Characterization of Representative Plasticizer Exposure Mixtures in NICU Neonates: A Non-negative Matrix Factorization Approach for Mixture Toxicity Assessment

Neonates hospitalized in neonatal intensive care units (NICUs) are exposed to complex mixtures of plasticizers released from medical devices during critical developmental periods. While biomonitoring studies typically assess individual metabolites, such approaches provide limited insight into realistic multiexposure patterns that are relevant for mixture toxicity assessment. Using urinary biomonitoring data from the ARMED-NEO cohort, we applied non-negative matrix factorization (NMF) to identify data-driven exposure mixtures in NICU neonates. This approach enabled the decomposition of high-dimensional metabolite data into a limited number of representative mixtures defined by their composition and contribution patterns. Three distinct exposure mixtures were identified, corresponding to coherent and clinically plausible multiexposure profiles associated with medical device use and care practices. These mixtures capture dominant co-occurrence patterns of plasticizer metabolites and provide a more realistic representation of neonatal exposure than single compound analyses. By translation of complex biomonitoring data into interpretable exposure profiles, this study offers a pragmatic framework for mixture-oriented toxicological research. These results support the design of targeted in vitro bioassays and contribute to improving the assessment of endocrine-disrupting mixture effects in vulnerable neonatal populations.