Weight, and the measurement of it, is a fundamental component of child and young people’s health, both in the assessment of acute illness and in surveillance of normal child health and chronic conditions. Paediatric teams spend much time measuring and referring to “underweight”, but the important concept of what underweight actually is is frequently forgotten! A useful definition of underweight is a low weight for height (conventionally using BMI), which is directly or indirectly detrimental to physical and psychological wellbeing, which in children potentially impacts on physical growth and development (including puberty), as well as neurocognitive development. In medical terms, detrimental features of underweight can be thought of as the acute features leading to medical instability which will be focused on here and the more chronic underweight or failure to thrive which may result in growth faltering.

For children, weight and height are traditionally compared to the population normal distribution using centiles, most commonly by plotting onto centile charts. This is in contrast to adults where standard body mass indices are used for overweight and underweight. In the UK, growth charts utilise the 1990 cross-sectional growth data taken from round 30,000 children of all ages and both sexes (Cole et al. 1995). Comparable data are available for other countries. More recently, body mass index (BMI: calculated as weight in Kg divided by the square of height in metres) has become the conventional way of measuring both underweight and overweight and both population-specific BMI centile charts and WHO global charts exist. The World Health Organization defines underweight using z-scores (standard deviations), below the mean for age and sex, with a BMI for age and sex −2 z-score (roughly equivalent to <2nd centile BMI) and severe underweight as under −3 z-score (Cole et al. 2007). In the eating disorder literature and clinical settings, an alternative measure called expected body weight or %median BMI is frequently used where a child’s BMI is expressed as a percentage of the median population BMI for age and sex (the 50th centile on a BMI growth chart). 90 %, 80 % and 70 % of the median BMI roughly equate to −1, −2 and −3 z-scores (or standard deviation scores (SDS)). % median BMI is frequently referred to as weight for height. This is potentially confusing, as there are a number of ways of calculating weight for height and should be avoided (Table 3.3).

Although % median BMI and degree of underweight are an important measure of risk, it is likely a poor proxy for medical instability, at least for early-onset eating disorders. In a UK surveillance study, 40 % of children less than 13 years old presenting with an eating disorder with signs of medical instability had a BMI above the 2nd centile/-2SD/80 % median BMI (Hudson et al. 2012b). Thus, %BMI at presentation is poorly sensitive to medical instability. In contrast, %median BMI or degree of underweight is quite specific, with children at lower %median BMI usually having clinical evidence of medical instability. There are two possible explanations for the unreliability of degree of underweight at presentation as a marker of medical instability. Firstly, BMI is normally distributed, and by using % of the median or centiles for individuals, we are comparing individuals to the population average. Many children will normally be healthy above the average and many below. As an example, a child who was premorbid at the 80th centile for BMI and loses weight and presents at the 50th centile will be at 100 % median BMI but may have lost sufficient weight for him or her to be medically compromised. In contrast, a child normally on the 3rd centile for BMI (just above 80 % median BMI and by definition 3 % of children at any age will be at or below this point normally) might lose a smaller amount of weight to become below the 80 % median BMI position but not be medically affected. Added to this is the fact that the population norms are not racially sensitive, and therefore, application to non-Caucasian and mixed-race children is potentially misleading. The second important consideration is rate of weight loss. In the early-onset eating disorder study, rate of weight loss was associated with presence of medical instability at presentation independent of actual BMI or %BMI (Hudson et al. 2012b). Typically, overweight patients who develop eating disorders lose significant amounts of weight quickly to reach a healthy weight by population standards but have medical instability on examination.

The key messages are that those working with underweight should understand its measurement conceptually, and, irrespective of %median BMI or BMI centile at presentation, all children and young people should have a thorough medical examination at presentation to look for signs of medical compromise. Research on paediatricians in the UK suggests that this might not happen due to lack of knowledge (Hudson et al. 2013) and highlights the importance of co-working between medical and mental health professionals.

The effects of starvation are seen most frequently in the cardiovascular system (Winick 1979; Peebles et al. 2006). As children starve, they rely on muscle metabolism to generate energy, and there is a paucity of protein for regeneration. The heart shrinks and this impacts the heart rate and its ability to function as a pump. Autopsies in starved children and young people show significantly reduced heart size (Kohn et al. 1998).

Presence of dehydration is variable in children and young people with AN, depending on whether restriction of oral intake has included fluids as well as food. Assessment and management of dehydration is a core skill of paediatricians. However malnutrition makes degree of hydration more challenging to assess as many of the tissue signs are unreliable (Mackenzie et al. 1989). History of fluid intake, losses and previous weights in children and young people who have been starved are also unreliable. Therefore, clinicians may need to rely more on urea and electrolytes. Fluid rehydration is always safer orally rather than via an intravenous route, so if dehydration is prominent, it is better to rehydrate with oral rehydration solution in an inpatient setting for 24 h, if this is tolerated and there is no shock present. As this contains few calories, it is usually considered acceptable by patients with AN. Presence of hypernatraemia should prompt slower rates of rehydration, and if corrected intravenously, isotonic fluids (e.g. 0.9 % NaCl) should be used. Dieticians should also factor fluid requirements into the meal plan as part of a team approach.

Despite being relatively common, surprisingly, little is known about both the aetiology and consequences of hypothermia in AN. It is likely a combination of low body fat content (leading to decreased capacity for insulation), low metabolic rate in starvation, central hypothalamic dysfunction and also potentially thyroid abnormalities. From a risk perspective, degree of hypothermia reflects severity of underweight, although again there is little evidence for what constitutes risk. Whether or not patients with hypothermia are at increased risk of pathologies relating to hypothermic illness is also unclear. The best practice is to provide nutrition to recover.