Background:  Cross-sectional HIV incidence assays often lack specificity, leading to inflated incidence estimates. We evaluated the BED-Capture EIA assay (BED) and an avidity assay for HIV incidence determination, varying the assay cut-offs and combining these assays with others in a multi-assay algorithm.

Methods:  BED was performed according to the manufacturer’s instructions, with a normalized optical density of <0.8 (standard cut-off) or <1.0 (BED screen). The avidity assay used the Genetic Systems HIV-1/HIV-2 + O EIA with 0.1M diethylamine as the chaotropic agent, using a cut-off of <40% (standard cut-off) or <80% (avidity screen). Samples were tested from adults with recent HIV infection (EXPLORE study, 217 infected ≤1 year, 141 infected ≤100 days) and adults with chronic HIV infection (ALIVE cohort, 284 infected 2 to 6 years, 488 samples, maximum 2 samples/subject). Logistic regression with general estimating equations were used to identify factors associated with misclassification.

Results:  Sensitivity—Using standard assay cut-offs, BED identified 125 of 144 (88.7%) subjects infected ≤100 days and 172 of 217 (79.3%) subjects infected ≤1 year as recent; avidity identified 107 of 141 (75.9%) subjects infected ≤100 days and 141 of 217 (65.0%) subjects infected ≤1 year as recent. Using the BED screen plus the avidity screen 125 of 144 (88.7%) subjects infected ≤100 days and 169 of 217 (77.9%) subjects infected ≤1 year were identified as recent. Specificity—Using standard assay cut-offs, BED misclassified 50/488 (10.2%) subjects with chronic infection as recent; misclassification was associated with viral load <400 copies/mL (odds ratio 2.7, 95%CI 1.3 to 4.9) and CD4 cell count <50 cells/µL (OR3.9, 95%CI 1.7 to 9.1). Gender, race, age, age of sample, hepatitis C virus (HCV) and herpes simplex virus 2 (HSV-2) antibody status, and antiretroviral use, were not significantly associated with misclassification; in contrast, avidity misclassified only 5 of 488 samples as recent, 3 of whom had a CD4 count <200. Using a multi-assay algorithm (BED screen + avidity screen + CD4>200 + viral load >400), only 4 of 488 (0.8%) of subjects with chronic infection were misclassified as recent; assay sensitivity was not compromised using this approach.

Conclusions:  The specificity for measuring HIV incidence increased from 89.8% for the standard BED assay alone, to 99.2% using a multi-assay algorithm, without compromising sensitivity. Use of the multi-assay algorithm described above can improve the precision of HIV incidence estimates.