Abstract
Background
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.
Methods
In this secondary analysis of the PEPaNIC RCT (n = 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.
Results
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 (P = 0.031), likelihood of earlier live PICU discharge (P = 0.030), and time to live weaning from mechanical ventilation (P = 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 (P = 0.038) revealing higher mortality with early-PN in patients with hypophosphatemia.
Conclusions
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.
Trial registration
ClinicalTrials.gov NCT01536275, registered February 2012.
Supplementary Information
The online version contains supplementary material available at 10.1186/s13054-026-05844-x.
Keywords: Clinical outcome, Critical illness, Hypophosphatemia, Children, Infection, Mechanical ventilation, Mortality, PICU, Parenteral nutrition
Background
Optimizing nutritional support of critically ill children remains a challenge, not only with regard to dosing and composition, but also with regard to timing [1, 2]. The Pediatric Early versus Late Parenteral Nutrition in Critical Illness (PEPaNIC) multicenter, randomized controlled trial (RCT) demonstrated that early supplementation of insufficient enteral nutrition with parenteral nutrition (early-PN) worsened clinical outcome of critically ill children, as compared with withholding PN for 1 week (late-PN) [3]. Patients in the early-PN group had a higher risk of acquiring a new infection during their stay in the pediatric intensive care unit (PICU) and showed a delayed recovery as compared with patients in the late-PN group. Thus, early full feeding with supplemental PN to age-adjusted targets appeared harmful, leading to current clinical practice guidelines to be more restrictive with nutritional support in the acute phase of critical illness [2, 4, 5]. However, it remains unclear when and for whom advancing artificial feeding through enteral or parenteral feeding could be beneficial. Availability of a quantifiable biomarker that would signal when a patient is ready or not ready yet to receive enhanced feeding would thus provide major progress.
Also in critically ill adults, early-PN was shown to be harmful as compared with late-PN in the Early versus Late Parenteral Nutrition in Critical Illness (EPaNIC) trial [6]. Since then, a few studies of adult critically ill patients have suggested that a decrease in serum/plasma phosphate levels may indicate metabolic intolerance to nutrition. First, the Refeeding RCT found in patients who developed hypophosphatemia within 72 h after starting artificial feeding that continuing and increasing nutritional support increased mortality risk as compared with temporary caloric restriction [7]. Similarly, a retrospective study has associated a high caloric intake with increased mortality in critically ill patients with hypophosphatemia, whereas no such association was observed in patients without hypophosphatemia [8]. A recent secondary analysis of the EPaNIC trial revealed that a decrease in plasma phosphate from day 1 to day 2 could identify patients who were harmed by early-PN. Indeed, in patients developing such decrease in plasma phosphate levels, patients receiving early-PN were less likely to be discharged alive from the intensive care unit (ICU) earlier as compared with patients in the late-PN group [9]. In contrast, the randomized intervention did not associate with impaired outcome for patients without an early phosphate decrease.
It remains unclear, however, whether these findings can be extrapolated to critically ill children. Moreover, it remains unclear whether hypophosphatemia in itself indicates metabolic intolerance to nutrition, or whether this only applies to hypophosphatemia that is induced after some minimum exposure to nutrition. We hypothesized that development of hypophosphatemia after initiation of nutritional support indicates metabolic intolerance to enhanced nutritional support in children, and hence, harm if such nutritional strategy is continued. We addressed this hypothesis in a secondary analysis of the PEPaNIC RCT.
Methods
Study design and participants
This study reports the results of a secondary analysis of the PEPaNIC RCT [3]. This multicenter RCT enrolled 1440 critically ill infants and children from full-term newborn to 17 years old with a medium to high risk of malnutrition and a central venous line in place for clinical purposes, who were admitted to the PICUs of Leuven (Belgium), Rotterdam (the Netherlands) or Edmonton (Canada) from June 2012 to July 2015 (ClinicaTtrials.gov, NCT01536275). The study investigated the impact of withholding parenteral nutrition for 1 week after PICU admission (late-PN) as compared with providing early parenteral nutrition to supplement insufficient enteral nutrition (early-PN). The study protocol was approved by the Institutional Review Boards of the participating centers (ML8052, NL49708.078, Pro00038098) and has been published [10]. Written informed consent was obtained from the parents or legal guardians.
In both randomization groups, enteral nutrition was initiated as soon as possible. In patients assigned to early-PN, PN was started within 24 h after PICU admission to supplement any insufficient enteral nutrition in order to reach local caloric and macronutrient targets early. PN was discontinued when patients received 80% of the caloric target through the enteral or oral route. Patients assigned to the late-PN group did not receive PN in the first week in the PICU. These patients received a mixture of dextrose 5% and saline (60/40 v/v) to match the fluid intake with that of the patients in the early-PN group, taking into account the volume of the enteral nutrition. If enteral nutrition still failed to provide 80% of the caloric target on the morning of day 8, PN was initiated to reach the target. Patients in both groups received intravenous trace elements, minerals, and vitamins early to prevent deficiencies and subsequent refeeding syndrome, and blood glucose was controlled with insulin according to local target ranges [3, 10].
Plasma (Leuven) or serum (Rotterdam, Edmonton) phosphate concentrations, further referred to as phosphate concentrations, were measured daily by routine clinical chemistry in the participating centers’ central hospital laboratories, at the time of the routine morning blood sampling (6 ± 2 a.m.). For this study, we focused on the phosphate concentrations on day 1 and day 2, which we classified as presence of hypophosphatemia (“hypophosphatemia”) or no presence of hypophosphatemia (“no hypophosphatemia”) on the respective days. Based on the age-dependent phosphate reference ranges, we defined hypophosphatemia as a phosphate concentration below 1.4 mmol/L for infants up to 30 days old, below 1.2 mmol/L for infants from 31 days old to younger than 1 year, below 1.05 mmol/L for children younger than 7 years and below 0.9 mmol/L for children aged 7 years or older.
Clinical outcomes
We here studied the risk of acquiring a new infection during the PICU stay and the duration of PICU dependency as primary endpoints, mortality at 90 days after randomization as the safety endpoint and duration of mechanical ventilation and duration of the hospital stay as secondary endpoints, as was done for the original RCT [3]. Acquisition of a new infection was defined as need for treatment with antimicrobial agents for more than 48 h, which was not for prophylaxis and not initiated prior to PICU admission or within the first 48 h of PICU admission. Duration of PICU dependency, mechanical ventilation and hospital stay were studied with time-to-event analyses censored at 90 days. To account for death as a competing risk in these analyses, non-surviving patients were censored beyond all survivors at 91 days.
Statistical analysis
Patient characteristics are reported as medians with interquartile ranges or as numbers with percentages. Wilcoxon Rank Sum and Chi square tests were used to compare continuous data and proportions among groups.
We first investigated whether there was a statistical interaction between randomization to early-PN versus late-PN and the occurrence of hypophosphatemia for the impact of the randomized intervention. To this end, we performed multivariable logistic regression and Cox proportional hazard analyses entering the randomized intervention, the occurrence of hypophosphatemia and their interaction term to the models, while adjusting for baseline risk factors. These baseline risk factors included treatment center, age, type of illness (admission diagnosis), severity of illness (pediatric index of mortality (PIM) 3 score and pediatric logistic organ dysfunction (PeLOD) score over the first 24 h) and severity of malnutrition risk (screening tool risk on nutritional status and growth in children (STRONGkids) score dichotomized for medium and high risk) [11–13]. For the outcomes for which an interaction was identified, we then assessed the impact of early-PN versus late-PN in patients with or without hypophosphatemia separately, adjusting for the same baseline risk factors. Phosphate concentrations can be affected by several treatments, including renal replacement therapy, diuretics, insulin, corticosteroids and red blood cell transfusion. Therefore, we performed sensitivity analyses, further adjusting for those treatments that were significantly different between patients with or without hypophosphatemia on the day of analysis. Results were reported as adjusted odds ratios (aOR) or hazard ratios (aHR) with the corresponding 95% confidence intervals.
These analyses were performed with the available day 1 and day 2 phosphate data. As a sensitivity analysis, we repeated the analyses for the total study population after imputation of missing data. To this end, missing data on day 1 and day 2 (for those patients still in the PICU on day 2) phosphate concentrations were imputed by the multiple imputation with chained equations (MICE) technique [14]. For patients admitted in Leuven and Rotterdam, less than 30% of data were missing, as cut-off for minimizing bias and instability of the imputation model. For patients included in Edmonton, insufficient phosphate data were available to allow imputation, hence these patients were excluded from imputation modelling. Predictors for imputing missing day 1 and/or day 2 phosphate data in Leuven and Rotterdam included center, demographics (age, sex, weight, height, BMI), characteristics upon PICU admission (STRONGkids risk of malnutrition category, first 24 h PeLOD score, PIM3 score, elective or emergency PICU admission, reason for PICU admission, infection upon PICU admission, mechanical ventilatory support upon PICU admission, mechanical hemodynamic assist device upon PICU admission, randomization to early-PN or late-PN), and clinical outcomes (PICU mortality, 90 day mortality, acquisition of a new infection, duration of PICU stay, duration of hospital stay, duration of mechanical ventilatory support, 2-year mortality and 4-year mortality). The number of imputation models was set at 35 to avoid the loss of statistical power. Pooled results were joined with the available Edmonton phosphate data. Imputation was performed with MICE (version 3.17.0) in R (version 3.5.3). All other statistical analyses were performed with use of JMP ©Pro17.0.0 (SAS Institute, Inc., Cary, NC). For interaction P-values, the threshold for statistical significance was set at 0.1 to avoid missing an interaction that could be clinically relevant, according to the significance threshold set in the protocol of the original PEPaNIC study [10]. For other analyses, a P ≤ 0.05 was considered statistically significant.
Results
Characteristics of the study population
Phosphate concentrations were available for 1060 patients on day 1, of whom 211 (19.9%) had hypophosphatemia, and for 883 patients on day 2, of whom 243 (27.5%) had hypophosphatemia (Fig. 1). Table 1 shows the baseline characteristics of patients with versus without hypophosphatemia on day 1 and for patients with or without hypophosphatemia on day 2, revealing some imbalances requiring adjustments in multivariable models for subsequent analyses. Patients with hypophosphatemia on day 1 were younger, weighed less, were less admitted after cardiac surgery and more for medical reasons, were more severely ill, and were more randomized to early-PN as compared with patients without hypophosphatemia on day 1. Patients with hypophosphatemia on day 2 were less admitted after cardiac surgery and more for medical reasons, and were more randomized to early-PN as compared with patients without hypophosphatemia on day 2. The patients’ phosphate profiles in the first week in the PICU are shown in Fig. 2. The phosphate concentrations of patients with hypophosphatemia on day 1 gradually recovered over time, returning to above the lower limit of normal in more than 50% by day 3 and in more than 75% by day 6 (Fig. 2A, B). Patients with hypophosphatemia on day 2 had experienced a decrease in plasma phosphate from day 1 toward day 2 (Fig. 2C). Phosphate concentrations subsequently rose, with resolution of the hypophosphatemia in more than 50% by day 4 and in more than 75% by day 7 (Fig. 2C, D). Patients with hypophosphatemia on day 1 or day 2 had a significantly higher intake of total calories, total protein, total glucose and total lipids on the day preceding the phosphate measurement as compared with those without hypophosphatemia (all p < 0.0001, except non-significant p = 0.19 for total lipids on day 1) (Table 2).
Fig. 1.
Flow chart of the study participants. Number of patients with phosphate measurements on day 1 or day 2 are shown in the top panel. The lower panel shows the number of patients included after imputation of missing phosphate data for the patients who had been recruited in Leuven or Rotterdam. Abbreviations: PICU: pediatric intensive care unit
Table 1.
Patients’ baseline characteristics
| Patient characteristic | Day 1 | Day 2 | ||||
|---|---|---|---|---|---|---|
| Hypophosphatemia | No hypophosphatemia | P | Hypophosphatemia | No hypophosphatemia | P | |
| n | 211 | 849 | 243 | 640 | ||
| Age (years), median (IQR) | 0.83 (0.05–4.76) | 1.62 (0.31–6.90) | 0.0003 | 1.14 (0.25–4.25) | 0.93 (0.20–5.98) | 0.45 |
| Male sex, no. (%) | 114 (54.0) | 495 (58.3) | 0.26 | 147 (60.5) | 368 (57.5) | 0.41 |
| Weight (kg), median (IQR) | 8.4 (3.8–17.0) | 10.3 (5.0–21.0) | 0.0004 | 8.9 (4.4–16.0) | 8.4 (4.4–20.0) | 0.41 |
| Center, no. (%) | 0.0001 | 0.24 | ||||
| Leuven | 121 (57.4) | 616 (72.6) | 143 (58.9) | 413 (64.5) | ||
| Rotterdam | 87 (41.2) | 227 (26.7) | 95 (39.1) | 219 (34.2) | ||
| Edmonton | 3 (1.4) | 6 (0.7) | 5 (2.1) | 8 (1.3) | ||
| Admission diagnosis, no. (%) | < 0.0001 | 0.0020 | ||||
| Surgical | ||||||
| Abdominal | 17 (8.1) | 48 (5.7) | 21 (8.6) | 34 (5.3) | ||
| Burns | 2 (1.0) | 6 (0.7) | 1 (0.4) | 7 (1.1) | ||
| Cardiac | 51 (24.2) | 430 (50.7) | 81 (33.3) | 302 (47.2) | ||
| Neurosurgery | 20 (9.5) | 66 (7.8) | 23 (9.5) | 32 (5.0) | ||
| Thoracic | 7 (3.3) | 39 (4.6) | 13 (5.4) | 20 (3.1) | ||
| Transplantation | 3 (1.4) | 9 (1.1) | 2 (0.8) | 12 (1.9) | ||
| Orthopedic surgery–trauma | 12 (5.7) | 39 (4.6) | 15 (6.2) | 12 (1.9) | ||
| Other | 3 (1.4) | 28 (3.3) | 3 (1.2) | 14 (2.2) | ||
| Medical | ||||||
| Cardiac | 6 (2.8) | 30 (3.5) | 8 (3.3) | 28 (4.4) | ||
| Gastrointestinal–hepatic | 1 (0.5) | 4 (0.5) | 1 (0.4) | 4 (0.6) | ||
| Oncologic–hematologic | 1 (0.5) | 7 (0.8) | 3 (1.2) | 6 (0.9) | ||
| Neurologic | 24 (11.4) | 44 (5.2) | 19 (7.8) | 53 (8.3) | ||
| Renal | 1 (0.5) | 0 (0.0) | 0 (0.0) | 1 (0.2) | ||
| Respiratory | 41 (19.4) | 66 (7.8) | 35 (14.4) | 78 (12.2) | ||
| Other | 22 (10.4) | 33 (3.9) | 18 (7.4) | 37 (5.8) | ||
| Severity of illness, median (IQR) | ||||||
| PIM3 score a | −3.0 (−4.1 to −1.9) | −3.8 (−4.4 to −2.7) | < 0.0001 | −3.4 (−4.2 to −2.3) | −3.2 (−4.2 to −2.3) | 0.57 |
| PIM3 probability of death (%), median (IQR) b | 4.9 (1.7–13.3) | 2.3 (1.2–6.6) | < 0.0001 | 3.3 (1.5–9.2) | 3.8 (1.5–9.5) | 0.57 |
| PeLOD score first 24 h in PICU c | 21 (11–32) | 22 (12–32) | 0.063 | 21 (12–32) | 22 (12–32) | 0.29 |
| STRONGkids risk, no. (%) d | 0.14 | 0.20 | ||||
| Medium | 187 (88.6) | 780 (91.9) | 213 (87.7) | 580 (90.6) | ||
| High | 24 (11.4) | 69 (8.1) | 30 (12.4) | 60 (9.4) | ||
| Randomization to early-PN, no. (%) | 143 (67.8) | 379 (44.6) | < 0.0001 | 145 (59.7) | 299 (46.7) | 0.0006 |
a Higher PIM3 scores indicate a higher risk of mortality. b PIM3 probability of death, ranging from 0% to 100%, with higher percentages indicating a higher probability of death in PICU. c PeLOD scores range from 0 to 71, with higher scores indicating more severe illness. d STRONGkids scores range from 0 to 5, with a score of 0 indicating a low risk of malnutrition, a score of 1 to 3 indicating a medium risk, and a score of 4 to 5 indicating a high risk
IQR: interquartile range, PeLOD: pediatric logistic organ dysfunction score, PN: parenteral nutrition, PICU: pediatric intensive care unit, PIM3: pediatric index of mortality 3 score, STRONGkids: screening tool for risk on nutritional status and growth in children
Fig. 2.
Phosphate profiles in the first week in the PICU. The figures show the phosphate profiles of patients in the first week of the PICU stay. (A) Phosphate profiles of patients with or without hypophosphatemia on day 1 for different age groups. (B) Proportion of patients with hypophosphatemia on day 1 in whom phosphate concentrations returned to above the lower limit of normal on the subsequent days. (C) Phosphate profiles of patients with or without hypophosphatemia on day 2 for different age groups. (D) Proportion of patients with hypophosphatemia on day 2 in whom phosphate concentrations returned to above the lower limit of normal on the subsequent days. In panels A and C, box plots depict medians with interquartile ranges (IQR), and whiskers are drawn to the furthest point within 1.5 times the IQR from the box. The dashed horizontal lines represent the cut-offs below which hypophosphatemia was defined per age group. Abbreviations: IQR: interquartile range, PICU: pediatric intensive care unit
Table 2.
Patients’ clinical outcomes and treatments on the day preceding the phosphate measurements
| Patient characteristic | Day 1 | Day 2 | ||||
|---|---|---|---|---|---|---|
| Hypophosphatemia | No hypophosphatemia | P | Hypophosphatemia | No hypophosphatemia | P | |
| n | 211 | 849 | 243 | 640 | ||
| Clinical outcomes | ||||||
| New infection, no. (%) | 34 (16.1) | 110 (13.0) | 0.23 | 40 (16.5) | 104 (16.3) | 0.93 |
| Days in PICU, median (IQR) | 5 (3–9) | 3 (2–6) | < 0.0001 | 5 (3–8) | 4 (2–8) | 0.32 |
| 90-day mortality, no (%) | 14 (6.6) | 49 (5.8) | 0.63 | 19 (7.8) | 41 (6.4) | 0.46 |
| Days MV, median (IQR) | 3 (2–7) | 2 (1–4) | < 0.0001 | 2 (1–6) | 2 (1–5) | 0.92 |
| Days in hospital, median (IQR) | 13 (7–29) | 8 (6–18) | < 0.0001 | 13 (7–25) | 10 (6–21) | 0.048 |
| Treatments on the day preceding hypoPi | ||||||
| Nutrition a | ||||||
| Calories (kcal/kg), median (IQR) | 16.7 (7.6–30.4) | 8.0 (4.1–22.3) | < 0.0001 | 37.8 (16.9–58.6) | 24.6 (8.7–50.4) | < 0.0001 |
| Protein (g/kg), median (IQR) | 0.4 (0.0–1.3.0.3) | 0.0 (0.0–0.9) | < 0.0001 | 1.2 (0.1–2.3) | 0.4 (0.0–1.9) | < 0.0001 |
| Glucose (g/kg), median (IQR) | 3.6 (1.8–6.1) | 1.9 (1.0–4.6) | < 0.0001 | 6.3 (3.1–10.1) | 3.8 (1.9–8.7) | < 0.0001 |
| Lipids (g/kg), median (IQR) | 0.0 (0.0–0.0) | 0.0 (0.0–0.0) | 0.19 | 0.2 (0.0–1.1) | 0.0 (0.0–0.8) | < 0.0001 |
| Insuline dose (IU per day) | 1.6 (0.0–11.0) | 3.0 (0.0–10.0) | 0.13 | 1.0 (0.0–14.9) | 2.0 (0.0–9.4) | 0.89 |
| Renal replacement therapy, no (%) | 3 (1.4) | 4 (0.5) | 0.16 | 3 (1.3) | 6 (1.0) | 0.69 |
| Diuretics, no (%) | 23 (10.9) | 79 (9.3) | 0.48 | 81 (34.8) | 314 (51.0) | < 0.0001 |
| Corticosteroids, no (%) | 36 (17.1) | 114 (13.4) | 0.18 | 58 (24.9) | 121 (19.6) | 0.098 |
| Transfusion, no (%) | 32 (15.2) | 123 (14.5) | 0.80 | 30 (12.9) | 85 (13.8) | 0.72 |
a Total nutrition administered between PICU admission and the morning thereafter for day 1 or nutrition administered during the next 24 h for day 2
HypoPi: hypophosphatemia, IQR: interquartile range
After imputation of missing phosphate data, prevalence of hypophosphatemia on day 1 (264/1352, 19.5%) and day 2 (307/1144, 26.8%) remained similar (Fig. 1). We also observed roughly similar differences in baseline characteristics and nutritional intake between patients with or without hypophosphatemia as shown in Table S1−2 of the Additional File 1. Regarding baseline characteristics for patients with phosphate data on day 2, age and weight became significantly different after imputation, with younger age and lower weight in patients with hypophosphatemia on day 2 than in those without.
Relation of early hypophosphatemia with impact of early-PN versus late-PN on outcome
Within the patient group with phosphate measurements on day 1, there was no interaction between the presence of hypophosphatemia on that day and the randomized intervention for any impact of the intervention on outcome (Table 3). Interaction analyses performed after imputation of missing phosphate data, which are shown in Additional File 1 Table S3, yielded similar results.
Table 3.
Interaction between randomization to early-PN and hypophosphatemia for impact on clinical outcome
| Clinical outcome | Patients with phosphate data on day 1 | Patients with phosphate data on day 2 | ||
|---|---|---|---|---|
| aOR/aHR (95% CI) | P | aOR/aHR (95% CI) | P | |
| Risk of acquiring a new infection in the PICU | ||||
| Randomization to early-PN versus late-PN | 1.941 (1.166–3.232) | 0.0085 | 3.277 (1.961–5.474) | < 0.0001 |
| Hypophosphatemia versus no hypophosphatemia a | 1.125 (0.669–1.891) | 0.65 | 0.728 (0.437–1.211) | 0.20 |
| Interaction term | - | 0.31 | - | 0.031 |
| Likelihood of earlier live PICU discharge | ||||
| Randomization to early-PN versus late-PN | 0.826 (0.697–0.980) | 0.030 | 0.727 (0.620–0.852) | < 0.0001 |
| Hypophosphatemia versus no hypophosphatemia a | 0.868 (0.731–1.031) | 0.10 | 1.051 (0.896–1.234) | 0.54 |
| Interaction term | - | 0.89 | - | 0.030 |
| 90-day mortality | ||||
| Randomization to early-PN versus late-PN | 1.697 (0.706–4.079) | 0.22 | 2.392 (1.055–5.422) | 0.028 |
| Hypophosphatemia versus no hypophosphatemia a | 0.513 (0.208–1.267) | 0.13 | 0.824 (0.367–1.850) | 0.63 |
| Interaction term | - | 0.96 | - | 0.15 |
| Likelihood of earlier live weaning from MV | ||||
| Randomization to early-PN versus late-PN | 0.886 (0.749–1.049) | 0.16 | 0.760 (0.648–0.890) | 0.0007 |
| Hypophosphatemia versus no hypophosphatemia a | 0.926 (0.780–1.101) | 0.38 | 1.045 (0.890–1.227) | 0.59 |
| Interaction term | - | 0.44 | - | 0.025 |
| Likelihood of earlier live hospital discharge | ||||
| Randomization to early-PN versus late-PN | 0.910 (0.767–1.079) | 0.28 | 0.793 (0.674–0.933) | 0.0056 |
| Hypophosphatemia versus no hypophosphatemia a | 0.874 (0.734–1.039) | 0.12 | 0.963 (0.817–1.135) | 0.65 |
| Interaction term | - | 0.22 | - | 0.47 |
a Phosphate status on day 1 for patients with phosphate data available at day 1 or phosphate status on day 2 for patients with phosphate data available at day 2. Hazard and odds ratios were adjusted for center, age, type of illness, severity of illness (PIM3 and PeLOD scores) and severity of malnutrition risk (dichotomized STRONGkids score)
aHR: adjusted hazard ratio, aOR: adjusted odds ratio, CI: confidence interval, MV: mechanical ventilation, PeLOD: pediatric logistic organ dysfunction score, PICU: pediatric intensive care unit, PIM3: pediatric index of mortality 3 score, PN: parenteral nutrition, STRONGkids: screening tool for risk on nutritional status and growth
In patients for whom day 2 phosphate data were available, the presence of hypophosphatemia on that day did show a significant interaction with randomization to early-PN versus late-PN for its impact on risk of new infection (P = 0.031), likelihood of earlier live PICU discharge (P = 0.030) and likelihood of earlier live weaning from mechanical ventilation (P = 0.025), as well as a trend for interaction for the impact on 90-day mortality (P = 0.15) (Table 3). No (trend for) interaction was observed for the impact of randomization on likelihood of earlier live hospital discharge (P = 0.47). The observed interactions were explained by particular harm by early-PN in patients with hypophosphatemia on day 2 as compared with those without (shown as forest plots in Fig. 3, with corresponding numeric data given in Additional File 1 Table S4). Indeed, in patients with hypophosphatemia on day 2, the increased risk of new infection with early-PN versus late-PN was more pronounced (aOR 4.347 (1.658–11.397), P = 0.0010) than in patients without hypophosphatemia on day 2 (aOR 1.933 (1.211–3.084), P = 0.0052). Also, lower likelihood of earlier live PICU discharge (aHR 0.551 (0.406–0.750), P = 0.0002), lower likelihood of earlier live weaning from mechanical ventilation (aHR 0.639 (0.475–0.861), P = 0.0035), and higher risk of 90-day mortality (aOR 17.451 (1.595–190.938.595.938), P = 0.0038) with early-PN versus late-PN were only observed in patients with hypophosphatemia on day 2. Similar results were obtained after imputation of missing phosphate data (Fig. 3, Additional File 1 Table S3-4), in which case the interaction for 90-day mortality became statistically significant (P = 0.038). As patients with or without hypophosphatemia on day 2 differed for treatment with diuretics on day 2, we performed sensitivity analyses further adjusting for this use of diuretics. As shown in Additional File 1 Table S5, this yielded similar results.
Fig. 3.
Impact of early-PN versus late-PN on outcome of patients with versus without hypophosphatemia. Diamonds represent hazard ratios or odds ratios, wheras whiskers represent the corresponding confidence intervals. All analyses were adjusted for center, age, type of illness, severity of illness (PIM3 and PeLOD scores) and severity of malnutrition risk (dichotomized STRONGkids score). Pint values are the p-values for the interaction of randomization to early-PN vs. late-PN and hypophosphatemia vs. no hypophosphatemia for impact of the randomized intervention on the different outcomes. Primary analyses were performed based on available phosphate measurements, whereas sensitivity analyses were performed after imputation of missing phosphate data for the patients recruited in Leuven or Rotterdam. Abbreviations: PeLOD: pediatric logistic organ dysfunction score, PICU: pediatric intensive care unit, PIM3: pediatric index of mortality 3 score, Pint: interaction p-value, PN: parenteral nutrition, STRONGkids: screening tool for risk on nutritional status and growth
Discussion
In this secondary analysis of the PEPaNIC RCT, development of hypophosphatemia on day 2, but not on day 1, revealed a significant interaction with randomization to early-PN or late-PN for the impact of the randomized intervention on patient-centered outcomes. These included the risk of new infections, the duration of PICU dependency, the dependency on mechanical ventilation, and 90-day mortality. As compared with patients without hypophosphatemia on day 2, patients with hypophosphatemia on day 2 suffered from significantly more harm from early-PN. In patients with day 2 hypophosphatemia, early-PN independently associated with an increased risk of new infections, a prolonged PICU dependency, a prolonged dependency on mechanical ventilation and an increased 90-day mortality risk. In contrast, in patients without hypophosphatemia, early-PN only associated with an increased infectious risk, with a lower effect size than in patients with day 2 hypophosphatemia. Similar results were obtained after imputing missing phosphate data.
The results of this study are in line with previous studies performed in adults which showed that development of hypophosphatemia may identify critically ill patients who were not ready yet for enhanced nutrition [7–9]. The current study is the first to provide similar evidence in critically ill children. In adult studies, however, metabolic intolerance to enhanced nutrition was identified for phosphate decreases to the moderate to severe hypophosphatemia range. We previously found that such decreases rarely occur in children, which is likely explained by the higher normal phosphate ranges for children as compared with adults [15]. Given that recent research in adults has shown that metabolic intolerance to enhanced nutrition may also be revealed by less severe phosphate decreases than those in the moderate to severe hypophosphatemia range [9], we here studied the interaction of the randomized nutritional intervention with any degree of hypophosphatemia, i.e., any phosphate measurement below the age-adjusted normal range.
Patients with hypophosphatemia were younger, weighed less, were less often admitted after cardiac surgery and more for medical reasons and were more often randomized to early-PN. The higher rate of early hypophosphatemia in the early-PN group as compared with the late-PN group reflects a form of refeeding hypophosphatemia. The statistical interaction between randomization to early-PN versus late-PN and the phosphate status on day 2 for its impact on outcome measures revealed the importance of such hypophosphatemia as indicator of metabolic intolerance to early-PN. We did not observe an interaction with the impact of the nutritional intervention for day 1 hypophosphatemia. This suggests that a minimum dose or duration of exposure to nutrition is needed before metabolic intolerance is revealed, a possibility that requires further investigation. In general, low phosphate levels returned to above the lower limit of normal over the next days. Future research should investigate further mechanisms of harm by macronutrients in patients with feeding-induced hypophosphatemia. Interestingly, an observational study in adults associated higher protein intake, but not higher glucose or lipid intake with increased mortality in patients with refeeding hypophosphatemia [16]. In line with this, a previous secondary analysis of the PEPaNIC RCT suggested that the harm by early-PN was explained by the amino acid doses, and not by the glucose or lipid doses [17].
Large RCTs in critically ill children and adults have shown harm by early enhanced nutrition, with harm being present in all studied subgroups, including critically ill neonates and undernourished children [3, 6, 18, 19]. Mechanistic studies have attributed harm to increased metabolic damage and suppressed cellular repair processes [5, 20–22]. Yet, prolonged fasting is likely to become deleterious at some time point. Currently, metabolic (in)tolerance and (un)readiness to nutrition cannot be monitored by validated markers in critically ill children. The current study puts forward development of hypophosphatemia as a biomarker of metabolic intolerance to enhanced nutrition. However, unlike what was shown for adults, critically ill children who did not develop hypophosphatemia after a minimum exposure to enhanced nutrition were also harmed by early-PN, although the effect size was smaller. Hence, awaiting further research, the results support the current recommendation to withhold early-PN for all PICU patients [3, 4]. Future research should investigate whether and how phosphate dynamics can be integrated into a metabolic monitor that integrates other aspects of the metabolic response to nutrition. Moreover, whereas the present study pinpointed hypophosphatemia as a “reactive” marker responding to enhanced nutrition, future research should also investigate whether the metabolic intolerance to nutrition can be predicted, which would avoid potentially harmful exposure both to premature enhanced nutrition and to excessive fasting.
Strengths of this study include the large sample size and the embedding in a multicenter RCT with detailed prospective data collection. Furthermore, it provides the first evidence to support nutrition-induced phosphate alterations as biomarker of metabolic intolerance of enhanced nutrition in critically ill children. Nevertheless, the study also has limitations. Although we adjusted for baseline risk factors, we cannot exclude residual confounding. We cannot exclude potential bias due to missing phosphate measurements, although similar results were obtained after imputing missing data. Significant interaction for 90-day mortality as outcome was only observed after imputation of missing day 2 phosphate data. However, since a trend for interaction was already observed for the analysis on the available day 2 phosphate data, this difference is more likely explained by increased statistical power after imputation rather than driven by the imputation model. Nevertheless, impact of early-PN versus late-PN on 90-day mortality of patients with hypophosphatemia on day 2 must be interpreted with caution in view of the low absolute number of non-survivors in this relatively small subgroup and the imprecise estimates of a potential treatment effect, as indicated by the wide confidence intervals. Finally, upon PICU admission phosphate data were lacking and we only studied the interaction of the nutritional intervention with phosphate changes on the first 2 days. Whether these findings can be extrapolated to alterations later during PICU stay, requires further study.
Conclusions
In this secondary analysis of the PEPaNIC RCT, hypophosphatemia on day 2 identified patients who were particularly harmed by the use of early-PN. These findings raise the possibility that a decrease in phosphate levels may represent metabolic intolerance to artificial nutrition. Clinically, this suggests that early phosphate changes may help recognize critically ill children who are vulnerable to metabolic intolerance and for whom aggressive nutrition may need to be postponed or adapted. This thus opens perspectives for integrating daily phosphate measurements into a broader metabolic monitoring tool, the clinical value of which requires further study in prospective validation studies.
Supplementary Information
Acknowledgements
We thank the research team members involved in the study for technical and administrative support. We also thank the children and their parents for their willingness to participate in the study.
Abbreviations
- aHR
adjusted hazard ratio
- aOR
adjusted odds ratio
- CI
Confidence interval
- EPaNIC
Early versus Late Parenteral Nutrition in Critical Illness
- ICU
Intensive care unit
- MICE
Multiple imputation with chained equations
- PeLOD
Pediatric logistic organ dysfunction
- PEPaNIC
Pediatric Early versus Late Parenteral Nutrition in Critical Illness
- PICU
Pediatric intensive care unit
- PIM3
Pediatric index of mortality 3
- PN
Parenteral nutrition
- RCT
Randomized controlled trial
- STRONGkids
Screening tool risk on nutritional status and growth in children
Author contributions
IV, GVdB and JG designed the study. FG, LM, SCV and KFJ gathered data. IV and NUS analyzed the data. IV, NUS, GVdB and JG interpreted the data. IV, NUS, GVdB and JG wrote the manuscript, which was reviewed and approved by all authors.
Funding
This work was supported by European Research Council Advanced Grants (AdG-2012-321670 from the Ideas Program of the European Union 7th framework program, AdG-2017-785809 from the Horizon 2020 Program and AdG-2023-101133276 from the Horizon Europe Program to GVdB - Views and opinions expressed are however those of the authors only and do not necessarily reflect those of the European Union. Neither the European Union nor the granting authority can be held responsible for them); by the Methusalem Program of the Flemish government (through the University of Leuven to GVdB and IV, METH14/06); by the Institute for Science and Technology, Flanders, Belgium (through the University of Leuven to GVdB, IWT-TBM150181 and IWT-TBM110685); by the Research Foundation-Flanders (senior clinical investigator fellowship 1842724 N to JG, and G029525N research grant to JG); by KU Leuven (starting grant STG/23/032 to JG); by the Sophia Research Foundation (SSWO to SV); by the Stichting Agis Zorginnovatie (to SV); by the Erasmus Trustfonds (to SV); and by an European Society for Clinical Nutrition and Metabolism (ESPEN) research grant (to SV).
Data availability
The datasets generated and/or analyzed during the current study are not publicly available due ethical constraints. Data sets and other documents may be shared on reasonable request under the format of future collaborative projects that further elaborate on this specific research topic. Proposals for collaborative projects must be directed to the corresponding author and will be considered for approval. The data set required to address the research question will then only be shared after signing of a data access agreement. Data are located in controlled access data storage at KU Leuven.
Declarations
Ethics approval and consent to participate
The participating sites’ institutional review boards approved the original and follow-up studies (Ethical Committee Research UZ/KU Leuven: ML8052; Medische Ethische Toetsingscommissie Erasmus MC: NL49708.078; Ethical Committee Stollery’s Children Hospital: Pro00038098), which were performed in accordance with the 1964 Declaration of Helsinki and its amendments. Written informed consent was acquired from parents or legal guardians.
Consent for publication
Not applicable.
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Greet Van den Berghe and Jan Gunst contributed equally to this work.
References
- 1.Mehta NM, Compher C, Directors, ASPENBo. A.S.P.E.N. clinical guidelines: nutrition support of the critically ill child. J Parenter Enteral Nutr. 2009;33:260–76. [DOI] [PubMed] [Google Scholar]
- 2.Joosten K, Embleton N, Yan W, Senterre T, nutrition, EEECwgopp. ESPGHAN/ESPEN/ESPR/CSPEN guidelines on pediatric parenteral nutrition: Energy. Clin Nutr. 2018;37:2309-14. [DOI] [PubMed]
- 3.Fivez T, Kerklaan D, Mesotten D, Verbruggen S, Wouters PJ, Vanhorebeek I, et al. Early versus late parenteral nutrition in critically ill children. N Engl J Med. 2016;374:1111–22. [DOI] [PubMed] [Google Scholar]
- 4.Tume LN, Valla FV, Joosten K, Jotterand Chaparro C, Latten L, Marino LV, et al. Nutritional support for children during critical illness: European Society of Pediatric and Neonatal Intensive Care (ESPNIC) metabolism, endocrine and nutrition section position statement and clinical recommendations. Intensive Care Med. 2020;46:411–25. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Gunst J, Vanhorebeek I, Verbruggen SC, Dulfer K, Joosten KF, den Van Berghe G. On how to feed critically ill children in intensive care: a slowly shifting paradigm. Clin Nutr. 2025;46:169–80. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Casaer MP, Mesotten D, Hermans G, Wouters PJ, Schetz M, Meyfroidt G, et al. Early versus late parenteral nutrition in critically ill adults. N Engl J Med. 2011;365:506–17. [DOI] [PubMed] [Google Scholar]
- 7.Doig GS, Simpson F, Heighes PT, Bellomo R, Chesher D, Caterson ID, et al. Restricted versus continued standard caloric intake during the management of refeeding syndrome in critically ill adults: a randomised, parallel-group, multicentre, single-blind controlled trial. Lancet Respir Med. 2015;3:943–52. [DOI] [PubMed] [Google Scholar]
- 8.Olthof LE, Koekkoek W, van Setten C, Kars JCN, van Blokland D, van Zanten ARH. Impact of caloric intake in critically ill patients with, and without, refeeding syndrome: a retrospective study. Clin Nutr. 2018;37:1609–17. [DOI] [PubMed] [Google Scholar]
- 9.Lauwers C, Langouche L, Wouters PJ, Wilmer A, den Van Berghe G, Casaer MP, et al. Early phosphate changes as potential indicator of unreadiness for artificial feeding: a secondary analysis of the EPaNIC RCT. Crit Care. 2025;29:48. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Fivez T, Kerklaan D, Verbruggen S, Vanhorebeek I, Verstraete S, Tibboel D, et al. Impact of withholding early parenteral nutrition completing enteral nutrition in pediatric critically ill patients (PEPaNIC trial): study protocol for a randomized controlled trial. Trials. 2015;16:202. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Hulst JM, Zwart H, Hop WC, Joosten KF. Dutch national survey to test the STRONGkids nutritional risk screening tool in hospitalized children. Clin Nutr. 2010;29:106–11. [DOI] [PubMed] [Google Scholar]
- 12.Leteurtre S, Martinot A, Duhamel A, Proulx F, Grandbastien B, Cotting J, et al. Validation of the paediatric logistic organ dysfunction (PELOD) score: prospective, observational, multicentre study. Lancet. 2003;362:192–7. [DOI] [PubMed] [Google Scholar]
- 13.Straney L, Clements A, Parslow RC, Pearson G, Shann F, Alexander J, et al. Paediatric index of mortality 3: an updated model for predicting mortality in pediatric intensive care*. Pediatr Crit Care Med. 2013;14:673–81. [DOI] [PubMed] [Google Scholar]
- 14.Wulff. Multiple imputation by chained equations in praxis: guidelines and review. Electron J Bus Res Methods. 2017;15:41–56. [Google Scholar]
- 15.Veldscholte K, Veen MAN, Eveleens RD, de Jonge RCJ, Vanhorebeek I, Gunst J, et al. Early hypophosphatemia in critically ill children and the effect of parenteral nutrition: a secondary analysis of the pepanic RCT. Clin Nutr. 2022;41:2500–8. [DOI] [PubMed] [Google Scholar]
- 16.Slingerland-Boot R, Rooijakkers E, Koekkoek K, van Blokland D, Arbous S, van Zanten A. Macronutrient intake and outcomes of ICU patients with refeeding hypophosphatemia. Clin Nutr ESPEN. 2023;55:191–9. [DOI] [PubMed] [Google Scholar]
- 17.Vanhorebeek I, Verbruggen S, Casaer MP, Gunst J, Wouters PJ, Hanot J, et al. Effect of early supplemental parenteral nutrition in the paediatric ICU: a preplanned observational study of post-randomisation treatments in the pepanic trial. Lancet Respir Med. 2017;5:475–83. [DOI] [PubMed] [Google Scholar]
- 18.van Puffelen E, Vanhorebeek I, Joosten KFM, Wouters PJ, den Van Berghe G, Verbruggen S. Early versus late parenteral nutrition in critically ill, term neonates: a preplanned secondary subgroup analysis of the PEPaNIC multicentre, randomised controlled trial. Lancet Child Adolesc Health. 2018;2:505–15. [DOI] [PubMed] [Google Scholar]
- 19.van Puffelen E, Hulst JM, Vanhorebeek I, Dulfer K, Van den Berghe G, Verbruggen S, et al. Outcomes of delaying parenteral nutrition for 1 week vs initiation within 24 hours among undernourished children in pediatric intensive care: a subanalysis of the pepanic randomized clinical trial. JAMA Netw Open. 2018;1:e182668. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Vanhorebeek I, Casaer M, Gunst J. Nutrition and autophagy deficiency in critical illness. Curr Opin Crit Care. 2023;29:306–14. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Vanhorebeek I, Gunst J, Derde S, Derese I, Boussemaere M, Guiza F, et al. Insufficient activation of autophagy allows cellular damage to accumulate in critically ill patients. J Clin Endocrinol Metab. 2011;96:E633-45. [DOI] [PubMed] [Google Scholar]
- 22.Hermans G, Casaer MP, Clerckx B, Guiza F, Vanhullebusch T, Derde S, et al. Effect of tolerating macronutrient deficit on the development of intensive-care unit acquired weakness: a subanalysis of the EPaNIC trial. Lancet Respir Med. 2013;1:621–9. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The datasets generated and/or analyzed during the current study are not publicly available due ethical constraints. Data sets and other documents may be shared on reasonable request under the format of future collaborative projects that further elaborate on this specific research topic. Proposals for collaborative projects must be directed to the corresponding author and will be considered for approval. The data set required to address the research question will then only be shared after signing of a data access agreement. Data are located in controlled access data storage at KU Leuven.