Advancing neurodevelopmental outcome prediction in children exposed to HIV using clinical features and placental imaging

About This Grant

PROJECT SUMMARY More than one million children who are HIV-exposed but uninfected (CHEU) are born to pregnant people with HIV (PPHIV) every year. CHEU have a higher risk of adverse early-life outcomes than HIV-unexposed peers, including neurodevelopmental deficits and a >2-fold risk of growth stunting. The pathophysiology of adverse CHEU outcomes is incompletely understood, but mounting evidence suggests that placental abnormalities play a key role. A better understanding of the causes and mechanisms of poor developmental outcomes in CHEU is essential to improve the care of PPHIV and their offspring. Our overarching goal is to determine which 1) clinical, 2) placental histological, and 3) placental stereologic features predict adverse CHEU neurodevelopmental and growth outcomes. Leveraging our ongoing multi-country (Uganda, South Africa) birth cohort (n=1,200) and linked placental biobank, we will perform state-of-the-art 3D placental stereology and build artificial intelligence (AI) classifier models to predict CHEU child health outcomes, employing causal inference and instrumental variable analysis to account for confounding. We will also perform mediation analysis to determine whether placental features mediate the relationship between clinical and laboratory features and child outcomes. Innovation: Distinct advantages of our proposed research include 1) simultaneous collection and comparison of CHEU and HIV/antiretroviral-unexposed placentas and children, 2) use of rich clinical data and complementary methods [3D stereologic imaging and histopathology] to evaluate associations between placental abnormalities and adverse CHEU neurodevelopmental and growth outcomes, and 3) use of causal inference and mediation analysis methods to identify key and modifiable features. Investigators: Our interdisciplinary team with expertise in placental collection and birth cohorts (Bebell, Gray), placental pathology and AI (Goldstein), bioinformatics, AI, and mediation analysis (Dreyfuss, Kawuma), placental ARV effects (Serghides), developmental psychology (Malcolm-Smith), and pediatric neurodevelopment (Donald) is well-poised to complete this work. Approach: We will leverage biobanked placental samples and extend follow-up of enrolled mother-child dyads in Dr. Bebell’s (R01HD11232) and Dr. Gray’s (R01HD102050) birth cohorts; Dr. Serghides’ laboratory infrastructure, Dr. Goldstein’s AI algorithms, and Dr. Dreyfuss’ mediation analysis and causal inference methods to elucidate the effects of HIV and specific ARV exposure on the placenta and child neurodevelopment and growth through age 5 years via these Specific Aims: 1a) Identify clinical and laboratory features that predict neurodevelopmental and growth outcomes in CHEU, 1b) Determine whether placental histologic diagnoses advance neurodevelopmental and growth outcome prediction, and 2) Incorporate placental stereology features into prediction models for neurodevelopmental and growth outcomes. Identifying HIV- and specific ARV-related placental abnormalities and associations with adverse CHEU outcomes has great potential to improve child health by informing ARV selection in pregnancy and early identification and intervention for at-risk children.