARE. Each critical abstract contains a brief summary of the review methods, results and conclusions followed by a detailed critical assessment on the reliability of the review and the conclusions drawn.