Purpose: Early diagnosis of HIV infection in exposed infants allows timely HIV treatment. Guidelines focus on polymerase-chain-reaction (PCR) as a testing method for infants <18mo, but implementation is expensive and logistically difficult in resource-poor settings. Rapid HIV Testing (RHT) is cheaper and simpler, but maternal antibodies prevent definitive HIV-positive diagnoses for infants <18mo. However, RHT might allow diagnosis of uninfected infants and avert unnecessary PCR testing. We evaluated the cost-effectiveness of conventional (PCR) and modified (PCR + RHT) algorithms for infant HIV diagnosis.

Methods: To calculate RHT sensitivity and specificity, we used data from 788 HIV-exposed 1.5-18 month old infants from Tororo District Hospital and Mulago Hospital, Uganda, tested with both RHT and DNA-PCR (gold-standard) from 08/05 to 02/07. We estimated costs through time-motion studies and collected additional data from project records. We developed a decision-analytic model comparing two modified testing algorithms to the conventional algorithm (PCR-Alone). The modified algorithms assumed negative results on a single RHT test (Single-RHT) or two sequential RHTs (Sequential-RHT) ruled out infection without the need for PCR. A positive result was treated as inconclusive and PCR performed immediately. We used the model to estimate costs, outcomes (HIV-infected infants correctly identified), and cost-effectiveness. The model considered programmatic factors (e.g. loss-to-follow-up) as well as test performance in calculating results.

Results: RHT sensitivity was 94% (95%CI: 89-97%). RHT specificity was age-dependent, rising from 8% (95%CI: 5-12%) at 1.5-3 months to 87% (95%CI: 74-95%) at 12-18 months. HIV prevalence was 16% (95%CI: 13-19%) and 41% (95%CI: 35-47%) in asymptomatic and symptomatic infants respectively. Unit costs for RHT were $0.85-$1.16 depending on the test used, and $31.34 for PCR. Loss-to-follow-up was estimated as 5% between appointments. Based on these data, the model estimated the percentage of HIV-infected infants correctly diagnosed would be 88.2% for Single-RHT, 93.9% for Sequential-RHT, and 94.4% for PCR-Alone algorithms respectively. The least-effective strategy, Single-RHT, cost $119 per outcome (HIV-positive infant identified). The incremental cost-effectiveness of Sequential-RHT compared to Single-RHT was $79 per outcome, and over $6000 per outcome comparing PCR-Alone to Sequential-RHT.

Conclusions: Infant HIV testing algorithms incorporating RHT appear to be highly cost-effective compared to conventional algorithms, especially in asymptomatic infants and infants over 6 months, and could facilitate wider implementation of infant HIV testing programs in resource-poor settings.