After several years of experience with body mass index (BMI)-for-age reference standards in the US, the UK, and elsewhere, reflection on ‘how things are going’ seems timely. In this issue, Reilly1 offers a summary of the evidence base for the diagnostic accuracy of BMI in youth and his perspective on what is achieved by a definition of overweight and obesity based on high BMI. To complement this, in our short review, we describe the BMI measure itself, the utility of a BMI z-score (s.d. score), their utility in cross-sectional and longitudinal applications in public health/surveillance, clinical and population-based research settings.
Body mass index defined
Body mass index is a measure of weight adjusted for height. It is calculated as weight in kilograms divided by the square of height in meters. Although BMI is an imperfect tool – it does not distinguish overweight due to excess fat mass from overweight due to excess lean mass – it is the most commonly used measure for assessing obesity in adults. Other methods of determining adiposity are more accurate,2 but have limited applicability to screening or studying large populations. The BMI is well correlated with these more direct fatness measures,3, 4 and weight and height are simple, inexpensive, non-invasive measurements that are recorded routinely in clinical and research settings.
Others have discussed the limitations of clinical screening for high adiposity by comparing weight centiles to height centiles,5 and the inaccuracy of assessing overweight by observation or ‘eye-balling’ has been established.6, 7 Therefore, for screening or for epidemiologic research, using a weight/height index to define obesity has advantages that outweigh its limitations. Despite the likelihood of misclassification of the small percentage of individuals whose high BMI is due to lean muscle mass (e.g. some professional athletes), the great majority of individuals with high BMI have excess body fat.
Use of body mass index in children and adolescents
The BMI is used to assess weight status in children and adolescents as well as adults, but whereas in adults the BMI cut points that define obesity and overweight are not linked to age and do not differ for males and females, in growing children BMI varies with age and sex. Thus, a 5-year-old boy with a BMI of 20 kg/m2 is likely to be overfat, but a 15-year-old boy with a BMI of 20 kg/m2 is likely to be lean. As a result, for BMI to be meaningful in children it must be compared to a reference-standard that accounts for child age and sex.
Choice of a reference standard
National and international BMI-for-age reference standards are available. The US BMI-for-age reference is based on nationally representative data from boys and girls ages 2–20 years collected between 1963 and 1980.8 National reference standards are also in use in the UK,9 and are under development elsewhere. An international BMI reference has been produced by the International Obesity Task Force (IOTF) with data from children in the US, UK, Hong Kong, the Netherlands, Singapore and Brazil.10 Controversy exists about whether and under what circumstances a national or international reference standard is best.11, 12, 13, 14
Body mass index z-scores
Body mass index z-scores, also called BMI standard deviation (s.d.) scores, are measures of relative weight adjusted for child age and sex. Given a child's age, sex, BMI, and an appropriate reference standard, a BMI z-score (or its equivalent BMI-for-age percentile) can be determined. It should be noted that BMI z-scores are calculated relative to an external reference (whether national or international) and not to an internal reference. The effect of this is that more than 5% of the population could be, for example, above the 95th percentile of BMI-for-age. Whereas, if an internal reference were used, exactly 5% of every sample would be above the 95th percentile, and the specific BMI cut point designating the 95th percentile would not be the same across samples. For this reason, BMI z-scores are based on an external reference.
Body mass index z-scores correspond to growth chart percentiles, and can be converted into their equivalent BMI-for-age percentiles by comparison to a normal distribution table, but this is unnecessary for clinical application. In clinical practice, BMI-for-age growth charts can be used to determine a child's BMI-for-age percentile and to track relative weight status through childhood and adolescence. Body mass index-for-age growth charts are available based on US15 and UK9 reference standards. The US charts can be downloaded from www.cdc.gov/growthcharts. The UK charts can be ordered from http://shop.healthforallchildren.co.uk/. Using the US BMI-for-age reference, a 5-year-old boy with a BMI of 20 kg/m2 has a BMI z-score of approximately 2.5 (BMI >99th percentile) and a 15-year-old boy with a BMI of 20 kg/m2 has a BMI z-score of approximately 0.0 (BMI=50th percentile). In the United States, BMI-for-age percentiles above the 95th percentile in children and adolescents are labeled ‘overweight’ and BMI-for-age percentiles between the 85th and 95th percentiles are labeled ‘at risk for overweight.’16 Therefore, although both the 5- and 15-year-old boys described above have a BMI equal to 20 kg/m2, only the 5-year-old would be considered ‘overweight.’ Consensus in the UK is awaited; at present different cutoff points are being used for surveillance and clinical application.
Considerations for research
As alluded to above, BMI z-scores and BMI percentiles are in one sense equivalent; one can be converted into the other using a simple mathematical transformation. In most research applications, either BMI z-scores or BMI-for-age percentiles can be used to determine cut points and classify weight status of children and adolescents. However, if a continuous measure of relative weight is required, such as to assess change in adiposity, BMI z-scores and BMI-for-age percentiles will not be equivalent, and BMI-for-age percentiles are poorly suited for statistical analysis.17 Analyses using BMI z-scores can be reported in BMI percentile units for interpretability, but BMI percentile should not be the analytic variable. BMI z-scores can be appropriately used to compare between group means and to model relative weight trajectories longitudinally. It is sometimes suggested that crude BMI be used as the analytic variable with adjustment for child age and sex, but this approach is likely to decrease the interpretability of results because an internal rather than external standardization for age and sex is used.
Implications for longitudinal research into adulthood
When research spans childhood and adulthood, discrepancies between the child/adolescent and adult definitions of overweight become apparent. In adults, except for those of Asian heritage,18 BMI cut points of >25 kg/m2 indicating overweight and >30 kg/m2 indicating obesity are widely accepted, as is the terminology of overweight for the lesser extreme and obesity for the greater extreme. Equivalent terminology of overweight and obesity is used for children and adolescents by the IOTF and UK BMI-for-age references. In the US, however, the terminology of at risk for overweight and overweight has been adopted in an effort to reduce stigmatizing language; an unfortunate by-product of this is the uncertainty of meaning now attached to the term overweight, which can refer to different levels of relative weight depending upon age and reference standard. Additionally, many countries (e.g. Canada, Chile, Australia and Mexico) are using the US BMI reference, but not the ‘at risk for overweight’ terminology.
Of the reference standards currently available, only the IOTF reference provides a smooth transition from the child/adolescent to the adult definition of overweight and obesity; the IOTF reference defines overweight as the centile of BMI-for-age that passes through a BMI of 25 kg/m2 at age 18 years, and obesity as the centile that passes through a BMI of 30 kg/m2. To address the discontinuity between the UK BMI-for-age reference and adult definitions, Chinn and Rona19 provide cutoff points that pass through BMI=25 and 30 kg/m2 at age 19.5 years. The US BMI reference is defined for youth 2–20 years; in girls the 85th percentile crosses a BMI of 25 kg/m2 at age16.5 years and the 95th percentile crosses a BMI of 30 kg/m2 at 17 years, and in boys these estimates are 17.5 years for the 85th percentile and 19.5 years for the 95th percentile. Thus, an 18-year-old girl with a BMI of 25.5 kg/m2 would be considered overweight by the adult definition and normal weight by the US child/adolescent definition (BMI percentile ∼80th); with BMI=27 kg/m2 she would be overweight by the adult and at risk for overweight by the child/adolescent definition (BMI percentile ∼90th), and with BMI=30 kg/m2 she would be obese by the adult and at risk for overweight by the child/adolescent definition (BMI percentile ∼94th). In longitudinal research including youth and adults, there is not a straight-forward definition of overweight and obesity that can be used at all ages. We20 and others21 have used the age-20 BMI z-score reference for individuals 20 years of age and older in order to have a consistent metric of relative weight across development that can be analyzed as a continuous variable. Alternatively, if establishing cut points, individuals could be considered obese if their BMI-for-age is above the 95th percentile or their BMI is above the 30 kg/m2 cut point.22
Considerations for public health surveillance
The public health arena needs childhood obesity criteria in order to quantify the extent of the problem, establish its prevalence in different geographic areas and among different subgroups, to monitor temporal trends, and to test the impact of interventions implemented to address the problem. Several requirements for a weight measure accompany these public health uses. First, one needs definitions based on a fixed cut-off point. In the US, the CDC 2000 growth reference was constructed using data pooled across several national surveys and excluded the weight data after 1980, due to concerns that a trend of increasing child weight was already evident. In the UK, where the BMI-for-age reference is based on data collected in 1990, it has been suggested that this standard be frozen to aid future comparisons.23 In adults, observations regarding the percentage body fat and health risks at different BMI levels in Asians led to the conclusion in a World Health Organization expert consultation that disease risk occurred at levels below the current definitions.24 We lack comparable information in children. To the extent that any measure means the same thing for population subgroups, comparisons using the same cut point will be valid. Second, valid comparison across population subgroups and across time requires standard measurement protocols for height and weight; these protocols and training materials to support their use are readily available.25, 26 Third, where valid comparisons across populations are desired, one must apply a common reference standard to populations within the same time frame using consistent age groupings. For example, Wang et al.27 examined trends in pediatric overweight prevalence based on two waves of data from four countries using IOTF criteria. The survey periods although fairly similar do not match: the earlier period was 1974, 1991, 1992 and 1971 and the later period was 1997, 1997, 1998 and 1994, for Brazil, China, Russian and the United States, respectively. Such inconsistencies in timing obscure similarities and differences. In another study, in collaboration with the World Health Organization, school surveys were conducted using standard data collection methodology across Europe, Israel and the US, a massive undertaking that provides critical information but is resource intensive and requires extraordinary coordination.28 In an alternative approach, Obesity (formerly Obesity Research) has established reporting guidelines to make more consistent the reporting of results, by requiring that IOTF criteria be used in addition to any national reference, and by defining standard age groupings. Other journals might consider a similar approach. Although consistent reporting will make it possible to directly compare across time and place, as discussed above, such comparisons will be valid only if one assumes that the cutoff points have the same meaning in terms of body composition and health consequences.
As Reilly1 concludes elsewhere in this issue, a clinically based guideline based on BMI-for-age is generally useful in pediatric practice, given its consistently high specificity. A recent review of screening for pediatric overweight undertaken by the US Preventive Services Task Force came to a similar conclusion.29 The Task Force based their assessment on the demonstrated association of childhood overweight, particularly during adolescence, with adult overweight.
Looking to the future
Height and weight are simple and ubiquitous measures and have historically formed the basis of growth monitoring. Body mass index, which shows reasonably good correlations with more direct measures of adiposity and consistent linkages with adult overweight- and obesity-related co-morbidities, will likely continue to be the main measure of weight status in children. It is doubtful that we will gain much by further refinement of algebraic manipulations of height and weight. A more direct measure that reflects adiposity would be preferable, but the current alternatives are poorly suited to clinical or population research applications. Bioelectric impedance, arguably the most appealing proxy measure of adiposity for field use, is sensitive to hydration status, may vary by ethnicity, and requires instrument- and population-specific equations.30
Addition of an anthropometric indicator of central adiposity has been suggested. Several reference standards for waist circumference and waist–hip ratio have been developed and evidence is accumulating to suggest that these measures in combination with BMI may have utility for identification of those children whose high BMI has greatest health impact.31, 32, 33, 34, 35 The waist:height ratio also shows some promise as useful measure of size-adjusted central adiposity.36, 37, 38 The addition of circumferential measures may help to address the major weakness of BMI: its inability to distinguish between elevated adiposity and elevated lean mass. From a public health perspective, the observation that from 1987 to ∼1997 waist circumferences of British children increased more than BMI33 suggests that surveillance by BMI alone may obscure important changes in body composition39 at the population level. Prospective studies that demonstrate that the addition of an indirect measure of central adiposity is more tightly linked to obesity-related health consequences than use of BMI-for-age alone would provide some of the justification needed to add further complexity to research applications and to weight-screening recommendations.
As the evidence accrues, BMI-for-age continues to offer a valid and readily available measure for use in clinical and population-based applications. Widespread adoption of BMI-for-age in all three sectors will depend upon continued efforts to train individuals in the appropriate use of national and international growth references.
Reilly JJ . Diagnostic ability of the BMI for age in pediatrics. Int J Obes Relat Metab Disord 2006; 30: 595–597.
Sopher A, Shen W, Pietrobelli A . Pediatric body composition methods. In: Heymsfield SB, Lohman TG, Wang ZM, Going SB (eds.). Human Body Composition. Human Kinetics: Champaign, IL, 2005, pp 129–140.
Willett WC . Anthropometric measures and body composition. Nutritional Epidemiology, 2nd edn. Oxford University Press: New York, 1998, pp 244–272.
Pietrobelli A, Faith MS, Allison DB, Gallagher D, Chiumello G, Heymsfield SB . Body mass index as a measure of adiposity among children and adolescents: a validation study. J Pediatr 1998; 132: 204–210.
Cole TJ . A chart to link child centiles of body mass index, weight and height. Eur J Clin Nutr 2002; 56: 1194–1199.
Cross JH, Holden C, MacDonald A, Pearmain G, Stevens MC, Booth IW . Clinical examination compared with anthropometry in evaluating nutritional status. Arch Dis Childhood 1995; 72: 60–61.
CDC Nutrition and Physical Activity. CDC Growth Chart Training. June 22, 2005. Available at: http://www.cdc.gov/nccdphp/dnpa/growthcharts/training/modules/index.htm Accessed December 10, 2005.
Kuczmarski RJ, Ogden CL, Grummer-Strawn LM, Flegal KM, Guo SS, Wei R et al. CDC Growth Charts: United States. National Center for Health Statistics: Hyattsville, Maryland, Dec 4, 2000.
Cole TJ, Freeman JV, Preece MA . Body mass index reference curves for the UK, 1990. Arch Dis Childhood 1995; 73: 25–29.
Cole TJ, Bellizzi MC, Flegal KM, Dietz WH . Establishing a standard definition for child overweight and obesity worldwide: international survey. BMJ 2000; 320: 1240–1243.
Janssen I, Katzmarzyk PT, Srinivasan SR, Chen W, Malina RM, Bouchard C et al. Utility of childhood BMI in the prediction of adulthood disease: a comparison of national and international references. Obes Res 2005; 13: 1106–1115.
Rona RJ, Chinn S . One cheer for the international definitions of overweight and obestiy. Arch Dis Childhood 2002; 87: 390–391.
Reilly JJ . Assessment of childhood obesity: national reference data or international approach? Obes Res 2002; 10: 838–840.
Deurenberg P . Universal cut-off BMI points for obesity are not appropriate. Br J Nutr 2001; 85: 135–136.
Centers for Disease Control and Prevention. CDC Growth Charts: United States. Available at: http://www.cdc.gov/growthcharts/ Accessed December, 2005.
Barlow SE, Dietz WH . Obesity evaluation and treatment: expert committee recommendations. Pediatrics 1998; 102: 1–11.
Cole TJ, Faith MS, Pietrobelli A, Heo M . What is the best measure of adiposity change in growing children: BMI, BMI %, BMI z-score or BMI centile? Eur J Clin Nutr 2005; 59: 419–425. Corregendum (vol 459, pg 807).
Obesity in Asia Collaboration. Ethnic comparisons of obesity in the Asia-Pacific region: protocol for a collaborative overview of cross-sectional studies. Obes Rev 2005; 6: 193–198.
Chinn S, Rona RJ . International definitions of overweight and obesity for children: a lasting solution? Ann Hum Biol 2002; 29: 306–313.
Anderson SE, Cohen P, Naumova EN, Must A . Association of depression and anxiety disorders with weight change in a prospective community-based study of children followed up into adulthood. Arch Pediatr Adolesc Med 2006; 160: (in press).
Wright CM, Parker L, Lamont D, Craft AW . Implications of childhood obesity for adult health: findings from thousand families cohort study. BMJ 2001; 323: 1280–1284.
Field AE . Epidemiology of the health of the health and economic consequences of pediatric obesity. In: Goran MI, Southern MS (eds.). Handbook of Pediatric Obesity: Etiology, Pathophysiology and Prevention. CRC Press, Taylor & Francis: Boca Raton, 2006, pp 1–19.
Wright CM, Booth IW, Buckler JMH, Cameron N, Cole TJ, Healy MJR et al. Growth reference charts for use in the United Kingdom. Arch Dis Childhood 2002; 86: 11–14.
WHO Expert Consultation. Appropriate body-mass index for Asian populations and its implications for policy and intervention strategies. Lancet 2004; 363: 157–163.
Lohman TG, Roche AF, Martorell R (eds) Anthropometric Standardization Reference Manual. Human Kinetics: Champaign, IL, 1988, pp 1–177.
US Dept of Health and Human Services. Accurately weighing and measuring infants, children and adolescents: technique. Available at: http://depts.washington.edu/growth/ Accessed December, 2005.
Wang Y, Monteiro C, Popkin BM . Trends of obesity and underweight in older children and adolescents in the United States, Brazil, China, and Russia. Am J Clin Nutr 2002; 75: 971–977.
Lissau I, Overpeck MD, Ruan WJ, Due P, Holstein BE, Hediger ML et al. Body mass index and overweight in adolescents in 13 European countries, Israel, and the United States. Arch Pediatr Adolesc Med 2004; 158: 27–33.
Whitlock EP, Williams SB, Gold R, Smith PR, Shipman SA . Screening and interventions for childhood overweight: a summary of evidence for the US preventive services task force. Pediatrics 2005; 116: E125–E144.
Mast M, Sönnichsen A, Langnäse K, Labitzke K, Bruse U, Preuß U et al. Inconsistencies in bioelectrical impedance and anthropometric measurements of fat mass in a field study of prepubertal children. Br J Nutr 2002; 87: 163–175.
Fernandez JR, Redden DT, Pietrobelli A, Allison DB . Waist circumference percentiles in nationally representative samples of African-American, European-American, and Mexican-American children and adolescents. J Pediatr 2004; 145: 439–444.
Fredriks AM, van Buuren S, Fekkes M, Verloove-Vanhorick SP, Wit JM . Are age references for waist circumference, hip circumference and waist–hip ratio in Dutch children useful in clinical practice? Eur J Pediatr 2005; 164: 216–222.
McCarthy HD, Jarrett KV, Crawley HF . The development of waist circumference percentiles in British children aged 5.0–16.9 year. Eur J Clin Nutr 2001; 55: 902–907.
Moreno LA, Fleta J, Mur L, Feja C, Sarria A, Bueno M . Indices of body fat distribution in Spanish children aged 4.0 to 14.9 years. J Pediatr Gastroenterol Nutr 1997; 25: 175–181.
Pietrobelli A, Fernandez JR, Redden JT, Allison DB . Waist circumference percentiles in nationally representative samples of black, white, and Hispanic children. Int J Obes Relat Metab Disord 2002; 26: S93.
Gustat J, Elkasabany A, Srinivasan S, Berenson GS . Relation of abdominal height to cardiovascular risk factors in young adults: The Bogalusa Heart Study. Am J Epidemiol 2000; 151: 885–891.
Kahn HS, Imperatore G, Cheng YJ . A population-based comparison of BMI percentiles and waist-to-height ratio for identifying cardiovascular risk in youth. J Pediatr 2005; 146: 482–488.
McCarthy HD, Ashwell M . A study of central fatness using waist-to-height ratios in UK children and adolescents over two decades supports the simple message – ‘keep your waist circumference to less than half your height. Int J Obesity Relat Metab Disord (advance online publication, 24 January 2006).
Wells JCK, Coward WA, Cole TJ, Davies PSW . The contribution of fat and fat-free tissue to body mass index in contemporary children and the reference child. Int J Obes Relat Metab Disord 2002; 26: 1323–1328.
About this article
Associations of extracurricular physical activity patterns and body composition components in a multi-ethnic population of UK children (the Size and Lung Function in Children study): a multilevel modelling analysis
BMC Public Health (2019)
Sleep and Breathing (2019)
BMC Pediatrics (2018)
BMC Medicine (2018)
Maternal Depressive Symptoms Mediate the Association between Socio-economic Status and Adolescent Weight Outcomes: A Longitudinal Analysis
Maternal and Child Health Journal (2018)