A cross-sectional analysis of factors associated with the development of refeeding

19 Background: Refeeding syndrome (RFS) is a life-threatening, underdiagnosed, and under-20 researched complication in treating children with severe acute malnutrition (SAM). This study 21 aimed to determine the incidence and onset of RFS and identify biochemical abnormalities, 22 clinical signs, and complications associated with RFS development in children 0–59 months 23 treated in a South African public hospital setting. 24 Methods: A retrospective cohort study was performed on hospital files of children diagnosed 25 with SAM at Rahima Moosa Mother and Child Hospital, Johannesburg, from 1/10/2014 to 26 31/12/2018. A total of 148 files could be retrieved from the hospital archives. The diagnosis of 27 SAM based on the World Health Organization definition was confirmed in 126 of these children, 28 and they were included in the study. The onset of RFS among the children included in the study 29 was diagnosed based on published criteria for RFS. Children who developed RFS and those who 30 did not were compared concerning biochemistry and clinical signs and symptoms on admission.


Introduction
Severe acute malnutrition is a term used by the World Health Organization in children under five years of age to indicate severe undernutrition and is diagnosed by the occurrence of either severe wasting, i.e., weight-for-length/height <-3 standard deviation (SD) on the WHO growth standards; a mid-upper arm circumference (MUAC) <11.5 cm in children between 6 and 59 months; or the presence of bilateral pitting edema [1].According to the WHO, following the WHO management guidelines for SAM may reduce the case fatality rate to below 10% [2].However, Tickell & Denno highlighted that these guidelines were developed mainly based on expert opinion rather than scientific evidence and that additional research is vital to improving the guidelines and children's outcomes [2].
Since World War II, it has been known that initiating feeds in a starved patient may potentially contribute to their fatality [3,4].When feeding is initiated in an individual after starvation, a sudden shift from catabolism to anabolism and a switch back to carbohydrate metabolism causes blood glucose and insulin levels to rise and phosphate, potassium, and magnesium to rise and shift into the intracellular space [5].Fluid retention may also increase the extracellular fluid volume [5].Thus, the syndrome is characterized by a drop in blood phosphate to below normal levels (hypophosphatemia) within five days [6] of initiating nutritional therapy.Clinical symptoms range from mild to severe and may be associated with cardiac arrhythmias, cardiac failure, renal failure, and death [6].Preventing the onset of RFS in patients treated with SAM may improve patient outcomes and reduce mortality.However, most RFS studies focus on adults, children with anorexia, and those in intensive settings.Very little research has focused on RFS in the context of SAM [7][8][9][10][11], while guidelines for treating SAM [1] and RFS [5,[7][8][9] have .been developed independently.Although the WHO guidelines for the treatment of SAM include progressing feeding slowly [1], a lack of research in this field makes it difficult to determine whether this approach is optimal for preventing RFS in this vulnerable population.
In the African context, only two published studies thus far have reported RFS incidence in children admitted with SAM [4,7].Mbethe & Mda reported on 104 children with SAM treated according to the WHO management protocol in a teaching hospital in Gauteng Province [4], while Okinyi studied 160 children with SAM in a Kenyan public hospital after initiating feeding with F-75 [7].Considering the paucity of data, this study aimed to determine the incidence of RFS and explore associated admission data that might indicate risk factors for the development of RFS in South African children aged 0 -59 months diagnosed with SAM and treated according to the WHO treatment guidelines in a public health setting.

Methods
Ethics approval was obtained from the Health Sciences Research Ethics Committee of the University of the Free State (UFS-HSD2018/0154/2602), and permission to do the study at the Rahima Moosa Mother and Child Hospital was granted by the Gauteng Department of Health and the hospital management.

Study design, setting, and study population
A retrospective cohort study of retrievable hospital files of children aged 0 -59 months admitted with SAM to Rahima Moosa Mother and Child Hospital, Coronationville, Johannesburg, South Africa, from 1 October 2014 to 31 December 2018, was conducted.
. According to electronic statistics kept by the Dietetics Department of the hospital, 592 children with SAM were admitted during the period under review; however, only 148 of the hospital files of these children (25% of the identified total) could be retrieved from the Hospital Archives Department.The diagnosis of SAM was verified based on the World Health Organization definition [1] according to the weight, length, MUAC, and presence of edema captured in these files.The diagnosis of SAM could be confirmed for 126 children who were then eligible for inclusion in the current study.Thus, the sample represented 21% of the children admitted to the hospital with SAM during the time under review.

Data collection
Anthropometry (growth indicators), age, gender, ethnicity, country of origin, and clinical outcomes were recorded from the hospital files.Biochemical values associated with SAM, RFS, or prognosis were recorded; these included blood levels of electrolytes and minerals (phosphorus, potassium, magnesium, calcium, and sodium), indicators of kidney function (urea and creatinine), inflammation and metabolic stress (C-reactive protein (CRP) and albumin), hemoglobin, coagulation (platelets and international normalized ratio (INR), and indicators of liver function (total bilirubin, alanine aminotransferase (ALT), alkaline phosphatase (ALP), gamma-glutamyl transferase (GGT) and aspartate aminotransaminase (AST).Values were interpreted according to published cut-offs.All clinical signs and medical complications recorded in the admission files were documented. .

Data analysis
The onset of RFS among the children included in the study was retrospectively diagnosed if a drop in blood phosphate levels by >0.16 mmol/L to below 0.65 mmol/L occurred after feeding was initiated.There is no universally accepted definition of RFS [6], but according to a recent systematic review [3], this definition by Marik and Bedigian [17] is one of the most frequently used.The sample was stratified into children who developed RFS (the RFS-positive group) and those who did not (the RFS-negative group).Recorded variables were compared between the two groups.

Statistical analysis
The data from the hospital files were captured in Microsoft Excel (2013) and analyzed using SAS® version 9.4, copyright© 2014 (SAS Institute Inc. SAS and all other SAS Institute Inc. product or service names are registered trademarks or trademarks of SAS Institute Inc., Cary, NC, USA.).
Medians and percentiles were used to describe numerical data, and frequencies and percentages to describe categorical data.The groups were compared using contingency tables applying the Kruskal-Wallis test for numerical data and Chi-square or Fisher's exact tests, as applicable, for categorical data.A value of p<0.05 was considered statistically significant.

Results
The study included 126 children in whom the diagnosis of SAM could be retrospectively confirmed, of which 11 (8.7%) developed RFS after feeding was initiated.The children had a median age of 34.2 months (IQR: 26.0 to 48.4 months).The sample was stratified according to children that developed RFS (RFS-positive group) and those that did not (RFS-negative group).
The anthropometry, length of hospital stay, and mortality are summarized in Table 2.The RFSpositive group was younger, shorter, and weighed less than the RFS-negative group, but the differences were not statistically significant.MUAC was recorded for fewer than half of the children (n=60; 47.6%) and did not differ between RFS positive and negative groups.Edema was present in 63.6% and 39.1% (p=0.11), and severe wasting in 45.5% and 60.9% (p=0.32) of the RFS positive and RSF negative groups, respectively.The median duration of hospitalization was 12 days (IQR: 8 to 17 days).The median stay in hospital was significantly longer for the RFSpositive group than for the RFS-negative group (18 vs 12 days; p=0.003).
Overall, children gained a median of 0.6 g/kg/day (IQR: 0.2 to 0.9 g/kg/day) during their hospital stay.Nineteen (19) children (15.1%) died in hospital, and the median time until death was nine days (IQR: 2 to 13 days).However, the mortality rate did not differ significantly between the RSF positive (n=2; 18.2%) and the RSF negative groups (n=17, 14.8%).with urinary tract infection (UTI), while 22.2% received nasogastric tube feeding.A significantly higher percentage of children who developed RFS presented with diarrhea (p=0.04),dehydration (p=0.02), and UTI (p=0.04) on admission than those that did not.The RFS positive group included significantly fewer children who were HIV exposed (had a negative PCR negative test but whose mother was positive) than the RFS negative group (p=0.03).

Discussion
Very few studies, and only one other South African study to date, have focused on RFS in children with malnutrition.This retrospective study among children with SAM treated in a South African public hospital setting recorded an incidence of RFS of 8.7% and found that children who developed RFS were significantly more likely to present with hypophosphatemia, hypokalemia, hyponatremia, dehydration, coagulopathy, and UTI on admission and stayed in the hospital for significantly longer than children who did not develop RFS.
In the current study, the incidence of RFS in children admitted to hospital with SAM was lower than the 15% incidence found by Mbethe and Mda [4] in a teaching hospital in Gauteng, South Africa, and the 21% found by Okinyi [7] in Kenya.However, comparing these three studies is complicated because they used different definitions to define the onset of hypophosphatemia.
The current study used the definition identified in a recent systematic review as one of the .most used, but its accuracy in predicting adverse outcomes has not been tested in children with SAM [4,5].Therefore, to effectively study associated risk factors and outcomes of RFS in the context of SAM, a universally accepted definition for hypophosphatemia to indicate the onset of RFS, which is currently lacking [6], needs to be developed.
Hypophosphatemia was present in 20% (n=25) of the 126 children in the current study (Table 3), which concurs with other studies that have reported low phosphate levels on hospital admission for children with SAM [9,18].In agreement with the findings of Mbethe and Mda [4], hypophosphatemia was significantly (p=0.04)associated with RFS in the current study [5].
Notably, the serum phosphate levels were done on admission before feeds had been initiated, whereas RFS usually occurs within five days of starting to refeed.Therefore, this low phosphate might be a prognostic marker for developing RFS.Moreover, several studies have found that low admission phosphate levels are associated with an increased risk of dying [19,20].Thus, phosphate assessment is indicated in all patients with SAM to allow for proper supplementation and referral [18,20].However, regular blood tests to detect a dropping serum phosphate may not be feasible in many settings where SAM is rife, so it is critical to identify risk factors for RFS.
In the current study, 30 (24%) children with SAM presented with severe hypokalemia and 25 (20%) with hyponatremia on admission.According to previous studies, hyponatremia and hypokalemia are the common electrolyte disturbances in malnourished children, made worse by diarrhea, vomiting, and dehydration [21,22].In addition, hypokalemia and hypomagnesemia frequently occur in children with SAM because of muscle loss and kidney dysfunction due to reductive adaptations during starvation [1,18].Diarrhea [4,19] and dehydration [18] are also .commonly associated with RFS or hypophosphatemia.In the current study, about half of the children presented with diarrhea and/or dehydration on admission, while children who developed RFS during their hospital stay presented with a significantly higher presence of diarrhea (p=0.04),dehydration (p=0.02),severe hypokalemia (p<0.001)and/or hyponatremia (p=0.004) on admission than those who did not.
Coagulopathy has not been described as a risk factor for RFS.However, a report by the WHO identified it as a poor prognostic factor in children admitted with kwashiorkor [23].De Maayer and Saloojee found that having a prolonged clotting time (INR >1.7) conferred the highest risk of death among children with SAM [24].In the current study, a significantly higher percentage of children in the RFS-positive group presented with an INR >1.7 on admission than the children in the RFS-negative group (p=0.049), which might reflect the severity of SAM, but further research is required.
Other clinical features commonly associated with RFS or hypophosphatemia are edema [4,18,19] and dermatosis [4,18], which also commonly occur in children with SAM [1].In the current study, 41% of the children presented with edema on admission.However, although the prevalence of edema was higher in the RFS-positive group than in the RFS-negative group (63.6% vs 31.9%), the difference did not reach statistical significance.Similarly, the prevalence of dermatosis, reported in 31% of the children, was not significantly different between the two groups.
Urinary tract infections are common in malnourished children, and the risk of UTI increases with the severity of malnutrition [25].Most studies of children with SAM have been conducted . in inpatient facilities and have reported a high prevalence of UTI; South Africa reporting among the highest at up to 42% [25].In the current study, UTI was present in around 20% of the children on admission and was significantly more prevalent (p=0.04) in the group that developed RFS than in those that did not.
In the Kenyan study [7], the prevalence of RFS was significantly associated with HIV infection (but not with dehydration status), while HIV infection was found in a South African study to be associated with an increased risk of death in children with SAM and RFS [24].The current study in which 18.3% of the children were HIV positive and 50% HIV exposed did not confirm the association with RFS.Moreover, for reasons that are not currently clear, RFS developed significantly more in children who were not HIV exposed than those who were.
The mortality rate amongst the hospitalized South African children with SAM was 20% to 30% in 2012 [26].Mbethe and Mda [4] reported a mortality rate of 9.5% and noted that most of the children in their study who died had developed RFS.Mortality rates in the current study were 15.1%: 18.2% and 14.5% in the children who developed RFS and those that did not, respectively.However, this may not be a true representation of the study population as 91.7% (n=24) of hospital files of children who demised could be obtained, compared to only 37.7% (n=124) of hospital files for those who were discharged from hospital, indicating retrieval bias.
The true mortality in both groups might thus be lower.The only other South African study on RFS in children with SAM reported that 6% of children who developed RFS died [4].RFS also resulted in a significantly longer duration of hospitalization (18 vs 12 days, p=0.003), which is similar to observations in older adults [12].The average weight gain during hospitalization for the children with RFS was 0.9 kg (or 7.7 g/kg/day) and 0.5 kg (7.2 g/kg/day) for those without .RFS.These differences were not statistically significant, and the trend for higher weight gain in the RFS-positive group may have been attributed to the longer length of hospitalization.

Limitations
A limitation of this study was the difficulty in obtaining files from the hospital archives, producing a small sample size with potential retrieval bias.In addition, the researchers had to rely on the information recorded in the hospital files and records completed by others.Thus, there was no way of knowing if measurement errors such as inaccurate weighing, measuring of length/height, and MUAC had been made by medical personnel on admission or if errors or omittance occurred when the data was originally captured in the hospital files.Nevertheless, the sample size and findings compare well with similar studies in Africa [4,7].

Conclusion and recommendations
Refeeding syndrome likely contributes to the high mortality rates experienced in children with SAM, although no direct association with mortality was found in the current study.The association between hypophosphatemia and mortality in children with SAM has been demonstrated previously [18,27].Therefore, children at risk for developing RFS need to be identified early and monitored closely.The authors recommend that children with SAM and hypophosphatemia, hypokalemia, hyponatremia, INR >1.7, diarrhea, or dehydration need a cautious approach to feeding, regular electrolyte monitoring, and replacement of deficient nutrients. .

Table 3
summarizes the abnormal biochemical findings, showing that on admission, a significantly larger percentage of children in the RFS-positive group presented with hypophosphatemia (p=0.04) and severe hypokalemia (0.0005), hyponatremia (p=0.004), and elevated INR (p=0.049)than in the RFS negative group.Table 4 summarizes the clinical signs and symptoms recorded on admission.Complications on admission were very common, and of the 126 children with SAM in this study, 46.8% presented with vomiting, 52.4% with diarrhea, 53.2% with acute gastroenteritis (AGE), 49.2% with edema, 31% with dermatosis, 10.3% with hypoglycemia, 32.5% with hyperglycemia, 1.6% with hypothermia, 36.5% with pneumonia, 42.9% with respiratory complications, 17.5% with sepsis, .