Background <p>The PEPaNIC RCT showed that early supplementation of insufficient enteral nutrition by parenteral nutrition (early-PN) worsened outcome of critically ill children as compared with withholding PN for 1 week (late-PN). The best timing to initiate nutritional support in the pediatric intensive care unit (PICU) remains unclear. In adults, declining phosphate levels may identify patients who are particularly harmed by early-PN. We therefore assessed whether early hypophosphatemia in critically ill children may indicate metabolic intolerance to nutrition.</p> Methods <p>In this secondary analysis of the PEPaNIC RCT (<i>n</i> = 1440), we investigated whether development of hypophosphatemia statistically interacts with the randomized intervention for its impact on clinical outcome, adjusting for baseline risk factors. Outcomes of interest included the incidence of new infections and the duration of PICU dependency as primary endpoints, 90-day mortality as safety endpoint, and duration of mechanical ventilation and of hospital stay as secondary endpoints. Subsequently, the impact of early-PN vs. late-PN in patients with and without early hypophosphatemia was assessed. Analyses were performed for phosphate abnormalities on PICU day 1 and 2, with and without imputing 20.3% (day 1) or 22.0% (day 2) missing phosphate data.</p> Results <p>Of patients with available phosphate measurements, 19.9% (211/1060) and 27.5% (243/883) developed hypophosphatemia on day 1 and day 2, respectively. Day 1 hypophosphatemia did not interact with the randomized intervention for any studied outcome. On day 2, hypophosphatemia interacted with randomization for risk of new infection (<i>P</i> = 0.031), likelihood of earlier live PICU discharge (<i>P</i> = 0.030), and time to live weaning from mechanical ventilation (<i>P</i> = 0.025). Harm by early-PN was more pronounced in patients with hypophosphatemia than in those without. Results after imputing missing phosphate data were similar, with an additional interaction for 90-day mortality (<i>P</i> = 0.038) revealing higher mortality with early-PN in patients with hypophosphatemia.</p> Conclusions <p>Development of hypophosphatemia may identify critically ill children who are particularly harmed by early-PN. This opens perspectives for its use as a biomarker of metabolic intolerance to enhanced nutrition, which requires further investigation.</p> Trial registration <p>ClinicalTrials.gov NCT01536275, registered February 2012.</p>