Finding Those in Needs for Antiretroviral: A Validation of an HIV Risk Prediction Model for Women Aged 15-49 Years and Evaluation of the Diagnostic Test Accuracy of Nucleic Acid Testing for HIV Early Infant Diagnosis

Abstract

Study 1--Temporal External Validation of a Risk Prediction Model for HIV-1 Acquisition Among Women Aged 15–49 years in Eastern and Southern African Countries Background: Women aged 15–49 in Eastern and Southern Africa (ESA) are at higher risk of acquiring HIV-1 and should be prioritized for prevention, including pre-exposure prophylaxis (PrEP). However, resources are limited, making efficient identification of those at highest risk essential. Rosenberg et al. developed a parsimonious risk prediction model from nationally representative data across 15 African countries (2015–2019), demonstrating favorable discrimination (cross-validated AUC 0.79) with two predictors: population viremia and having a partner outside home. Since then, HIV-1 incidence has declined and epidemiological shifts may have occurred. This study aimed to evaluate the model's performance in guiding PrEP prioritization in ESA using more recent survey data. Methods: We conducted a temporal validation using Population-based HIV Impact Assessment (PHIA) surveys from nine ESA countries (2020–2023), comprising 69,030 women aged 15–49 who were HIV-1 seronegative or had recent infection. Predicted probabilities were generated by applying the original coefficients without re-estimation. Discrimination was assessed using survey-weighted AUC, and calibration through calibration-in-the-large (CITL) and calibration slope. PrEP prioritization utility was evaluated at fixed sensitivity thresholds using the proportion of women prioritized and number needed to treat (NNT). The model was redeveloped using a similar approach to the original to characterize differences in predictor importance. Results: The model demonstrated moderate discrimination (AUC 0.73; 95% CI: 0.69–0.76), lower than the development AUC of 0.79. Evaluating Eastern and Southern Africa separately yielded AUCs of 0.77 (95% CI: 0.73–0.80) and 0.69 (95% CI: 0.64–0.75), respectively. Miscalibration was observed (CITL 0.57; slope 2.49), though estimates should be interpreted cautiously given sparse events. At approximately 33% sensitivity, 10.9% of women were prioritized with an incidence rate of 1.58 per 100 person-years (NNT: 63). Redevelopment retained the same two predictors but with different relative importance. Conclusions: The model maintained moderate discrimination among ESA women surveyed in 2020–2023. Whether recalibration or redevelopment should be considered will depend on the intended use of the model, the practical impact on PrEP delivery, and data availability. Study 2--Diagnostic Accuracy of Nucleic Acid Tests for Human Immunodeficiency Virus Early Infant Diagnosis: A Systematic Review and Meta-Analysis Background. Early infant diagnosis (EID) of HIV relies on nucleic acid testing (NAT) because maternal antibodies persist in HIV-exposed infants for up to 18 months. Since the last systematic review in 2017, the landscape of NAT for EID may have advanced with new assays, including point-of-care (PoC) platforms. No review has since evaluated the diagnostic accuracy of individual NAT assays or explored potential sources of heterogeneity in their performance. Methods: We searched PubMed, Embase, Cochrane Library, and Global Index Medicus for studies published between 2015 and 2025 evaluating NAT assays for HIV EID in infants aged 0–18 months and included the five studies from previous review. Pooled sensitivity and specificity were estimated using bivariate random-effects models for each assay evaluated in three or more studies. To assess heterogeneity, all assays were pooled irrespective of individual platform, and testing timing and PoC versus laboratory-based platform were assessed through meta-regression. Certainty of evidence was assessed using the GRADE framework. Results: Of 1,355 records screened, 29 studies were included, evaluating seven individually pooled NAT assays. All assays demonstrated high diagnostic accuracy, with pooled sensitivity ranging from 0.96 to 1.00 and specificity from 0.99 to 1.00. Overall pooled sensitivity was 0.99 (95% CI: 0.98–1.00) and specificity was 1.00 (95% CI: 1.00–1.00). Testing at birth was associated with lower sensitivity compared with testing at four weeks or older (relative sensitivity: 0.94; 95% CI: 0.89–1.00; p = 0.04), with no effect on specificity. PoC and laboratory-based assays demonstrated equivalent diagnostic accuracy. Deeks' test indicated small-study effects (p = 0.01). Certainty of evidence was very low for overall assay accuracy, low to moderate for the effect of testing timing, and moderate for PoC equivalence. Conclusion: NAT assays for HIV EID demonstrate consistently high diagnostic accuracy, with no evidence favoring one assay over another. PoC assays perform comparably to laboratory-based assays, supporting their use in resource-limited settings. The choice of assay may be guided by operational considerations rather than diagnostic performance.