South Africa - An early-infant HIV-risk score for targeted HIV testing at birth

Background

Early HIV testing is needed to guarantee early HIV treatment success for very young infants, but universal testing is expensive. In this study, we examined the feasibility of using an early-infant HIV-risk score for targeted PCR testing and early diagnosis of HIV.

Methods and findings

We reviewed the maternal and infant characteristics of a cross-sectional sample of HIV exposed newborns at Kalafong Provincial Tertiary Hospital, Gauteng, South Africa. Infants were clinically evaluated and tested for HIV infection by PCR within 72 hours of birth. We quantified associations between HIV infection and individual parameters by fitting univariate and multivariate logistic regression models. We determined sensitivity and specificity for various cut-points of the derived risk scores.

From August 2014 to December 2016, of 15 175 live births, 3356 (22.12%) were born to HIV-infected mothers. We screened 1911 infants, and enrolled 1759 (92%) of these. Mothers who had no antenatal care visits (5.7% (97/1688)) were more likely to give birth to babies who tested PCR positive (p=0.0005). Most mothers (98.8%) knew their HIV status before delivery and were on cART (1626/1704, 95.4%). Virological control varied with HIV viral loads not detectable in 595 (60.15%) of mothers. One in five mothers (217/990, 21.9%) had viral loads greater than 1000 copies/µL. More than a quarter of babies (432/1655, 26.1%) were born at a gestational age <38weeks. Low birth weight (<2.5kg) was documented in 398/1598 (24.55%) and 13/31(40.63%) of the PCR negative and positive infants, respectively (p=0.0329). Fewer than 15% of babies were clinically symptomatic at birth. Growth restriction or small for gestational age were documented in 204/1689 (12.08%) babies, of whom six (6/37, 16.22%) were PCR positive. Symptomatic newborns more frequently tested HIV positive (p=0.0042). The newborn HIV PCR positivity rate was 1.8% (31/1759).

The most significant risk factors for HIV infection in very young infants were detectable maternal HIV viral load, maternal cART duration of <1 month, and an infant that was symptomatic at birth. We included small-for-gestational-age with the above three characteristics in multivariate analyses and developed a two-, three-, and four-risk model, with a predictive probability score of a newborn PCR positive test at 0.28, 0.498, and 0.57 respectively. Sensitivities of the three- and four-risk scores as a probability of 0.02 and 0.04 are 80% and 76%, respectively.

Conclusion

Targeted PCR testing to diagnose HIV infection in very young infants requires access to maternal viral load testing. Even if risk models include other parameters such as maternal cART history, infant birthweight, gestation estimates and symptoms, one-in-five infected infants will not be targeted for testing. At present; we support universal PCR testing at birth within the South African PMTCT context.