Parenteral nutrition (PN) is fundamental in the management of premature infants hospitalized in neonatal intensive care units (NICUs). Traditionally, standardized PN (SPN) has been recommended for most newborns due to safety, faster initiation, and reduced risk of prescribing errors. However, individualized PN (IPN) remains necessary in cases of metabolic instability or prolonged PN. Recently, AI-based models, such as the TPN2.0 algorithm, have emerged as potential tools to enhance precision, safety, and clinical outcomes in PN by combining standardization and personalization. The aim of this review is to summarize and critically discuss current evidence regarding standardized, individualized, and AI-based approaches to PN in premature infants hospitalized in NICUs. Despite guidelines, practice varies, and the emerging evidence for AI needs critical synthesis. This narrative review is based on a focused, critical appraisal of the available literature. A nonsystematic search of PubMed/MEDLINE, Wiley Online Library, ScienceDirect, and Google Scholar was conducted to identify relevant articles. Additional sources included international clinical guidelines and consensus statements. The search focused on publications addressing standardized, individualized, and AI-supported PN in neonatal and pediatric intensive care settings. Selected studies were analyzed qualitatively with emphasis on clinical outcomes, nutritional adequacy, safety profiles, and practical implementation aspects. Due to the heterogeneity of study designs and outcomes, a formal systematic review methodology and meta-analysis were not performed. Across multiple studies, SPN has been found to provide adequate macronutrient and electrolyte intake for most NICU patients, enabling faster initiation of PN and reducing prescribing errors. SPN often resulted in improved early protein, glucose, calcium, and phosphate delivery compared with IPN, with fewer electrolyte disturbances and comparable or better growth outcomes. Evidence supporting IPN benefits has been inconsistent and limited primarily to metabolically unstable or complex cases. AI-based PN (TPN2.0) demonstrated promising results, with physicians rating its recommendations higher than standard prescriptions. In infants whose clinically prescribed PN differed most from AI recommendations, higher morbidity, including NEC, was observed compared with AI-guided formulations. The algorithm reduced formulation subjectivity, streamlined workflow, and enabled rapid, guideline-adherent PN prescription. Early evidence suggests a potential association with reduced rates of mortality, cholestasis, sepsis, and NEC with the use of AI-supported PN strategies. SPN remains the safest and most efficient first-line approach for most premature infants, ensuring rapid initiation and reducing prescription-related risks. IPN continues to be essential for selected high-risk patients with complex metabolic needs. Emerging AI-based systems such as TPN2.0 may bridge these approaches by delivering personalized yet SPN formulations, improving safety, efficiency, and potentially clinical outcomes. Further high-quality prospective trials are needed to validate these findings and support the integration of AI into routine NICU nutritional practice.
Walendziak et al. (Mon,) studied this question.