Refeeding syndrome (RS) is an acute metabolic disturbance that occurs upon reintroduction of oral, enteral nutrition (EN), or parenteral nutrition (PN) after prolonged fasting or suboptimal feeding (1). The body transitions from a catabolic to an anabolic state. This results in intracellular demand for inorganic phosphorus (P), potassium (K), magnesium (Mg), and thiamine (vitamin B1), causing hypophosphatemia, hypokalemia, and/or hypomagnesemia (1–3). RS can manifest itself as a mild electrolyte disorder with no associated clinical symptoms or as a severe electrolyte disorder leading, without supplementation, to severe organ failure, such as respiratory and cardiac failure, musculoskeletal weakness, and Wernicke’s encephalopathy (4–6). The mortality associated with RS ranges from 0 to 71% (7). In the pediatric intensive care unit (PICU), a mortality rate of 6% was observed among severely malnourished South African children who developed RS (8).

Children at risk of developing RS are those who have significantly reduced their energy intake for seven to ten days before the reintroduction of nutrition, and those who are undernourished (4–6, 9). The PICU population is thus at risk of RS at the time of refeeding (4), considering the high prevalence of malnutrition in this population, ranging from 15–25% (10–14), and the recommendation of introducing early EN within 24 hours of PICU admission (15).

The lack of consensus diagnostic criteria and the frequency of electrolyte disorders not associated with diet makes the diagnosis of RS complex in critically ill children (9). Therefore, data regarding its incidence is lacking. To our knowledge, only the American study by Dunn et al. (1999), conducted in children receiving PN in intermediate care and in the PICU, showed that 9% of the children included were at risk of RS but did not provide a clear incidence of RS (16). The authors described an electrolyte drop in 27% of subjects one day after PN, the most common being hypophosphatemia (16).

In March 2020, the American Society for Parenteral and Enteral Nutrition (ASPEN) published a consensus recommendation on, among other things, diagnostic criteria for RS for adults and children (4). These were: “A decrease in any 1, 2, or 3 of serum phosphorus, potassium, or magnesium levels by 10%-20% (mild RS), 20%-30% (moderate RS), or > 30% (severe RS), and/or organ dysfunction resulting from a decrease in any of these and/or due to thiamin deficiency (severe RS); And occurring within 5 days of reinitiating or substantially increasing energy provision” (without specifying what a significant increase is).

The consequences of refeeding syndrome can be serious in critically ill children, who are a vulnerable population. Therefore, the main aim of this study was to measure the incidence of RS according to the ASPEN definition in critically ill children with nutritional support (NS), hospitalized in the PICU. The secondary objectives were to define the degree of severity of RS and to compare children with RS to others.

The primary outcome of the study was the incidence of RS among critically ill children with NS admitted to the PICU. We decided to measure two incidences: one among the overall sample and one among the undernourished children. The number of children at each stage of the RS assessment is expressed as absolute (n) and relative (%) frequencies. The incidence of RS was calculated using the following formulas, with the results expressed as relative frequencies (%):

-Number of children with probable RS divided by the number of children in the baseline sample.

-Number of children with probable RS divided by the number of children at risk for RS (BMI z-score ≤ –2 SD).

Secondary outcomes were to describe children with probable RS and to compare them to all children without RS and to children at risk who did not developed RS. The following data were extracted from the studies by Valla et al. (10, 17) or from electronic patient records and compared among groups: child characteristics (age, sex, type and diagnosis of admission, Pediatric Logistic Organ Dysfunction I (PELOD I) and Pediatric Index of Mortality Score II (PIM II) severity scores (10, 17), respiratory support (duration of invasive ventilation and non-invasive ventilation), length of stay, acquired infections as defined by the Center for Disease Control (10) and mortality), electrolyte determination, nutritional support including supplementation (P, K, and/or Mg), and confounding factors.

To describe children with probable RS, categorical variables (including levels of RS severity) were expressed as absolute (n) and relative (%) frequencies. The distribution of quantitative variables was assessed by means of their mean, standard deviation, median, interquartile range, skewness, kurtosis, histogram, and box plot. Because the quantitative variables were not normally distributed, they were described by their median (p50) and interquartile range (Q1–Q3).

To compare children with probable RS to at-risk children who did not develop RS, the chi-square statistical test of homogeneity was used for categorical variables. However, no inference could be made on the categorical variables, which were too small to meet the conditions for the application of this statistical test. Quantitative variables were compared between the groups using the non-parametric Wilcoxon Mann-Whitney rank sum test. No statistical tests could be performed to compare children with probable RS and children without RS, because the sample size was so imbalanced between these two groups.

Statistical tests were conducted in a two-sided manner and the significance level was set at 5% (p < 0.05).

Statistical analyses were performed using StataIC^® version 16 software (StataCorp, College Station, TX, United States).

The characteristics of the participants are described in Table 1. Children were classified into three groups: those who developed probable RS (“RS child”), those without RS (“without RS”), and those who were at risk of RS but who did not developed RS (“at RS risk”). The PELOD I and PIM II scores were significantly higher in “RS child” than in the “at RS risk” (p = 0.003 and p = 0.011). The length of PICU stay was the longest in “RS child” than in “at RS risk” (p < 0.001). The frequency of acquired infection during the hospital stay was also the highest in “RS child” than in “at RS risk” (p = 0.002). Compared to the “at RS risk” group, “RS child” were more frequently mechanically ventilated (p = 0.001).

The frequency of significant electrolyte disorders (P, K and/or Mg) was 67.3% in the baseline sample (see Supplementary Figure 1).

Forty-four children (47.3%) of “RS child” had a significant disorder in one of the three electrolytes (P, K, or Mg), 42 (45.1%) in two electrolytes, and 7 (7.6%) in all. Seventy-four children (79.6%) had a significant phosphate disorder, the majority of which was severe (55.4%). A total of 61 children (65.6%) had a significant potassium disorder, the majority of which was also severe (39.3%). Fourteen children (15.1%) had a significant Mg disorder, the majority of which were mild (64.3%) (Figure 3).

The percentage of electrolyte disorders according to the number of determinations in children with probable RS (“RS child”) is presented in Table 2. Figure 4 shows the frequency of electrolyte determination by group. In “RS child,” Mg was measured more often than “at RS risk” (p = 0.005).

The types of nutritional support and supplementation provided to the children in the different groups are shown in Table 3. Most children were enterally fed (91.8%) and few received parenteral nutrition (15.8%). “RS child” received PN more frequently than “at RS risk” (p = 0.019). They also received a combination of EN and PN more frequently than “at RS risk” (p = 0.023). They received more supplementation with electrolytes than “at RS risk” (p = 0.001), mainly with P (p = 0.001).

Energy intake via NS for “RS child,” expressed as a percentage of the energy target, is shown for each day in Figure 5. The energy target was already covered by day 2 (101.2%; IQR:58.9-142.5). Significant energy intake are described in Figure 6.

The frequencies of confounding factors for the children in the three groups are shown in Table 4. Most children had two and more confounding factors in “RS child” group, while many children in “at RS risk” had none or one confounding factors (p = 0.041). The frequency of diuretic use was also the highest in “RS child” (“at RS risk,” p < 0.001).

When we compared the children with probable RS with the “at RS risk,” we did not observe any significant differences in age, gender or reason for admission to the PICU. However, we observed that they had a longer PICU length of stay. The latter was almost 50% longer than that of children at risk of RS. This was consistent with the fact that they had factors that could impact their PICU length of stay: they had significantly higher severity scores, were more frequently on mechanical ventilation, and had higher rates of acquired infections. This is consistent with the ASPEN (4) which defines acquired immunodeficency syndrome as a clinical condition associated with an increased risk of RS.

The mortality rate in children with probable RS was 6.5%. This was consistent with the study by Mbethe and Mda (8), which found a 6% mortality rate among severely malnourished children in the PICU who developed RS. In this study, we were not able to statistically compare the results with those of the children at risk of RS due to small sample size.

Most of the children in our study were enterally fed (91.4–94.4%). This figure is in line with guidelines advocating that EN should be preferred in critically ill children (37). Regarding the management of NS for children with probable RS, the median energy intake on day 1 was high (75% of the energy target), and the energy target was already covered by day 2 (101.2%). For almost all children (95.7%), intake at day 1 was considered significant according to our criteria (> 25% of the energy target). These results could partly explain the incidence of RS, and the severity of cases. However, the protocol of RS management of the Hospices Civils de Lyon reached the energy target after three to five days of refeeding. Several factors could explain this discrepancy. Children with RS were more likely to receive a combination of EN and PN than others. The addition of PN during the first five days significantly increases energy intake (37) and may increase the risk of overfeeding. However, there is no consensus regarding the energy target during the first days of refeeding (5). The ASPEN recommends that energy intake should not exceed 40–50% of the target (4). Others have described an intake of 20–75% of the target energy during the first few days of refeeding as “safe” (1, 5) (i.e., a range of one to more than three times). Therefore, the criteria chosen in this study to define significant energy intake were strict compared to others. Finally, children with probable RS were supplemented more than others with P, K, and/or Mg, which is consistent with the protocol for identifying and correcting electrolyte disorders.

The incidence of RS among critically ill children at risk was high (46.7%), and severe most of the time. The PICU length of stay and acquired infections were greater in children who developed RS than in those at risk who did not. This underscores the importance of screening children for RS in the PICU, particularly in children with higher severity scores, in order to diagnose, prevent, treat (including thiamine supplementation) and monitor (with systematic P, Mg and K assays) RS adequately. Our study also revealed that the lack of consensus on precise refeeding recommendations may lead to excessive intake and may promote RS. This finding demonstrates the importance of establishing new progressive refeeding recommendations for the first few days after PICU admission. Finally, RS is difficult to assess in the PICU, and prospective studies using all ASPEN risks and diagnostic criteria are needed.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by Scientific and Ethical Committee of Hospices Civils de Lyon, France. Written informed consent was obtained from the individual(s), and minor(s)’ legal guardian/next of kin, for the publication of any potentially identifiable images or data included in this article.

FV, CCJ, SB, and TV contributed to the study conception. FV and CCF recruited the patients and collected the data in previous studies [study references: (10, 18)]. FV and FB extracted additional data from electronic patient records. SB and TV calculated nutritional intake and performed the statistical analyses and analyzed the results in collaboration with FV and CCJ. SB, TV, FV, CCJ, and LT interpreted the data. LT edited the English. SB and TV wrote the draft manuscript, which was reviewed and approved by all the authors. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.