What is Normal, Healthy Growth? Global Health, Human Biology, and Parental Perspectives

Abstract

Objectives:

The widespread variation seen in human growth globally stands at odds with the global health perspective that young child growth should not vary across populations if nutritional, environmental and care needs are met. This paper: 1) evaluates the idea that a single standard of ‘healthy’ growth characterizes children under age 5, 2) discusses how variation from this standard is viewed in global health, in human biology and by parents, and 3) explores how views of ‘normal’ growth shape biomedical and parental responses.

Methods:

This paper reviews the anthropological, public health and clinical literature on the nature of child growth and the applicability of World Health Organization Multicenter Growth Reference Study growth standards across contexts.

Results:

The considerable variability in child growth across contexts makes it unlikely that any one framework, with issues of sample selection and representativeness, can serve as the model of healthy growth. Global health, human biology and parents differ in the emphasis they place on heredity vs. environmental context in understanding this variability, but human biologists and parents tend to view a wider range of growth as ‘normal.’ Since both biomedicine and parents base their care decisions on their perceptions of normal, healthy growth, the comparative framework used has important implications for medical treatment and feeding practices.

Conclusions:

A more nuanced approach that incorporates the biology of growth and its association with health outcomes across contexts is critical to identify patterns of healthy growth and to avoid over-reliance on a single standard that may pathologize variability.

Introduction

Body size in infancy and early childhood is linked to short-term health outcomes, such as morbidity and mortality from common childhood illness in children under 5 (Black et al. 2008); (Olofin et al. 2013), and longer term outcomes, such as work productivity, reproductive outcomes and the development of cardiometabolic diseases, in adulthood (Adair et al. 2013; Martorell 2017). Given these associations between child size and current and future health, child growth is a widely used marker of nutritional status and health by global health practitioners, clinicians and parents alike. In global health, linear growth is considered the “best overall indicator of child well-being” and poor growth is attributed to inadequate diets, pathogenic environments or care practices (Perumal et al. 2018; Scheffler et al. 2020; de Onis and Branca 2016). Consequently, measures of height-for-age are used to identify populations in need of intervention (Perumal et al. 2018; Frongillo et al. 2019). Clinically, length, weight and head circumference are routinely monitored during pediatric visits. Abnormal patterns, such as growth faltering or rapid weight gain, are viewed as indicators of underlying medical and/or social conditions requiring further investigation (Christesen et al. 2016). Within families, caretakers interpret child growth and weight gain as signs of healthy development and may seek medical care for growth they view as abnormal (Lucas et al. 2007).

In each of these domains, public health, clinical and parental, assessment that a child is growing normally or is small or large for their age requires a comparative framework for what a “normal” child should look and grow like. While parents, and even health practitioners, may judge children against others in their families and communities (Thompson et al. 2014), formal evaluations of poor growth --stunting, growth faltering or overweight-- are made by comparing children to age- and sex-specific growth charts for length/height, weight, and body mass index. Growth monitoring using these charts is considered an essential tool in the public health and pediatric toolkit for evaluating the degree to which children’s nutritional and physiological needs are being met, to diagnosis underlying growth and metabolic disorders, and for timely medical intervention (Rogol and Hayden 2014; Christesen et al. 2016). However, these assessments, that a child’s growth is normal or abnormal or that a population is short or thin, are highly dependent on the growth charts used (Turck et al. 2013).

In over 125 countries as of 2016 (Heude et al. 2017), this comparative framework is provided by the WHO Multicentre Growth Reference Study (WHO MGRS) growth charts. Developed based on samples of relatively affluent, breastfed infants from six locations (the United States, Norway, Oman, Ghana, Brazil and India), these charts serve as growth standards, a framework for how children should grow (WHO Multicentre Growth Reference Study Group 2006). Underlying this claim is the assumption that growth is shaped primarily by nutrition and environmental exposures, particularly those that contribute to repeated infection, and that all children under the age of 5 grow similarly if exclusively breastfed, free of illness and living in socioeconomically advantaged environments. Thus, departures from the WHO MGRS curves are viewed as evidence of abnormal growth stemming from deprivation.

However, this claim that a single standard defining the average and normative boundaries of growth variability (henceforth referred to as ‘norm’) represents healthy child growth runs counter to decades, if not centuries, of work within human biology and auxology documenting considerable variability in child growth and adult size globally. Understanding this variability is central to theoretical and empirical research within human biology (e.g. (Eveleth and Tanner 1990). Variation in growth rates and timing of transitions between periods of growth, such as from infant to childhood growth phases, is viewed as a hallmark of our species’ life history and as evidence that growth represents an adaptation, allowing individuals to alter their growth trajectories in response to environmental context (Bernstein and Dufour 2017). Further, examinations of growth biology document the complex feedback loops integrating environmental, genetic, and hormonal signals into the timing and magnitude of linear growth (Lampl and Schoen 2017).

This seeming paradox between the assertion of the WHO MGRS that a single norm exists for healthy growth and the wide variation seen in child growth across settings raises several key questions about the nature of human growth and variation: 1) is there a universal pattern of ‘normal,’ healthy growth? 2) if this single norm exists, should individual and population-level variability be viewed as signs of pathology? 3) how do views of ‘normal’ shape biomedical and parental responses to variation? This paper evaluates the claim that the WHO standards represent a universal norm of growth and explores how variability is interpreted in global health, by human biologists and by parents. Understanding the underlying assumptions, sample selection and construction is critical for evaluating whether the MGRS standards serve as the model of “normal” human growth and for interpreting variation in individuals and populations that do not “follow the lines.” This understanding is particularly important given the large-scale use of growth charts in global health programs and doctors’ offices to identify growth problems, social deprivation, and individuals and populations in need of intervention, as is demonstrated in recent debates over the many uses and interpretations of stunting.

How is “normal” defined?

The definition of “normal” human growth relies on the comparative framework used. Until the mid-2000s, the mostly commonly used framework was the NCHS/WHO growth reference, based on the longitudinal growth measures of predominantly White, formula-fed infants participating in the Fels Longitudinal Growth Study in Ohio for children from birth to age 2 and then on nationally-representative cross-sectional samples of children aged 2 and older participating in the US National Health and Nutrition Examination Surveys from 1963–1994 (Kuczmarski 2000). Perhaps unsurprisingly, the growth of children in other settings, particularly low- and middle-income countries (LMIC), tended to be slower than these reference children leading to concerns that breastfed children would be seen as faltering in clinical settings (WHO Working Group on the Growth Reference Protocol and WHO Task Force on Methods for the Natural Regulation of Fertility 2000). The longer and heavier size of the NCHS/WHO infants also led to concerns that prevalence estimates of malnutrition in many public health programs would be flawed. The poor fit between this ‘norm’ and the growth seen in children globally, led to the creation of the WHO MGRS growth charts.

Based on the growth of healthy, exclusively breastfed, socioeconomically advantaged infants in six locations, these charts are more representative of the growth of infants and young children globally. The consistency seen in the patterns of linear growth between the sites -- 3% of the variance in length/height is seen between sites compared to a 70% variance between individuals within sites-- led the WHO to conclude that the growth of infants and young children does not vary across populations when nutritional, environmental and care needs are met (WHO Multicentre Growth Reference Study Group 2006). They argue that these inter-site similarities provide strong evidence that growth potential is universal until at least 5 years of age and that the MGRS curves provide the “best description of physiological growth and should be applied to children everywhere, regardless of ethnicity, socioeconomic status and type of feeding.”

Consequently, unlike the NCHS/WHO charts, which were a reference, a statistical description of child size, the WHO growth charts serve as a standard, a prescriptive model for how healthy children should grow. By arguing that they show how children should grow regardless of setting, the WHO MGRS standard purports to describe the “optimal” human condition rather than the patterns of growth seen in a particular time and place (de Onis et al. 2004; Garza and de Onis 2004). Further, calling the observed growth of socioeconomically advantaged, exclusively breastfed, healthy children an optimal pattern links their achieved size to the definition of “healthy” growth. As one paper evaluating the strengths and limitations of the charts for global use states, a key strength of the charts is that they describe “growth as it should be to optimize long-term health” (Ziegler and Nelson 2012). Thus, the movement from a reference to a standard explicitly adds a value-judgment to infant and young child growth patterns or, as the WHO states, “the WHO standards serve a symbol of children’s right to achieve their genetic growth potential” (de Onis and Branca 2016), i.e. as the norm for healthy child growth.

How well does this norm fit?

Despite the widespread adoption of the WHO MGRS standards, the assumption that a universal, “normal” underlying pattern of growth characterizes all children under the age of 5 has not gone unquestioned (e.g. (Heude et al. 2017; van Buuren and van Wouwe 2008). As of 2020, at least 70 papers have been published assessing whether the WHO growth charts fit the growth patterns of healthy children better than nationally- or regionally- specific charts (Heude et al. 2017). Many of these studies document patterns of growth that differ from the MGRS curves even in socioeconomically advantaged children. Well-off children in Hong Kong, for example, were significantly shorter than the WHO MGRS children at age 3 (Hui et al. 2008), while European children from studies conducted in Iceland, the Netherlands, and Germany were significantly longer/taller than the MGRS infants and children (Natale and Rajagopalan 2014).

The question of how much variability is seen between the samples included in the WHO MGRS and between the MGRS samples and other groups of children is an important one, since limited variability is necessary, first, to justify the pooling of the sites for a single standard and, second, to support the use of the standards as a global pattern of child growth. In terms of intra-sample variability, WHO reports of only 3.4% variation in length between sites with no obvious or consistent differences across sites (WHO Multicentre Growth Reference Study Group 2006). This limited variability was used to support the pooling of samples into a single mean (Cole 2007). However, other studies have pointed out that inter-site variability in the WHO MGRS increases with age, from 1% at 12 months to 7% at 60 months (Turck et al. 2013; van Buuren and van Wouwe 2008) and differs between sites (Hackman and Hruschka 2020). HAZ differences between the Brazilian and Omani samples, for example, were 0.63–0.73 standard deviations (SD) from the pooled SD at 24 to 48 months (Hackman and Hruschka 2020).

Along with this intra-study variability, considerable inter-sample variability has been documented in comparisons between the MGRS and other global samples. Comparison of the linear growth of children from 55 countries to the MGRS found that, while most of the included countries had a mean length/height for age that fell within 0.5 SD of the WHO MGRS mean, the cut-off used by the WHO to evaluate variability between sites, 20% of the means were outside this range and another 44–48% of samples were at least 0.25 SD from the MGRS mean at 4 or more time points (Natale and Rajagopalan 2014). Along with these differences in the central tendency of child length/height from the MGRS, other studies have found differences in the range of variability. Using a sample of all children under 5 in LMIC participating in Demographic and Health Surveys (DHS) from “ideal home environments,” Karra and colleagues (2017) found that, while the mean HAZ in this global sample was very close to the MGRS mean, the distribution of HAZ was more disperse, with a SD of 1.33, and more children fell into the tails of the distribution (+/−2 SD). While these differences could stem from measurement error between the studies as noted by the authors, they may also reflect greater variability in infant and young child growth than that included in the MGRS sample.

This inter- and intra-site variability calls into question the claim that WHO charts are a “neutral, value-free yardstick” (Ziegler and Nelson 2012) free from geographic and temporal influence (Butte et al. 2006). As is the case in any study, the sample selected for inclusion into the MGRS and the time the study was conducted impact the data collected. The MGRS sites were chosen to be representative of broad regions --North America, South America, Europe, Africa, the Middle East and South Asia-- but other regions with known differences in growth, such as East Asia or Oceania (WHO Working Group on the Growth Reference Protocol and WHO Task Force on Methods for the Natural Regulation of Fertility 2000), were excluded. Perhaps even more critically, the rigid inclusion criteria –no known environmental constraints on growth, exclusive breastfeeding until 4 months and continued breastfeeding to 12 months, no maternal smoking, single, term birth and no significant morbidity (de Onis et al. 2004) -- and intensive measurement schedule −−21 visits over the first 2 years in the longitudinal subsample-- limited the sample to a relatively homogeneous group of mothers living in a small number of urban neighborhoods within each of these countries. These infants, therefore, likely do not capture the social, behavioral, and genetic diversity of their own countries, much less the full global range of genetic and environmental variation (Karra et al. 2017).

Further, with the exception of Norway, which may have reached a plateau in height, many of the included MGRS countries are likely still undergoing the secular trend (Natale and Rajagopalan 2014; Hui et al. 2008). Unless the participating families had been living free from environmental and dietary constraints for a number of generations, the growth of even these socioeconomically advantaged infants may not represent an ‘optimal growth pattern’ due to epigenetic and intergenerational effects on infant and child growth (Wells 2017). Thus, despite claims that the MGRS can be used “to assess children everywhere” (De Onis 2016), the MGRS growth standards, like other growth charts, nevertheless describe how children living at different stages of secular change were growing at a specific point in time (Natale and Rajagopalan 2014). Consequently, as a number of authors have noted (e.g. (Karra et al. 2017; Heude et al. 2017)(van Buuren and van Wouwe 2008; Wells 2017), the extent to which the MGRS sample reflects ‘global human growth’ independent of time and location is unknown.

The impact of sample selection and retention is even more clear when examining variability in other MGRS measures, such as weight and head circumference. While the site-specific weight and head circumference data have not been published by the WHO, the more widely spaced centiles compared to the HAZ charts stem from greater inter-individual and inter-site variability (Cole 2007). Yet, despite this higher variability within the MGRS sample, these charts show worse fit than the HAZ charts in cross-national comparisons. In their comparison, Natale and Rajagopalan (2014) found that, overall, 31% of the 54 national mean WAZ scores were at least 0.5 SD and 62% were at least 0.25 SD from the WHO mean. These differences in WAZ may stem from the children ultimately included in the sample. Just under half (49%) of infants enrolled in the longitudinal sample were excluded due to ‘non-compliance’ with study criteria, including the feeding recommendations, morbidity affecting child growth, or attrition (Turck et al. 2013). Since smaller size tends be associated with the discontinuation of exclusive breastfeeding (Kramer et al. 2011), these exclusion criteria may have led to selective drop out and increased weight in the remaining participants (Ziegler and Nelson 2012). Analysis of the growth of children within the sample suggests that those who were retained have higher early weight velocity (Binns et al. 2008), providing evidence for selective drop out. In addition, to avoid influence of “unhealthy weights,” children with WAZ of +/− 3 SD were removed from the longitudinal sample and children with WAZ >2 SD were removed from the cross-sectional sample due to the highly right-skewed distribution of the data. The removal of these data points, necessary to produce a chart representing “optimal” growth patterns, nevertheless raises questions of what is meant by universal growth if even children in this “optimal” sample are removed due to excessive variability. Together these issues of representativeness, sample inclusion and exclusion, and statistical considerations highlight the important issue that the standard for ‘normal’ growth is highly contingent on the data collected.

How do we understand variability in child growth?

Underlying many of these critiques is a fundamental question about the nature of variability in growth: If a universal norm represents healthy growth for infants and young children of all ethnicities and locations, then do departures from this norm indicate pathology? This question of how to distinguish variability from pathology is a long-standing one in public health and human biology and has been the subject of numerous symposia and special issues across several decades (e.g. Human Organization vol 48(1), Spring 1989 and American Journal of Human Biology, vol 19(5), 2007). While nearly all public health and human biology researchers agree that nutrition and environmental factors play an important role in shaping variability in child growth globally, the extent to which adaptation to environmental stressors and underlying genetic variability contribute to population differences in growth is more contentious.

Proponents of a universal pattern of growth place much greater emphasis on the contribution of poor diets and environmental conditions, particularly those that contribute to repeated infection, to these population differences (Butte et al. 2006; Habicht et al. 1974)(Martorell et al. 1975), and question whether the degree of genetic isolation and the selective strength of ancestral environments were sufficient to shape population differences in linear growth (Butte et al. 2006; Martorell 2017). They support this position with evidence showing that growth is similar among children of differing ethnicities living in “middle class” environments, in places such as the US, the UK, Australia, Japan and Columbia (Habicht et al. 1974), and that growth differs between children of similar ethnicities living in different socioeconomic conditions (Habicht et al. 1974; Bogin and Loucky 1997). The conclusion drawn from these studies is that “any racial or ethnic effect [on growth] is small compared with environmental effects” (Bogin and Loucky, 1997). Consequently, from this perspective population differences in child height, weight or body composition are viewed as evidence that environmental conditions, nutrition, and/or care practices are insufficient to allow children to follow the ‘normal’ growth pattern. Departures from the norm should be assessed as potentially pathological, or, at the very least, as requiring examination of environmental and socioeconomic context.

Historically, human biologists have placed more emphasis on adaptation to environmental stressors, such as high altitude, cold climates, or chronic undernutrition in tropical environments, as important sources of population variability in size (Leonard 2018). As Bernstein and Dufour describe in their special issue on the anniversary of Eveleth and Tanner’s volume, Worldwide Variation in Human Growth, variability in growth has been viewed as “the most critical ‘strategy’ that any one individual can employ in charting a successful course to survival and reproduction.” (Bernstein and Dufour 2017). From this perspective, variability in growth patterns stems not only from the resources available or constraints faced within individuals’ current environments, but also from population histories of adaptation to environmental stressors and other demographic processes, such as genetic drift or gene flow, that led to population differences in growth potential (Hruschka and Hadley 2016; Wells 2017). Variation from this view would not necessarily reflect pathology or even a response to current environmental constraints.

However, how to disentangle ‘beneficial’ variability in growth that could reflect successful adaptation to local conditions from pathological variability stemming from social and economic disadvantage has been and remains hotly debated (Wells 2017; Ellison and Jasienska 2007). In the 1980’s, this debate centered around the “Small but Healthy” hypothesis, in which economist David Seckler (Seckler 1982) proposed that the short stature of individuals in LMIC represented an adaptation allowing for increased energetic efficiency and reduced daily energy needs. Widely criticized by both public health and anthropology (Pelto and Pelto 1989; Martorell 1989), this hypothesis highlighted the fallacy of assuming that variability was prima facie evidence for adaptability.

Similar arguments about the adaptative nature of size variability are at the center of more recent debates around the whether the plasticity of human growth seen in response to undernutrition in utero represents an adaptive response (Bogin et al. 2018; Wells 2017), as proposed by the thrifty phenotype (Hales and Barker 2001) and predictive adaptive response (Gluckman et al. 2007) hypotheses. Key to these theoretical models are the assumptions that size differences at birth represent altered growth trajectories in response to the predicted adult environment, that this facultative adjustment is adaptative, and that plasticity is usually beneficial. Both hypotheses have been thoroughly critiqued in the literature for these assumptions (e.g., (Wells 2012; Ellison and Jasienska 2007; Bogin et al. 2007). For the purpose of understanding variability in growth and distinguishing ‘normal’ from pathological variation, these debates underscore the importance of understanding the processes leading to individual and population differences in size.

Within human biology, the shift from an emphasis on adaptability to life history theory as a theoretical framework and the growing understanding of the genetics and epigenetic processes contributing to patterns of growth highlight the limitations of focusing on size outcomes and viewing variability as only pathological or only adaptive (Leonard 2018; Goodman 2013). As Alfonso-Durrity and Valeggia (2016) describe, life history theory, with its focus on resource allocation strategies that represent time-integrated trade-offs, call in to question a dichotomous normal-pathological model for understanding human growth variability. Rather, this approach suggests that documenting and examining differences in the magnitude and tempo of growth, i.e. the process of growth leading to size variability, should be the focus of scientific inquiry.

Certainly, recent work on the genetics of human growth supports the importance of a genetic contribution to differing growth patterns between populations and also highlights the need to explore gene by environment interactions in shaping growth trajectories. While past estimates of genetic differences in height between populations were relatively low, ranging from <4% to 15% depending on the methods used and populations surveyed, more recent genome wide association studies (GWAS) suggest that these figures may be an underestimate (Wood et al. 2014). A comparison of GWAS data from 14 European countries found that many independent loci contribute to population differences in height, accounting for 24% genetic variance between populations (Wood et al. 2014). GWAS comparison of twins from Europe, Australia, the US and East Asia similarly attributed “an important proportion” of the differences in height across childhood and adolescence between populations to genetic variation (Jelenkovic et al. 2016). Importantly, these researchers also documented gene by environment interactions that were relatively similar in magnitude between regions, suggesting that populations have similar potential for growth to be shaped by environmental conditions. From a life history perspective, these gene by environment interactions may be part of the selection for growth as a “robust, but plastic, process” that provides individuals with the chance to reach adulthood in diverse and dynamic environmental conditions (Alfonso-Durruty and Valeggia 2016).

More recent understandings of the biology of linear growth also highlight the importance of gene by environment interactions in shaping individual growth. Many of the gene variants associated with height in GWAS studies are linked to proteins with numerous roles in growth-related cellular processes (Wood et al. 2014). Exploring how the products of these gene variants interact with environmental context is critical for understanding individual and population variation in growth. As recently described by Lampl and Schoen (2017), linear growth is controlled by numerous feedback loops between gene products and environmental signals of nutrition, energetics, and stress at the growth plate. Attained height reflects the interactions of a complex, dynamic system with numerous cross-talking inputs, resulting in highly variable and non-linear growth processes (Lampl, 2005; (Lampl and Schoen 2017). From this systems perspective, growth patterns are unlikely to be universal, since the complex interplay of genes and environments will vary across individuals and populations.

While the relative contribution of the environment vs. genetic inheritance has been the focus of much of the disagreement about the universality of human growth between global health practitioners and human biologists, parents tend to rely on local models that incorporate both heredity and environmental context to define normal growth and distinguish between variation and pathology. Parents in settings as disparate as the United States, rural Mexico, Bangladesh, Malawi and Tanzania all view ‘normal’ growth as a sign of health (Flax et al. 2016; Turnbull et al. 2009; Thompson et al. 2014; Mchome, Bailey, Darak and Haisma 2019; Hossain et al. 2018) and attribute abnormal growth to poor diets and environmental exposures leading to infections and allergies (Reifsnider et al. 2000; Turnbull et al. 2009). Across settings, parents rely on both growth charts and local markers, such as the size of relatives and children in the community and clothing and diaper sizes, to judge whether their children are growing “normally” (Mchome, Bailey, Darak, Kessy, et al. 2019; Lucas et al. 2007; Thompson et al. 2014). Parents’ views of whether their children are growing ‘normally’ also depends on their child’s age and may change across time, as parents judge their children against the ‘expected’ growth velocity for different developmental stages, such as infancy vs. early childhood (Moffat 2000).

However, parental assessments of size are often distinct from their views on growth. Unlike growth, which they attribute to diet and environmental factors, parents across studies viewed their child’s size as stemming from heredity or ‘God’s will’ (Mchome, Bailey, Darak and Haisma 2019). In a systematic review of parents’ views of infant growth and size, Lucas and colleagues (2007) documented that across a number of settings, parents viewed size, particularly small size, as stemming from inherited differences or children “simply going to be the size they are going to be.” Unlike growth, which can be influenced by parental care practices, size is seen as stemming from genetic and/or other inherent characteristics of the child. For example, American mothers of children with growth faltering viewed good diets as important for growth and felt gratified when their children gained weight or grew in height; however, they did not believe they could influence their child’s size (Reifsnider et al. 2000). Similarly, parents in Tanzania separated the process of growth from the attainment of size, viewing ‘growing well’ and ‘eating well’ as signs of health, but not viewing height as linked to nutrition or something that parents could control (Mchome, Bailey, Darak, Kessy, et al. 2019). While distinct from both the global health and human biology models of growth and child size, these widely shared parental models of growth and child size are also important for understanding how parents view normal growth and how they interpret and act on variability.

Why does this matter for global health, medicine, and parents?

These debates about the universality of growth represented in the WHO charts, how to interpret departures from this norm, and the underlying sources of growth variation are important since they inform global health intervention, clinical practice, and parental health-seeking decisions. For each of these groups, global health policy makers, health professionals, and parents, disentangling ‘normal’ variation from pathology is a critical concern. In both global health and biomedical settings, normalcy of growth is judged against the curves of the growth chart and the assumption is made that, for all children and populations, individuals failing to track along a growth channel (i.e. the 25^th percentile or the 60^th percentile) or fall outside the bounds of the curves are suspected of ill health (Zemel 2017). Two key assumptions of growth monitoring are that 1) individual children grow following the lines of the population-derived chart and 2) position along the curves or deviations from them have the same meaning for all children and populations (i.e. being in 5^th percentile for height in the United States, for example, carries the same risk as being in the 5^th percentile for height in India). Research addressing this first assumption has shown that the biology of growth does not follow the statistically-smoothed lines represented in the growth chart (Lampl and Thompson 2007); rather, child growth follows an episodic model with both genetic and environmental determinants (e.g. (Lampl et al. 1992; Lampl 2005; Lampl and Schoen 2017). Less research has examined this second assumption and, consequently, what population differences in growth potential may mean for the interpretation of child health remains an important question.

The WHO contends that universal cut-points, +/− 2 SD from the mean, can be used as biomarkers to identify children at risk for elevated morbidity and mortality (de Onis and Branca 2016). The idea that growth potential may vary between populations, however, suggests that a single cut-point may also not indicate risk similarly in all populations. Hruschka and colleagues have argued that observed z-scores should be partitioned into two components, accrued, which reflects the influence of environmental factors and morbidity, and basal, which reflects an “underlying” state (Hruschka and Hadley 2016; Hackman and Hruschka 2020; Hadley and Hruschka 2017). When partitioned, substantial differences in basal z-scores unrelated to infant and child mortality are seen between countries after controlling for the differences in accrued HAZ associated with poor diet and environmental conditions. In support of this argument, they compare the stunting prevalence derived from observed HAZ in Haiti (6%) and Guatemala (14%) to child health metrics (Hackman and Hruschka 2020). They show that, despite what seems like a growth advantage, Haitian children have a greater basal HAZ and experience considerably greater deficits in accrued HAZ, contributing to higher mortality and morbidity despite their higher observed HAZ. These findings suggest that the relationship between ill health and body size may be population-specific and that using the WHO standards may not identify populations at risk.

Concerns about who is missed when compared to a single norm are also central in the clinical literature. In clinical practice, departure from the growth chart canal is often one of main signals of growth disorders and/or underlying disease (Zemel 2017). Thus, the underlying identification of diseases requiring early intervention for improved outcomes, such as Turner syndrome, cystic fibrosis (CF), and Crohn’s disease, are dependent on the growth charts used (Usatin et al. 2017; Christesen et al. 2016; Savage et al. 2016). In girls with Turner Syndrome, for example, growth failure may be one of the only early signs of the condition and the sensitivity of detection is significantly different using the WHO vs. population-based growth charts, 36% vs. 72%, respectively (Saari et al. 2013). Similarly, in conditions like CF where larger size is associated with better lung function and long-term survivorship, clinical advice and intervention may rely on the growth charts used with fewer children viewed as “at risk” on the WHO charts with its lower WHZ than the CDC charts (Usatin et al. 2017). As a recent review concluded, the accumulating evidence that the heights of healthy children may differ from the normal ranges of the WHO MGRS limits their use for accurate screening for growth disorders in many settings, including Canada, Belgium, Denmark, Norway and Turkey (Christesen et al. 2016), and local, national charts may be preferred.

In LMIC, similar concerns have been raised about the applicability of the WHO MGRS standard when all children suffer some degree of faltering. For example, in Bangladesh, 60% of children would have a length or height <−2 SD and the vast majority would fall below the mean (van Buuren and van Wouwe 2008). While this distribution may accurately reflect constraints on growth, it has also led to concerns that doctors and parents could perceive the MGRS standards as unattainable and the charts, therefore, as not applying or relevant for Bangladeshi children. This question of whether charts are meaningful for the population under study is an important one (Moffat 2000) and may depend on the goal of growth monitoring, generating comparative data on the prevalence of stunting or identifying risk factors within a local setting (Martin et al. 2019). Likewise, the use of national vs. international charts may depend on the resources available for both implementation and training on new charts and also for evaluating and treating the larger number of children who could be identified for intervention (Duggan 2010).

Despite the better predictive power of national growth charts for identifying growth disorders and children at greatest risk within their local context (Savage et al. 2016; Heude et al. 2017), national charts have several important limitations. Secular trends lead to charts quickly become outdated (Christesen et al., 2016). In particular, the increasing prevalence of obesity has led to concerns that overweight and obesity have become the ‘norm’ in many national charts further driving obesity rates (Cole 2007). Practical concerns are also important; the cost and time needed for data collection and reference construction may be out of reach for many LMIC (Huede et al., 2019). Recently, big data approaches have been proposed to capitalize on the increasing use of electronic medical records to more easily and frequently create national references (Huede et al., 2019). While this approach would cut down on the cost, the need for electronic tracking and widespread routine pediatric care may still pose significant limitations for many LMIC; however, such approaches may be possible on regional levels.

While little research examines how specific cut-points and/or choice of growth charts influence parental health seeking or other care behaviors, concerns have been raised that the mismatch between local norms and globally-derived charts may negatively impact parents and their children. Since parents in numerous settings view growth as a sign of health and adequate parenting, smaller size in comparison to the WHO MGRS cut-points compared to locally- or nationally-derived charts may contribute to anxiety about child diet and feelings of inadequacy or failure when children’s growth is slower (Thompson et al. 2014; Reifsnider et al. 2000; Wright and Weaver 2007). Concerns have been raised, for example, that the more rapid weight gain of WHO infants in the first six months of life compared to other growth references may, counter to the goals of the WHO, lead to earlier cessation of breastfeeding in smaller infants as mothers and health professionals become concerned that infants are not getting enough (Ziegler and Nelson 2012; Binns et al. 2008). Parents in LMIC, particularly those in regions undergoing rapid economic development, may alter their infant and young child feeding practices to achieve the “Western” linear growth standards of the WHO MGRS, potentially driving the development of overweight (Flax et al. 2020). Importantly, while parents do value growth chart assessment, they also base their decisions about their child’s health and their decision to seek care on multiple models of ‘normal growth’ (Mchome, Bailey, Darak, Kessy, et al. 2019; Thompson et al. 2014; Flax et al. 2016). Low-income African American mothers, for example, do incorporate growth charts into their assessments of healthy growth, but also rely on their child’s eating, temperament and prior growth, creating individualized definitions of “normal” that may change over time (Thompson et al. 2014). What parents decide to do when biomedical and personal models of ‘normal’ come into conflict remains an important and understudied question.

Applications of normalcy: Defining stunting

One place where concerns over defining ‘normal’ vs. pathology has received considerable recent attention is in the treatment and prevention of child stunting. While the causes and consequences of stunted linear growth have long received attention in global health and human biology, this measure has increasingly become the focus of large-scale global health efforts with the inclusion of stunting eradication as one the Sustainable Development Goals of the United Nations. At the same time, the understanding and use of stunting has also evolved over the past several decades from a general indicator of “living and welfare” in a population (Eveleth and Tanner 1990) to an individual-level condition or disease state, a direct reflection of undernutrition, and an outcome metric to evaluate the effectiveness of public health and social programs (Perumal et al. 2018).

These expanding uses of stunting have been criticized for a number of reasons that parallel concerns about defining a universal norm of human growth. First, researchers have criticized the over-reliance on <−2 HAZ as a biomarker for poor growth (Perumal et al. 2018). While low HAZ is indisputably associated with increased morbidity and mortality (Olofin et al. 2013; Black et al. 2008), little evidence supports a threshold effect for increased morbidity or mortality at −2 z-scores. Rather, pooled analysis of prospective growth studies has documented increasing risk of mortality from all causes with every 0.5 SD decrease in HAZ below −1 SD without evidence of an inflection point in risk at −2 HAZ (Olofin et al. 2013). Similarly to Hruschka and colleagues concerns over population differences in basal HAZ, global health researchers have raised concerns about the applicability of this single cut-point in populations where the entire distribution of HAZ may be shifted downwards, indicating that nearly all children have a height deficit relative to their growth potential (Perumal et al. 2018; Roth et al. 2017). In these cases, prevalence measures of stunting may dramatically underestimate the number of children who are affected by inadequate growth environments.

Similar to debates around the underlying causes of population variation in growth, the recent shift to interpreting stunting as a direct indicator of undernutrition has been critiqued for ignoring the multifactoral determinants of stunting (Perumal et al. 2018). First, the argument that stunting necessarily reflects malnutrition relies on a simplistic understanding of the biology of growth, namely that growth in early life is primarily driven by dietary intake in an energy in/growth out manner. This understanding has been criticized as ignoring the complex biology of linear growth (e.g. (Frongillo et al. 2019; Lampl and Schoen 2017), with its underlying gene by environment interactions that shape the timing and frequency of growth spurts. This understanding of growth also ignores other important determinants of stunting, such as exposure to unclean water, limited sanitation or lack of vaccination for common childhood illnesses, that increase the risk of infection, increase energy needs for immune activation, and food absorption through damage to the gut (Vilcins et al. 2018; Martorell 1989). This assumption that growth is primarily driven by diet places the responsibility for poor child growth on the child’s caretakers. For example, a 2014 WHO colloquium report calls caretakers “the main protagonists of healthy child growth and development (providing appropriate feeding, care and simulation).” Such a categorization makes short stature or poor child growth the “fault” of caretakers and implies that better parental decision-making or more proper behavior would improve child growth. Locating the “fault” for stunting within the household ignores population-wide, community-level exposures, such as lack of water treatment facilities or inadequate health care services, that contribute to nearly ubiquitous growth faltering among children in some LMIC settings (Roth et al. 2017).

Indeed, the reliance of externally generated growth curves like the MGRS can not identify individual-, maternal- or household factors that shape growth within communities where stunting is ubiquitous and attributable to community-wide factors. Martin and colleagues, for example, document age-related deviations from the WHO MGRS curves among the Tsimane that could only be identified using locally derived growth charts (Martin et al. 2019). For the Tsimane, relying on a universal norm for growth would not identify biologically relevant factors contributing to growth variation between children. In a strongly worded critique, Scheffler and colleagues (2020) similarly question the appropriateness of the use of global growth standards derived from high SES and mainly westernized populations to classify children in mainly low SES, remote areas of former European colonies as undernourished and unhealthy without examining other indicators of energy balance or nutritional adequacy. Certainly, from this perspective, the WHO MGRS standards cannot be viewed unquestioningly as a “neutral, value-free yardstick” of normal child growth that should apply to all children globally without a better understanding of the range of variability and health consequences of this variability.

Conclusion

Considerable variability exists in child growth across populations and interpretations of this variability rely on the comparative framework used. However, frameworks, such as the WHO MGRS standard, depend on the representativeness of the sample selected, the inclusion/exclusion criteria used and the statistical methods applied to represent ‘normal’ growth. This single norm therefore cannot represent healthy growth for all children in all locations at all times; yet, variation from this norm is often taken as evidence of pathology. Underlying the view of variability as pathology are debates within public health and human biology about the amount of variation that can be attributed to hereditary vs. environmental factors. Historically, public health has tended to view growth variability as stemming from poor nutrition, pathogenic environments and inadequate parental care. Human biology, by contrast, has placed more emphasis on adaptation as a source of population variation. However, both perspectives are limited by a normal-pathological model that ignores that child growth is a dynamic process shaped by genetics, population histories of adaptation, and exposures within the current physical and social environment. Parents’ perceptions of ‘normal’ growth also permit a more integrated model that also judges individual children’s trajectories in comparison to others in their local context and more broadly to growth charts and other developmental markers. These views of ‘normal’ are important for both biomedicine and parents, since they shape referrals for care, treatment strategies, and feeding practices, influencing subsequent child growth and health outcomes.

As human biologists often working among lower income and remote populations in former European colonies (Scheffler et al., 2020), it seems important that we take a more cautious and contextualized approach to defining normal child growth. What we define as “normal” is a historical and statistical artifact shaped by the referent group chosen and the social, political, and economic context of the included samples. As others reviewing the use of growth references have noted, for many of the world’s populations we do not know what growth would look like under ‘ideal’ conditions because few children have been able to grow up in socioeconomically advantaged conditions (Eveleth and Tanner 1990; Stinson 2012). Without these examples, we cannot know what the true “universal” norm of human growth looks like or, from a theoretical perspective, understand individual and population-level adaptability. Over-reliance on a single universal model of healthy human growth that does not include these children may pathologize “normal” variation and, at the same time, miss children in need of intervention because they fall within the cut-point for “normal.” While this paper has mainly focused on small body size, issues of defining normal and intervening on poor growth apply to other child growth issues and child development more broadly, such as overweight and obesity, growth hormone treatment of idiopathic short stature, and achievement of developmental milestones. In all of these cases, a more nuanced understanding of the biology of growth, its underlying determinants, and its association with health is critical for identifying poor growth in individuals and populations, shaping successful interventions, and addressing parental concerns about the “normalcy” and health of their children.