OBJECTIVE

The carbohydrate-insulin model (CIM) claims that chronic exposure to hyperinsulinemia induced by dietary carbohydrates explains development of obesity via direct effects of insulin and/or low postprandial metabolic fuel levels. We aimed at testing whether indices of hyperinsulinemia and postprandial glucose levels can predict increases in the degree of obesity over time.

RESEARCH DESIGN AND METHODS

Children and adolescents with obesity attending a pediatric obesity clinic performed oral glucose tolerance tests (OGTTs) and received standard obesity management. Indices of hyperinsulinemia and insulin secretion were derived from the OGTT and evaluated in the face of changes in the degree of obesity over time.

RESULTS

A total of 591 children (217 males and 374 females) participated, and the mean follow-up was 1.86 ± 1.29 years. OGTT-derived area under the curve of insulin, peak insulin, fasting insulin, the insulinogenic index, or insulin at 30 min were not associated with greater changes in the degree of obesity in univariate or multivariate analyses (adjusted for baseline age, BMI z score, sex, and ethnicity). Low postprandial glucose <75 mg/dL was not associated with greater changes in the degree of obesity in univariate or multivariate analyses. In a subsample of 104 participants with a follow-up >4 years, none of these parameters was associated with greater increases in the degree of obesity.

CONCLUSIONS

In children and adolescents with obesity, exposure to hyperinsulinemia, greater insulin secretion, or low postprandial glucose is not associated with greater increases in the degree of obesity over 2–4 years. The CIM should be evaluated in children with lower BMI and for longer follow-up periods.

Obesity among children and adults is a major public health concern and is considered an epidemic of the 21st century. The genetic causes of obesity have been studied in large populations and provide ∼20% of the variation in BMI among adults (1,2). As the obesity epidemic emerged only in recent decades, it is obvious that a strong environmental component is involved. The main model explaining weight gain is based on energy balance disequilibrium in which increased caloric intake and reduced energy expenditure result in a positive energy balance, leading to adipose tissue deposition (3). An alternative model is the carbohydrate-insulin model (CIM), which posits that “not all calories are equal,” such that consuming food with a greater glycemic load (calculated as the product of carbohydrates and their glycemic index) results in a hormonal response governed by insulin that promotes increased adipose tissue deposition, which drives a positive energy balance (4). Specifically, following a high-glycemic-load meal, an anabolic state is induced by hyperinsulinemia that suppresses glucagon and by a high-glucose–dependent insulinotropic polypeptide, all of which promote uptake of nutrients to liver, muscle, and adipose tissue (5,6). The transition from anabolism to catabolism in the postprandial period does not coincide with nutrient delivery from the gut, which is mostly completed at 3 h following the meal (7) while the hormonal anabolic profile is still dominant. This may result in low nutrient concentrations in the blood (specifically of glucose) (8), perceived by the brain as energy deprivation of peripheral tissues and leading to food cravings, specifically of carbohydrates (9), in order to relieve this metabolic stress. This creates a vicious cycle of repeated carbohydrate consumption, leading to a positive energy balance and weight gain.

In order to test whether the CIM model can be used to explain obesity dynamics in children and adolescents, we tested whether exposure to hyperinsulinemia, greater insulin secretion, or a low glucose level 3 h following an oral glucose load can predict changes in the degree of obesity over time. Based on the CIM model, those with greater insulin secretion or those who experience low glucose levels following a high glycemic load will increase their degree of obesity over time more than those who do not manifest these metabolic responses.

Children and adolescents with obesity referred to the Yale Childhood Obesity Clinic between 1998 and 2016 participated in a study aimed at defining “metabolic markers and predictors of childhood obesity” and were included in the current analysis (ClinicalTrials.gov NCT01966627). Age at recruitment was between 6 and 19 years. All participants received standard care of clinic-based behavioral modification therapy aimed at the child and the family. Participants visited the clinic biannually and received diet and exercise guidance along with brief psychosocial counseling. Subjects who were >7 years of age, performed more than one oral glucose tolerance test (OGTT) as previously described (10), and had all available relevant data were included in the current analysis. Some of the participants of this analysis have been described in previous publications (1113). In addition to subject assent and parental consent, thorough physical examinations were obtained from each participant. The Human Investigation Committee at Yale University School of Medicine approved the study.

Weight and height were measured on the morning of the OGTT, and BMI was calculated. Percent body fat was determined by a body fat analyzer (TBF-300; Tanita Corp of America, Inc, Arlington Heights, IL). Baseline fasting blood samples were obtained for measurements of glucose along with additional biochemical tests not described in this analysis. All participants performed a standard OGTT as previously described (14).

Plasma glucose was determined with a YSI 2700 Analyzer (Yellow Springs Instruments, Yellow Springs, OH). Plasma insulin levels were measured using a double-antibody radioimmunoassay from Millipore (Burlington, MA).

Calculations

To measure the degree of obesity, an age- and sex-dependent BMI SD score (BMI z score) was calculated (15). The primary outcome of the analysis was BMI z-score change between the two OGTTs. We chose three different parameters to test the impact of hyperinsulinemia on BMI z-score change:

  1. The area under the curve of insulin during the 180-min OGTT (AUCinsulin) was used to reflect postprandial hyperinsulinemia. The AUC was calculated using the trapezoidal rule.

  2. Fasting hyperinsulinemia was used to reflect exposure to elevated insulin concentrations in between meals and during the nighttime.

  3. Peak insulin was used to reflect what may be described as “insulin hypersecretion” following the oral glucose challenge that may lead to exposure of tissues, such as the brain, to hyperinsulinemia for a brief period yet is still able to induce a hormonal and autonomic response (16).

We further tested two surrogates of acute insulin secretion:

  1. Insulin at 30 min of the OGTT: this parameter has been used as a surrogate of insulin secretion and shown to be associated with increased weight gain among adults consuming high-glycemic-load diets (17).

  2. The insulinogenic index, calculated as (Insulin30min − Insulin0min)/(Glucose30min − Glucose0min), which has been shown to correlate with the acute insulin response (18).

We further divided the cohort into those who had a 3-h glucose <75 mg/dL compared with those who did not meet this criterion. The concentration of 75 mg/dL was chosen as a glucose level at which the hormonal counterregulatory response is already detectable (19).

Statistical Analysis

Parameters are presented as means ± SD. Nonnormally distributed variables were log transformed for the sake of analysis. We performed univariate analyses for determinants of the primary outcome of the analysis (BMI z-score change between the two OGTTs). We performed linear regression analyses using the primary outcome as a dependent variable and incorporating baseline parameters (age, BMI z score, and indices of hyperinsulinemia), sex, racial background, and time between the studies as independent variables. The analysis was performed using SPSS 24.0 statistical software.

Description of the Multiethnic Cohort of Children and Adolescents With Obesity

A total of 591 children and adolescents (217 males and 374 females) with obesity were part of this analysis (Table 1). The mean age was 12.33 ± 2.80 years, and all were by definition considered obese at baseline (mean BMI z score 2.41 ± 0.33 [range 1.65–3.28]). A total of 407 had normal glucose tolerance, while 184 had prediabetes (impaired glucose tolerance, impaired fasting glucose, or both). The mean time between the two oral glucose tolerance tests was 1.86 ± 1.29 years, and the mean BMI z change was −0.05 ± 0.27 (range −1.6 to 0.63).

Table 1

Baseline and follow-up characteristics of study participants

BaselineFollow-up
Mean ± SDRangeMean ± SDRange
Sex (male/female) 217/374    
Ethnicity (Caucasian/African American/Hispanic) 240/178/173    
Age (years) 12.33 ± 2.80 8–19 14.65 ± 2.57 9.30 ± 21.9 
Height (cm) 156.41 ± 13.58 114–195 158.62 ± 11.74 125–195 
Weight (kg) 86.26 ± 25.30 49.95–199.20 100.03 ± 26.30 53,20–211.40 
BMI (kg/m234.58 ± 6.63 23.41–62.49 36.90 ± 7.78 23.60–72.25 
BMI z score 2.41 ± 0.33 1.66–3.28 2.32 ± 0.43 1.53–3.57 
Fasting glucose (mg/dL) 92 ± 8 61–120 93 ± 9 58–132 
2-h glucose (mg/dL) 126 ± 24 68–198 125 ± 27 58–240 
3-h glucose (mg/dL) 97 ± 25 42–195 95 ± 27 43–232 
AUCinsulin (μU/mL/min) 200 ± 156 38–1,556 184 ± 142 32–1,563 
Fasting insulin (μU/mL) 36 ± 27 3–130 37 ± 22 4–144 
Peak insulin (μU/mL) 339 ± 279 31–2,775 316 ± 256 28–3,075 
BaselineFollow-up
Mean ± SDRangeMean ± SDRange
Sex (male/female) 217/374    
Ethnicity (Caucasian/African American/Hispanic) 240/178/173    
Age (years) 12.33 ± 2.80 8–19 14.65 ± 2.57 9.30 ± 21.9 
Height (cm) 156.41 ± 13.58 114–195 158.62 ± 11.74 125–195 
Weight (kg) 86.26 ± 25.30 49.95–199.20 100.03 ± 26.30 53,20–211.40 
BMI (kg/m234.58 ± 6.63 23.41–62.49 36.90 ± 7.78 23.60–72.25 
BMI z score 2.41 ± 0.33 1.66–3.28 2.32 ± 0.43 1.53–3.57 
Fasting glucose (mg/dL) 92 ± 8 61–120 93 ± 9 58–132 
2-h glucose (mg/dL) 126 ± 24 68–198 125 ± 27 58–240 
3-h glucose (mg/dL) 97 ± 25 42–195 95 ± 27 43–232 
AUCinsulin (μU/mL/min) 200 ± 156 38–1,556 184 ± 142 32–1,563 
Fasting insulin (μU/mL) 36 ± 27 3–130 37 ± 22 4–144 
Peak insulin (μU/mL) 339 ± 279 31–2,775 316 ± 256 28–3,075 

The table includes baseline anthropometrics as well as the parameters of interest for the analysis.

The majority of participants were pubertal at baseline (Table 2). Of note, prepubertal participants were more obese (P < 0.001) and more insulin-sensitive (P = 0.001) and thus had a lower acute insulin response (P = 0.01). Changes in the degree of obesity over time were comparable between prepubertal and pubertal participants. Most participants had normal glucose tolerance at baseline. While those with normal versus impaired glucose tolerance were similarly obese, those with impaired glucose tolerance had, as expected, lower insulin sensitivity (P < 0.001) and greater indices of insulinemia and insulin secretion (AUCinsulin, fasting insulin, peak insulin, and insulinogenic index, P < 0.001 for all). Of note, those with impaired glucose tolerance reduced their degree of obesity more than those with normal glucose tolerance (P = 0.01; data not shown).

Table 2

Study participants by pubertal status and glucose tolerance at baseline

BMI zAUC insulinFasting insulinPeak insulinInsulin sensitivityInsulinogenic indexΔBMI z score
Pubertal status        
 Prepubertal (n = 128) 2.54 ± 0.36 184 ± 139 29 ± 18 326 ± 226 1.94 ± 1.11 4.32 ± 3.22 −0.09 ± 0.31 
 Pubertal (n = 463) 2.37 ± 0.31 205 ± 161 39 ± 22 341 ± 261 1.55 ± 0.88 5.36 ± 5.54 −0.04 ± 0.26 
P value (t test) <0.001 0.14 <0.001 0.52 0.001 0.01 0.17 
Glucose tolerance        
 Normal (n = 407) 2.40 ± 0.32 170 ± 119 34 ± 20 302 ± 201 1.85 ± 1.00 5.53 ± 5.72 −0.03 ± 0.25 
 Impaired (n = 184) 2.42 ± 0.34 268 ± 202 43 ± 23 415 ± 330 1.16 ± 0.60 4.11 ± 3.22 −0.10 ± 0.30 
P value (t test) 0.66 <0.001 <0.001 <0.001 <0.001 <0.001 0.01 
BMI zAUC insulinFasting insulinPeak insulinInsulin sensitivityInsulinogenic indexΔBMI z score
Pubertal status        
 Prepubertal (n = 128) 2.54 ± 0.36 184 ± 139 29 ± 18 326 ± 226 1.94 ± 1.11 4.32 ± 3.22 −0.09 ± 0.31 
 Pubertal (n = 463) 2.37 ± 0.31 205 ± 161 39 ± 22 341 ± 261 1.55 ± 0.88 5.36 ± 5.54 −0.04 ± 0.26 
P value (t test) <0.001 0.14 <0.001 0.52 0.001 0.01 0.17 
Glucose tolerance        
 Normal (n = 407) 2.40 ± 0.32 170 ± 119 34 ± 20 302 ± 201 1.85 ± 1.00 5.53 ± 5.72 −0.03 ± 0.25 
 Impaired (n = 184) 2.42 ± 0.34 268 ± 202 43 ± 23 415 ± 330 1.16 ± 0.60 4.11 ± 3.22 −0.10 ± 0.30 
P value (t test) 0.66 <0.001 <0.001 <0.001 <0.001 <0.001 0.01 

Data are presented as means ± SD.

Fasting insulin and peak insulin expressed in microunits per milliliter; AUC insulin expressed in microunits per milliliter per minute; insulin sensitivity (Matsuda index) expressed as units; and insulinogenic index expressed in microunits per milliliter/micrograms per deciliter. Boldface P values indicate statistical significance.

Baseline Postprandial Hyperinsulinemia (AUCinsulin) and Change in the Degree of Obesity

The correlation coefficient of AUCinsulin and change in BMI z score over time was negative and significant (r = −0.08; P = 0.04) (Fig. 1A). We divided the cohort into tertiles of AUC of insulin during the glucose tolerance test as a measure of postprandial hyperinsulinemia. Of note, the BMI z change was comparable across the AUC insulin tertiles (P ANOVA = 0.10) (Fig. 2A). We used the BMI z change as a dependent parameter in a linear regression model and introduced sex, ethnicity, time between the OGTTs, baseline age, baseline BMI z score, and AUCinsulin tertiles as independent variables. The model was significant with a modest R2 of 0.036, with male sex being the only significant predictor of BMI z change (β = 0.067; P = 0.003), while AUCinsulin tertile was not significantly associated with the outcome (P = 0.22) (Fig. 2B), including post hoc comparisons between tertiles (tertile 1 vs. tertile 3, P = 0.46). Introducing AUCinsulin as a continuous variable and not as tertiles into this model led to similar results (P = 0.29).

Figure 1

Correlations of BMI z-score changes and AUCinsulin (A), peak insulin (B), fasting insulin (C), and Glucose180min (D) of the OGTT.

Figure 1

Correlations of BMI z-score changes and AUCinsulin (A), peak insulin (B), fasting insulin (C), and Glucose180min (D) of the OGTT.

Close modal
Figure 2

BMI z-score changes by AUCinsulin (A and B), fasting insulin (C and D), peak insulin (E and F), and Glucose180min (G and H) of the OGTT. Simple comparisons are from univariate comparisons. Values presented as means ± SEs. Adjusted values are derived from the linear regression models. The values for AUCinsulin tertiles are: tertile 1 <156; 156 ≤ tertile 2 ≤ 206; tertile 3 >206 (AUCinsulin expressed in microunits per milliliter per minute). The values for fasting insulin tertiles are: tertile 1 <26; 26 ≤ tertile 2 ≤ 39; tertile 3 >39 (fasting insulin expressed in microunits per milliliter). The values for peak insulin tertiles are: tertile 1 <203; 203 ≤ tertile 2 ≤ 358; tertile 3 >358 (peak insulin expressed in microunits per milliliter). Glucose180min categories compared were <75 or >75 mg/dL.

Figure 2

BMI z-score changes by AUCinsulin (A and B), fasting insulin (C and D), peak insulin (E and F), and Glucose180min (G and H) of the OGTT. Simple comparisons are from univariate comparisons. Values presented as means ± SEs. Adjusted values are derived from the linear regression models. The values for AUCinsulin tertiles are: tertile 1 <156; 156 ≤ tertile 2 ≤ 206; tertile 3 >206 (AUCinsulin expressed in microunits per milliliter per minute). The values for fasting insulin tertiles are: tertile 1 <26; 26 ≤ tertile 2 ≤ 39; tertile 3 >39 (fasting insulin expressed in microunits per milliliter). The values for peak insulin tertiles are: tertile 1 <203; 203 ≤ tertile 2 ≤ 358; tertile 3 >358 (peak insulin expressed in microunits per milliliter). Glucose180min categories compared were <75 or >75 mg/dL.

Close modal

Baseline Fasting Insulin and Change in the Degree of Obesity

The correlation coefficient of fasting insulin and change in BMI z score over time was negative and not significant (r = −0.04; P = 0.32) (Fig. 1B). We divided the cohort into tertiles of fasting insulin as a measure of hyperinsulinemia in between meals. The BMI z change was comparable between the fasting insulin tertiles (P ANOVA = 0.47) (Fig. 2C). We used the BMI z change as a dependent parameter in a linear regression model and introduced sex, ethnicity, time between the OGTTs, baseline age, baseline BMI z score, and fasting insulin tertiles as independent variables. Similar to the AUCinsulin analysis, the model was significant, with a modest R2 of 0.040 with male sex being the only significant predictor of BMI z change (β = 0.069; P = 0.003), while fasting insulin tertiles were not significantly associated with the outcome (P = 0.11) (Fig. 2D), including post hoc comparisons between tertiles (tertile 1 vs. tertile 3, P = 0.18). Introducing fasting insulin as a continuous variable and not as tertiles into this model led to similar results (P = 0.54).

Baseline Peak Insulin and Change in the Degree of Obesity

The correlation coefficient of peak insulin and change in BMI z score over time was negative and not significant (r = −0.05; P = 0.22) (Fig. 1C). We divided the cohort into tertiles of peak insulin during the OGTT as an additional measure of postprandial hyperinsulinemia. The BMI z-score change was comparable across the peak insulin tertiles (P ANOVA = 0.14) (Fig. 2E). We used the BMI z-score change as a dependent parameter in a linear regression model and introduced sex, ethnicity, time between the OGTTs, baseline age, baseline BMI z score, and peak insulin tertiles as independent variables. The model was significant, with a modest R2 of 0.036 with male sex being the only significant predictor of BMI z change (β = 0.073; P = 0.003), while peak insulin tertile was not significantly associated with the outcome (P = 0.56) (Fig. 2F), including post hoc comparisons between tertiles (tertile 1 vs. tertile 3, P = 0.47). Introducing peak insulin as a continuous variable and not as tertiles into this model led to similar results (P = 0.70).

Baseline Indices of Insulin Secretion and Change in the Degree of Obesity

The correlation coefficient of Insulin30min and change in BMI z score over time was negative and not significant (r = −0.01; P = 0.70). We divided the cohort into tertiles of Insulin30min during the OGTT as a measure of acute insulin secretion. The BMI z-score change was comparable across the Insulin30min tertiles (P ANOVA = 0.91). We used the BMI z-score change as a dependent parameter in a linear regression model and introduced sex, ethnicity, time between the OGTTs, baseline age, baseline BMI z score, and Insulin30min tertile as independent variables. The model was significant, with a modest R2 of 0.028 with male sex again being the only significant predictor of BMI z change (β = 0.074; P = 0.001) while Insulin30min tertile was not significantly associated with the outcome (P = 0.74), including post hoc comparisons between tertiles (tertile 1 vs. tertile 3, P = 1.0). Introducing Insulin30min as a continuous variable and not as tertiles into this model led to similar nonsignificant results (P = 0.36).

Similarly, the correlation coefficient of the insulinogenic index with BMI z-score change was negative and nonsignificant (r = −0.03; P = 0.47). Upon dividing the cohort into tertiles of the insulinogenic index, the BMI z-score change was comparable across the insulinogenic index tertiles (P ANOVA = 0.72). We used the BMI z-score change as a dependent parameter in a linear regression model and introduced sex, ethnicity, time between the OGTTs, baseline age, baseline BMI z score, and insulinogenic index tertile as independent variables. The model was significant, with a modest R2 of 0.030 with male sex emerging again as the only significant predictor of BMI z change (β = 0.074; P = 0.002), while insulinogenic index tertile was not significantly associated with the outcome (P = 0.78), including post hoc comparisons between tertiles (tertile 1 vs. tertile 3, P = 0.99). Introducing the insulinogenic index as a continuous variable and not as tertiles into this model led to similar nonsignificant results (P = 0.55).

Low Baseline Post-OGTT Glucose and Changes in the Degree of Obesity

The correlation coefficient of glucose at 180 min of the OGTT (Glucose180min) and change in BMI z score over time was negative and not significant (r = −0.06; P = 0.14) (Fig. 1D). We divided the cohort into those that had a 180-min glucose <75 mg/dL (n = 123) and those who did not (n = 468). The BMI z change was comparable between the two groups (P = 0.48) (Fig. 2G). We used the BMI z change as a dependent parameter in a linear regression model and introduced sex, ethnicity, time between the OGTTs, baseline age, baseline BMI z score, and the two 180-min glucose categories as independent variables. Similar to the previous analyses, the model was significant, with a modest R2 of 0.031 with male sex being the only significant predictor of BMI z change (β = 0.072; P = 0.002), while having a 180-min glucose <75 mg/dL on the OGTT was not significantly associated with the outcome (P = 0.76) (Fig. 2H).

Using a stricter definition of Glucose180min <65 mg/dL (n = 44; 23 males and 21 females) yielded similar nonsignificant results in the univariate and multivariate analyses. Upon comparing those whose Glucose180min was lower than fasting glucose to those whose Glucose180min did not return to fasting levels, no difference was demonstrated in BMI z-score change over time in univariate or multivariate analyses (data not shown for both).

Impact of AUCinsulin, Fasting Insulin, Peak Insulin, and Low Glucose180min on BMI z Change Over a Longer Follow-up

A subsample of 104 participants who repeated annual OGTTs (38 males and 66 females; 36 Caucasians, 25 African American, and 43 Hispanic) had a mean follow-up of 4.25 ± 1.81 years. We repeated the analyses of AUCinsulin tertiles, fasting insulin tertiles, peak insulin, Insulin30min, the insulinogenic index, and subjects with Glucose180min <75 mg/dL as described above using the same regression models. None of these six independent parameters emerged as a predictor of BMI z change over the longer follow-up period in univariate or multivariate analyses (data not shown). As in the whole cohort, male sex was a significant predictor of larger BMI z-score change in all models.

Changes in Percent Body Fat and Indices of Insulin Secretion, Hyperinsulinemia, and Post-OGTT Glucose

As an alternative to changes in BMI z score, we evaluated the same parameters used earlier as predictors of changes in percent body fat using the same analytical strategy. We had complete data for 479 participants (167 male and 312 female). We introduced percent body fat change as a dependent variable in a linear regression model with baseline age, percent body fat, sex, race, time between studies, and the potential predictors of interest individually (AUCinsulin, fasting insulin, and peak insulin as surrogates of exposure to hyperinsulinemia, Insulin30min and the insulinogenic index as surrogates of insulin secretion, and Glucose180min as a surrogate of postprandial glycemia) as independent variables. None of the parameters of interest emerged as significant predictors of changes in percent body fat in these models (P = 0.49, P = 0.91, P = 0.56, P = 0.75, P = 0.68, and P = 16 for AUCinsulin, fasting insulin, peak insulin, Insulin30min, the insulinogenic index, and Glucose180min, respectively).

There is an active debate in the literature regarding the environmental origins of the obesity epidemic, with the CIM trying to challenge the traditional energy balance model (4,2022). Both models are difficult to test empirically in controlled conditions in humans over long periods; thus, the majority of the current evidence relies on short interventions, mostly in rodents. In this analysis, we used a multiethnic cohort of children and adolescents with obesity and tested whether baseline exposure to fasting or postprandial hyperinsulinemia, greater insulin secretion, or low post-OGTT glucose levels predicts greater changes in the degree of obesity over time, as predicted by the CIM. We found no significant effects of baseline and post-OGTT–derived hyperinsulinemia or greater insulin secretion or late postprandial relative hypoglycemia on obesity dynamics over ∼2 years in the entire cohort and over ∼4 years in a smaller subsample.

Management of obesity, specifically in children and adolescents, is challenging, and the long-term results of most lifestyle modification–oriented interventions is modest at best. The standard strategy has traditionally been aiming at reduction of caloric intake along with increasing energy expenditure via voluntary physical activity. This approach uses reduction of intake using a “balanced diet” without specific emphasis on specific components of the diet and their metabolic and hormonal effects and considers all calories equal. Importantly, based on the energy balance model, obesity leads to insulin resistance driving compensatory hyperinsulinemia. The CIM is based on the unique metabolic effects of carbohydrates via insulin and other hormones, compared with other dietary elements, on lipid deposition and energy metabolism and aims at reducing carbohydrate intake in order minimize the anabolic and behavioral effects it induces. The CIM claims that hyperinsulinemia induced by carbohydrate intake comes first and drives an anabolic milieu conducive to further caloric intake, thus playing a causal role in the development of weight gain and related insulin resistance (23). Based on the CIM theory, chronic exposure to hyperinsulinemia should drive greater weight gain over time. There are some observations in the literature that can provide support to this claim: Sigal et al. (24) have shown that the acute insulin response is associated with greater weight gain over time in adults, specifically in those who are insulin-sensitive. Several groups have shown in prepubertal children at risk for development of obesity that fasting insulin correlated with the rate of weight gain in both girls and boys (25,26). Using a surrogate derived from the OGTT (Insulin30min), mixed results have been shown regarding the association of insulin secretion and weight gain in adults (17,27). By using Mendelian randomization analyses, higher genetically determined Insulin30min was strongly associated with higher BMI, consistent with a causal role in development of obesity (28). Moreover, insulin secretion magnitude has been shown to modify the weight loss response to specific dietary interventions in adults (29). Complementing these observations, preclinical studies suggest that a mild reduction of hyperinsulinemia may prevent or reduce obesity (30,31). Our analysis provides three different measures of exposure to hyperinsulinemia (fasting insulin, postprandial [AUCinsulin], and peak insulin) and two measures of insulin secretion (Insulin30min and the insulinogenic index) yet, in contrast to some of the above-mentioned publications, failed to demonstrate any relation of these parameters with the change in the degree of obesity. Of note, the ranges of AUCinsulin, peak insulin, fasting insulin, Insulin30min, and the insulinogenic index in this analysis are very broad and vary from very low values to those at least one to two orders of magnitude greater (Table 1). Despite the wide range in insulin from low to very high insulin levels, the BMI z-score changes were comparable across insulin tertiles. Strengthening these observations, we repeated our analyses with an index of body composition (change of percent body fat) and showed the exact same results as those shown for BMI z score.

Importantly, unlike the above-cited articles, our study participants were already with obesity at baseline, and the question remains whether they could have been identified based on their insulin profiles prior to becoming obese. Based on the findings described in this study, the exposure to hyperinsulinemia or presence of increased insulin secretion does not predict greater degree of obesity development in those who already have high BMI z scores at baseline.

Postprandial hyperinsulinemia may result in a hormonal imbalance, leading to circulating low nutrient concentrations (glucose and free fatty acids) 3–4 h following the consumption of the high glycemic load (8). This may be perceived by specific brain nuclei as a signal of low energy status in critical peripheral tissues such as the liver and lead to a counterregulatory response aimed at increasing peripheral availability of energy sources (32). It is well established that decreasing glucose concentrations induces initiation of feeding, which can be prevented by glucose infusion (33,34). Low peripheral glucose levels, even slightly above the hypoglycemic definition, may also induce a hunger response that favors food-seeking behavior (35). The seemingly low peripheral energy availability may result in reduced energy expenditure and improved mechanical muscular efficiency (36,37). Taken together, a repeated exaggerated response to a high glycemic load should induce a metabolic milieu that theoretically favors weight gain over time by way of hormonal, autonomic, and behavioral responses. In our analysis, we chose to compare those with a relatively low (<75 mg/dL) postprandial glucose level to other participants and found no such effect in regards to changes in the degree of obesity (with a similar finding in those with postprandial glucose <65 mg/dL). It is unknown whether the OGTT response, in regards to low postprandial glucose, is reproducible in repeated studies (the correlation of glucose at 180 min between baseline and follow-up was r = 0.38; P < 0.001), so that perhaps concentrations of other metabolic fuels such as free fatty acids, not assessed in this analysis, may better predict changes in the degree of obesity.

The analysis presented in this article was aimed at testing specific predictions of the CIM model in regards to weight gain in children with obesity and failed to prove them. This may be due to several reasons: first, our study participants are already youth with obesity, and their metabolic responses may differ from their counterparts who have not acquired a significant amount of weight. Second, we did not test concentrations of other nutrients and hormones that have a unique dynamic profile following a high carbohydrate load, such as free fatty acids and gut hormones, both of which can perhaps serve as stronger predictors (compared with glucose and insulin) of weight gain. Third, we did not accurately assess the dietary patterns of our participants. One can confidently assume that the vast majority of them consume a typical Western diet rich in carbohydrates and ultra-processed food, and this reduces the possibility of an effect modification of the diet on BMI dynamics. Fourth, the follow-up of ∼2 years for the majority of participants and of ∼4 years for the subgroup may be too short for evaluation of the long-term impact of high insulin secretion and postprandial hyperinsulinemia. In contrast, short-term results of comparing a high- versus a low-carbohydrate diet in controlled conditions for a short period of time do not support the CIM and are in agreement with our findings (38).

We conclude that in children and adolescents with obesity, fasting as well as postprandial hyperinsulinemia, greater insulin secretion, or relatively low postprandial glucose levels are not associated with increasing degrees of obesity over a follow-up of 2–4 years. Additional studies performed in children and adolescents who are leaner at baseline and have longer follow-ups are needed to further investigate the relevance of the CIM model in the pediatric population.

Clinical trial reg. nos. NCT00000112 and NCT01967849, clinicaltrials.gov

See accompanying article, p. 1303.

Acknowledgments. This article is in loving memory of Bridget Pierpont, the research coordinator of the Yale Pathophysiology of Type 2 Diabetes in Youth Study cohort, who died unexpectedly during the preparation of this manuscript. The authors thank the participants and families for their willingness and cooperation.

Funding. This study was supported by the National Institutes of Health, Eunice Kennedy Shriver National Institute of Child Health and Human Development (grants R01-HD-40787, R01DK111038, R01-HD-28016, and K24-HD-01464 to S.C.), the National Center for Research Resources (Clinical and Translational Science Award UL1-RR-0249139 to S.C.), the American Diabetes Association (Distinguished Clinical Scientist Award to S.C.), and the National Institute of Diabetes and Digestive and Kidney Diseases (grants R01-DK-111038 to S.C. and R01-DK-114504-01A1).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. R.H. and R.W. performed the analysis and wrote the manuscript. A.G. took part in the clinical care and added to the discussion. S.C. and R.W. designed the study, took part in the clinical care and the analysis, and added to the discussion. All authors read, edited, and approved the manuscript. R.W. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

1.
Robinson
MR
,
Hemani
G
,
Medina-Gomez
C
, et al
.
Population genetic differentiation of height and body mass index across Europe
.
Nat Genet
2015
;
47
:
1357
1362
2.
Locke
AE
,
Kahali
B
,
Berndt
SI
, et al.;
LifeLines Cohort Study
;
ADIPOGen Consortium
;
AGEN-BMI Working Group
;
CARDIOGRAMplusC4D Consortium
;
CKDGen Consortium
;
GLGC
;
ICBP
;
MAGIC Investigators
;
MuTHER Consortium
;
MIGen Consortium
;
PAGE Consortium
;
ReproGen Consortium
;
GENIE Consortium
;
International Endogene Consortium
.
Genetic studies of body mass index yield new insights for obesity biology
.
Nature
2015
;
518
:
197
206
3.
Schwartz
MW
,
Seeley
RJ
,
Zeltser
LM
, et al
.
Obesity pathogenesis: an Endocrine Society scientific statement
.
Endocr Rev
2017
;
38
:
267
296
4.
Ludwig
DS
,
Aronne
LJ
,
Astrup
A
, et al
.
The carbohydrate-insulin model: a physiological perspective on the obesity pandemic
.
Am J Clin Nutr
2021
;
114
:
1873
1885
5.
Pfeiffer
AFH
,
Keyhani-Nejad
F
.
High glycemic index metabolic damage - a pivotal role of GIP and GLP-1
.
Trends Endocrinol Metab
2018
;
29
:
289
299
6.
Wolever
TM
.
Dietary carbohydrates and insulin action in humans
.
Br J Nutr
2000
;
83
(
Suppl. 1
):
S97
S102
7.
Ferrannini
E
,
Bjorkman
O
,
Reichard
GA
Jr
, et al
.
The disposal of an oral glucose load in healthy subjects. A quantitative study
.
Diabetes
1985
;
34
:
580
588
8.
Shimy
KJ
,
Feldman
HA
,
Klein
GL
,
Bielak
L
,
Ebbeling
CB
,
Ludwig
DS
.
Effects of dietary carbohydrate content on circulating metabolic fuel availability in the postprandial state
.
J Endocr Soc
2020
;
4
:
bvaa062
9.
Strachan
MW
,
Ewing
FM
,
Frier
BM
,
Harper
A
,
Deary
IJ
.
Food cravings during acute hypoglycaemia in adults with type 1 diabetes
.
Physiol Behav
2004
;
80
:
675
682
10.
Sinha
R
,
Fisch
G
,
Teague
B
, et al
.
Prevalence of impaired glucose tolerance among children and adolescents with marked obesity [published correction appears in N Engl J Med 2002;346:1756]
.
N Engl J Med
2002
;
346
:
802
810
11.
Hagman
E
,
Hecht
L
,
Marko
L
, et al
.
Predictors of responses to clinic-based childhood obesity care
.
Pediatr Diabetes
2018
;
19
:
1351
1356
12.
Weiss
R
,
Taksali
SE
,
Tamborlane
WV
,
Burgert
TS
,
Savoye
M
,
Caprio
S
.
Predictors of changes in glucose tolerance status in obese youth
.
Diabetes Care
2005
;
28
:
902
909
13.
Weiss
R
,
Cali
AM
,
Dziura
J
,
Burgert
TS
,
Tamborlane
WV
,
Caprio
S
.
Degree of obesity and glucose allostasis are major effectors of glucose tolerance dynamics in obese youth
.
Diabetes Care
2007
;
30
:
1845
1850
14.
Weiss
R
,
Dziura
J
,
Burgert
TS
, et al
.
Obesity and the metabolic syndrome in children and adolescents
.
N Engl J Med
2004
;
350
:
2362
2374
15.
Kuczmarski
RJ
,
Ogden
CL
,
Guo
SS
, et al
.
2000 CDC Growth Charts for the United States: methods and development
.
Vital Health Stat 11
2002
;
246
:
1
190
16.
Ludwig
DS
.
The glycemic index: physiological mechanisms relating to obesity, diabetes, and cardiovascular disease
.
JAMA
2002
;
287
:
2414
2423
17.
Chaput
JP
,
Tremblay
A
,
Rimm
EB
,
Bouchard
C
,
Ludwig
DS
.
A novel interaction between dietary composition and insulin secretion: effects on weight gain in the Quebec Family Study
.
Am J Clin Nutr
2008
;
87
:
303
309
18.
Phillips
DI
,
Clark
PM
,
Hales
CN
,
Osmond
C
.
Understanding oral glucose tolerance: comparison of glucose or insulin measurements during the oral glucose tolerance test with specific measurements of insulin resistance and insulin secretion
.
Diabet Med
1994
;
11
:
286
292
19.
Jones
TW
,
Porter
P
,
Sherwin
RS
, et al
.
Decreased epinephrine responses to hypoglycemia during sleep
.
N Engl J Med
1998
;
338
:
1657
1662
20.
Hall
KD
,
Guyenet
SJ
,
Leibel
RL
.
The carbohydrate-insulin model of obesity is difficult to reconcile with current evidence
.
JAMA Intern Med
2018
;
178
:
1103
1105
21.
Speakman
JR
,
Hall
KD
.
Carbohydrates, insulin, and obesity
.
Science
2021
;
372
:
577
578
22.
Ludwig
DS
,
Ebbeling
CB
.
The carbohydrate-insulin model of obesity: beyond “calories in, calories out”
.
JAMA Intern Med
2018
;
178
:
1098
1103
23.
Templeman
NM
,
Skovsø
S
,
Page
MM
,
Lim
GE
,
Johnson
JD
.
A causal role for hyperinsulinemia in obesity
.
J Endocrinol
2017
;
232
:
R173
R183
24.
Sigal
RJ
,
El-Hashimy
M
,
Martin
BC
,
Soeldner
JS
,
Krolewski
AS
,
Warram
JH
.
Acute postchallenge hyperinsulinemia predicts weight gain: a prospective study
.
Diabetes
1997
;
46
:
1025
1029
25.
Odeleye
OE
,
de Courten
M
,
Pettitt
DJ
,
Ravussin
E
.
Fasting hyperinsulinemia is a predictor of increased body weight gain and obesity in Pima Indian children
.
Diabetes
1997
;
46
:
1341
1345
26.
Chen
YY
,
Wang
JP
,
Jiang
YY
, et al
.
Fasting plasma insulin at 5 years of age predicted subsequent weight increase in early childhood over a 5-year period-the Da Qing Children Cohort Study
.
PLoS One
2015
;
10
:
e0127389
27.
Gardner
CD
,
Trepanowski
JF
,
Del Gobbo
LC
, et al
.
Effect of low-fat vs low-carbohydrate diet on 12-month weight loss in overweight adults and the association with genotype pattern or insulin secretion: the DIETFITS randomized clinical trial
.
JAMA
2018
;
319
:
667
679
28.
Astley
CM
,
Todd
JN
,
Salem
RM
, et al
.
Genetic evidence that carbohydrate-stimulated insulin secretion leads to obesity
.
Clin Chem
2018
;
64
:
192
200
29.
Ebbeling
CB
,
Leidig
MM
,
Feldman
HA
,
Lovesky
MM
,
Ludwig
DS
.
Effects of a low-glycemic load vs low-fat diet in obese young adults: a randomized trial
.
JAMA
2007
;
297
:
2092
2102
30.
Page
MM
,
Johnson
JD
.
Mild suppression of hyperinsulinemia to treat obesity and insulin resistance
.
Trends Endocrinol Metab
2018
;
29
:
389
399
31.
Templeman
NM
,
Clee
SM
,
Johnson
JD
.
Suppression of hyperinsulinaemia in growing female mice provides long-term protection against obesity
.
Diabetologia
2015
;
58
:
2392
2402
32.
Jais
A
,
Brüning
JC
.
Arcuate nucleus-dependent regulation of metabolism - pathways to obesity and diabetes mellitus
.
Endocr Rev
.
7 September 2021 [Epub ahead of print]. DOI: 10.1210/endrev/bnab025
33.
Louis-Sylvestre
J
,
Le Magnen
J
.
Fall in blood glucose level precedes meal onset in free-feeding rats
.
Neurosci Biobehav Rev
1980
;
4
(
Suppl. 1
):
13
15
34.
Campfield
LA
,
Brandon
P
,
Smith
FJ
.
On-line continuous measurement of blood glucose and meal pattern in free-feeding rats: the role of glucose in meal initiation
.
Brain Res Bull
1985
;
14
:
605
616
35.
Dunn-Meynell
AA
,
Sanders
NM
,
Compton
D
, et al
.
Relationship among brain and blood glucose levels and spontaneous and glucoprivic feeding
.
J Neurosci
2009
;
29
:
7015
7022
36.
Goldsmith
R
,
Joanisse
DR
,
Gallagher
D
, et al
.
Effects of experimental weight perturbation on skeletal muscle work efficiency, fuel utilization, and biochemistry in human subjects
.
Am J Physiol Regul Integr Comp Physiol
2010
;
298
:
R79
R88
37.
Pereira
MA
,
Swain
J
,
Goldfine
AB
,
Rifai
N
,
Ludwig
DS
.
Effects of a low-glycemic load diet on resting energy expenditure and heart disease risk factors during weight loss
.
JAMA
2004
;
292
:
2482
2490
38.
Hall
KD
,
Bemis
T
,
Brychta
R
, et al
.
Calorie for calorie, dietary fat restriction results in more body fat loss than carbohydrate restriction in people with obesity
.
Cell Metab
2015
;
22
:
427
436
Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at https://www.diabetesjournals.org/journals/pages/license.