A reduced-carbohydrate and lactose-free formulation for stabilization among hospitalized children with severe acute malnutrition: A double-blind, randomized controlled trial

Background Children with medically complicated severe acute malnutrition (SAM) have high risk of inpatient mortality. Diarrhea, carbohydrate malabsorption, and refeeding syndrome may contribute to early mortality and delayed recovery. We tested the hypothesis that a lactose-free, low-carbohydrate F75 milk would serve to limit these risks, thereby reducing the number of days in the stabilization phase. Methods and findings In a multicenter double-blind trial, hospitalized severely malnourished children were randomized to receive standard formula (F75) or isocaloric modified F75 (mF75) without lactose and with reduced carbohydrate. The primary endpoint was time to stabilization, as defined by the World Health Organization (WHO), with intention-to-treat analysis. Secondary outcomes included in-hospital mortality, diarrhea, and biochemical features of malabsorption and refeeding syndrome. The trial was registered at clinicaltrials.gov (NCT02246296). Four hundred eighteen and 425 severely malnourished children were randomized to F75 and mF75, respectively, with 516 (61%) enrolled in Kenya and 327 (39%) in Malawi. Children with a median age of 16 months were enrolled between 4 December 2014 and 24 December 2015. One hundred ninety-four (46%) children assigned to F75 and 188 (44%) to mF75 had diarrhea at admission. Median time to stabilization was 3 days (IQR 2–5 days), which was similar between randomized groups (0.23 [95% CI −0.13 to 0.60], P = 0.59). There was no evidence of effect modification by diarrhea at admission, age, edema, or HIV status. Thirty-six and 39 children died before stabilization in the F75 and in mF75 arm, respectively (P = 0.84). Cumulative days with diarrhea (P = 0.27), enteral (P = 0.42) or intravenous fluids (P = 0.19), other serious adverse events before stabilization, and serum and stool biochemistry at day 3 did not differ between groups. The main limitation was that the primary outcome of clinical stabilization was based on WHO guidelines, comprising clinical evidence of recovery from acute illness as well as metabolic stabilization evidenced by recovery of appetite. Conclusions Empirically treating hospitalized severely malnourished children during the stabilization phase with lactose-free, reduced-carbohydrate milk formula did not improve clinical outcomes. The biochemical analyses suggest that the lactose-free formulae may still exceed a carbohydrate load threshold for intestinal absorption, which may limit their usefulness in the context of complicated SAM. Trial registration ClinicalTrials.gov NCT02246296.


Introduction
Children with complicated severe acute malnutrition (SAM) are admitted to the hospital because they are severely ill or unable to feed sufficiently [1]. In African hospitals, their risk of inpatient death ranges between 10%-30% [2][3][4]. In contrast, children with SAM who are clinically stable, i.e., without signs of illness and who have an appetite (uncomplicated SAM), are usually treated in community-based programs and have a substantially lower mortality risk, ranging between <1%-7% [5][6][7]. In hospital settings, mortality may be reduced to some extent by adhering to World Health Organization (WHO)-recommended management [2,4,8,9], but this may not address the full spectrum of infections and metabolic abnormalities of these seriously ill children [10][11][12]. In the 2013 Lancet Maternal and Child Nutrition series, improving management of SAM was identified as having the greatest likely impact on child mortality amongst nutritional interventions [13].
Current guidelines for the nutritional management of SAM in the hospital define 3 phases of treatment [14]: 1) the "stabilization phase," during which children are fed a liquid diet (standard F75 [F75]) with a relatively low-protein (approximately 9 g/l) and relatively low-energy content (75 kcal/100 ml). F75 was designed to meet the estimated nutritional requirements to restore physiological and metabolic functions and to prevent refeeding syndrome while medical conditions stabilize. No weight gain is expected during this phase of treatment; 2) the "transition phase," during which higher protein and energy through either F100 formula or readyto-use therapeutic foods (RUTFs) are started with supplemental F75 formula; and 3) the "rehabilitation phase," with an increased daily intake of F100 or RUTFs in order to achieve catch-up growth. Once a child has stabilized and tolerates RUTFs, WHO guidelines recommend discharge from hospital care, with continuation of the rehabilitation phase continued in the community [14,15].
The original F75 formulation was designed based on what was known about the pathophysiology of SAM at that time, before separation of treatment of complicated and uncomplicated SAM [16]. Today's inpatient children with complicated SAM have a very different clinical profile from uncomplicated SAM treated in the community. Around 65% of calories in the F75 formula is derived from carbohydrates (maltodextrin, lactose, and sucrose). Once ingested, disaccharides such as lactose or sucrose are hydrolyzed into monosaccharides by disaccharidases found at the tip of small intestinal villi. Released monosaccharides such as glucose and galactose are then transported across the apical membrane through Na + -dependent glucose transporters, whereas fructose is taken up through a facilitative fructose transporter [17]. However, evidence suggests that children with SAM have impaired absorption of mono-and disaccharides, regardless of the presence of gastroenteritis [18,19]. Limited histological evidence also shows intestinal villous atrophy [18], which is consistent with clinical signs suggestive of carbohydrate malabsorption [20]. Thus, we hypothesized that revising the current F75 into a modified, lactose-free, and reduced-carbohydrate F75 formulation could reduce osmotic diarrhea and thereby lower the number of days to complete the first phase of treatment.
Early deterioration in children with SAM may also be related to refeeding syndrome, comprising severe metabolic derangements driven by insulin excretion after a sudden shift from a catabolic to an anabolic state [21]. Refeeding syndrome is characterized by hypophosphatemia, hypokalemia, and hypomagnesemia, which may impair cardiac, pulmonary, and neurological function and can result in (sudden) death. As protein synthesis is stimulated in anabolism, increased production of adenosine triphosphate (ATP) leads to a higher cellular demand for phosphate [21]. Furthermore, pancreatic insulin secretion induces the cellular uptake of glucose and electrolytes, including phosphate and potassium, and hypophosphatemia is common during nutritional rehabilitation of malnourished children and is associated with mortality [22][23][24]. We also hypothesized that a reduction in carbohydrate content of the modified F75 formula could lower the risk of refeeding syndrome. Together, reformulated F75 could plausibly improve early clinical outcomes among hospitalized children with SAM.
We conducted a randomized, double-blind controlled trial evaluating a modified F75 formula with substantially reduced carbohydrate content and without lactose versus the currently recommended formulation of F75 among hospitalized children with SAM in Kenya and Malawi.

Study design and participants
The study was a randomized, double-blind controlled trial conducted at two Kenyan hospitals (Kilifi County Hospital and Coast General Hospital, Mombasa) and one Malawian hospital (Queen Elizabeth Central Hospital, Blantyre). All children admitted to these study hospitals were screened for complicated SAM. The study was discussed in detail with parents or carers of potential participants in their local language, and written informed consent was sought. In patients for which re-establishing feeding was urgent, initial verbal assent was sought and followed up with written consent after treatment initiation.
Inclusion criteria were • age 6 months to 13 years; • SAM defined as mid-upper arm circumference (MUAC) < 11.5 cm or weight-for-height Z score < −3 if younger than 5 years of age, BMI Z score < −3 if older than 5 years, or edematous malnutrition at any age; • being admitted to the hospital because of medical complications or failing an appetite test, as defined by WHO guidelines [16]; • weight-for-length/height or BMI, weight-for-age, and length/height-for-age Z scores were calculated using WHO 2006 and 2007 references.
Exclusion criteria were • lack of informed consent, • known allergy to milk products

Investigational product
We compared F75 to the modified F75 (mF75) formulation (Table 1). Osmolarity was 298 mOsm/L and 232 mOsm/L for F75 and mF75, respectively. The formulations were isocaloric, and the reduced energy from carbohydrates was compensated by increased mediumchain triglycerides (MCTs), while protein and micronutrient composition were unaltered. Both formulations were manufactured by Nutriset (Nutriset, Malaunay, France) and dispensed by trained staff according to WHO recommendations: 95 kcal/kg per day, divided into 8 feeds per 24 hours.

Randomization and masking
Sequential study numbers were computer generated using Stata Statistical Software: Release 12 (College Station, TX: StataCorp LP) with random block sizes for each site prior to the trial. An independent statistician was responsible for computerized sequence generation. Children were allocated study numbers sequentially at each site using sealed numbered envelopes. Group allocation was blinded to participants and all trial personnel, including trial coordinators and the principal investigator. The study products were identical in appearance and packaged in color-coded sachets.

Procedures
Children were managed in the hospital as per Malawian or Kenyan national guidelines, both of which are based on WHO recommendations [15,16]. HIV testing by rapid antibody test was offered to all participants according to national guidelines, with appropriate counseling, follow-up tests, and referrals offered depending on results. Children were examined daily, and clinical status was recorded on a standardized proforma during ward rounds. If a child deteriorated after stabilization, they were returned to the stabilization phase and received F75 or mF75 as originally allocated. Blood and fecal samples were collected at admission and after 3 days of hospitalization.
The clinical team was blinded for the allocated formula and decided whether a child was fit for discharge based on criteria such as being in rehabilitation phase, clinical condition, ability to finish RUTF, and appetite level. RUTF, nutritional counseling, and follow-up in a nutrition clinic were provided at discharge, as per standard of care at each institution.

Primary and secondary outcomes
The primary outcome was based on the purpose of F75, which is clinical stabilization, after which a child can progress to higher protein and energy feeds to promote catch-up growth. We therefore chose the number of days between admission and first stabilization as the primary outcome, defined as having reached the transition phase of treatment and switched to another feed type. Stabilization was based on the WHO guideline [15]: 1. absence of any WHO "danger" or emergency signs: obstructed breathing, respiratory distress, cyanosis, shock (delayed capillary refill plus fast and weak pulse plus temperature gradient), severe anemia (Hb < 5 g/dl), congestive cardiac failure, impaired consciousness, convulsions, severe dehydration, profuse watery diarrhea or vomiting, hypothermia; 2. loss of edema (if present on admission), defined as improving from a severe +++ edema (severe: generalized bilateral pitting edema) to ++ edema (moderate: no upper arm or upper leg edema and no facial edema), or from ++ edema to + edema (mild: only feet/ankle edema), or none; 3. and tolerating the full prescribed volume of F75 feeds and observed to be completing the feeds.
Study clinicians received additional training on the WHO guidelines as well as bedside and scenario-based training on the criteria for transition to improve standardization across sites. In order to include children who died before stabilization in the analyses, they were classified as "not stabilized." Although there is some subjectivity in determining a child's medical condition that is "improving," we chose stabilization as a main outcome rather than episodes of diarrhea because 1) it is the purpose of F75; 2) once a child switches to F100 or RUTFs, earlier transition could conceivably result in an increase in diarrhea unrelated to mF75 or F75; 3) diarrhea is very difficult to quantify objectively since maternal recall is not accurate in identifying episodes or severity [25]; 4) diapers may not distinguish every stool episode, and their weight is affected by urine; and 5) refeeding syndrome may be critical to early outcomes.
Other prespecified secondary outcomes were 1. mortality in hospital; 2. time to discharge from hospital in days; 3. the total number of days spent in the stabilization phase, including periods when children deteriorated and returned to the stabilization phase; 4. percentage change in weight between admission and day 5; 5. serum sodium, potassium, magnesium, calcium, phosphate, and albumin at day 3 and changes between admission and day 3; 6. stool osmotic gap and pH at day 3.

Laboratory analyses
Blood was collected at admission and day 3, from which serum was obtained and stored at −80 C. For analysis, samples were thawed overnight at 4˚C, vortexed, and centrifuged at 1,500 g for 3 minutes. Albumin, alanine aminotransferase (ALT), creatinine, calcium, magnesium, potassium, phosphate, sodium, and urea were batch analyzed using the P-module on a Roche Modular (Roche, Indianapolis, United States) at a single laboratory (University Medical Centre Groningen, the Netherlands). Fecal samples were collected at both time points and stored at −80˚C. Fecal biochemistry was batch analyzed only amongst patients with clinical diarrhea in the same laboratory. These samples were thawed overnight at 4˚C and weighed, diluted 3.5 times with distilled water, homogenized (Precellys 24 Homogenizer, Bertin, Rockeville, USA), and centrifuged at 16,100 g/rcf for 5 minutes. Fecal water was extracted, and osmolality was measured by freezing point depression (Osmo Station OM-6050, Arkray, Tokyo, Japan). In addition, sodium, potassium, and chloride were analyzed using the ISE900-module on Cobas 8000 (Roche, Indianapolis, USA). pH was determined using a standard pH meter. The stool osmotic gap was calculated by the equation 290 − 2 × (stool Na + stool K).

Statistical analyses
A statistical analysis plan was reviewed and approved by an independent Data and Safety Monitoring Committee (DSMC) before the treatment allocation was unmasked, and the research database was locked. The DSMC oversaw the trial and reported to the Trial Steering Committee (TSC). An interim unblinded analysis was conducted by the DSMC during the trial after 50% of participants were recruited. A recommendation to the sponsor to discontinue recruitment, in all patients or in selected subgroups, would be made by the TSC on advice from the DSMC if the data provide proof beyond reasonable doubt that one of the treatment arms is better in terms of the primary outcome or safety guided by the Haybittle-Peto criteria. The Haybittle-Peto boundary states that if an interim analysis shows a probability of less than 0.001 that the treatments are different, then the trial should be stopped early.
For the primary endpoint of time to first stabilization, we regarded it as essential to include children who died before stabilization in the analysis. Therefore, the time to stabilization of children who died was entered as "999," ensuring that they ranked bottom in the rank sum test and were included in the longest quartile of the IQR. An intention-to-treat analysis was conducted including all available data, including from children given the wrong formula at one or more feeds (primary endpoint evaluable) and children who withdrew or absconded, i.e., left the hospital without notifying the medical team or were transferred prior to stabilization (primary endpoint nonevaluable). We calculated the median (IQR) time to stabilization in the 2 randomized groups and compared their distributions using a Wilcoxon rank sum test and incidence rate ratio using Cox regression. We also conducted prespecified subgroup analyses to assess effect modification by site, HIV antibody status, edema status, and age using likelihood ratio tests. In addition, using the cmprsk R-package, we conducted competitive risk analysis to compare the cumulative incidence functions of the mutually exclusive risks of either dying prior to stabilization or achieving stabilization. Differences in these incidence functions indicate whether the cumulative probability of either dying before stabilization or achieving stabilization differs between groups as treatment progresses.
For secondary outcomes, we compared the number of days with prespecified clinical features between allocated groups using regression with a zero-inflated negative binomial distribution, since many children had no days with these problems. To compare the distributions of continuous variables, we used Student t test or Wilcoxon's rank sum test if the variables were skewed. We compared categorical variables between randomized groups using χ 2 or Fisher's exact tests. Trends in proportions across ordered groups were assessed by a nonparametric test for trend. HIV test results were analyzed with "not done" as a separate category, as some children who died before testing or whose parents refused testing may have been more likely to have HIV infection or exposure. Blood and stool biochemical variables between intervention arms were compared while adjusting for site, whereas models comparing admission and day 3 included a random intercept for patients. All statistical analyses were done using either Stata

Sample size
Based on prior trial experience [28], we estimated that 50% of children allocated to F75 would stabilize by day 5, and we calculated the sample size required to demonstrate a 10% absolute increase in percentage stabilized at day 5 to 60% in the intervention arm (equivalent to a hazard ratio of 1.35). With an alpha of 5% and power of 80%, the number of patients required was 381 per arm, and to allow for 10% anticipated losses, we aimed to recruit 420 per arm.

Ethical considerations
The trial was registered at ClinicalTrials.gov, number NCT02246296. Ethical
Of a total of 7,098 child days of hospitalization, 2,615 were prior to stabilization, and 75/843 (8.9%) children died prior to stabilization. Nine children voluntarily withdrew from the study prior to stabilization, and 6 absconded from the hospital (as defined above). Seven children were transferred to another hospital, 4 of whom because of hospital closure due to a health worker strike. Nine children (1.0%) received the wrong formula at one or more feeding times and were included in the primary analysis.

Primary outcome
The median time to stabilization was 3 days (IQR 2-6 days) (Fig 2, Table 3, and S1 Fig), and there was no evidence for difference in time to stabilization between randomized groups; absolute difference 0.23 (95% CI −0.13 to 0.60, P = 0.59). Prespecified subgroup analyses did not suggest effect modification by the presence of diarrhea at admission, age, edema, or site (S1 Table). There was evidence of an effect of HIV status (P < 0.0001) on both time to stabilization and time to mortality but no interaction between HIV status and effect of allocated feeds (P � 0.33, S2 Table).

Secondary clinical outcomes
There was no effect of the intervention on mortality before stabilization (P = 0.84) or during hospitalization overall (P = 0.44, Table 3). As presented in S3 Table, 117 (14%) children experienced a serious adverse event before stabilization without evidence of difference between randomized groups (P = 0.59). Diarrhea was common; of the 416 children without diarrhea at admission, 176 (38%) developed diarrhea in hospital. The number of days with diarrhea prior to stabilization and during the whole hospitalization period were similar in both groups (Table 3). There were no significant differences in the number of days with vomiting or receiving rehydration fluids (ReSoMal or IV fluids). The types of other feeds children received (nonstandard milks) and routes (nasogastric or oral) were not affected by randomized allocation, except breast milk, which was lower in the mF75 group (P = 0.04). Hypoglycemia (<3.0 mmol/l) rarely occurred, and the number of days with an episode of hypoglycemia was not influenced by the intervention ( Table 3). The number of days receiving antibiotics was also not affected by the intervention and neither was the mean (±SD) percent of weight change between admission and day 5 (3.1% ± 5.8% in nonoedematous, P = 0.78; −3.4% ± 7.1% in oedematous patients, P = 0.63). Although uncommonly diagnosed, the incidence of cliniciandefined congestive heart failure was lower in the mF75 group (P = 0.007).  Table 4 shows the serum and fecal biochemistry. There were no differences between the overall distributions of biochemical concentrations in serum between the standard and mF75 on day 3. Hyponatremia was common, both at admission and after 3 days (Table 5), without differences between the groups, whereas hypernatremia was rare. The prevalence of abnormal potassium concentrations was not different between the groups. Hypokalemia was frequently found but decreased in prevalence between admission and day 3, whereas hyperkalemia increased. Hypermagnesemia was more common than hypomagnesaemia at admission, but not on day 3, and no differences were found between the F75 and mF75 group. Hypophosphatemia was present in 10% of children at admission, declining to 5.9% at day 3, with higher prevalence in the mF75 group (P = 0.009). Among children with diarrhea on day 3, when stool collection was undertaken, 88% had a fecal osmotic gap >100 mOsm/kg, indicating osmotic diarrhea (Table 6). There were no differences between the randomized groups in the prevalence of low <50 mOsm/kg or high >100 mOsm/kg stool osmotic gap at day 3.

Discussion
We tested a significantly modified formulation of F75 among hospitalized children with SAM. There was no evidence of effect on the primary outcome of time to stabilization. The median time to stabilization was 3 days, which was shorter than anticipated, possibly due to careful provision and timing of F75 feeds but within the range suggested by WHO guidelines [14]. Combined with the fact that there was also no evidence of differences in any of the secondary outcomes, we conclude that an approximately 30% reduction in carbohydrate load without lactose is not more effective than standard F75. Similarity in the primary and secondary Modified formulation for children with severe acute malnutrition outcomes between groups does not suggest that the result was due to a lack of statistical power. The overall inpatient case fatality ratio of 15% was similar to estimates reported before by the participating sites [12,28].
There is strong evidence from observational studies that carbohydrate digestion and absorption is impaired in severely malnourished children [18,19]. Consequently, ingesting carbohydrates beyond the intestinal capacity for absorption can lead to osmotic diarrhea. Diarrhea has been associated with mortality in severely malnourished children [29,30], and cases of osmotic diarrhea often present with clinical complications of electrolyte disturbances and dehydration. However, there is limited evidence from intervention studies. Among children with SAM, Kerpel-Fronius and colleagues gave boluses of 2-2.5 g/kg of different types of carbohydrates and found that lactose administration especially lowered stool pH, a marker of carbohydrate malabsorption [31]. These sugar boluses were in a physiological range that is similar to a 3-hourly feed of F75 and suggested that changing lactose to sucrose might aid absorption. Kukurozovic and colleagues evaluated the effects of different lactose-free formulas in malnourished Australian aboriginal children presenting with diarrhea and found significant differences in the duration of diarrhea between the different formulas, potentially related to their osmolality [32]. The modification of energy from carbohydrates in our study was substantial (approximately 30% reduction) and suggests no linear relationship between the amount of carbohydrates ingested and stool frequency or consistency in this population. This finding is consistent with mF75 still exceeding a threshold for intestinal absorption among children who had diarrhea. Most experience on feeding children with severely impaired intestinal function has been gained from treating short bowel syndrome, for which carbohydrates can be a limiting factor in the ability to advance feeds, thus use of complex carbohydrates is commonly advised [33]. The carbohydrates used in the F75 formula apart from lactose are, for the most part, maltodextrin, which is a complex carbohydrate but shorter and with more loosely bound links than starches. Limited evidence has suggested that starches might be beneficial in patients with short bowel syndrome, which could also be relevant for children with SAM [34]. However, a more radical reformulation of F75 would have to ensure supply of energy in a form that prevents hypoglycemia. We are aware of an ongoing study evaluating whether lactose-free formula affects the outcome of severely malnourished children (www.isrctn.com/ ISRCTN98124592). Diarrhea in SAM is multifactorial; infectious causes and intestinal inflammation as well as impaired digestion of fat and protein related to exocrine pancreas and hepatobiliary dysfunction likely impact the overall nutrient absorptive capacity, and these pathways would not be directly ameliorated by our intervention [12,35,36]. Most cases of diarrhea were osmotic in nature rather than secretory, suggesting that bacterial infections such as enterotoxin-producing Escherichia coli or rotavirus were likely not contributing substantially to the diarrhea at day 3. We did not determine diarrhea etiology to estimate the contribution of infectious causes.
Biochemical findings suggest that refeeding syndrome was relatively uncommon on day 3 in our study. The incidence of refeeding syndrome during treatment of severely malnourished children is not well established due to the lack of clear defining criteria and serial biochemical measurements. The prevalence of hypophosphatemia (<0.70 mmol/l) on day 3, an indicator of refeeding syndrome, was relatively low at 5.9% in our cohort but significantly higher in the mF75 group, although the median serum phosphate did not differ. We found no other differences in biochemical features of refeeding syndrome. Kimutai and colleagues reported an increase in hypophosphatemia (<1.20 mmol/l) from 6% on admission to 22% on the second day of hospitalization of children with SAM [37]. A recent study in Uganda reported hypophosphatemia on the second day of hospitalization in 58% of children with SAM who died and 13% in the children who survived [24]. The cutoff was much higher than in our study at 1.6 mmol/l for infants and 1.1 mmol/l in older children. In the latter study by Rytter and colleagues, children with diarrhea were given a rice porridge, which has a higher carbohydrate content than F75, and although not a trial, mortality was reported to be higher in this group. Another possible explanation for the low prevalence of severe biochemical signs of refeeding syndrome in SAM might also be related to the observation that insulin response is impaired in SAM children [38]. We choose to replace the carbohydrates with lipids, in particular with MCTs, for a number of reasons. First, MCTs are more rapidly hydrolyzed in the intestinal lumen than long-chain triglycerides, and MCTs do not require pancreatic lipase or bile for intestinal absorption. Secondly, MCTs are more easily absorbed, as they bypass the lymphatic system. MCTs require no or minimal carnitine to enter mitochondria, which could be beneficial if carnitine levels are deficient. Finally, lipid malabsorption is thought not to lead to dehydration or electrolyte disturbances. Despite these potential theoretical beneficial effects for critically ill malnourished children, our study did not suggest that increased MCTs impact overall clinical outcome.
Strengths of this trial are the study size; the double-blind and multicenter design, thereby enhancing its generalizability; and it being carried out in routine practice conditions. We captured detailed daily clinical information, which gave us unique insight into clinical evolution. There has been a long-standing debate on how to best transition children from the F75 formula to RUTFs or the F100, with similar concerns about increasing carbohydrate load and the risk of exacerbating or new onset diarrhea [15]. This study also directly informs this debate and suggests that reducing the carbohydrate content during transition or rehabilitation might not impact clinical outcome. There is one recent trial from our group that supports this, although more and larger trials are needed for generalizability [39]. The main limitation of the study is that the main outcome, stabilization with transition to the next phase of therapy, was based on WHO guidelines, comprising clinical evidence of recovery from acute illness as well as metabolic stabilization evidenced by recovery of appetite. We chose this main outcome as opposed to diarrhea for a number of reasons. Firstly, stabilization is the purpose of using F75 milk formula. Diarrhea is difficult to assess objectively, particularly in our setting. Maternal recall is not accurate in identifying diarrhea or assessing its severity [25]. Alternatively, the systematic observation of diapers could have been introduced; however, considering the added clinical load, the implementation in our setting was not feasible. For example, determining stool frequency would have required very frequent checks over 24-hour periods by already short-staffed research nurses. Importantly, the lack of improvement in individual daily clinical signs in our trial strongly suggests that the similar results obtained from the different formulations on the primary outcome is not due to subjective variation in the clinical decision of transitioning a child from the stabilization phase.
In conclusion, using a reduced-carbohydrate, lactose-free F75 did not improve the time to stabilization or other clinical outcomes of children with complicated severe malnutrition. The results support the ongoing use of the current F75 formulation and suggest that empiric use of lactose-free formula for children with diarrhea is not needed. It is important to explore whether our findings are replicated in an Asian context, as there could be Africa/Asia differences in SAM and associated infections. Finally, the carbohydrate content and composition are not the only determinants of F75 efficacy, and other aspects of treatment are also likely to be important.

S1 CONSORT checklist.
(PDF) S1 Protocol. Randomized controlled trial of a reduced-carbohydrate formulation of F75 therapeutic milk among children with SAM. SAM, severe acute malnutrition. (DOCX) S1 Table. Primary endpoint and prespecified subgroup analyses. All data are median (IQR) or geometric means (95% CI), as indicated. (DOCX) S2 Table. Competitive risk models. All data are presented as median (IQR). (DOCX) S3 Table. List of serious adverse events. (DOCX) S1 Fig. Probability of stabilization and death between the two treatment groups. Competing risk analysis simultaneously compares the estimated cumulative incidence curves of mutually exclusive events: first stabilization (solid line) or death prior to stabilization (dashed lines). Withdrawals and absconded cases were censored. Differences in cumulative incidence functions between F75 (blue line) and mF75 (black line) and all subgroup analysis models were compared using Gray's test. Significance threshold, P < 0.05. F75, standard F75; mF75, modified F75. (TIF)