Clinical and Biochemical Markers of Risk in Uncomplicated Severe Acute Malnutrition

The OptiDiag study followed a multicentric design and was conducted in Bangladesh, Burkina Faso, and Liberia. We present here data obtained at the time of enrollment, before therapeutic refeeding was begun.

The geographic, demographic, and epidemiological profiles of the study areas are detailed in the Supplemental Information.

Children with SAM were identified and referred to treatment by active and passive screening by community health workers and health staff (see Supplemental Information). Nonedematous, uncomplicated children with SAM aged 6 to 59 months were recruited into the study at the time of admission to therapeutic feeding programs. Exclusion criteria were bilateral pitting edema; medical complications requiring inpatient care, including inability to drink or breastfeed, bloody diarrhea, uncontrolled vomiting, convulsions, lethargy, or unconsciousness; known peanut and/or milk allergies; congenital malformations that affect food intake; and plans to leave the catchment area within the next 6 months.

Our recruitment process (Supplemental Information) was aimed at ensuring a balanced representation of each anthropometric phenotype over the 1-year implementation period to account for seasonal variability.

Anthropometric measurements were taken in duplicate and followed standard WHO recommendations for children in this age group (Supplemental Information).^11,12  WHZ and height-for-age z scores (HAZ) were calculated by using WHO growth standards.¹³ 

Leptin levels were compared among the 3 experimental groups and interpreted in comparison to levels in normal, healthy infants and children.¹⁴  We also compared the percentage of children in each group with serum leptin levels <35 pg/mL because this cutoff had high predictive value for mortality among hospitalized patients with SAM.^9,10 

We measured retinol-binding protein (RBP), serum ferritin, and soluble transferrin receptor (sTfR) as biomarkers of vitamin A and iron status. Because RBP and ferritin are known to inaccurately reflect nutrition status in the presence of subclinical inflammation, correction factors were calculated to adjust their values.¹⁵  Thresholds for defining abnormal biochemical parameters in blood are described in the Supplemental Information.

All analytes were measured by using enzyme-linked immunosorbent assays and, with rare exception, were performed in duplicate.

During clinical examinations, nurses measured vital signs, including temperature and heart and respiratory rates. Presence or absence of respiratory infection and changes in skin and hair were recorded. Visible severe wasting, pneumonia, hydration state, and clinical evidence of vitamin A and iron deficiencies were assessed by using WHO guidelines for the integrated management of childhood illness (IMCI) and the WHO Training Course on the Management of Severe Malnutrition: Principles of Care.^16,17 

Caretakers were asked to report recent changes in their children’s general state of health, body weight, and food intake.

The study was exploratory in nature. Therefore, we employed the general rule of thumb that a minimum sample size of 30 is sufficient for normal approximation.¹⁸  To account for possible difficulties in collecting biological samples, this number was increased to a minimum of 50 patients per anthropometric phenotype in each study site.

Characteristics of the study population were summarized as percentages, means ± SDs, or medians and interquartile ranges (IQRs) if not normally distributed. Differences in clinical features, health and nutritional histories, and biochemical data among the 3 phenotypes were assessed by using χ² and Kruskal-Wallis tests; when statistical differences were detected (P < .05), we compared the following groups: (1) WHZ only versus MUAC only; (2) both low WHZ and low MUAC versus MUAC only, which tests the effect of adding low WHZ to low MUAC; and (3) all MUAC (MUAC <115 mm with or without WHZ <–3) versus children with low WHZ alone, who would be denied eligibility to SAM treatment under a program abandoning the use of WHZ as an independent criterion of detection and admission.

Construction of bivariate and multivariate logistic models is described in the Supplemental Information. Confounding and/or effect modification by site, age, sex, severe stunting, and presence of a low sitting/standing ratio z score (SSRZ) was assessed.¹⁹  All analyses were performed by using Stata version 13 (Stata Corp, College Station, TX); P values <.05 were considered statistically significant.

A total of 473 patients were recruited between March 2017 and June 2018. Refusal of caregivers for participation of their children in the study was rare. Twenty-two patients (4.7%) were excluded from analysis because of inclusion error: 4 patients (0.9%) were younger than 6 months at the time of admission, and 18 patients (3.8%) were only moderately malnourished. The distribution of patients across phenotype groups was as follows: 138 (30.6%), low WHZ only; 152 (33.7%), both low WHZ and low MUAC; 161 (35.7%), low MUAC only. In total, 142 patients were from Bangladesh, 165 were from Burkina Faso, and 144 were from Liberia. Each of the 3 phenotypes represented approximately one-third of the sample size in each study site; there was no statistically significant difference (P = .681 from Pearson’s χ² test) in phenotype proportion among study sites (Table 1).

Statistically significant differences in age, sex, stunting, and SSRZ were detected among phenotypes (Table 1). Compared with children with low WHZ only, children with low MUAC only were younger, more stunted, and more often girls and had a lower proportion of body length in the legs. Children with both low WHZ and low MUAC fell in the range between children with either low WHZ or low MUAC.

Biochemical data, clinical features, and health and nutritional histories among phenotypes at admission are shown in Table 2. Significant differences were observed and prompted pairwise comparisons, beginning with leptin, the primary outcome measure. Unadjusted and adjusted analyses of continuous data are presented in Table 3. Unadjusted bivariate pairwise analyses of categorical data are presented in Table 4 and adjusted multivariate models that included age, sex, stunting, proportion of body length in the legs, and country as covariates are presented in Table 5.

Among the 3 anthropometric phenotypes, median serum leptin levels were lowest in patients with both low WHZ and low MUAC (180.10 pg/mL; IQR, 87.7 to 346.8), followed by those with low WHZ only (215.80 pg/mL; IQR, 122.0 to 362.6), and was highest in those with low MUAC only (331.50 pg/mL; IQR, 159.0 to 560.1). Unadjusted and adjusted models for pairwise comparisons confirmed the statistical significance of the observed differences between the patients with low MUAC only and the 2 other groups (Table 3). As illustrated in Fig 1, this pattern was consistent across the 3 study sites. There were relatively few children with severe hypoleptinemia (≤35 pg/mL); however, the proportions of children with leptin levels below this cutoff followed a similar trend, with the highest rate of severe hypoleptinemia in children with both low WHZ and low MUAC (Table 2). Leptin levels among all patients correlated far more strongly with WHZ (r = 0.287, P < .001) than with MUAC (r = 0.102, P = .03; Fig 2).

Similar patterns were found for a wide range of clinical and biochemical indicators of nutritional deficit, hydration, and iron balance, as well as caretaker-reported changes in nutritional and health status (Table 4). Conversely, children with low MUAC only were 3 times more likely to present with malarial infection (unadjusted odds ratio [UOR], 0.33; P = .003) than children with low WHZ only.

Relative to the combined cohort of all children with low MUAC (MUAC <115 mm with and without WHZ <–3), children with low WHZ only presented with comparable deficits in most clinical and biochemical indicators. For example, median serum leptin levels were not statistically different between these 2 phenotypes (Tables 4 and 5). However, children with low WHZ alone had lower body iron stores and comparable or higher rates of visible severe wasting and dehydration (Table 4).

Additional clinical indicators are presented in Supplemental Table 6. Unadjusted and adjusted analyses of continuous and categorical data, corrected for false discovery rates, are presented in Supplemental Tables 7–9.

The pathophysiology and functional severity associated with different anthropometric phenotypes of SAM have heretofore not been well characterized.⁸  Yet policy makers and government stakeholders need robust scientific evidence to generate evidence-based guidelines to appropriately target and prioritize malnourished children for treatment. Although the WHO recommends measurement of both WHZ and MUAC for assessment of SAM, the use of MUAC as a single anthropometric assessment tool for case finding and admission to therapeutic feeding programs and the abandonment of WHZ are increasingly promoted and applied.^4–6  This is in part because children presenting with WHZ <–3 but MUAC ≥ 115 mm have been assumed to be at lower risks of morbidity and mortality than children with low MUAC.⁴  By characterizing the clinical and biochemical features of separate cohorts of children with low WHZ alone, low MUAC alone, and both low WHZ and low MUAC, we take a novel approach to addressing this controversy.

We present 3 new observations of clinical import; results were consistent across all 3 countries, lending credibility to our findings. First, a range of clinical and biochemical indicators reveal that children with low WHZ alone have deficits in nutritional status, hydration, and iron balance equal to or more severe than those in children with low MUAC alone, and they have lower levels of leptin, a critical marker of mortality risk in SAM.^9,10  The only complication detected more frequently in children with low MUAC only than in those with low WHZ only was malaria prevalence, a finding consistent with observations made in a previous study.²⁰  Second, children with both low WHZ and low MUAC have the most severe nutritional deficits and lowest leptin levels, suggesting that they are at highest risks of acute and postdischarge morbidity and death. Finally, children with low WHZ alone display vulnerabilities at least as great as the combined cohort of all children with low MUAC.

In previous investigations^9,21  we suggested potential mechanisms that may explain the relationship between hypoleptinemia and mortality in SAM. Leptin is produced by white adipocytes, and its levels in circulation reflect the mass of subcutaneous and, to a lesser extent, visceral white adipose tissue. A low level of leptin suggests white adipose tissue depletion. Because white adipose tissue represents the major bodily storage form of mobilizable energy, a low level of leptin likely reflects a deficiency of body energy stores. Under conditions of severe malnutrition and other severe illnesses and stress, a lack of white adipose tissue may limit the ability to sustain energy production for cardiopulmonary function and gluconeogenesis and thereby increase the risk of death.^9,10,21  Leptin also promotes immune cell development and function; a lack of leptin production or action in patients with mutations in leptin or the leptin receptor is associated with increased risks of infection.^22,23  Thus, hypoleptinemia in infants and children with SAM may impair immune function and predispose them to mortality from sepsis and other major infectious diseases.^24–26 

Our study has some limitations. First, heterogeneity in screening processes among study sites likely contributed to slight differences in age range, anthropometric deficits, and clinical profiles between countries. The active screening network in Bangladesh was highly developed, and a majority of participants there were identified in their communities. In Liberia, a majority of participants were identified by passive screening during medical management. Only in Burkina Faso were participants passively screened during growth monitoring activities. Yet active community screening, growth monitoring and promotion, and passive screening are all major routes that lead children with SAM to medical care. We consider heterogeneities between study sites to reflect the widespread variability of screening procedures used for SAM and in fact validate the generalizability of our results. Second, our study was focused on uncomplicated SAM. Recommendations that propose the abandonment of the use of WHZ as an independent detection and admission criterion have thus far not been applied to inpatient management of complicated SAM. To avoid selection bias in recruiting patients, we conducted our study in primary health care units, where uncomplicated cases are routinely managed, rather than in hospitals. In these settings, we had to rely on low-invasive procedures, such as capillary blood sampling. The limitation of blood volume precluded us from measuring potentially valuable analytes, including albumin, micronutrients, and amino acids. Finally, leptin levels may decrease with time since feeding.²⁷  In our study, children were assessed in the morning and were administered an appetite test and/or food challenge as per routine procedures ∼1.5 hours before blood sampling. Therefore, we think it unlikely that differences in time since feeding explain the striking differences in leptin levels among the various groups.

Previous analyses of receiver operating characteristic curves in unselected cohorts of children, in community settings where treatment of SAM was not available, revealed that MUAC has greater discriminatory performance for subsequent mortality than WHZ.²⁸  This observation has been used by some investigators as a rationale for prioritizing low MUAC and discounting or omitting low WHZ for identification and management of SAM.⁴  Receiver operating characteristic curves, however, do not quantify relative risk of death or even probability of death of children below specific cutoffs. To assess this more precisely, we recently calculated mortality hazard ratios associated with different phenotypes of SAM in available data sets of past community cohort studies.²⁹  We found similar hazard ratios in malnourished children with low MUAC only and low WHZ only and higher mortality risk in children with combined deficits. Importantly, these findings from unselected and untreated cohorts of children are supported by the few clinical studies that compared the vulnerability of different subpopulations of children with SAM on admission to treatment^20,30  and by a recent reanalysis of case fatality rates observed in large historical cohorts of malnourished patients.³¹  Moreover, WHO experts have noted that the association between absolute MUAC and mortality is subject to age bias (younger children have lower MUAC and higher risks of death) and have questioned its causal relationship with mortality risk.³²  The results reported here are consistent with and extend previous findings: compared with children with low MUAC, children with low WHZ have lower leptin levels and higher rates of severe iron deficiency, dehydration, and visible signs of severe wasting, all of which are associated with higher risks of acute and postdischarge mortality. In our study, all patients were carefully managed with WHO recommendations, regardless of anthropometric criteria. As expected in well-treated, uncomplicated SAM, short-term mortality was low; 1 death occurred in the group with low WHZ only; a second was in the group with both low WHZ and low MUAC.

Together, these observations indicate that malnourished children with low WHZ require intensive nutritional assessment and intervention, even if MUAC exceeds 115 mm. Highest priority should be assigned to children with combined deficits in WHZ and MUAC, who are at greatest risk of life-threatening illness and death. It is clear now that abandonment of WHZ as a diagnostic criterion for SAM effectively excludes from treatment a large number of children with high risks of morbidity and mortality.³³  The analyses of representative survey data indicate that children with low WHZ only may represent >40% of the SAM caseload in most countries, with higher proportions in high-burden and acute-crisis contexts^3,34 ; increasing the MUAC threshold for admission would increase the number of these children who would be eligible for treatment but would dramatically increase the program target at the expense of specificity.³⁵  In summary, the weight of evidence suggests that both WHZ and MUAC should be measured in all children evaluated for malnutrition.

A research priority should be to harness technological capacity to develop innovative diagnostic methods to identify children with low WHZ in the community. Until then, both WHZ and MUAC should be retained as independent diagnostic criteria for SAM, in line with WHO recommendations.

We thank Huaxia Cui from the Duke Molecular Physiology Institute for her assistance with leptin assays. Additionally, we thank Suvi T. Kangas, Victor Nikièmia, and all OptiDiag research staff and nurses for their contributions to this study.

HAZ

height-for-age z score

IMCI

Integrated Management of Childhood Illness

IQR

interquartile range

MUAC

mid–upper arm circumference

RBP

retinol-binding protein

SAM

severe acute malnutrition

SSRZ

sitting/standing ratio z score

sTfR

soluble transferrin receptor

UOR

unadjusted odds ratio

WHO

World Health Organization

WHZ

weight-for-height z score