Impact of Severe Obesity on Cardiovascular Risk Factors in Youth

Abstract

Objective

to compare cardiovascular risk factor clustering (CVRFC) in severely obese youth to those with lower degrees of obesity.

Study design

We divided a childhood obesity clinic derived cohort into the degrees of obesity (class I, II and III) while adding a “class IV” category corresponding to >160% of the 95^th centile of BMI for age and sex. In a cross-sectional analysis we investigated the presence of CVRFC in 2244 participants; in 621 who were followed longitudinally, we investigated the determinants of endpoint CVRFC.

Results

Class IV obesity was associated with increased risk for CVRFC compared with overweight (OR=17.26, P < .001) at a similar magnitude to class III obesity (OR=17.26, p<0.001). Males were at greater risk for presence of CVRFC (OR=1.57, p=0.03) compared with females. Adiponectin (OR=0.90, p<0.001) and leptin levels (OR=0.98, p=0.008) were protective, independent of degree of obesity. Baseline class IV obesity was associated with increased risk compared with overweight of having CVRFC at follow-up (OR=5.76, p=0.001), to a similar extent as class III obesity (OR=5.36, p=0.001). Changes in the degree of obesity were significant predictors of CVRFC on follow-up (OR=1.04, p<0.01 per percent BMI change).

Conclusions

The metabolic risk associated with obesity in childhood is conferred prior to reaching class IV obesity. An individualized risk stratification approach in children with severe obesity should be based on presence of complications rather than simple BMI cutoffs.

Although the increase in the prevalence of childhood obesity in the US seems to have recently reached a plateau¹, the prevalence of severe obesity is still on the rise². In adults the degrees of obesity are classified according to body mass index (BMI) cutoffs (30–34.9 kg/m² considered “mild” or class I, 35–39.9 kg/m² “moderate” or class II and BMI>40 kg/m² considered “severe” or class III ^3–5); similar definitions of childhood obesity are lacking. Higher obesity categories in adults have been shown to predict mortality⁶. The indications for an intensive approach towards obesity management in adults, namely surgical (bariatric) interventions, are usually based on the presence of severe obesity or moderate obesity with significant co-morbidities ^7–10. In contrast, the thresholds for performing such interventions in youth according to most ^{11, 12} but not all ¹³ authorities are set at higher BMI thresholds representing “extreme obesity” (BMI>50 kg/m² with mild co-morbidities or BMI>40 kg/m² with major co-morbidities). This raises the question whether extremely obese children are at greater risk for the presence and/or development of obesity driven metabolic co-morbidities justifying reserving intensive surgical interventions only for this high risk obesity category. In other words, although the presence of obesity associated morbidity such as pre-diabetes and cardiovascular risk factor clustering (CVRFC, defined by some as “the metabolic syndrome”) is common and has been shown to significantly rise with greater degrees of obesity in childhood,^14–16 it is not clear whether this risk increment plateaus above a certain BMI threshold or continues to rise in cases of extreme obesity. This question is important because this should dictate a strategy for defining the optimal timing for surgical and perhaps pharmacological interventions in morbidly obese children and adolescents¹⁷.

We evaluated the presence of obesity associated CVRFC in children and adolescents across the spectrum of obesity. Using a cross sectional approach – we evaluated whether such risk factors are more prevalent in extremely obese children compared with their counterparts with lower degrees of obesity. In a subsample that was followed longitudinally, we evaluated the dynamics of anthropometric and biochemical CV risk factors over time in children with various degrees of obesity at baseline. We hypothesized that children and adolescents with extreme obesity will manifest greater CVRFC that will worsen over time than those with milder forms of obesity.

Methods

Subjects aged 7–20 years were recruited to the Yale Pathophysiology of Type 2 Diabetes in Youth Study, a long-term, multiethnic cohort aimed at studying early alternations in glucose metabolism in obese children (NCT01967849). All were referred to the Yale Pediatric Obesity Clinic and received an intervention based on behavioral modification as previously described^{18, 19}. None were actively participating in other weight loss oriented programs. Follow up included bi-annual clinic visits in which behavioral modification was attempted via health monitoring, nutritional guidance and recommendations for increased physical activity along with family oriented counseling. Participants of this analysis did not take any medications at baseline or during follow-up that may affect blood pressure, glucose, lipid metabolism or weight gain/loss and had a normal functioning thyroid at baseline and throughout the follow-up. None of the participants underwent surgical interventions for obesity. Parental written consent was obtained before entering the study as was assent from the children. The Yale University School of Medicine human investigation committee approved the study.

At every clinic visit anthropometric measurements and biochemical markers were documented and an oral glucose tolerance test (OGTT) was performed at baseline and at two year intervals. The OGTT test was performed with the administration of 1.75 g of glucose per kilogram of body weight (maximal dose 75 g). Baseline blood samples were obtained from the children while they were fasting prior to the OGTT for measurement of levels of glucose, insulin, lipids, adipocytokines and C-reactive protein. Weight and height were measured to the nearest 0.1 cm and 0.1 kg respectively and BMI was calculated. Blood pressure was measured 3 times while seated, and the average of the last 2 measurements was used for the analysis. The physical examination included determination of the stage of puberty according to the criteria of Tanner²⁰. Participants were defined as pre-pubertal or pubertal (tanner stage >1).

Plasma glucose was determined with a YSI 2700 Analyzer (Yellow Springs Instruments) and lipid levels (cholesterol, triglycerides and HDL-cholesterol) were measured with an AutoAnalyzer (Model 747–200, Roche-Hitachi). Insulin and total adiponectin levels were measured using double antibody RIAs from Millipore.(insulin intra- and inter-assay coefficients of variation are 6.8 and 11.6%, respectively; adiponectin intra- and inter-assay coefficients of variation are 7.1 and 9.5%, respectively). C-reactive protein (CRP) levels were measured using the ultrasensitive assay (Kamiya Biomedical) with an intra-assay coefficient of variation that is no greater than 3.0% and inter-assay coefficient lower than 11.6%)

Elevated systolic or diastolic blood pressure was defined as a value that exceeded the 95th percentile for age and sex ²¹. Abnormalities in fasting levels of triglycerides and high-density lipoprotein (HDL) cholesterol were adjusted for age, sex, and race (>95th percentile for triglycerides; <5th percentile for HDL cholesterol)²². Impaired glucose tolerance was defined as a glucose level greater than 140 mg per deciliter (7.8 mmol per liter) but less than 200 mg per deciliter (11.1 mmol per liter) at two hours of the OGTT ²³. The participants were classified as meeting the criteria of CVRFC if they met 3 or more of the following criteria: BMI z score>2.0; triglyceride level above the 95th percentile for age and sex; HDL-cholesterol level below the 5th percentile for age and sex; systolic or diastolic blood pressure above the 95th percentile for age and sex; impaired glucose tolerance on the OGTT¹⁴. This definition of CVRFC has been shown to be associated with intimal medial thickness in obese youth²⁴. Insulin sensitivity was determined using the whole body insulin sensitivity index (WBISI)²⁵ calculated using all glucose and insulin measurements from the OGTT. For assessment of degrees of obesity in the participants we followed the approach used by Skinner et al who created a classification similar to that used in adults.¹⁵ This defines “overweight” as being between the 85^th and 95^th centile for age and sex, “mild obesity” (class I) as being between 100–120% of the 95^th centile for age and sex, “moderate obesity” (class II) as being between 120–140% of the 95^th centile for age and sex and “severe obesity” (class III) as being above the 140% of the 95^th centile for age and sex. In order to assess the impact of the extremes of obesity in childhood, we defined the “extreme-obesity” category (“class IV”) as being greater than the 160% of the 95^th centile for age and sex.

Variables are presented as means ± standard deviations. Variables that were not normally distributed were log transformed for the analysis, yet the results are presented as the original values for ease of interpretation. Group comparisons between obesity categories in the analyses were performed using ANOVA with post-hoc Bonferonni corrections for multiple comparisons. Logistic regression models were performed to identify the predictors of the presence of CV risk clustering in the cross sectional and longitudinal analyses using the overweight group as a reference. All analyses were performed using SPSS 19.0 for Windows.

Results

At baseline there were 2244 participants in the cross sectional analysis (Table I). Subjects were classified into obesity categories with an addition of the class IV obesity (extreme) category. In total, 40% were males and 60% were females. There was a higher prevalence of females in the lower compared with the higher obesity categories (pχ²<0.001). The race breakdown between obesity categories showed a significant predominance of African Americans in the super and severe obesity categories (pχ²<0.001). The overweight participants were slightly older compared with other obesity categories (p<0.01 vs. all other obesity categories).

Table 1.

Baseline patient characteristics by obesity category

Characteristics		Overweight N(%)	Class I N(%)	Class II N(%)	Class III N(%)	Class IV N(%)	P χ2
Sex	Male	80 (34.2)	186 (36.8)	255 (38.2)	199 (43.4)	187 (49.5)	<0.001
	Female	154 (63.2)	320 (63.2)	413 (61.8)	259 (56.6)	191 (50.5)
Race	Caucasian	127 (54.3)	283 (55.9)	321 (48.1)	210 (45.9)	151 (39.9)	<0.001
	African-American	48 (20.5)	101 (20.0)	198 (29.6)	155 (33.8)	159 (42.1)
	Hispanic	59 (25.2)	122 (24.1)	149 (22.3)	93 (20.3)	68 (18.0)
		Mean ±SD	Mean ±SD	Mean ±SD	Mean ±SD	Mean ±SD	p ANOVA	P Class III. vs. Class IV
Age (years)		14.2 ±3.1	13.4 ±2.7	13.0 ±2.8	12.7 ±2.7	13.2 ±3.4	<0.01	0.18
Weight (Kg)		59.4 ±15.2	74.4 ±17.2	85.3 ±20.5	97.0 ±22.9	119.1 ±31.9	<0.001	<0.001
Height (cm)		159.4 ±13.8	159.1 ±12.4	158.8 ±13.4	159.4 ±13.5	160.1 ±13.9	0.69	1.00
Waist circumference(cm)		80.1 ±11.7	94.5 ±11.8	103.2 ±12.7	112.3 ±13.0	125.4 ±16.2	<0.001	<0.001
Hip circumference(cm)		93.6 ±11.4	103.1 ±12.6	110.5 ±12.7	122.1 ±63.5	131.9 ±15.5	<0.001	0.001
Waist hip ratio		0.86 ±0.08	0.94 ±0.44	0.94 ±0.07	0.95 ±0.12	0.95 ±0.08	<0.001	1.00
DBP(mmHg)		67 ± 8	68 ± 9	70 ±9	71 ± 9	71 ±11	<0.001	0.66
SBP(mmHg) *		113 ±10	117 ±11	120 ±12	123 ±12	124 ±13	<0.001	0.83
Cholesterol (mg/dL)		160±43	158 ±35	159 ±33	159 ±31	151 ± 28	0.007	0.06
LDL(mg/dL)		90 ± 36	92 ±30	94 ±28	95 ±26	90 ±24	0.86	0.09
TG(mg/dL)		96 ± 84	117 ±89	113±96	114 ±65	102 ±61	0.007	0.34
HDL(mg/dL)		50 ±12	43.24 ±11	43 ±10	41 ±9	41 ±9	<0.001	1.00
Fasting glucose (mg/dL)		91± 10	92 ±8	93 ±9.51	93 ±9	93 ±10	0.006	1.00
2 hour glucose (mg/dL) *		115± 28	122 ±28	124 ±29	125 ±28	123 ±31	<0.001	0.99
Insulin Sensitivity (WBISI)		2.96 ±1.73	2.11 ±1.19	1.81 ±1.02	1.59 ±0.97	1.47 ±0.83	<0.001	0.27
CRP (mg/dL) *		1.1 ±2.0	2.1 ±3.5	3.1 ±4.2	3.9 ±5.2	6.9 ±7.5	<0.001	<0.001
Adiponectin (μg/dL)		9.4 ±4.3	8.0 ± 4.1	8.0 ±4.1	7.4 ±3.6	7.4 ±3.5	<0.001	0.99
Leptin (ng/dL) *		13.0 ±9.8	21.9 ±11.5	28.0 ±11.2	33.7 ±12.8	43.9 ±15.9	<0.001	<0.001
ALT (u/L) *		17 ±15	24 ±4	25±25	29 ±33	26 ±25	<0.001	0.99

P ANOVA vales presented for comparisons of all obesity categories along with specific p values for the comparisons of class III and class IV obese categories. These comparisons were performed using the post hoc Bonferroni correction including all obesity categories.

^*parameters that were log transformed for the analysis, presented as original values for ease of interpretation.

As expected, weight increased by obesity category from ~59 kg in the overweight group to ~119 kg in the class IV group. Similarly, waist circumference, hip circumference and waist-to-hip ratio significantly increased across obesity categories (p ANOVA<0.001 for all). Diastolic and systolic blood pressure increased by obesity category and HDL-cholesterol tended to decrease (p<0.001 for all). Triglycerides tended to increase across obesity categories (p=0.007) and LDL-cholesterol was comparable among the obesity categories (p=0.86). Fasting and 2-hr glucose tended to increase across obesity categories (p =0.006 and <0.001 respectively) as did ALT (p<0.001). In post hoc comparisons, significant differences were detected between class III and class IV obese participants in waist circumference and waist-to-hip ratio.

Importantly, CRP increased while adiponectin decreased across obesity categories (p <0.001 for both) and leptin increased with growing degrees of obesity (p<0.001). Both leptin and CRP were significantly greater in class IV compared with class III obese participants (p<0.001 for both). Insulin sensitivity significantly decreased (p<0.001) with no significant difference between severe and extremely obese participants.

Components of the metabolic syndrome, namely fasting and 2-hr glucose, triglycerides, HDL-cholesterol and systolic blood pressure were all associated with the degree of obesity (Figure 1, A–E). Fasting, 2-hr glucose and systolic blood pressure were also associated with pubertal status. Specifically, fasting glucose and systolic blood pressure were higher in pre-pubertal children compared with their pubertal counterparts while 2-hr glucose was greater in pubertal adolescents. Post-hoc comparisons revealed no differences in any of these variables between class III and class IV obese participants.

Figure 1. Relation of metabolic parameters, obesity category and pubertal status.

White bars- pre-pubertal, black bars – pubertal, * p<0.001 for class IV vs. class III obesity category

ALT, leptin, adiponectin, C-reactive protein and insulin sensitivity (Figure 1, F–J) were significantly associated with the degree of obesity. ALT, leptin and WBISI (whole body insulin sensitivity index) were also associated with pubertal status showing higher ALT and insulin sensitivity and lower leptin in pre-pubertal children compared with pubertal adolescents. Importantly, only CRP and leptin were significantly higher in class IV compared with the class III participants in both pubertal categories (p<0.001 for both) while no other metabolic factor demonstrated a similar difference between these 2 obesity categories.

We used a logistic regression model to identify predictors of the presence of the CVRFC (Table 2). Model 1 included pubertal category, degree of obesity, race and sex as independent variables. In this model, increasing degrees of obesity were associated with significantly greater risk of meeting criteria for the CVRFC. Importantly, the class III and class IV obesity groups showed significantly increased risk compared with overweight participants yet of very similar magnitudes (OR=17.26 and OR=19.30 for class III and class IV respectively, p <0.001 for both). Obese males were at greater risk for CVRFC compared with females (OR=1.57, p=0.03). In model 2 we added ALT (as a surrogate of hepatic steatosis), CRP, leptin and adiponectin as independent variables. Although obesity categories remained significant predictors of the presence of the presence of CVRFC, leptin and adiponectin showed a protective effect (OR=0.98 and OR=0.90 respectively, p=0.008 and p<0.001 respectively). ALT seemed to have a small risk increasing effect that in this model reached marginal significance (p=0.06). As in the previous model, male sex increased the risk of CVRFC compared with females (OR=1.58, p=0.02). Again, both class III and class IV categories were at significantly greater risk for CVRFC in this model yet the odds ratios were of similar magnitudes (18.86 and 20.32 respectively, p<0.001 for both).

Table 2.

Determinants of the presence of CVRFC

		Model 1			Model 2			Model 3
Variable		OR	95% C.I.	p-value	OR	95% C.I.	p-value	OR	95% C.I.	p-value
Puberty category	Pre-pubertal	1			1			1
	Pubertal	0.97	0.63–1.48	0.88	1.22	0.49–3.01	0.66	0.84	0.25–2.87	0.79
Obesity Category	Overweight	1			1
	Class I	6.92	3.30–14.50	<0.001	7.80	1.75–34.62	0.007
	Class II	12.89	6.23–26.66	<0.001	13.32	3.04–58.25	0.001
	Class III	17.26	8.28–35.96	<0.001	18.86	4.14–85.55	<0.001	1
	Class IV	19.30	9.20–40.46	<0.001	20.32	4.18–98.77	<0.001	1.29	0.71–2.32	0.40
Race	Caucasian	1			1			1
	African-American	0.72	0.58–0.91	0.006	0.64	0.42–0.99	0.04	1.14	0.54–2.39	0.72
	Hispanic	0.91	0.71–1.17	0.49	1.07	0.66–1.71	0.77	0.53	0.24–1.17	0.54
Sex	Female	1			1			1
	Male	1.57	1.02–2.42	0.03	1.58	1.07–2.33	0.02	1.13	0.34–3.76	0.83
ALT (per 1 IU)					1.01	1.00–1.02	0.06	1	0.98–1.01	1
CRP (per 1 mg/dl)					1.02	0.98–1.05	0.39	1	0.96–1.05	0.70
Leptin (per 1 ng/dl)					0.98	0.96–0.99	0.008	0.96	0.94–0.98	0.001
Adiponectin (per 1 μg/dL)					0.90	0.85–0.94	<0.001	0.90	0.83–0.98	0.02

Model 1 includes pubertal status, obesity category, race and sex as independent variables. Model 2 includes the variables of model 1 with addition of ALT, CRP, leptin and adiponectin. Model 3 has the variables of model 2 yet is limited to only class III and class IV obese participants.

Limiting this analysis to participants older than 14 years (n=829) who are potentially bariatric surgery candidates, resulted in similar results. In model 1, the ORs for class III and class IV adolescents of CVRFC were 39 and 31 respectively (p<0.001 for both) and male sex had an OR of 1.96 compared with females (p<0.001). In model 2, similar ORs for both the extreme obesity categories were demonstrated along with a protective effect of leptin (OR=0.95, CI=0.92–0.97, p=0.001) and adiponectin (OR=0.87, CI=0.78–0.97, p=0.01). Upon limiting the analysis only to the class III and class IV participants (model 3), class IV participants were not at increased risk for CVRFC compared with class III participants while leptin and adiponectin maintained their protective effect.

Upon defining pubertal status as pre-, mid-, and post-pubertal, pubertal status remained a non-significant predictor of CVRFC while not affecting any of the other significant variables. Six hundred and twenty one participants were followed up conservatively for a mean of 34.3±23.7 months. We similarly divided them by baseline degree of obesity categories. Upon comparing the cross sectional and longitudinal subsample derived from it, there was a difference in the distribution of sex (pχ²=0.03 with a higher proportion of females in the longitudinal cohort, 65 vs. 63%), race (pχ²=0.02 with a higher proportion of Hispanics compared with Caucasians in the longitudinal cohort) and for baseline age (0.76 years younger in the longitudinal cohort, p <0.01). The follow up time amongst obesity categories increased with growing degrees of obesity (33±21 vs. 41± 27 months for overweight vs. class IV respectively, p=0.01).

Weight change (Figure 2, A) ranged from 10.79 kg in the overweight group to 28.95 kg in the class IV category (p<0.001). Importantly, weight change in the class IV subjects was significantly greater vs. all other obesity categories (p<0.001 for all). Similarly, BMI (Figure 2, B) increased to a greater extent with growing degrees of obesity with a more pronounced increase among the class IV participants compared with all other obesity categories (p=0.003 vs. class III, p<0.001 vs. all others). The percent change in BMI was 10.16±14.70, 7.21±14.45, 10.24±15.88, 9.31±16.23 and 13.26±16.31 percent for overweight, class I, II, III and IV baseline obesity categories respectively (p ANOVA=0.06). In post-hoc analysis, Class IV obesity had a significantly greater percent BMI change than class I (p=0.03). Follow up anthropometric indices tended to increase with growing degrees of baseline obesity (waist and hip circumference) as did systolic blood pressure and fasting triglycerides (p<0.001 and p=0.02 respectively). Follow-up CRP, leptin and insulin sensitivity showed trends similar to those demonstrated earlier for the entire cohort (greater follow up leptin and CRP and lower insulin sensitivity with greater degrees of baseline obesity, p<0.001 for all).

Figure 2. Change in BMI and weight by baseline obesity categories.

* p<0.001 for class IV obese vs. all other obesity categories. ** p=0.003 vs. class III obese, p<0.001 vs. all other obesity categories)

We performed logistic regression models for determining the predictors of the presence of CVRFC at follow up. In model 1, age, baseline obesity category, race, and sex were included as independent variables. The odds of CVRFC on follow up increased significantly by baseline obesity category from an OR of 2.64 (p= 0.06) in class I obese to 5.76 (p=0.001) for class IV obese compared with those that were overweight at baseline. Males had more than double the risk than females of CVRFC (OR=2.45, p<0.001). In this model, Hispanics had greater odds of CVRFC compared with Caucasians (OR=1.57, p=0.05). In model 2, we added presence of CVRFC at baseline, time of follow up and BMI percent change from baseline as independent variables to the model. Although of slightly lower magnitude, baseline class III and IV obesity remained significant predictors of similar magnitude for the presence of CV risk clustering on follow up (OR=4.87, p=0.01 and OR=4.60, p=0.02 respectively). As expected, the presence of CVRFC at baseline was a strong predictor of its presence on follow up (OR=5.10, p<0.001). Importantly, the percent change of BMI from baseline was a significant predictor of CVRFC on follow up (OR=1.04, CI =1.02–1.07, p<0.001). Limiting the analysis to those aged 14 years or older yielded different results as baseline obesity categories were no longer predictors of CVRFC on the follow up and only having baseline presence of CVRFC (OR=14.72 CI:4.36–49.64, p<0.001) and change in percent change in BMI from baseline (OR=1.11 CI 1.04–1.19, p=0.001) remained significant predictors of follow up CVRFC. Limiting the analysis to only class III and IV obese participants (model 3) demonstrated no impact of degree of obesity on the development of CVRFC over time while baseline CVRFC, male sex and percent change in BMI from baseline remained significant predictors.

Discussion

“Class IV obesity” represents a threshold for applying intensive surgical interventions in adolescence yet the definition of this threshold is unclear. In our cross sectional analysis we show that the prevalence of CVRFC in obese youth indeed rises with greater degrees of obesity yet seems to plateau above the threshold of class III obesity in most of the variables tested. In addition, male sex in obese youth is an independent risk factor for the presence of CVRFC and higher levels of specific adipocytokines have a small yet significant protective effect. Our longitudinal analysis shows that baseline class IV obesity confers a risk of similar magnitude to that of class III obesity for CVRFC on follow up. Moreover, a strong determinant of CVRFC developed over time is the percent change BMI change which was comparable across baseline obesity categories. These results indicate that the postulated linear dose-response of increasing degree of obesity on the presence of individual or clustered CV risk factors in childhood seems to plateau above the “class III obesity” threshold (corresponding roughly to an adult BMI>40 kg/m²).

Categorization of degrees of obesity according to BMI thresholds is a simple approach applicable in primary care and specialist clinics alike. The caveat of this approach is that fat distribution and metabolic factors such as hormone and cytokine concentrations, which significantly impact metabolic health, are neglected. It has been shown in adults and children alike that the determinants of metabolic health are not “how obese you are” rather “where the fat is stored”²⁶. This means that a normal weight individual can have an adverse pattern of lipid partitioning that confers a riskier metabolic profile than an obese individual with a favorable lipid partitioning pattern²⁷. A favorable lipid partitioning profile indicates that the majority of fat is deposited in the subcutaneous depot while the adverse profile indicates the deposition of fat in the intra-abdominal compartment ^{28, 29} and in insulin-responsive tissues such the liver and skeletal muscle^{30, 31}. We show that obesity driven metabolic derangements are already present in class III obese youth and suggest that waiting for such children to reach severe obesity thresholds in order to apply pharmacological or surgical interventions may not be justified.

Our results suggest that higher concentrations of specific adipocytokines per given degree of obesity confer protection from the presence of CVRFC. Greater concentrations of adiponectin are associated with insulin sensitivity as the hormone activates fat oxidation in the liver and skeletal muscle thus reducing concentration of lipids within these tissues^32–34. We have previously shown ³⁵ the inverse relation of adiponectin and markers of inflammation and the current analysis sheds further light on the potential protective effects of this adipocytokine. Leptin, on the other hand, increases in proportion to fat mass³⁶ and the finding described here (higher leptin associated with protection from CVRFC) may be explained by higher leptin concentrations being a surrogate of greater amounts of subcutaneous fat, indicating a favorable lipid partitioning profile ³⁷. Greater leptin and adiponectin per given BMI ^{38 39} may thus serve as potential markers for risk stratification in obese youth⁴⁰. Importantly, some metabolic variables tested seem to show a non-linear relation with the degree of obesity (greater insulin sensitivity and HDL-cholesterol in the overweight participants and greater CRP in class IV obesity compared with other obesity categories can be seen in Figure 2). This may imply that insulin sensitivity steeply declines in early stages of obesity in this age group nearly reaching a low plateau in severely obese youth. On the other hand, CRP rises steeply in class IV obesity, perhaps indicating that the inflammatory components of this unique phenotype are indeed related somewhat to absolute amounts of adipose tissue.

Male sex emerged in this analysis as a risk predictor for the presence of CVRFC. This finding is in line with previous observations indicating that male sex is a risk factor for hypertension⁴¹, dyslipidemia⁴² and non-alcoholic fatty liver disease^{42, 43} in childhood. Weight dynamics also emerged as significant predictors of the worsening of CVRFC. This observation is not surprising and has been previously described⁴⁴ yet the novelty of the present analysis is that the severely obese participants tended to gain more weight and increase their BMI even more than their less obese counterparts. Importantly, the vast majority of participants in this study did not achieve major degrees of change in their degree of obesity during the follow-up, emphasizing the limitations of standard-of-care conservative obesity management in this age group. African American participants were protected in some of the models from CVRFC compared with Caucasians, probably due to a more favorable lipid profile⁴⁵. Importantly, models for presence or development of CVRFC explained only a modest portion of the variability indicating that additional factors such as genetic background, physical fitness and dietary components should be taken into account with less emphasis on the degree of obesity.

Skinner et al have recently published an analysis highlighting the increased metabolic risk of rising degrees of obesity in childhood¹⁵. Importantly, in that analysis less than 5% of participants were within the class III obesity category, thus inference regarding subgroups within this category (class III vs. class IV obesity) was not performed. Our analysis focuses on the severe obesity categories and highlights the class IV group. As our population is derived from an obesity referral clinic (in contrast to a population based cohort), we were able to perform an in-depth analysis looking into sub-categories of severe obesity. Our sample focuses on the very high risk group of obese children – the obvious candidates for intensive interventions such as bariatric surgery. A limitation of the present analysis is the broad age groups of the participants. As puberty has a major yet transient impact on insulin sensitivity and thus potentially on the presence of dyslipidemia, hypertension and altered glucose metabolism ^{46, 47}, it would be useful in future studies to categorize participants in the longitudinal assessments by their pubertal dynamics rather than by the follow-up time. We tried to account for this factor by analyzing participants based on puberty categories (pre-pubertal and pubertal) and by repeating the models only in those older than 14 years of age which yielded similar results. Additionally, our reference group of overweight subjects may represent a greater risk profile compared with subjects of similar anthropometric measures derived from the population as they were referred to an obesity clinic, yet this may suggest that the high odds ratios we describe for the obese underestimate the actual risk. To overcome this concern, we tested our models in a subsample that included only obese subjects which did not impact our main findings. We did not have robust data regarding specific dietary habits and level of fitness of our study participants, both important determinants of CV risk at any age. This is an important caveat of this analysis that needs further investigation.

The current analysis emphasizes that categorization by BMI thresholds may be a simplistic and insufficient approach to identify metabolic risk, specifically in severely in obese children. Specifically, the notion that “the bigger you are - the worse you are metabolically” may be true for overweight, class I and II obese children where risk seems to increase linearly yet is less accurate in the class III and IV obese children who share CVRFC when they are heavier than the 140% of 95^th centile for their age and sex appropriate BMI, regardless of how much heavier they are. Future studies are warranted to test a metabolic risk stratification methodology of pediatric patient selection for conservative as well as surgical anti-obesity interventions longitudinally.

Table 3.

Anthropometric and metabolic characteristics of the longitudinal cohort by baseline obesity category.

		Baseline obesity category
		Overweight N(%)		Class I N(%)		Class II N(%)		Class III N (%)		Class IV N (%)		p-value
Sex	Male	27 (32.9)		37 (28.7)		70 (38.9)		35 (34.3)		55 (43.0)		0.15
	Female	55 (67.1)		92 (71.3)		110 (61.1)		67 (65.7)		73 (57.0)
Race	Caucasian	44 (57.3)		67 (51.9)		79 (43.9)		33 (32.4)		43 (33.6)		<0.01
	African-American	16 (19.5)		30 (23.3)		53 (29.4)		37 (36.3)		50 (39.1)
	Hispanic	22 (26.8)		32 (24.8)		48 (26.7)		32 (31.4)		35 (27.3)
Follow up time (months)		32.9 ±20.9		30.6 ±21.8		32.7 ±22.4		35.6 ±25.0		41.1 ± 26.8		0.01
		Mean ±SD		Mean ±SD		Mean ±SD		Mean ±SD		Mean ±SD
		Baseline	F/U	Baseline	F/U	Baseline	F/U	Baseline	F/U	Baseline	F/U	Baseline	F/U
Age (years)		13.0 ±2.7	15.8 ±2.7	13.1 ±2.7	16.6 ±2.7	12.5 ±2.8	15.3±2.5	12.4 ±2.7	15.4 ±2.4	11.4 ±2.9	14.9 ±3.1	<0.01	0.18
Weight (Kg)		56.7 ±14.1	70.2 ±16.7	72.1 ±17.3	82.4 ±16.5	82.2 ±20.7	98.5 ±19.3	94.7 ±22.9	111.1 ±20.4	105.2 ±28.5	134.2 ±31.2	<0.001	<0.001
BMI (kg/m2)		22.61 ±2.85	25.85 ±5.45	28.52 ±3.00	30.40 ±4.02	32.66 ±3.74	35.7 ±4.67	37.32 ±3.97	40.50 ±5.27	43.27 ±5.75	48.86 ±8.67	<0.001	<0.001
BMI-z score		1.08 ±0.65	1.01 ±0.87	1.97 ±0.15	1.85 ±0.38	2.36 ±0.11	2.30 ±0.31	2.61 ±0.14	2.52 ±0.28	2.84 ±0.15	2.81 ±0.33	<0.001	<0.001
Height (cm)		157 ±14	164 ±10	158 ±13	164 ±10	157 ±14	165 ±10	157 ±14	165 ±10	154 ±14	166 ±10	0.34	0.72
Waist circ. (cm)		79 ±10	88 ±13	93 ±11	99±12	102 ±11	110 ±13	111 ±13	117 ±12	120±14	131 ±18	<0.001	<0.001
DBP(mmHg)		66 ±9	68± 8	69 ±10	70 ±9	70 ±9	70 ±8	71 ±10	71 ±9	72 ±11	70 ±11	<0.01	0.33
SBP(mmHg)		115 ± 11	117±11	118 ±10	118±11	119 ±13	119±11	123 ±13	120±12	125 ±14	125±12	<0.01	<0.001
Cholesterol (mg/dL)		163±43	155 ±36	158±34	156 ±33	158±38	154 ±36	160±31	156 ±31	150±26	144 ±28	0.18	0.08
LDL(mg/dL)		90 ±35	88±31	91 ±29	89±26	94 ±33	91 ±31	95 ±26	91 ±31	89 ±23	85 ±22	0.40	0.45
Triglyceride (mg/dL)		115±102	87 ±47	125±97	119 ±107	108±61	100 ±57	119±75	114 ±82	106±67	94 ±48	0.30	0.02
HDL(mg/dL)		50 ±14	49 ±13	42 ±11	43 ±11	43 ±9	45 ±29	41 ±9	42 ±10	41 ±11	43 ±12	<0.01	0.17
Fasting glucose (mg/dL)		93 ±10	92 ±10	92 ±8	92 ±9	95 ±11	94 ±11	95 ±10	94 ±11	94 ±9	95 ±11	0.16	0.14
2 hours glucose (mg/dL)		127 ±27	120±28	127 ±28	122±29	128 ±32	123±29	133 ±32	128±31	130 ±33	128±30	0.48	0.19
Insulin sensitivity (WBISI)		2.54 ±1.55	2.78 ±1.80	2.03 ±1.24	2.23 ±1.56	1.70 ±1.05	1.81 ±1.20	1.36 ±0.77	1.56 ±1.03	1.34 ±0.71	1.60 ±1.01	<0.01	<0.001
CRP(mg/dL)		0.81 ±1.29	1.92 ±6.27	1.71 ±2.14	2.98 ±5.47	2.85 ±2.48	3.22 ±4.28	4.20 ±6.82	3.72 ±5.07	7.44 ±9.34	6.07 ±6.65	<0.01	0.04
Adiponectin (μg/dL)		8.85 ±4.68	9.09 ±4.26	8.29 ±4.03	8.78 ±5.05	8.26 ±4.55	8.41 ±4.77	6.88 ±2.89	7.62 ±2.98	7.43 ±3.91	8.11 ±4.95	0.03	0.55
Leptin(ng/dL)		15.58 ±10.27	18.86 ±12.51	21.50 ±9.50	24.15 ±15.71	29.19 ±11.98	32.24 ±15.20	33.35 ±13.11	36.71 ±15.76	40.44 ±12.68	45.71 ±15.66	<0.01	<0.001
ALT(u/L)		18 ±17	19 ±15	24 ±25	22 ±22	28 ±33	24 ±21	34 ±47	31 ±35	25 ±15	25 ±19	0.05	0.01

Table 4.

Predictors of the presence of CV risk clustering on follow up

		Model 1			Model 2			Model 3
Variable		OR	95% C.I.	p-value	OR	C.I.	p-value	OR	C.I.	p-value
Age		1.01	0.94–1.09	0.74	1.09	0.99–1.20	0.06	1.09	0.95–1.24	0.21
Baseline BMI Category	Overweight	1			1
	Class I	2.64	0.93–7.44	0.06	3.27	0.90–11.91	0.05
	Class II	3.10	1.14–8.44	0.02	2.82	0.79–10.03	0.07
	Class III	5.36	1.94–14.79	0.001	4.87	1.35–17.57	0.01	1
	Class IV	5.76	2.02–16.45	0.001	4.60	1.23–17.23	0.02	1.03	0.54–1.98	0.91
Race	Caucasian	1			1			1
	African-American	0.96	0.59–1.56	0.86	1.02	0.60–1.74	0.84	0.69	0.32–1.48	0.34
	Hispanic	1.57	1.01–2.49	0.05	1.70	1.02–2.84	0.03	1.38	0.63–2.97	0.42
Sex	Female	1			1
	Male	2.45	1.65–3.64	<0.001	2.47	1.56–3.71	<0.001	2.58	1.36–4.87	0.003
CVD risk clustering at baseline					5.10	3.25–8.01	<0.001	3.68	1.95–6.94	<0.001
Time of follow up					0.98	0.98–1.00	0.72	0.99	0.97–1.01	0.19
BMI percent change from baseline					1.04	1.02–1.07	<0.001	1.03	1.01–1.06	0.01
Adjusted R2		0.126			0.285			0.23

Model 1 has baseline age, degree of obesity, race and sex as independent variables. Model 2 uses the independent variables of model 1 with addition of baseline presence of CVD risk clustering, time of the follow up and percent change in BMI from baseline during the follow up. Model 3 uses the independent variables of model 2 yet is limited to class III and class IV obese participants.

List of abbreviations

CVRFC

cardio-vascular risk factor clustering

BMI

body mass index

OGTT

oral glucose tolerance test