Abstract
BACKGROUND AND OBJECTIVES: Accurate HIV incidence estimates among blood donors are necessary to assess the effectiveness of programs aimed at limiting transfusion-transmitted HIV. We assessed the impact of undisclosed HIV status and antiretroviral (ARV) use on HIV recency and incidence estimates using increasingly comprehensive recent infection testing algorithms. MATERIALS AND METHODS: Using 2017 donation data from first-time and lapsed donors, we populated four HIV recency algorithms: (1) serology and limiting-antigen avidity testing, (2) with individual donation nucleic amplification testing (ID-NAT) added to Algorithm 1, (3) with viral load added to Algorithm 2 and (4) with ARV testing added to Algorithm 3. Algorithm-specific mean durations of recent infection (MDRI) and false recency rates (FRR) were calculated and used to derive and compare incidence estimates. RESULTS: Compared with Algorithm 4, progressive algorithms misclassified fewer donors as recent: Algorithm 1: 61 (12.1%); Algorithm 2: 14 (2.8%) and Algorithm 3: 3 (0.6%). Algorithm-specific MDRI and FRR values resulted in marginally lower incidence estimates: Algorithm 1: 0.19% per annum (p.a.) (95% confidence interval [CI]: 0.13%-0.26%); Algorithm 2: 0.18% p.a. (95% CI: 0.13%-0.22%); Algorithm 3: 0.17% p.a. (95% CI: 0.13%-0.22%) and Algorithm 4: 0.17% p.a. (95% CI: 0.13%-0.21%). CONCLUSION: We confirmed significant misclassification of recent HIV cases when not including viral load and ARV testing. Context-specific MDRI and FRR resulted in progressively lower incidence estimates but did not fully account for the context-specific variability in incidence modelling. The inclusion of ARV testing, in addition to viral load and ID-NAT testing, did not have a significant impact on incidence estimates.