Performance of the Maxim and Sedia Limiting Antigen Avidity assays for population-level HIV incidence surveillance

Abstract Background Two manufacturers, Maxim Biomedical and Sedia Biosciences Corporation, supply US CDC-approved versions of the HIV-1 Limiting Antigen Avidity EIA (LAg assay) for detecting ‘recent’ HIV infection in cross-sectional incidence estimation. This study assesses and compares the performance of the Maxim and Sedia LAg assays for incidence surveillance. Methods We ran both assays on a panel of 2,500 well-characterized HIV-1-infected specimens, most with estimated dates of (detectable) infection. We analysed concordance of assay results, assessed reproducibility using repeat testing and estimated the critical performance characteristics of a test for recent infection—mean duration of recent infection (MDRI) and false-recent rate (FRR)—for a range of normalized optical density (ODn) recency discrimination thresholds, alone and in combination with viral load thresholds. We further defined three hypothetical surveillance scenarios and evaluated overall performance for incidence surveillance, defined as the precision of incidence estimates, by estimating context-specific performance characteristics. Results The Maxim assay produced lower ODn values than the Sedia assay on average, largely as a result of higher calibrator readings (mean calibrator OD of 0.749 vs. 0.643). Correlation of non-normalized OD readings was greater (R2 = 0.938) than those of ODn readings (R2 = 0.908), and the slope was closer to unity (1.054 vs. 0.899). Reproducibility of repeat testing of three blinded control specimens (25 replicates each) was slightly better for the Maxim assay (CV 8.9% to 14.8% vs. 13.2% to 15.0%). The MDRI of a Maxim-based algorithm at the ‘standard’ recency discrimination threshold in combination with viral load (ODn ≤1.5 & VL >1,000) was 201 days (95% CI: 180,223) and for Sedia was 171 days (95% CI: 152,191). Commensurately, the Maxim algorithm had a higher FRR in treatment-naive subjects (1.7% vs. 1.1%). We observed statistically significant differences in MDRI using the ODn alone (≤1.5) and in combination with viral load (>1,000). Under three fully-specified hypothetical surveillance scenarios (comparable to South Africa, Kenya and a concentrated epidemic), recent infection testing algorithms based on the two assays produced similar precision of incidence estimates. Conclusions Differences in ODn measurements between the Maxim and Sedia LAg assays on the same specimens largely resulted from differences in the reactivity of calibrators supplied by the manufacturers. Performance for surveillance purposes was extremely similar, although different ODn thresholds were optimal and different values of MDRI and FRR were appropriate for use in survey planning and incidence estimation.