Pediatric undernutrition defined by body composition—are we there yet?

See corresponding article on page 1456.

Despite widespread acceptance of growth standards for infants and young children in recent decades (1), universally accepted diagnostic criteria for childhood undernutrition in pediatrics have remained elusive. For decades and in concert with the Waterlow criteria (2), the WHO has historically expressed child undernutrition as deficits in weight for age (underweight), length/height for age (stunting), or weight for length/height (wasting) (3). Guidelines by the WHO, UNICEF, Academy of Nutrition and Dietetics (AND), and American Society for Parenteral and Enteral Nutrition (ASPEN) have included reduced weight for length and/or low BMI, and low mid-upper arm circumference as additional anthropometric measures (all of which have been associated with increased risk of death), as well as reduced length/height for age (4–6). Low weight for age z score is absent in the AND/ASPEN criteria (4).

As useful as these anthropometric criteria have been for the study of undernutrition (especially in large groups of children), however, their shortcomings in measuring body composition are apparent. This has been noted in children with undernutrition (7), obesity (8, 9), as well as a wide variety of chronic illnesses (10–13) and across countries (14). Despite the acknowledgement that the quality and function of body composition influence outcomes, neither have been incorporated as diagnostic criteria (3, 15–18). This contrasts with definitions of adult undernutrition, which include objective measures of strength, subjective physical examination assessment of lean and fat mass, and/or quantitative assessment of lean mass using body composition techniques (19, 20).

In this issue of The American Journal of Clinical Nutrition, Lara-Pompa et al. (21) take a major step forward in assessing the impact of body composition on clinical outcomes in children in a vulnerable population of inpatients. In this well-designed study, the authors enrolled 118 participants in a tertiary care facility in the United Kingdom with heterogeneous underlying illnesses. Traditional anthropometrics, body composition by DXA, and 3 validated malnutrition screening tools were assessed. The authors defined malnutrition based on height, weight, BMI, and lean mass or fat mass standard deviation scores (SDS) < −2 as compared with published UK-specific reference ranges. The relations between malnutrition diagnostic indicators, screening tools, and clinical outcomes were quantified.

Not surprisingly, they observed high rates of undernutrition, with 13.6% of patients with low height, 8.5% with low weight, 16.9% with low lean mass, and 5.9% with low fat mass—despite only 4.2% presenting with low BMI. BMI is routinely used in clinical practice to classify nutrition status, but missed 80% of cases of undernutrition defined by low lean mass and 71.4% defined by low fat mass. Importantly, no relation of BMI with outcomes [prolonged length of stay (LOS) or complications] was observed, whereas lean mass SDS < −2 increased the odds of prolonged LOS by 4.53-fold and low fat mass SDS < −2 by 6.07-fold. These thought-provoking pediatric data suggest that if we continue with our present use of BMI for undernutrition diagnosis, we will routinely miss abnormalities in body composition that may be more important clinical determinants of health outcomes. Are these sufficient data to add reduced lean and fat mass to our pediatric malnutrition diagnostic criteria?

Low lean and fat mass were defined as SDS < −2 based on age- and gender-normative values from the UK reference cohort, and it is critical to understand several limitations of this definition. The UK reference curves for body composition were obtained from a relatively small cohort of ∼500 children (22). Although this reference tool may be appropriate for the demographics of this current cohort (race and ethnicity were not reported), these cutoffs may not be generalizable to ethnically diverse populations. The reference curves do not address the known influence of pubertal status (23). For example, should a 14-y-old Tanner 5 male have the same threshold of unadjusted lean mass for the diagnosis of malnutrition as a 14-y-old Tanner 2 or 3 male? Lastly, the distribution of body composition at the extremes of weight may not be well defined by the UK reference group, because the lowest BMI z score assessed was −2.99 in males and −3.33 in females (22).

The authors opted to define undernutrition based on unadjusted lean mass. In this age range, lean mass increases exponentially with height (24). The cohort studied was remarkably low in stature compared to the reference population (14% had a height SDS score < −2), which may explain the high rates of lean mass depletion. If the population is enriched in patients in whom genetic height potential—regardless of nutrition status—is reduced, then our estimates of malnutrition by lean mass may not be accurate. Importantly, and reflecting the characteristics of an inpatient population, 19% were receiving either parenteral or enteral nutrition, 11% were wheelchair users, 13% were fluid restricted, and 34% of the population was following dietary restrictions, which was a significant predictor of low height. This last point requires additional investigation into the indication and type of restriction, and its potential effect on body composition. Notably, adult malnutrition diagnostic criteria recommend expressing lean mass as a function of height (e.g., fat-free mass index or skeletal muscle index in kg/m²) and low fat-free mass index has been shown to correlate with longer LOS (25).

There are several important considerations for the interpretation of the relation of undernutrition with outcomes in this study, all of which could suggest a stronger effect than observed. First, “increased LOS” was based at least in part on subjective physician assessment of anticipated LOS. The treatment of this variable as a dichotomous outcome could weaken an objective difference in LOS if analyzed as a continuous variable. Second, the definition of complications might explain the lack of association with body composition. The authors defined complications as 1) intensive care unit transfer; 2) transfer to another hospital instead of home; 3) unplanned, increased use of parenteral or enteral nutrition; or 4) episodes of fever or infection treated with antibiotics. These may be questioned as overly broad definitions; transfer to home institution could represent decreased reliance on tertiary care due to an improvement in health status, and the unplanned use of parenteral or enteral nutrition may actually be an appropriate intervention for this undernourished population. In addition, other malnutrition-related entities such as diarrhea, pressure ulcers, or mortality, or complications that may occur after discharge (e.g., 30-d readmission rates), were not included in the definition. In addition, the relation between anthropometrics or body composition and outcomes may not be linear. For example, in adult males, lean mass demonstrates a U-shaped relation with mortality, whereas BMI demonstrates a J-shaped curve (26), and obesity increases the risk of complications in hospitalized pediatric patients (27). Including the patients with overnutrition in the same category of normal weight may not reflect the subtleties of this relation, especially because 12% were over-nourished based on BMI SDS > 2. Finally, as the authors note, children at particularly high risk of both malnutrition and adverse outcomes (children under age 5 y and the sickest children) were not included in the study.

The authors selected 3 validated screening tools to aid in identifying an “at-risk” population. Of the 3, the Pediatric Yorkhill Malnutrition Score (PYMS) had the best sensitivity for detecting low BMI, whereas all are known to have poor positive predictive value compared to standard references (28). Of this cohort, 25% screened positive on the PYMS, which was the only independent predictor for increased LOS. Screening Tool for the Assessment of Malnutrition in Pediatrics (STAMP) scores were high risk in 35.5%, and showed the overall best sensitivity at detecting low weight, height, and lean mass. PYMS and Screening Tool for Risk of Impaired Nutritional Status and Growth (STRONG_kids) both independently identified increased risk of complications. Compared with PYMS and STAMP, STRONG_kids was positive in only 18.4% of the cohort. This is somewhat surprising because physical examination findings of fat and lean stores are part of this screening tool (in contrast to BMI as found in STAMP and PYMS). In addition, like STAMP, STRONG_kids includes underlying diseases that were highly represented in this cohort (gastrointestinal diagnoses, anticipated major surgery, and cystic fibrosis). Taken together with existing data, no single screening tool appeared to confer major advantage over the others, and certainly none was more strongly associated with longer LOS as were the measures of body composition.

Before body composition can be used to guide pediatric clinical practice, there are important knowledge gaps to be filled. We will need substantially more data defining body composition normative values in a racially diverse population, clarifying the best method of assessing body composition (e.g., adjusting lean mass for height or total mass, or ratio of lean to fat mass), and determining the role of accounting for pubertal status. Access to DXA will need to be enhanced, and the benefit compared with the trivial risk of radiation exposure will need to be discussed on a population level. The use of nonradioactive bedside measures of body composition should also be considered (29).

Furthermore, we need to know how to act on this information. In a child with low lean mass and low BMI, without significant inflammation, the answer may be as simple as increasing calories. But how do we preferentially “grow” lean tissue in a child with normal BMI, and avoid development of obesity, especially during admission to hospital? And what is the “target” lean or fat mass for an individual to optimize outcomes? This is an area ripe for future research and the findings by Lara-Pompa et al. have propelled this field forward in the quest for outcomes-based definitions of malnutrition.

Notes

Supported in part by NIH grants K24 DK104676 and P30 DK040561 (to CPD).