Measuring severe obesity in pediatrics: A cohort study

doi:10.21203/rs.3.rs-2715882/v1

Purpose: To examine cross-sectional and longitudinal relationships between body mass index (BMI)-derived metrics for measuring severe obesity (SO) using the Centers for Disease Control and Prevention (CDC) and World Health Organization (WHO) references and cardiometabolic risk factors in children and adolescents.

Methods: In this cohort study completed from 2013 to 2021, we examined data from 3- to 18-year-olds enrolled in the CANadian Pediatric Weight management Registry. Anthropometric data were used to create nine BMI-derived metrics based on the CDC and WHO references. Cardiometabolic risk factors were examined, including dysglycemia, dyslipidemia, and elevated blood pressure. Analyses included intraclass correlation coefficients (ICC) and receiver operator characteristic area-under-the-curve (ROC AUC).

Results: Our sample included 1,288 participants (n=666 [51.7%] girls; n=874 [67.9%] white), with SO of 59.9–67.0%. ICCs revealed high tracking (0.90–0.94) for most BMI-derived metrics. ROC AUC analyses showed CDC and WHO metrics discriminated the presence of cardiometabolic risk factors, which improved with increasing numbers of risk factors. Overall, most BMI-derived metrics rated poorly in identifying presence of cardiometabolic risk factors.

Conclusion: CDC BMI percent of the 95^th percentile and WHO BMIz performed similarly as measures of SO, suggesting both can be used for clinical care and research in pediatrics. The latter definition may be particularly useful for clinicians and researchers from countries that recommend using the WHO growth reference.

Obesity

Anthropometry

Pediatric

Growth

Canada

Cardiometabolic Risk Factors

Severe obesity in pediatrics is a global health issue.
Few reports have evaluated BMI-derived metrics based on the WHO growth reference.

What is New

Our analyses showed that CDC BMI_p95 and WHO BMIz performed similarly as evidence-based measures of severe obesity in pediatrics.
WHO BMIz should be particularly useful for clinicians and researchers from countries that recommend using the WHO growth reference.

Obesity is complex [1], chronic [2], and affects millions of children and adolescents worldwide [3]. Although some data suggest the prevalence of pediatric obesity has stabilized [4, 5], severe obesity (SO) has continued to increase [6]. Since negative health consequences (e.g., dyslipidemia, high blood pressure, insulin resistance) are common with SO [7, 8], effective approaches are needed for screening and management.

Most children and adolescents enrolled in obesity management have SO [8–12]. To evaluate anthropometric changes [13], clinicians calculate body mass index (BMI) and compare individual measurements to a reference population. BMI is recommended for assessing weight status [14] since height and weight can be measured reliably and accurately [15] and because BMI is associated with obesity-related health risks [16]; however, it is not clear how to best express BMI-derived metrics for those with SO, which hinders our ability to understand the magnitude of the issue and health impacts.

In the US, the Centers for Disease Control and Prevention (CDC) BMI Growth Charts are used to define obesity [17]. Due to sparse reference data and compression > 97th percentile, CDC BMI z-score (BMIz) is a poor metric for assessing obesity severity [18–20]. Alternatively, BMI as a percentage of the 95th percentile (CDC BMI_p95) is used often to define SO and has emerged as the preferred measure of SO for children and adolescents [21–23].

In Canada, World Health Organization (WHO) criteria are used to define pediatric obesity [24]. It is important to note that the WHO and CDC BMI references are not interchangeable [21]. To mitigate the asymptotic plateau beyond Z = 3, the WHO expresses high BMIz values in terms of SD₂₃ (the distance between z = 2 and z = 3), providing a continuous, longitudinal z-score measure. Given differences between the CDC and WHO references and the lack of an international standard for measuring SO, our objective was to examine relationships between several CDC and WHO BMI-derived metrics and cardiometabolic risk factors for clinical utility in children and adolescents with obesity.

Study Design and Participants

Our study included data from the CANadian Pediatric Weight management Registry (CANPWR). Completed from 2013–2021, CANPWR was an observational cohort study to evaluate determinants of health outcomes from children and adolescents from nine multidisciplinary pediatric obesity management clinics in Canada. Several CANPWR reports have been published [8, 25–28]. Overall, of the 1,907 individuals approached in these clinics to participate in CANPWR, 1,334 (70.0%) enrolled in CANPWR. For the current study, we included data collected at baseline, 6 months, and 12 months follow-up for cross-sectional and longitudinal analyses. Participants were included if they had data collected, including age, sex, height, weight, blood pressure, and laboratory testing. Children provided assent and their parent or legal guardian provided informed, written consent prior to data collection. Each of the nine participating weight management clinics obtained ethical approval from their local institutional research ethics boards.

We examined several BMI-derived metrics in our analyses, including untransformed BMI (kg/m²) along with, based on the CDC reference, BMIz (CDC BMIz), BMI percent of the 95th percentile (CDC BMI_p95), BMI percent of the median standardized for age 20y (CDC BMI_pmed), modified BMIz (CDC BMIz_{_mod}), and extended BMIz (CDC BMIz_{_ext}), all of which were described in a recent report [29]. We included several companion metrics based on the WHO reference, including WHO BMIz (with SD₂₃ correction for Z ≥ 3 or Z ≤ -3), BMI percent of the 97th percentile (WHO BMI_p97), and BMI percent of the median standardized for age 19y (WHO BMI_pmed). In total, we generated and evaluated nine alternative metrics based on the CDC and WHO references.

Data Collection & Measurements

CANPWR data were collected using harmonized procedures across centres [30] with de-identified data entered into a secure, web-enabled software application (iDataFax 2014.1) for central management in the Population Health Research Institute (McMaster University, Hamilton, Ontario, Canada) [30]. Height and weight were measured twice for reliability [30] with a venous blood draw to assess cardiometabolic risk factors, which included abnormal values for fasting plasma glucose (≥ 6.1 mmol/L; [31]), hemoglobin A1C (≥ 5.6% [32] and ≥ 6.0% [31]), total cholesterol (> 5.18 mmol/L), HDL cholesterol (> 1.04 mmol/L), LDL cholesterol (≥ 3.37 mmol/L), and triglycerides (≥ 1.13 mmol/L for ≤ 9 year olds; ≥1.47 mmol/L for > 9 year olds) [33]. Systolic (SBP) and diastolic blood pressure (DBP) were measured and percentiles determined according to recommendations [34]. Each visit was assigned a blood pressure category consistent with normal, elevated, stage 1 or stage 2 hypertension [34], with stages 1 and 2 defined as elevated BP. Diagnoses were combined with documented use of medications to manage dysglycemia, dyslipidemia, or elevated blood pressure, or a clinical diagnosis assigned by primary healthcare providers. Questions from the Canadian Health Measures Survey [35] were used to document family sociodemographic characteristics.

Sample Size Calculations

Sample size calculations were based on the number of cases required to identify a statistically significant receiver operator characteristic area-under-the-curve (ROC AUC) using methods described by Obuchowski and McClish [36, 37]. ROC plots are used to assess discrimination performance for binary diagnoses – here presence or absence of cardiometabolic risk factors. A plot of sensitivity on the y axis vs 1-specificity on the x-axis is made as the diagnostic threshold is varied, and the ROC AUC serves as a summary measure, with AUC = 0.5 no better than chance and a value of 1.0 in the case of perfect discrimination. A 95% confidence interval that crosses 0.5 is therefore non-diagnostic. At baseline and 12 months, a minimum detectable AUC = 0.60 could be achieved with 80% power and α = 0.05. At 6 months, minimum detectable AUC = 0.62. For the longitudinal cohort, participants required complete clinical and laboratory data for ≥ 2 visits. We defined an adverse outcome as ≥ 1 risk factor (dysglycemia, dyslipidemia, or elevated blood pressure). Defining an adverse outcome as ≥ 1 risk factor that was confirmed on ≥ 2 occasions, minimum detectable effect size was 0.60 with 80% power and α = 0.05 In the absence of simple formulae to assess our ability to distinguish between two different AUC values, Monte Carlo simulations were performed on synthetic populations that shared the adverse outcome profile of our cohort, with population ROC AUCs of 0.60 and 0.70, respectively. Power to distinguish between them was calculated using the methods of Obuchowski and McClish [37] as the sample size varied (n = 200–800), with mean power calculated using 500 replicates for each. Approximately 690 participants could distinguish between a ROC AUC of 0.60 and 0.70 with power of 80% and α = 0.05.

Statistical Analysis

We performed statistical analyses in R v3.5.0 [38] with p < 0.05 deemed statistically significant. CDC and WHO BMIz scores and alternate metrics were calculated using our publicly available R-Shiny tools (https://canpwrgrowth.shinyapps.io/CANPWR_WHO/; https://canpwrgrowth.shinyapps.io/CANPWR_CDC/). For descriptive purposes, we calculated the prevalence of SO using an age- and sex-specific CDC BMI ≥ 120% of the 95th percentile [17] and WHO BMI ≥ 99.9th percentile (WHO BMI_p99.9, WHO BMIz ≥ 3.0) [39]. Definitions of SO were extended to all participants, although the WHO definition was not originally intended for children < 5 years of age [39], which included 46/1,288 (~ 3.6%) in our sample. We used Pearson correlation coefficients to measure cross-sectional relationships between the BMI-derived metrics at each time point. Tracking performance was assessed by intra-class correlations (ICC; 95% confidence intervals [CI]), calculated with the Shrout-Fleiss algorithm [40], which required complete data at all 3 time points (n = 836). Though utility may be limited with short-term follow-up, ICC is deemed useful when tracking metrics over time, contrasting the similarity of an individual child’s values with differences across children; Freedman et al. completed similar analyses in a large sample of children with high BMIs [18]. ROC AUCs were calculated with the pROC package [41], with pairwise comparisons using the Delong test for correlated curves [42]. ROC AUC values from ≤ 0.5, > 0.5 to < 0.7, ≥ 0.7 to < 0.8, ≥ 0.8 to < 0.9, and ≥ 0.9 were considered as providing no, poor, acceptable, excellent, and outstanding, respectively, discrimination between individuals based on the presence of risk factors [43]. We completed cross-sectional analyses at baseline, 6 months, and 12 months using ≥ 1, ≥2, or 3 risk factors to define an adverse outcome at each time point. A longitudinal sub-group included complete data at baseline with a follow-up visit at 6 months and/or 12 months.

We had complete anthropometric data for 1,288 individuals at baseline (females: n=666 [51.7%]; white ethnicity: n=874 [67.9%]; age: 12.4 years old [inter-quartile range {IQR}: 9.8, 15.0]; height: 157.4 cm [143.7, 166.5], weight: 80.0 kg [57.8, 101.7]; BMI: 34.3 kg/m² [28.6, 39.0]). At 6 and 12 months post-baseline, some anthropometric, blood pressure, and laboratory data were missing due to participant drop out and because some clinics collected blood pressure and laboratory data every 12 months only (Table 1). We had complete anthropometric data for ≥2 time points for 394 participants in our longitudinal cohort. The prevalence of SO at all time points was similar based on CDC (BMI_p95) and WHO (BMI_p99.9) references (59.9–67.0%); estimates were slightly higher using the CDC reference. At baseline (n=723), the prevalence of laboratory risk factors ranged from 3.0% (for glucose) to 39.1% (for triglycerides). In total, 25.3%, 48.4%, 23.9%, and 2.4% had 0, 1, 2, and 3 risk factors at baseline; similar prevalence levels were observed at 6 and 12 months (data not shown).

Our baseline Pearson correlation coefficient matrix (Table 2) revealed most BMI-derived metrics were positively and significantly correlated. Similar r values for correlation coefficients were calculated at 6 and 12 months (not shown). Tracking of BMI-derived metrics from baseline, 6 and 12 months was assessed using ICCs (Table 3). For the 836 participants with complete anthropometric data, ICCs were high for most metrics (0.90–0.94). When we compared CDC BMIz (ICC: 0.90; 95% CI: 0.89–0.91) and WHO BMIz (ICC: 0.94; 95% CI: 0.93–0.94), they differed (p<0.001). When we compared CDC BMIz to all CDC BMI-derived metrics, all alternative metrics were superior (data not shown), although this has limited clinical significance since all ICCs ≥0.90.

Figure 1 (Panel A) highlights differences between CDC and WHO BMIz in histogram form, with CDC BMIz truncated at >3.0 and WHO BMIz showing a long right tail, the result of SD₂₃ correction and a leaner reference population. A histogram for weight z-scores is seen in Figure 1 (Panel B), revealing a similar discrepancy. As shown in Figure 1 (Panel C), truncated tail values were corrected with CDC BMI_p95 or WHO BMI_p97. Figure 1 (Panel D) plots CDC versus WHO BMIz with superimposed loess smooth to illustrate the asymptotic plateau in CDC BMIz.

Figure 2 shows a representative CANPWR growth trajectory on CDC and WHO BMI growth charts. As shown in the accompanying table, BMIz scores were not interchangeable – truncated CDC BMIz values did not capture changes in anthropometry, which were more accurately depicted by other metrics. The WHO growth charts have a slightly higher upper limit than the CDC version (39 versus 36 kg/m²), but neither is useful for plotting BMI values >40 kg/m², a limitation that was addressed by the new CDC growth charts for children with very high BMIs that enable plotting values up to 60 kg/m² [29].

In our cross-sectional analyses (Table 4), the ROC AUC values for CDC and WHO BMI-derived variables ranged from 0.55–0.73 and were statistically significant (95% CIs did not cross the null hypothesis value of 0.5). Discrimination was generally poor, with a few deemed acceptable, though values tended to increase with the number of risk factors used to define an adverse outcome. None ranked as excellent or outstanding. This was confirmed by pairwise comparisons showing minimal differences between (i) CDC BMIz or (ii) WHO BMIz and their alternative metrics. Most comparisons were non-significant (p>0.05). For the few that achieved statistical significance, all included small differences, with ROC AUC values in the poor category.

In our longitudinal analyses (Table 5), the ROC AUC values for CDC and WHO BMI-derived metrics ranged from 0.59–0.72. Like our cross-sectional analyses, most ranked as poor, few were acceptable, and values tended to increase with increasing number of risk factors. In pairwise comparisons between CDC BMIz and derived metrics, the latter were superior statistically (all p<0.05). WHO BMIz also outperformed CDC BMIz (3 co-morbidities: CDC BMIz [ROC AUC: 0.60] versus WHO BMIz [ROC AUC: 0.67]; p<0.004), although both values were deemed poor. In contrast to the CDC BMI-derived metrics, none of the alternative WHO BMI-derived metrics were superior statistically to WHO BMIz (all p>0.05). A small number of metrics were acceptable, but none were excellent or outstanding.

Our analyses revealed that most BMI-derived metrics were positively and significantly related with one another, with most tracking well. CDC BMIz performed poorly as a measure of SO, and our ROC AUC analyses revealed that both CDC and WHO BMI-derived metrics had poor abilities to distinguish cardiometabolic risk factor profiles.

As in previous reports [20, 44], CDC BMIz suffered from compression when BMI values exceeded the 97th percentile; WHO BMIz did not suffer from this limitation and performed similarly to CDC BMI_p95. The impact of asymptotic compression on the right tail of the BMIz distribution was first noted during development of the WHO growth reference charts [24], where distances between z-scores increased progressively further from the median. Gulati et al., [19] noted similar issues with the CDC growth charts, concluding that “the transformed data does not allow for modelling in the extreme tails of the distribution”. At least two different solutions were offered: Although the CDC and WHO both used US National Center for Health Statistics (NCHS) data, the WHO truncated theirs at 1973 to avoid the obesity epidemic, excluded 3% of children with ‘unhealthy’ weights for heights, and applied the SD₂₃ correction [45]. In 2016, the CDC adopted BMI_p95 to assess SO (BMI ≥ 120% of the 95th percentile) and the degree of obesity severity, which was added to their SAS macros that same year as their preferred metric [46].

While most CDC and WHO BMI-derived metrics were significantly correlated with one another, none emerged as a superior measure of SO to identify cardiometabolic risk. Aside from BMI percentile, the tracking performance of all CDC and WHO BMI-derived metrics was good. CDC BMI_p95 has been evaluated extensively [20, 44, 47], informing the decision by the CDC to include it as their preferred option to define SO [46]. Since we did not find meaningful differences between WHO BMIz and alternative metrics, we compared WHO BMIz to CDC BMI_p95. WHO BMIz is advantageous as a continuous, easily interpreted measure of BMI throughout its range. Arguably, the metric is more familiar to clinicians than many of the alternatives. In considering the CDC BMI_p95 and WHO BMIz, it is worth noting that (i) the two metrics were highly and significantly correlated, (ii) the percentage of individuals classified as having SO was similar between metrics, (iii) both the CDC and WHO references were based on US children and adolescents collected by the NCHS, and (iv) the CDC has already adopted the WHO growth standard for 0–2y [48]. WHO BMIz could therefore be used universally as a measure SO in children and adolescents, particularly in countries including Canada and many European countries that recommend using the WHO classification system. Additional research and knowledge translation activities are needed to inform discussions regarding such an endeavour.

None of the BMI-derived metrics performed well in differentiating children and adolescents based on cardiometabolic risks. This highlights the limited utility of BMI-derived metrics as proxy measures of obesity-related risk factors for SO. Other anthropometric metrics of abdominal adiposity, including waist circumference (WC) or waist-to-height ratio, may be better for predicting risk[49–51]. Collecting accurate and reliable anthropometric data in pediatric obesity can be a challenge. Several of our CANPWR sites did not measure WC due to difficulty in identifying anatomical landmarks or physical/emotional discomfort. While interest is increasing to move beyond BMI as a singular measure of obesity [52], evidence-based metrics like CDC BMI_p95 and WHO BMIz represent reliable, available, and familiar options for screening and evaluating anthropometric changes over time.

Our study was not without limitations. First, most participants enrolled in CANPWR identified as white, and all were referred by clinicians to multidisciplinary management of pediatric obesity, so our results may differ from non-clinical samples drawn from the broader community. In a previous report [25], we found that CANPWR participants differed from a nationally representative sample of children enrolled in the Canadian Health Measures Survey (CHMS). Specifically, CANPWR participants had higher anthropometric (e.g., BMI z-score, waist circumference) and cardiometabolic risk factor (e.g., blood pressure, triglycerides) values versus their CHMS peers. Second, our sample included a relatively small number of younger children (n = 46 < 5 years). This issue limits the generalizability of our data to this stage, which is notable since health risks related to obesity, including cardiometabolic risk factors, tend to be less common in younger children versus older children (i.e., 12–18 years of age). Third, our follow-up data collection decreased over time as participants discontinued their clinical care or were lost to follow-up. Many factors lead to attrition in managing pediatric obesity [53], so it is possible that participants who contributed longitudinal data differed from those who did not. Nevertheless, comparisons of demographic data from our cohorts at 0, 6, and 12 months (as shown in Table 1) were consistent with data missing at random. Although this is never provable, the similarities in our results across visits suggest this assumption is reasonable.

Overall, CDC BMI_p95 and WHO BMIz performed similarly as measures of SO in pediatrics. Although several BMI-derived metrics had advantages, CDC BMIz as well as CDC and WHO BMI percentiles had limited utility. Future research should extend our findings to other groups to evaluate the potential to establish a universal measure of SO in pediatrics.

BIC (Bayesian Information Criteria)

BMI (body mass index)

BMIz (body mass index z-score)

CANPWR (CANadian Pediatric Weight management Registry)

CDC (Centers for Disease Control and Prevention)

DBP (diastolic blood pressure)

HDL (high density lipoprotein)

ICC (intraclass correlation coefficients)

LDL (low density lipoprotein)

ROC AUC (receiver operator characteristic area-under-the-curve)

SBP (systolic blood pressure)

SO (severe obesity)

WHO (World Health Organization)

Competing Interests

Geoff Ball has received research funding from the Canadian Institutes of Health Research, Alberta Health Services, the Women and Children’s Health Research Institute (University of Alberta), and the Public Health Agency of Canada. He has served as a consultant for Novo Nordisk Canada. The other authors have no example conflicts of interest to disclose.

Acknowledgements

The authors wish to thank all the families who contributed their data to the CANPWR study as well as all the research, clinical, and administrative team members across our sites for their assistance in collecting and managing study data.

Funding

This research was funded by the Stollery Children’s Hospital Foundation and the Alberta Women’s Health Foundation through the Women and Children’s Health Research Institute at the University of Alberta.

Author Contributions

Dr. Geoff Ball conceptualized and designed the study, drafted the initial manuscript, and critically reviewed and revised the manuscript.

Dr. Atul Sharma helped to conceptualize and design the study, planned, and conducted data analysis, and critically reviewed and revised the manuscript.

Dr. Sarah A Moore helped to conceptualize and design the study, and critically reviewed and revised the manuscript.

Dr. Dan L Metzger assisted with data analysis, and critically reviewed and revised the manuscript for important intellectual content.

Dr. Doug Klein helped to conceptualize and design the study, and critically reviewed and revised the manuscript.

Dr. Katherine Morrison helped to conceptualize and design the study, led the collection and management of study data from the multi-centre CANadian Pediatric Weight management Registry (CANPWR) study, and critically reviewed and revised the manuscript.

Drs. Annick Buchholz, J-P Chanoine, Jill Hamilton, MD, Josephine Ho, Anne-Marie Laberge, Laurent Legault, Lehana Thabane, Mark S Tremblay, and Ian Zenlea are CANPWR Investigators who critically reviewed and revised the manuscript.

All authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.

Consent Statement

Patient consent was required for our study.

Avis JL, Bridger T, Buchholz A, et al (2014) It’s like rocket science…only more complex: challenges and experiences related to managing pediatric obesity in Canada. Expert Rev Endocrinol Metab 9:223–229. https://doi.org/10.1586/17446651.2014.897605
Singh AS, Mulder C, Twisk JWR, et al (2008) Tracking of childhood overweight into adulthood: a systematic review of the literature. Obes Rev 9:474–488. https://doi.org/10.1111/j.1467-789X.2008.00475.x
NCD Risk Factor Collaboration (NCD-RisC) (2017) Worldwide trends in body-mass index, underweight, overweight, and obesity from 1975 to 2016: a pooled analysis of 2416 population-based measurement studies in 128·9 million children, adolescents, and adults. Lancet 390:2627–2642. https://doi.org/10.1016/S0140-6736(17)32129-3
Hales CM, Fryar CD, Carroll MD, et al (2018) Trends in Obesity and Severe Obesity Prevalence in US Youth and Adults by Sex and Age, 2007-2008 to 2015-2016. JAMA 319:1723–1725. https://doi.org/10.1001/jama.2018.3060
Rodd C, Sharma AK (2017) Prevalence of overweight and obesity in Canadian children, 2004 to 2013: Impact of socioeconomic determinants. Paediatr Child Health 22:153–158. https://doi.org/10.1093/pch/pxx057
Pinhas-Hamiel O, Hamiel U, Bendor CD, et al (2022) The Global Spread of Severe Obesity in Toddlers, Children, and Adolescents: A Systematic Review and Meta-Analysis. Obes Facts 15:118–134. https://doi.org/10.1159/000521913
Skinner AC, Perrin EM, Moss LA, Skelton JA (2015) Cardiometabolic Risks and Severity of Obesity in Children and Young Adults. N Engl J Med 373:1307–1317. https://doi.org/10.1056/NEJMoa1502821
Hadjiyannakis S, Ibrahim Q, Li J, et al (2019) Obesity class versus the Edmonton Obesity Staging System for Pediatrics to define health risk in childhood obesity: results from the CANPWR cross-sectional study. Lancet Child Adolesc Health 3:398–407. https://doi.org/10.1016/S2352-4642(19)30056-2
Buchholz A, Howard AL, Baldwin K, et al (2020) Health trajectories of children with severe obesity attending a weight management program. Paediatr Child Health 25:439–446. https://doi.org/10.1093/pch/pxz088
Kumar S, King EC, Christison AL, et al (2019) Health Outcomes of Youth in Clinical Pediatric Weight Management Programs in POWER. J Pediatr 208:57-65.e4. https://doi.org/10.1016/j.jpeds.2018.12.049
Perez AJ, Yaskina M, Maximova K, et al (2018) Predicting Enrollment in Multidisciplinary Clinical Care for Pediatric Weight Management. J Pediatr 202:129–135. https://doi.org/10.1016/j.jpeds.2018.06.038
Hagman E, Danielsson P, Lindberg L, et al (2020) Paediatric obesity treatment during 14 years in Sweden: Lessons from the Swedish Childhood Obesity Treatment Register-BORIS. Pediatr Obes 15:e12626. https://doi.org/10.1111/ijpo.12626
Styne DM, Arslanian SA, Connor EL, et al (2017) Pediatric Obesity-Assessment, Treatment, and Prevention: An Endocrine Society Clinical Practice Guideline. J Clin Endocrinol Metab 102:709–757. https://doi.org/10.1210/jc.2016-2573
Cole TJ, Lobstein T (2012) Extended international (IOTF) body mass index cut-offs for thinness, overweight and obesity. Pediatr Obes 7:284–294. https://doi.org/10.1111/j.2047-6310.2012.00064.x
Carsley S, Parkin PC, Tu K, et al (2019) Reliability of routinely collected anthropometric measurements in primary care. BMC Med Res Methodol 19:84. https://doi.org/10.1186/s12874-019-0726-8
Freedman DS, Sherry B (2009) The validity of BMI as an indicator of body fatness and risk among children. Pediatrics 124 Suppl 1:S23-34. https://doi.org/10.1542/peds.2008-3586E
Kuczmarski RJ, Ogden CL, Grummer-Strawn LM, et al (2000) CDC growth charts: United States. Adv Data 1–27
Freedman DS, Davies AJG, Kompaniyets L, et al (2021) A Longitudinal Comparison of Alternatives to Body Mass Index Z-Scores for Children with Very High Body Mass Indexes. J Pediatr 235:156–162. https://doi.org/10.1016/j.jpeds.2021.02.072
Gulati AK, Kaplan DW, Daniels SR (2012) Clinical tracking of severely obese children: a new growth chart. Pediatrics 130:1136–1140. https://doi.org/10.1542/peds.2012-0596
Freedman DS, Butte NF, Taveras EM, et al (2017) BMI z-Scores are a poor indicator of adiposity among 2- to 19-year-olds with very high BMIs, NHANES 1999-2000 to 2013-2014. Obesity (Silver Spring) 25:739–746. https://doi.org/10.1002/oby.21782
Valerio G, Maffeis C, Balsamo A, et al (2013) Severe obesity and cardiometabolic risk in children: comparison from two international classification systems. PLoS One 8:e83793. https://doi.org/10.1371/journal.pone.0083793
Flegal KM, Wei R, Ogden CL, et al (2009) Characterizing extreme values of body mass index-for-age by using the 2000 Centers for Disease Control and Prevention growth charts. Am J Clin Nutr 90:1314–1320. https://doi.org/10.3945/ajcn.2009.28335
Dietz WH (2017) Time to Adopt New Measures of Severe Obesity in Children and Adolescents. Pediatrics 140:e20172148. https://doi.org/10.1542/peds.2017-2148
de Onis M, Onyango AW, Borghi E, et al (2007) Development of a WHO growth reference for school-aged children and adolescents. Bull World Health Organ 85:660–667. https://doi.org/10.2471/blt.07.043497
Tremblay MS, Feng M, Garriguet D, et al (2015) Canadian Pediatric Weight Management Registry (CANPWR): baseline descriptive statistics and comparison to Canadian norms. BMC Obes 2:29. https://doi.org/10.1186/s40608-015-0060-6
Damanhoury S, Morrison KM, Mian R, et al (2021) Metabolically healthy obesity in children enrolled in the CANadian Pediatric Weight management Registry (CANPWR): An exploratory secondary analysis of baseline data. Clin Obes e12490. https://doi.org/10.1111/cob.12490
Clairman H, Dettmer E, Buchholz A, et al (2019) Pathways to eating in children and adolescents with obesity. Int J Obes (Lond) 43:1193–1201. https://doi.org/10.1038/s41366-018-0271-2
McPhee PG, Zenlea I, Hamilton JK, et al (2021) Individual and family characteristics associated with health indicators at entry into multidisciplinary pediatric weight management: findings from the CANadian Pediatric Weight management Registry (CANPWR). Int J Obes (Lond). https://doi.org/10.1038/s41366-021-00959-3
Hales CM, Freedman DS, Akinbami L, et al (2022) Hales CM, Freedman DS, Akinbami L, Wei R, Ogden CL. Evaluation of alternative body mass index (BMI) metrics to monitor weight status in children and adolescents with extremely high BMI using CDC BMI-for-age growth charts. National Center for Health Statistics. Vital Health Stat 2(197). 2022. DOI: https://dx.doi.org/10.15620/cdc:121711.
Morrison KM, Damanhoury S, Buchholz A, et al (2014) The CANadian Pediatric Weight management Registry (CANPWR): study protocol. BMC Pediatr 14:161. https://doi.org/10.1186/1471-2431-14-161
Diabetes Canada Clinical Practice Guidelines Expert Committee, Punthakee Z, Goldenberg R, Katz P (2018) Definition, Classification and Diagnosis of Diabetes, Prediabetes and Metabolic Syndrome. Can J Diabetes 42 Suppl 1:S10–S15. https://doi.org/10.1016/j.jcjd.2017.10.003
American Diabetes Association Professional Practice Committee (2022) 2. Classification and Diagnosis of Diabetes: Standards of Medical Care in Diabetes-2022. Diabetes Care 45:S17–S38. https://doi.org/10.2337/dc22-S002
Expert Panel on Integrated Guidelines for Cardiovascular Health and Risk Reduction in Children and Adolescents, National Heart, Lung, and Blood Institute (2011) Expert panel on integrated guidelines for cardiovascular health and risk reduction in children and adolescents: summary report. Pediatrics 128 Suppl 5:S213-256. https://doi.org/10.1542/peds.2009-2107C
Flynn JT, Kaelber DC, Baker-Smith CM, et al (2017) Clinical Practice Guideline for Screening and Management of High Blood Pressure in Children and Adolescents. Pediatrics 140:e20171904. https://doi.org/10.1542/peds.2017-1904
Tremblay M, Wolfson M, Connor Gorber S (2007) Canadian Health Measures Survey: rationale, background and overview. Health Rep 18 Suppl:7–20
Obuchowski NA (2003) Receiver operating characteristic curves and their use in radiology. Radiology 229:3–8. https://doi.org/10.1148/radiol.2291010898
Obuchowski NA, McClish DK (1997) Sample size determination for diagnostic accuracy studies involving binormal ROC curve indices. Stat Med 16:1529–1542. https://doi.org/10.1002/(sici)1097-0258(19970715)16:13<1529::aid-sim565>3.0.co;2-h
Core Team R (2020) R: A language and environment for statistical computing.
Dietitians of Canada, Canadian Paediatric Society, College of Family Physicians of Canada, et al (2010) Promoting optimal monitoring of child growth in Canada: using the new WHO growth charts. Can J Diet Pract Res 71:e1-3. https://doi.org/10.3148/71.1.2010.54
Shrout PE, Fleiss JL (1979) Intraclass correlation: uses in assessing rater reliability. Psychological Bulletin 86:420–428
Robin X, Turck N, Hainard A, et al (2011) pROC: an open-source package for R and S+ to analyze and compare ROC curves. BMC Bioinformatics 12:1–8. https://doi.org/10.1186/1471-2105-12-77
DeLong ER, DeLong DM, Clarke-Pearson DL (1988) Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics 44:837–845
Applied Logistic Regression. In: Wiley Online Library. http://onlinelibrary.wiley.com/doi/book/10.1002/9781118548387. Accessed 29 Sep 2022
Freedman DS, Berenson GS (2017) Tracking of BMI z Scores for Severe Obesity. Pediatrics 140:e20171072. https://doi.org/10.1542/peds.2017-1072
Turck D, Michaelsen KF, Shamir R, et al (2013) World Health Organization 2006 child growth standards and 2007 growth reference charts: A discussion paper by the committee on Nutrition of the European Society for Pediatric Gastroenterology, Hepatology, and Nutrition. J Pediatr Gastroenterol Nutr 57:258–264. https://doi.org/10.1097/MPG.0b013e318298003f
SAS Program ( ages 0 to < 20 years ) | Resources | Growth Chart Training | Nutrition | DNPAO | CDC. https://www.cdc.gov/nccdphp/dnpao/growthcharts/resources/sas.htm. Accessed 7 Oct 2022
Freedman DS, Butte NF, Taveras EM, et al (2017) BMI z-Scores are a poor indicator of adiposity among 2- to 19-year-olds with very high BMIs, NHANES 1999-2000 to 2013-2014. Obesity (Silver Spring) 25:739–746. https://doi.org/10.1002/oby.21782
A SAS Program for the WHO Growth Charts (ages 0 to <2 years) | DNPAO | CDC. https://www.cdc.gov/nccdphp/dnpao/growthcharts/resources/sas-who.htm. Accessed 7 Oct 2022
Ejtahed H-S, Mahmoodi Z, Qorbani M, et al (2021) A comparison between body mass index and waist circumference for identifying continuous metabolic syndrome risk score components in Iranian school-aged children using a structural equation modeling approach: the CASPIAN-V study. Eat Weight Disord 26:1609–1616. https://doi.org/10.1007/s40519-020-00971-y
Tee JYH, Gan WY, Lim PY (2020) Comparisons of body mass index, waist circumference, waist-to-height ratio and a body shape index (ABSI) in predicting high blood pressure among Malaysian adolescents: a cross-sectional study. BMJ Open 10:e032874. https://doi.org/10.1136/bmjopen-2019-032874
Sharma AK, Metzger DL, Daymont C, et al (2015) LMS tables for waist-circumference and waist-height ratio Z-scores in children aged 5-19 y in NHANES III: association with cardio-metabolic risks. Pediatr Res 78:723–729. https://doi.org/10.1038/pr.2015.160
Hadjiyannakis S, Buchholz A, Chanoine J-P, et al (2016) The Edmonton Obesity Staging System for Pediatrics: A proposed clinical staging system for paediatric obesity. Paediatr Child Health 21:21–26. https://doi.org/10.1093/pch/21.1.21
Dhaliwal J, Nosworthy NMI, Holt NL, et al (2014) Attrition and the management of pediatric obesity: an integrative review. Child Obes 10:461–473. https://doi.org/10.1089/chi.2014.0060

Table 1. Demographic, anthropometric, and cardiometabolic risk factors of study participants at baseline, 6 months post-baseline, and 12 months post-baseline.

	Baseline (n=723)	6 Months Post-Baseline (n=282)	12 Months Post-Baseline (n=415)
*Sex (females, males) (n; %)*	382; 52.8, 341; 47.2	141; 52.8, 133; 47.2	209; 50.4, 206; 49.6
*Ethnicity (non-white, white) (n; %)*	204; 28.2, 519; 71.8	75; 26.6, 207; 73.4	116; 28.0, 299; 72.0
Age (y)	12.7 (10.4, 15.3)^a	13.0 (11.3, 16.2)	14.0 (11.6, 16.3)
Height (cm)	159.2 (146.4, 167.2)	163.2 (152.7, 169.0)	163.6 (152.3, 170.3)
Weight (kg)	86.0 (63.1, 107.7)	96.7 (75.2, 113.9)	94.1 (72.2, 113.1)
Body mass index (BMI; kg/m²)	33.3 (28.6, 39.0)	35.6 (30.8, 41.4)	34.6 (29.9, 40.4)
CDC BMI z-score	2.42 (2.15, 2.65)	2.46 (2.23, 2.68)	2.41 (2.13, 2.64)
CDC BMI percentile	99.2 (98.4, 99.6)	99.3 (98.7, 99.6)	99.2 (98.4, 99.6)
WHO BMI z-score	3.40 (2.82, 4.19)	3.51 (3.00, 4.37)	3.41 (2.82, 4.16)
WHO BMI percentile	100.0 (99.8, 100.0)	100.0 (99.8, 100.0)	100.0 (99.8, 100.0)
Glucose (mmol/L)	4.9 (4.6, 5.2)	4.9 (4.6, 5.2)	4.9 (4.6, 5.2)
*Abnormal: >6.1 (n; %)*	22; 3.0	10; 3.5	19; 4.6
Hemoglobin A1C (%)	5.3 (5.1, 5.5)	5.3 (5.1, 5.5)	5.4 (5.1, 5.5)
*Abnormal: >6.0 (n; %), >5.6 (n; %)*	34; 4.7, 158; 21.9	19; 6.7, 68; 24.1	25; 6.0, 91; 21.9
Total cholesterol (mmol/L)	4.20 (3.74, 4.71)	4.26 (3.76, 4.84)	4.24 (3.79, 4.81)
*Abnormal: >5.18 (n; %)*	84; 11.6	40; 14.2	53; 12.8
HDL cholesterol (mmol/L)	1.13 (0.99, 1.28)	1.10 (0.94, 1.29)	1.12 (0.98, 1.32)
*Abnormal: >1.04 (n; %)*	241; 33.3	111; 39.4	141; 34.0
LDL cholesterol (mmol/L)	2.44 (2.06, 2.91)	2.49 (2.07, 3.00)	2.43 (2.02, 2.94)
*Abnormal: ≥3.37 (n; %)*	84; 11.6	32; 11.3	45; 10.8
Triglycerides (mmol/L)	1.22 (0.89, 1.75)	1.28 (0.91, 1.82)	1.20 (0.92, 1.64)
*Abnormal: ≥1.13 (≤9 years); ≥1.47 (>9 years) (n; %)*	283; 39.1	115; 40.8	152; 36.6
Systolic Blood Pressure (SBP; percentile)	0.82 (0.61, 0.93)	0.80 (0.60, 0.92)	0.79 (0.56, 0.92)
Diastolic Blood Pressure (DBP; percentile)	0.75 (0.51, 0.91)	0.81 (0.54, 0.92)	0.77 (0.55, 0.90)
*Abnormal SBP or DBP (n; %):^b* *Elevated blood pressure* *Stage 1 hypertension* *Stage 2 hypertension*	150; 20.7 194; 26.8 45; 6.2	57; 20.2 82; 29.1 9; 3.2	93; 22.4 120; 28.9 8; 1.9

CDC (Centers for Disease Control and Prevention)

WHO (World Health Organization)

^aContinuous variables were presented as median (inter quartile range) unless otherwise specified.

^bAbnormal SBP or DBP was based on clinical practice guidelines from the American Academy of Pediatrics (see Flynn et al., Pediatrics, 2017) and included individuals with (i) documented use of medications to manage dysglycemia, dyslipidemia, or elevated blood pressure or (ii) a clinical diagnosis assigned by their primary healthcare provider.

Table 2. Pearson correlation matrix of anthropometric- and BMI-derived variables using the Centers for Disease Control and Prevention (CDC) and World Health Organization (WHO) reference charts at baseline (n=1,288).

	BMI (kg/m²)	CDC BMIz	CDC BMI_p95	CDC BMI_pmed	CDC BMIz_{_mod}	CDC BMIz_{_ext}	WHO BMIz	WHO BMI_p97	WHO BMI_pmed
BMI (kg/m²)	-	0.45*	0.81	0.73	0.59*	0.67	0.54*	0.81	0.75
CDC BMIz	0.45*	-	0.75	0.79	0.90*	0.82	0.87*	0.72	0.74
CDC BMI_p95	0.81	0.75	-	0.98*	0.91	0.95*	0.88	1.00*	0.99*
CDC BMI_pmed	0.73	0.79	0.98*	-	0.94*	0.98*	0.93*	0.97*	0.98*
CDC BMIz_{_mod}	0.59*	0.90*	0.91	0.94*	-	0.97*	0.98*	0.89	0.91
CDC BMIz_{_ext}	0.67	0.82	0.95*	0.98*	0.97*	-	0.95*	0.94*	0.96*
WHO BMIz	0.54*	0.87*	0.88	0.93*	0.98*	0.95*	-	0.87	0.90
WHO BMI_p97	0.81	0.72	1.00	0.97	0.89	0.94	0.87	-	0.99
WHO BMI_pmed	0.75	0.74	0.99*	0.98*	0.91	0.96*	0.90	0.99*	-

* p<0.05

BMI (body mass index)

CDC BMIz (CDC BMI z-score)

CDC BMI_p95 (CDC BMI percentage of the 95^th percentile)

CDC BMI_pmed (CDC BMI as percentage distance from the median standardized for age 20y)

CDC BMIz__mod(CDC BMI z-score modified)

CDC BMIz_{_ext}(CDC BMI z-score extended)

WHO BMIz (WHO BMI z-score, with SD₂₃ correction for Z outside of [-3, 3])

WHO BMI_p97 (WHO BMI percentage of the 97^th percentile)

WHO BMI_pmed (WHO BMI as a percentage distance from the median standardized for age 19y)

Table 3. Intraclass correlations (ICC) for BMI-derived metrics based on Centers for Disease Control and Prevention (CDC) and World Health Organization (WHO) reference charts at baseline (n=836).

	ICC	95% CI
CDC BMIz	0.90	0.89–0.91
CDC BMI_p95	0.94	0.93–0.95
CDC BMI_pmed	0.94	0.93–0.94
CDC BMIz_{_mod}	0.94	0.93–0.94
CDC BMIz_{_ext}	0.93	0.93–0.94
WHO BMIz	0.94	0.93–0.94
WHO BMI_p97	0.94	0.93–0.94
WHO BMI_pmed	0.94	0.93–0.94

ICC (intraclass correlation coefficient)

CI (confidence interval)

BMI (body mass index)

CDC BMIz (CDC BMI z-score)

CDC BMI_p95 (CDC BMI percentage of the 95^th percentile)

CDC BMI_pmed (CDC BMI as percentage distance from the median standardized for age 20y)

CDC BMIz__mod(CDC BMI z-score modified)

CDC BMIz_{_ext}(CDC BMI z-score extended)

WHO BMIz (WHO BMI z-score, with SD₂₃ correction for Z outside of [-3, 3])

WHO BMI_p97 (WHO BMI percentage of the 97^th percentile)

WHO BMI_pmed (WHO BMI as a percentage distance from the median standardized for age 19y)

Table 4. Cross-sectional receiver operator characteristic area-under-the-curve (ROC AUC) for Centers for Disease Control and Prevention (CDC) and World Health Organization (WHO) BMI-derived metrics at baseline with either ≥1, ≥2, or 3 cardiometabolic risk factors (n=723).

	ROC AUC (95% CI) ≥1 abnormal CRF	ROC AUC (95% CI) ≥2 abnormal CRF	ROC AUC (95% CI) 3 abnormal CRF
CDC BMIz	0.62 (0.57–0.66)	0.63 (0.58–0.67)	0.65 (0.53–0.78)
CDC BMI_p95	0.62 (0.58–0.67)	0.64 (0.59–0.68)	0.68 (0.56–0.80)
CDC BMI_pmed	0.62 (0.57–0.66)	0.62 (0.58–0.67)	0.67 (0.55–0.78)
CDC BMIz_{_mod}	0.62 (0.57–0.66)	0.63 (0.58–0.68)	0.66 (0.53–0.78)
CDC BMIz_{_ext}	0.62 (0.57–0.66)	0.63 (0.58–0.67)	0.66 (0.54–0.78)
WHO BMIz	0.61 (0.56–0.66)	0.62 (0.58–0.67)	0.65 (0.53–0.77)
WHO BMI_p97	0.62 (0.56–0.66)	0.64 (0.59–0.68)	0.68 (0.56–0.80)
WHO BMI_pmed	0.61 (0.57–0.66)	0.63 (0.58–0.67)	0.66 (0.53–0.78)

ROC AUC (receiver operator characteristic area-under-the-curve)

CI (confidence interval)

CRF (cardiometabolic risk factor)

Abnormal CRFs included dysglycemia, dyslipidemia, and elevated blood pressure

CDC BMIz (CDC BMI z-score)

CDC BMI_p95 (CDC BMI percentage of the 95^th percentile)

CDC BMI_pmed (CDC BMI as percentage distance from the median standardized for age 20y)

CDC BMIz__mod(CDC BMI z-score modified)

CDC BMIz_{_ext}(CDC BMI z-score extended)

WHO BMIz (WHO BMI z-score, with SD₂₃ correction for Z outside of [-3, 3])

WHO BMI_p97 (WHO BMI percentage of the 97^th percentile)

WHO BMI_pmed (WHO BMI as a percentage distance from the median standardized for age 19y)

Table 5. Longitudinal receiver operator characteristic area-under-the-curve (ROC AUC) for Centers for Disease Control and Prevention (CDC) and World Health Organization (WHO) BMI-derived metrics at baseline and either 6 months post-baseline or 12 months post-baseline with either ≥1, ≥2, or 3 confirmed cardiometabolic risk factors (n=394).

	ROC AUC (95% CI) ≥1 abnormal CRF	ROC AUC (95% CI) ≥2 abnormal CRF	ROC AUC (95% CI) 3 abnormal CRF
CDC BMIz	0.59 (0.53–0.64)	0.66 (0.60–0.73)	0.60 (0.44–0.75)
CDC BMI_p95	0.60 (0.54–0.66)	0.70 (0.63–0.76)	0.70 (0.55–0.85)
CDC BMI_pmed	0.60 (0.54–0.66)	0.69 (0.62–0.75)	0.72 (0.58–0.86)
CDC BMIz_{_mod}	0.59 (0.53–0.65)	0.65 (0.59–0.71)	0.69 (0.57–0.80)
CDC BMIz_{_ext}	0.60 (0.54–0.65)	0.68 (0.62–0.75)	0.68 (0.52–0.83)
WHO BMIz	0.60 (0.54–0.65)	0.68 (0.62–0.74)	0.67 (0.53–0.81)
WHO BMI_p97	0.60 (0.54–0.66)	0.70 (0.63–0.76)	0.70 (0.55–0.85)
WHO BMI_pmed	0.60 (0.54–0.66)	0.69 (0.62–0.75)	0.69 (0.53–0.84)

ROC AUC (receiver operator characteristic area-under-the-curve)

CI (confidence interval)

CRF (cardiometabolic risk factor)

Abnormal CRFs included dysglycemia, dyslipidemia, and elevated blood pressure

CDC BMIz (CDC BMI z-score)

CDC BMI_p95 (CDC BMI percentage of the 95^th percentile)

CDC BMI_pmed (CDC BMI as percentage distance from the median standardized for age 20y)

CDC BMIz__mod(CDC BMI z-score modified)

CDC BMIz_{_ext}(CDC BMI z-score extended)

WHO BMIz (WHO BMI z-score, with SD₂₃ correction for Z outside of [-3, 3])

WHO BMI_p97 (WHO BMI percentage of the 97^th percentile)

WHO BMI_pmed (WHO BMI as a percentage distance from the median standardized for age 19y)

No competing interests reported.

Measuring severe obesity in pediatrics: A cohort study

Status:

Journal Publication

Version 1

Abstract

Figures

What Is Known

Introduction

Materials & Methods

Study Design and Participants

Data Collection & Measurements

Sample Size Calculations

Statistical Analysis

Results

Discussion

Conclusion

Abbreviations

Declarations

References

Tables

Additional Declarations

Status:

Journal Publication

Version 1