Cross sectional, ‘Recent Infection Testing Algorithm’ (RITA)-based estimation of national HIV incidence is being implemented in several African countries in 2015-17. The sample sizes (SS) and associated implementation costs required for precise estimation of incidence (I) are driven by the prevalence (P), I, the mean duration of recent infection (MDRI), and the false recent rate (FRR), which vary with HIV subtype and ART coverage. We sought to evaluate the potential applicability of a RITA combining the limiting antigen immunoassay (LAg) and viral load (VL) to assess the impact of population-level prevention interventions (PLPI) nationally and in key populations (KP). We also describe features of new RITAs that would enable broader implementation for these and other applications.
SS were calculated using publicly available spreadsheet tools and R customizations, for assessment of the impact of an intervention that reduces I by 50%, with power 0.8 at alpha 0.05, with correction for design effect. We used an MDRI of 130 days (d) and FRR of 0.25% for LAg+VL (cutoff 1000 copies/ml), and compared this to MDRI 160 d (LAg+VL cutoff 100 copies/ml) and 240 or 280 d as aspirational targets. Surveys were considered feasible if SS was below 30,000 (national) or 5000 (KP). National P and I data for 2014 as reported by UNAIDS were used.
Target SS for African countries with P>4% or I>0.3% for PLPI, and for one KP are shown in Table 1. With LAg+VL1000, PLPI surveys were feasible in 6 countries. If the MDRI could be increased to 280 d, SS would be reduced by approximately 60%, and surveys were feasible in an additional 5 countries. In countries with non-C subtypes, such as Cameroon, Kenya, Tanzania and Uganda, SS may be underestimated due to FRR>0.25%, but the impact of elevated FRR is dampened at longer MDRI. KP surveys in young women in South Africa (a high prevalence/incidence KP example) required SS >5000 but were feasible with MDRI of ≥240 d.
The currently available RITA of LAg+VL has limited applications for national PLPI surveys unless SS larger than 30,000 are considered; outside of high I settings it may generate imprecise estimates that are not able to detect large reductions in I. Similarly, LAg+VL surveys may not be feasible in KP surveys due to SS requirements. Development of new incidence assays with longer MDRI (≥240 d) and low FRR are needed to enable broader and more cost-effective use in national surveys, as well as for applications in PLPI and KP surveys.