Background: The proportion of HIV-exposed infants and young children infected with HIV in South Africa (SA) has declined markedly over the past decade as a result of the country's comprehensive prevention of mother-to-child transmission programme. This decrease has in turn reduced the positive predictive value (PPV) of diagnostic assays, necessitating review of early infant diagnosis (EID) algorithms to ensure improved accuracy.
Objectives: To evaluate the performance of the GeneXpert HIV-1 qualitative assay (Xpert EID) as a consecutive test for infants with an 'HIV-detected' polymerase chain reaction screening test at birth.
Methods: We retrospectively analysed a longitudinal cohort of HIV-exposed infants on whom birth testing was performed, using whole-blood ethylenediaminetetra-acetic acid samples, from four tertiary sites in Gauteng Province between June 2014 and December 2019. Birth samples from all infants with a Cobas AmpliPrep/Cobas TaqMan HIV-1 Qualitative Test v2.0 (CAP/CTM v2.0) HIV-detected screening test, a concurrent Xpert EID test and a subsequent confirmatory CAP/CTM v2.0 test on a separate specimen were included. Performance of the Xpert EID in predicting final HIV status was determined as proportions with 95% confidence intervals (CIs). A comparison of indeterminate CAP/CTM v2.0 results, as per National Health Laboratory Service resulting practice, with discordant CAP/CTM v2.0 v. Xpert EID results was performed.
Results: Of 150 infants who met the inclusion criteria, 6 (3.9%) had an Xpert EID result discordant with final HIV status: 5 (3.3%) were false negatives and 1 (0.7%) was false positive. As a consecutive test, the Xpert EID yielded a sensitivity of 96.5% (95% CI 92 - 98.9), specificity of 85.7% (95% CI 42.1 - 99.6), PPV of 99.3% (95% CI 95.7 - 99.9), negative predictive value of 54.5% (95% CI 32.5 - 74.9) and overall accuracy of 96.1% (95% CI 91.5 - 98.5). Using discordant CAP/CTM v2.0/Xpert EID results as criteria to verify indeterminate results instead of current practice would have reduced the number of indeterminate screening results by 42.1%, from 18 (12.6%) to 11 (7.2%), without increasing the false-positive rate.
Conclusions: Addition of the Xpert EID as a consecutive test for specimens with an HIV-detected PCR screening result has the potential to improve the PPV and reduce the indeterminate rate, thereby reducing diagnostic challenges and time to final status, in SA's EID programme.