OBJECTIVE

In the Treatment Options for Type 2 Diabetes in Adolescents and Youth (TODAY) study, metformin plus rosiglitazone (M + R) maintained glycemic control better than metformin alone (M) or metformin plus lifestyle (M + L) in youth with type 2 diabetes (T2D). We hypothesized that changes in visceral adipose tissue (VAT) and subcutaneous adipose tissue (SAT) would explain the differential treatment effects on glycemia.

RESEARCH DESIGN AND METHODS

In 626 youth ages 11–17 years with T2D duration <2 years, VAT and SAT were estimated by DXA at baseline and at 6 and 24 months. Changes from baseline were analyzed in linear mixed models.

RESULTS

Baseline mean age was 13.9 years, 66.4% were female, 72.2% were Hispanic/non-Hispanic black, and 20.3% were non-Hispanic white (NHW). Mean BMI was 33.7 kg/m2. VAT increased more in M + R (13.1%) than M + L (3.9%, P = 0.0006) or M (6.5%, P = 0.0146). SAT also increased more in M + R (13.3%) than in M + L (5.4%, P < 0.0001) or M (6.4%, P = 0.0005), indicating no significant fat redistribution in M + R. In NHWs, VAT increased more in M + R than M (P = 0.0192) and M + L (P = 0.0482) but did not explain the race-ethnicity differences in treatment effects on glycemic control among treatment groups. VAT and SAT increases correlated with higher HbA1c, lower insulin sensitivity, and lower oral disposition index (all P < 0.05), but associations did not differ by treatment group.

CONCLUSIONS

In contrast to the existing reports in adults with T2D, in TODAY, M + R resulted in the most VAT accumulation compared with M + L or M. Differential effects on depot-specific indirect measures of adiposity are unrelated to treatment effects in sustaining glycemic control. Additional studies are needed to understand the clinical markers of metabolic risk profile in youth with T2D on rosiglitazone.

Obesity is prevalent among children and adolescents, thus contributing to significant increases in the annual incidence of type 2 diabetes (T2D) among youth in the U.S. (1). The differential impact of specific fat compartments on insulin sensitivity and lipid levels is well described, with excess visceral adipose tissue (VAT) imparting a greater risk of insulin resistance and hyperlipidemia than excess subcutaneous adipose tissue (SAT) (2,3). However, little is known about the effects of diabetes treatment on these metabolic and disease outcomes in adolescents. The Treatment Options for Type 2 Diabetes in Adolescents and Youth (TODAY) study was a multicenter, double-blind, randomized clinical trial that assembled the largest group of children and adolescents with T2D to date (4). The primary objective of TODAY was to compare the efficacy of three treatment arms (metformin monotherapy [M], metformin plus rosiglitazone [M + R], and metformin plus intensive lifestyle intervention [M + L]) to achieve glycemic control. Approximately one-half of the cohort (n = 319, 45.6%) reached the primary outcome of glycemic failure after an average follow-up duration of 3.9 years (range 2–6.5 years). Regarding glycemic control, the M + R combination was superior to M (P = 0.006), and M + L was intermediate, but not different, from M or M + R (5).

Rosiglitazone, a potent thiazolidinedione (TZD), improves insulin sensitivity. TZDs act as ligands of peroxisome proliferator–activated receptor-γ receptors, predominantly expressed in adipose tissue, and promote expression of lipogenic genes (6,7). This action enhances preadipocyte growth and differentiation, with a consequential increase in total body adiposity. Prior work demonstrated that treatment group differences in whole-body adiposity measures in TODAY are generally small and unrelated to treatment effects in sustaining glycemic control (8). The redistribution of fat stores has been suggested to explain the alleviation of peripheral insulin resistance by TZDs (9). The current report presents an in-depth analysis of the 1) differential treatment effects on redistribution of adipose tissue and 2) evaluation of the relationships among VAT, SAT, and insulin sensitivity; β-cell function; and metabolic risk factors.

TODAY Randomized Clinical Trial

The rationale, design, and methods of the TODAY trial have been reported in detail (4). In brief, after initial screening, eligible participants entered a 2–6-month run-in period with goals of weaning from nonstudy diabetes medications, tolerating metformin up to a dose of 1,000 mg twice daily but no less than 1,000 mg/day, attaining HbA1c <8.0% (<64 mmol/mol) for at least 2 months on metformin alone, and demonstrating adherence to study medications and visit attendance. After the run-in phase, 699 participants were randomly assigned to M, M + R, or M + L. Eligibility criteria included youth 10–17 years of age with T2D according to American Diabetes Association criteria and diabetes duration <2 years, BMI ≥85th percentile, negative diabetes-associated autoantibodies, liver transaminases ≤2.5 times the upper limit of normal, fasting C-peptide >0.6 ng/mL, and an adult caregiver willing to support study participation. All participants took two capsules (consisting of metformin and placebo or metformin and rosiglitazone) twice daily after randomization. Both the investigators and the participants were masked to the pharmacologic treatment group. The primary end point was time to glycemic failure, defined as HbA1c ≥8% (≥64 mmol/mol) for 6 months or metabolic decompensation requiring insulin therapy that could not be discontinued for 3 months (5). After treatment failure, rosiglitazone was stopped, if applicable, and participants continued on metformin plus insulin for diabetes management. Of the 699 TODAY participants, 626 (n = 203 M; n = 206 M + R; n = 217 M + L) with DXA scans collected at baseline were included in these analyses. Comparison of the 626 participants included in the sample with the 73 excluded showed no difference by race-ethnicity; however, at baseline, excluded participants were more likely to be older (15.0 ± 1.9 vs. 13.9 ± 2.0 years, P < 0.0001), heavier (BMI 45.5 ± 10.5 vs. 33.7 ± 6.2 kg/m2, P < 0.0001), and male (50.7% vs. 33.6%, P = 0.0037) with a longer duration of diabetes (9.2 ± 6.5 vs. 7.7 ± 5.8 months, P = 0.0150). The study was approved by the institutional review boards for the protection of human subjects of each participating institution. All participants provided informed consent, and minor children confirmed assent according to local guidelines.

Data Collection Measures

Assessments of anthropometric and laboratory values were obtained at baseline and at months 6 and 24. Demographic information was collected at randomization (4). Race-ethnicity was self-reported using 2000 U.S. Census–based questions. Participants checked Hispanic/Latino ethnicity yes or no and then checked as many racial categories as needed. All physical measurements were made by trained staff according to a study-wide protocol (4). Height was measured without shoes using a stadiometer. Weight was measured twice using a seca scale (model 882; seca North America, Hanover, MA), with a third measurement taken if the first two differed by >0.2 kg, and measurements were averaged. BMI was calculated as weight in kilograms divided by height squared in meters. Waist circumference was measured at the iliac crest using a nonstretch, nontension fiberglass Gulick II tape measure. All blood samples were obtained and processed immediately according to standardized procedures and shipped on dry ice to the TODAY central laboratory (Northwest Lipid Research Laboratory, University of Washington, Seattle, WA) (4). HbA1c was measured with a dedicated high-performance liquid chromatography system (TOSOH Biosciences, South San Francisco, CA). Oral glucose tolerance tests were conducted after a 10- to 14-h overnight fast. Blood samples were analyzed for glucose, insulin, and C-peptide. Surrogate markers of insulin sensitivity, β-cell function, and oral disposition index (oDI), a measure of β-cell function relative to insulin sensitivity, were calculated as previously reported (10). Briefly, insulin sensitivity was calculated as 1/fasting insulin (1/IF), C-peptide index (ΔCpep30/ΔGlucose30) as the ratio of the incremental C-peptide and glucose responses over the first 30 min of the oral glucose tolerance test, and C-peptide oDI as the product of insulin sensitivity multiplied by the C-peptide index (1/IF × ΔCpep30/ΔGlucose30) (10). Total and high-molecular-weight adiponectin (HMWA) were assayed using latex beads–based adiponectin assay reagents (Otsuka Pharmaceutical, Tokyo, Japan; distributed by MedTest DX, Cortland Manor, NY) on a Modular P Chemistry Analyzer (Roche, Indianapolis, IN) as previously described (11).

DXA Outcomes

Whole-body DXA scans were performed at baseline and at months 6 and 24. Whole-body adiposity DXA measures were obtained from existing densitometry systems (models QDR4500A, Discovery A, Discovery W/Wi, Delphi W, and Delphi A [Hologic, Marlborough, MA], models Prodigy, iDXA, and DPX-IQ [GE Healthcare, Madison, WI]) at each clinical center according to study-specific guidelines for subject positioning. All scans were standardized across the various DXA systems, and quality assurance procedures, including phantom cross-calibration and longitudinal monitoring, were applied as previously described (8). The scans were analyzed centrally at the TODAY DXA central reading center (University of California, San Francisco) by study-trained personnel using software according to manufacturer guidelines. Whole-body measures obtained included percent fat, fat mass, lean mass, and total mass.

Hologic and GE Healthcare scans were reanalyzed using manufacturer-specific software (Apex 4.0 and Prodigy 14.1, respectively) that estimates VAT and SAT from the total abdominal fat volume, where VAT = total abdominal fat − SAT, and SAT is estimated from manufacturer-specific models using the fat projected outside the abdominal walls. This approach is used by both Hologic and GE Healthcare DXA systems. DXA VAT has been validated against either MRI or CT in both adults (12) and children (13). DXA VAT estimates from GE Healthcare and Hologic are highly correlated, and intermanufacturer pooling was performed using equations from Fan et al. (14). Estimates of VAT and SAT mass, area, and volume were obtained, but only results for mass are presented in the current study because measures of area and volume provided comparable results throughout. VAT:SAT ratios (as a surrogate indicator of the relative amounts of visceral to subcutaneous fat) were also computed.

Statistics

Analyses were performed using SAS 9.4 for Windows (SAS Institute, Cary, NC). All analyses were considered exploratory, with statistical significance defined as P < 0.05 and no adjustment for multiple testing. Data are presented as mean ± SD or percent. Analyses were performed on data collected at baseline (n = 626), month 6 (n = 516), and month 24 (n = 344), which were the major outcome data collection visits that included DXA. Only data collected before the occurrence of the primary outcome (failure to maintain glycemic control) were included because management with insulin, which was started after treatment failure, is known to change measures of body composition. Variables with a skewed distribution (insulin sensitivity, C-peptide oDI, and HMWA) were transformed as appropriate. F tests or χ2 tests were used to compare baseline variables among the treatment, sex, age, pubertal stage, and race-ethnicity groups. If the overall test was significant, pairwise comparisons were performed.

Longitudinal data were analyzed using linear mixed models for repeated measures accounting for the multiple observations per participant (SAS PROC MIXED). Models examining body composition differences over time (as measured by the percent change from baseline) by treatment group were evaluated. Treatment group, visit, a treatment-by-visit interaction, sex, age at baseline, race-ethnicity, and height were included as covariates. In addition, all models were adjusted for study site to account for the densitometry system type (Hologic, GE Healthcare) used to collect the data. Effect modification of sex and race-ethnicity-by-treatment group was explored by adding interaction terms in the models. Stratified results were presented only if an overall significant sex-by-treatment or race-ethnicity-by-treatment interaction was found. Regression slopes between body composition measures (change from baseline) with the diabetes-specific measures (HbA1c, insulin sensitivity, and β-cell function) were assessed in similar models, and overall associations and slope differences by treatment group were evaluated. Testing with the diabetes-specific measures was performed on the natural log-transformed values to approximate a normal distribution. Assumption of linearity between body composition measures and the log-transformed diabetes-specific measures was met.

Baseline Characteristics

Characteristics of the TODAY participants at baseline are as follows: mean age 13.9 ± 2.0 years, mean duration of diabetes 7.7 ± 5.8 months, 66.4% female, 31.1% non-Hispanic black (NHB), 41.1% Hispanic, 20.3% non-Hispanic white (NHW), and 7.5% other (5.9% American Indian and 1.6% Asian) (Table 1). Mean BMI was 33.7 kg/m2, mean BMI z score 2.2, and mean waist circumference 106.5 ± 14.6 cm. Participant characteristics at baseline were similar among treatment groups. BMI, VAT, and SAT were highest in the M group but not statistically significantly different from the other treatment groups (data not shown). VAT:SAT ratio was also similar in all treatment groups. BMI, waist circumference, and SAT were highly intercorrelated at baseline: All Pearson correlation coefficients (ρ) were >0.75. VAT showed moderate correlation with BMI, waist circumference, and VAT:SAT ratio (all ρ >0.50). Similar baseline correlations were observed in each of the three treatment groups (data not shown).

Table 1

Demographic and baseline characteristics

CharacteristicValue
Participants (n626 
Female sex 66.4 
Age (years) 13.9 ± 2.0 
Diabetes duration (months) 7.7 ± 5.8 
Race-ethnicity  
 NHB 31.1 
 Hispanic 41.1 
 NHW 20.3 
 Other 7.5 
Tanner stage  
 4–5 87.9 
 <4 12.1 
HbA1c (%) 6.0 ± 0.7 
HMWA (ng/mL) 2,931.0 ± 1,810.6 
Height (cm) 164.3 ± 9.3 
Waist circumference (cm) 106.5 ± 14.6 
BMI (kg/m233.7 ± 6.2 
BMI z score 2.2 ± 0.4 
VAT mass (g) 643.8 ± 309.9 
SAT mass (g) 2,379.5 ± 737.8 
VAT:SAT ratio 0.28 ± 0.11 
CharacteristicValue
Participants (n626 
Female sex 66.4 
Age (years) 13.9 ± 2.0 
Diabetes duration (months) 7.7 ± 5.8 
Race-ethnicity  
 NHB 31.1 
 Hispanic 41.1 
 NHW 20.3 
 Other 7.5 
Tanner stage  
 4–5 87.9 
 <4 12.1 
HbA1c (%) 6.0 ± 0.7 
HMWA (ng/mL) 2,931.0 ± 1,810.6 
Height (cm) 164.3 ± 9.3 
Waist circumference (cm) 106.5 ± 14.6 
BMI (kg/m233.7 ± 6.2 
BMI z score 2.2 ± 0.4 
VAT mass (g) 643.8 ± 309.9 
SAT mass (g) 2,379.5 ± 737.8 
VAT:SAT ratio 0.28 ± 0.11 

Data are mean ± SD or % unless otherwise indicated.

Treatment Effects on Redistribution of Adiposity

Table 2 shows adiposity measures (BMI, VAT, SAT, and VAT:SAT ratio) by study visit (baseline and adjusted percent change from baseline at months 6 and 24) and by treatment group. Supplementary Fig. 1 shows the trends for VAT and SAT graphically. During the first 6 months of treatment, SAT declined in the M + L group, increased in the M group, and remained fairly stable in the M + R group. Significant differences in SAT changes were found when comparing the M (P = 0.0129) and M + R (P = 0.0348) groups with the M + L group. Similar effects were observed for VAT changes, but none of the changes at month 6 differed by treatment group. All the changes were of small clinical significance.

Table 2

Adiposity measurements across study visits by treatment group

Baseline mean ± SD and % change from baseline
Pairwise comparison P value from adjusted models
MM + RM + LM vs. M + RM vs. M + LM + R vs. M + L
BMI (kg/m2      
 Baseline 34.2 ± 6.0 33.7 ± 6.2 33.2 ± 6.3    
 % ∆ 0–6 0.5 1.4 −1.3 NS 0.0213 0.0003 
 % ∆ 0–24 3.8 7.8 3.2 <0.0001 NS <0.0001 
VAT mass (g)       
 Baseline 683.2 ± 382.9 639.6 ± 295.3 611.1 ± 235.4    
 % ∆ 0–6 1.0 −0.4 −2.0 NS NS NS 
 % ∆ 0–24 6.5 13.1 3.9 0.0146 NS 0.0006 
SAT mass (g)       
 Baseline 2,421.2 ± 729.3 2,370.0 ± 707.1 2,349.4 ± 774.8    
 % ∆ 0–6 0.8 0.1 −3.4 NS 0.0129 0.0348 
 % ∆ 0–24 6.4 13.3 5.4 0.0005 NS <0.0001 
VAT:SAT ratio       
 Baseline 0.29 ± 0.13 0.28 ± 0.11 0.27 ± 0.10    
 % ∆ 0–6 0.5 −0.1 1.5 NS NS NS 
 % ∆ 0–24 0.9 0.5 −1.0 NS NS NS 
Baseline mean ± SD and % change from baseline
Pairwise comparison P value from adjusted models
MM + RM + LM vs. M + RM vs. M + LM + R vs. M + L
BMI (kg/m2      
 Baseline 34.2 ± 6.0 33.7 ± 6.2 33.2 ± 6.3    
 % ∆ 0–6 0.5 1.4 −1.3 NS 0.0213 0.0003 
 % ∆ 0–24 3.8 7.8 3.2 <0.0001 NS <0.0001 
VAT mass (g)       
 Baseline 683.2 ± 382.9 639.6 ± 295.3 611.1 ± 235.4    
 % ∆ 0–6 1.0 −0.4 −2.0 NS NS NS 
 % ∆ 0–24 6.5 13.1 3.9 0.0146 NS 0.0006 
SAT mass (g)       
 Baseline 2,421.2 ± 729.3 2,370.0 ± 707.1 2,349.4 ± 774.8    
 % ∆ 0–6 0.8 0.1 −3.4 NS 0.0129 0.0348 
 % ∆ 0–24 6.4 13.3 5.4 0.0005 NS <0.0001 
VAT:SAT ratio       
 Baseline 0.29 ± 0.13 0.28 ± 0.11 0.27 ± 0.10    
 % ∆ 0–6 0.5 −0.1 1.5 NS NS NS 
 % ∆ 0–24 0.9 0.5 −1.0 NS NS NS 

% ∆ denotes the percent change from baseline at months 6 or 24 for continuous variables. The three treatment groups are M, M + R, and M + L. P values were calculated from generalized linear mixed models testing for pairwise treatment differences as a function of time, study site (as a surrogate for DXA machine type), sex, age at baseline, height, race-ethnicity, treatment, and an interaction term for time by treatment. Nonsignificant P > 0.05.

After 24 months of treatment, both VAT and SAT increased relative to baseline in all treatment groups. VAT increased the most in M + R (13.1%) compared with M (6.5%, P = 0.0146) or M + L (3.9%, P = 0.0006). SAT also increased more in M + R (13.3%) than in M (6.4%, P = 0.0005) or M + L (5.4%, P < 0.0001). There were no statistically significant differences in VAT or SAT between M and M + L at month 24. Between-group differences in VAT:SAT ratio were not significant at any time point but showed a trend toward a greater decline in M + L than in the other two groups at month 24.

Effects of Sex and Race-Ethnicity

At baseline, female participants had higher SAT (2,519.6 ± 681.3 vs. 2,101.7 ±767.7 g, P < 0.0001) but no difference in VAT (637.2 ± 269.1 vs. 657.0 ± 378.3 g, P not significant) compared with male participants. VAT:SAT ratio was lower in females than in males (0.25 ± 0.09 vs. 0.32 ± 0.13, P < 0.0001). Over time, greater increases in SAT (as measured by the percent change from baseline) were observed in female compared with male participants (12.8% vs. 6.5% at month 24, P < 0.0001). No difference in changes over time by sex was identified for VAT. However, sex did not moderate treatment effects on any adiposity measures (sex-by-treatment interaction was not significant) (data not shown).

At baseline, NHBs had higher SAT (2,499.7 ± 725.9 vs. 2,315.2 ± 810.1 vs. 2,332.3 ± 700.2 g, P = 0.0274) and lower VAT (517.6 ± 208.6 vs. 718.6 ± 394.5 vs. 709.1 ± 309.8 g, P < 0.0001) than NHWs and Hispanics. Over time, the only treatment effects by race-ethnicity interactions observed were for VAT mass and VAT:SAT ratio. In NHBs, greater increases in VAT occurred in M than in M + L (P = 0.0362) at month 6, and in NHWs, greater increases occurred in M + R than in M (P = 0.0192) or in M + L (P = 0.0482) at month 24 (Fig. 1 and Supplementary Table 1). Within each treatment group, no race-ethnicity difference in VAT changes was identified. A small, but significant change by race-ethnicity was observed in VAT:SAT ratio at month 6 among Hispanic participants (greater increases in M + L vs. M [P = 0.0364] and M + R [P = 0.0053]) (Fig. 1). Nonetheless, no effect of treatment by race-ethnicity was observed in the VAT:SAT ratio at month 24, suggesting no race-ethnicity–specific differences in overall fat redistribution with treatment.

Figure 1

Adjusted mean percent change from baseline in VAT mass and VAT:SAT ratio at months 6 and 24 by treatment group and race-ethnicity. Adjusted mean percent change from baseline for VAT mass in NHBs (A), Hispanics (B), and NHWs (C), and VAT:SAT ratio in NHBs (D), Hispanics (E), and NHWs (F) at months 6 and 24 by treatment. The three treatment groups are M, M + R, and M + L. P values were calculated from generalized linear mixed models that tested for pairwise treatment differences in the percent change from baseline during the first 2 years of the study as a function of time, study site (as a surrogate for DXA machine type), sex, age at baseline, height, treatment, and an interaction term for time by treatment. Significant treatment group differences for the 6- and 24-month percent changes from baseline are indicated within the figure.

Figure 1

Adjusted mean percent change from baseline in VAT mass and VAT:SAT ratio at months 6 and 24 by treatment group and race-ethnicity. Adjusted mean percent change from baseline for VAT mass in NHBs (A), Hispanics (B), and NHWs (C), and VAT:SAT ratio in NHBs (D), Hispanics (E), and NHWs (F) at months 6 and 24 by treatment. The three treatment groups are M, M + R, and M + L. P values were calculated from generalized linear mixed models that tested for pairwise treatment differences in the percent change from baseline during the first 2 years of the study as a function of time, study site (as a surrogate for DXA machine type), sex, age at baseline, height, treatment, and an interaction term for time by treatment. Significant treatment group differences for the 6- and 24-month percent changes from baseline are indicated within the figure.

Close modal

Associations With HbA1c, Insulin Sensitivity, and Insulin Secretion

VAT and SAT increase from baseline correlated with higher HbA1c, lower insulin sensitivity, and lower C-peptide oDI at each time point. Regression analysis showed that in an unadjusted model, there was a more positive association between VAT accumulation on HbA1c at month 6 in the M + R than in the M + L group (P = 0.0344), but the between-group difference in slope was no longer significant after adjusting for age, sex, race-ethnicity, and height (Fig. 2A). No other treatment group difference was found in the associations between VAT and SAT changes with HbA1c (Figs. 2A and B and 3A and B). With regard to insulin sensitivity, at 6 months, M + L treatment resulted in lower insulin sensitivity with VAT accumulation than treatment with M (P = 0.0305) (Fig. 2C). In contrast, at 6 months, treatment with M + R significantly blunted the effect of visceral adiposity change on C-peptide oDI relative to that seen with M + L (P = 0.0427, M + R vs. M + L) (Fig. 2E). A similar slope difference between the M + R and M + L groups was found for the association between SAT change from baseline with C-peptide oDI at month 6 (P = 0.0143) (Fig. 3E). While these differential treatment effects were noted at month 6, no slope difference among the treatment groups was identified at month 24 for any of the associations. The adjusted β-coefficient values of the multivariable linear regression models depicted in Fig. 2 are provided in Supplementary Table 2.

Figure 2

Slopes of HbA1c, insulin sensitivity, β-cell function (as measured by C-peptide oDI), and HMWA versus change from baseline in VAT mass at months 6 and 24 by treatment group. Regression lines (slopes) of HbA1c (A and B), insulin sensitivity (C and D), C-peptide oDI (E and F), and HMWA (G and H) values at months 6 and 24 versus the respective change from baseline in VAT mass by treatment group. The three treatment groups are M, M + R, and M + L. P values were calculated from linear regression models testing for pairwise treatment differences in the slopes at months 6 and 24. Models were adjusted for study site (as surrogate for DXA machine type), sex, age at baseline, height, race-ethnicity, and treatment. Data are plotted on the original unit scale, but testing was performed on the log-transformed values of insulin sensitivity, C-peptide oDI, and HMWA because of lack of normality. Overall associations and significant P values for slope differences by treatment group are indicated within the figure.

Figure 2

Slopes of HbA1c, insulin sensitivity, β-cell function (as measured by C-peptide oDI), and HMWA versus change from baseline in VAT mass at months 6 and 24 by treatment group. Regression lines (slopes) of HbA1c (A and B), insulin sensitivity (C and D), C-peptide oDI (E and F), and HMWA (G and H) values at months 6 and 24 versus the respective change from baseline in VAT mass by treatment group. The three treatment groups are M, M + R, and M + L. P values were calculated from linear regression models testing for pairwise treatment differences in the slopes at months 6 and 24. Models were adjusted for study site (as surrogate for DXA machine type), sex, age at baseline, height, race-ethnicity, and treatment. Data are plotted on the original unit scale, but testing was performed on the log-transformed values of insulin sensitivity, C-peptide oDI, and HMWA because of lack of normality. Overall associations and significant P values for slope differences by treatment group are indicated within the figure.

Close modal
Figure 3

Slopes of HbA1c, insulin sensitivity, β-cell function (as measured by C-peptide oDI), and HMWA versus change from baseline in SAT mass at months 6 and 24, by treatment group. Regression lines (slopes) of HbA1c (A and B), insulin sensitivity (C and D), C-peptide oDI (E and F), and HMWA (G and H) values at months 6 and 24 versus the respective change from baseline in SAT mass by treatment group. The three treatment groups are M, M + R, and M + L. P values were calculated from linear regression models testing for pairwise treatment differences in the slopes at months 6 and 24. Models were adjusted for study site (as surrogate for DXA machine type), sex, age at baseline, height, race-ethnicity, and treatment. Data are plotted on the original unit scale, but testing was performed on the log-transformed values of insulin sensitivity, C-peptide oDI, and HMWA because of lack of normality. Overall associations and significant P values for slope differences by treatment group are indicated within the figure.

Figure 3

Slopes of HbA1c, insulin sensitivity, β-cell function (as measured by C-peptide oDI), and HMWA versus change from baseline in SAT mass at months 6 and 24, by treatment group. Regression lines (slopes) of HbA1c (A and B), insulin sensitivity (C and D), C-peptide oDI (E and F), and HMWA (G and H) values at months 6 and 24 versus the respective change from baseline in SAT mass by treatment group. The three treatment groups are M, M + R, and M + L. P values were calculated from linear regression models testing for pairwise treatment differences in the slopes at months 6 and 24. Models were adjusted for study site (as surrogate for DXA machine type), sex, age at baseline, height, race-ethnicity, and treatment. Data are plotted on the original unit scale, but testing was performed on the log-transformed values of insulin sensitivity, C-peptide oDI, and HMWA because of lack of normality. Overall associations and significant P values for slope differences by treatment group are indicated within the figure.

Close modal

Adiponectin and Its Association With Depot-Specific Adiposity Measures

There was an overall treatment group difference for HMWA over time. While an increase in adiponectin was noted in all treatment arms, the highest mean percent change in HMWA at 6 months was in M + R compared with M (9.1% vs. 0.2%, P < 0.0001) and M + L (9.1% vs. 0.5%, P < 0.0001). Similar temporal patterns in HMWA were observed in the longer term (at 24 months) among the three treatment groups. As previously reported, those in the M + R group had larger increases over time compared with the M and M + L groups, irrespective of race-ethnicity (11).

At baseline, HMWA weakly correlated with VAT and SAT (ρ <0.2 for both). Over time, VAT and SAT accumulation were associated with lower HMWA, independent of other measures of adiposity. After adjusting for other covariates, the association between VAT accumulation with HMWA levels at month 24 was more positive in the M + R group than in the M + L group (P = 0.0355) (Fig. 2), but no difference in the slopes by treatment group was found at month 6. The association between SAT accumulation with HMWA levels at months 6 and 24 was more positive in the M + R group than in the M + L group (P = 0.0203 and P = 0.0438, respectively) (Fig. 3 and Supplementary Table 2). Further analysis showed that HMWA increases from baseline correlated with higher insulin sensitivity and C-peptide oDI and lower HbA1c levels at all time points (all P < 0.05). However, these associations did not differ by treatment group. We evaluated the relationships between change in VAT (or SAT) and change in adiponectin and found significant associations at each time point. The relationships remained significant even after further adjustment for change in insulin sensitivity (all P < 0.05).

The treatment group differences reported in the primary outcome analysis (superiority of M + R in sustaining glycemic control compared with M and with M + L intermediate) of the TODAY trial (5) were not explained by concordant differences in BMI or whole-body adiposity. The current study advances previous observations by demonstrating that these treatment group differences are also not attributable to changes in VAT and SAT. In agreement with the weight gain associated with M + R treatment, data here indicate greater fat deposition in both visceral and subcutaneous depots with M + R than M or M + L treatment. It is well described that rosiglitazone increases SAT and decreases VAT in adults in line with the known redistribution effects of the TZDs (1518). Metformin is also known to cause reduction in visceral fat (16). It is notable that we found a substantial increase in both VAT and SAT with M + R in TODAY. Furthermore, youth in TODAY had an increase in VAT that was greater in M + R than in M + L and M. Comparisons between exact combination therapies used in TODAY have not been reported in adults. Nonetheless, these findings in TODAY youth are in contrast to the well-known compartment-specific effects of rosiglitazone on decline in VAT previously demonstrated in adults (1518). Changes in VAT and SAT closely followed the changes in BMI previously reported (8), indicating that despite its superiority in maintaining glycemic control, treatment with M + R was associated not only with weight gain but also with more fat deposition compared with other treatments (M and M + L).

It is established that VAT and SAT negatively affect insulin sensitivity and secretion, with greater VAT more predictive of insulin resistance (1921). Given this knowledge, the greater increase in VAT in M + R should lead to inferior glycemic control compared with M or M + L. However, such was not the case in this study. Indices of adiposity (VAT and SAT accumulation) were associated with higher HbA1c and lower insulin sensitivity and secretion, but these associations (slope) did not differ by treatment group. This observation suggests that the beneficial effect of TZDs on glycemia in youth with T2D is related to an improvement in insulin sensitivity, but this effect is not mediated through reduced visceral adiposity. Although the underlying reason for superior glycemic control of M + R is unclear, it appears as though M + R ameliorates the effect of increased visceral adiposity on glycemic control in this study population. This implies that in youth with T2D on rosiglitazone, monitoring waist circumference may not be truly reflective of insulin sensitivity in that subgroup, and other clinical markers should be used.

Some studies suggest that the VAT:SAT ratio is a correlate of cardiometabolic risk above and beyond BMI and VAT (22,23). A lower VAT:SAT ratio has been reported after treatment with TZDs (15). However, we did not detect any difference in VAT:SAT ratio between treatment groups, suggesting the absence of redistribution of fat and also as well as suggesting that the effect of TZDs on insulin sensitivity was mediated by factors other than quantitative changes in indices of adiposity.

Treatment effects on indices of adiposity in the current study were not moderated by sex. It is noteworthy that our study population comprised adolescents in whom body composition is changing rapidly partly because of increases in body size (24,25). During growth, both SAT and VAT accumulate. It is possible that treatment with M and M + L are more efficacious to offset growth-related rises in visceral and subcutaneous depots than is M + R treatment. We previously reported that adherence to treatment did not account for the difference in treatment outcomes in TODAY youth.

In the primary analysis, the treatment failure rate for NHBs (52.9%) was greater than that for NHWs (36.9%) or Hispanics (44.8%), with the highest failure rates observed in NHB participants treated with M alone (5). Racial-ethnic differences in visceral fat have been previously described in youth and adults. In line with these observations, NHB participants in TODAY had lower visceral fat compared with NHW participants. A greater increase in VAT in NHBs could explain the differences in glycemic control observed in TODAY. However, the rates of fat accumulation in NHB participants were no greater than those in Hispanic or NHW participants, irrespective of the treatment arm. Indeed, the increase in VAT was higher in NHWs with M + R treatment. From the current analysis, it is evident that race-ethnicity differences in changes in VAT or fat redistribution (VAT:SAT ratio) do not account for the race-ethnicity differences in glycemic control by treatment group observed in TODAY.

Adiponectin is a hormone exclusively expressed and secreted by adipose tissue. Low levels of adiponectin have been associated with T2D. Adiponectin is negatively related to visceral fat accumulation (26) as well as insulin resistance (2729) and exerts potent positive effects on pancreatic β-cell survival and functionality (29,30). In TODAY, we previously demonstrated that NHB youth with T2D had significantly lower HMWA, but baseline HMWA was not a predictor of glycemic failure (11). We observed a greater increase in adiponectin levels over time with M + R compared with M + L or M, irrespective of race-ethnicity. Other studies in T2D have also shown a similar pronounced effect of rosiglitazone on adiponectin, which could potentially explain the increase in insulin sensitivity (3134). In the current study, both VAT and SAT accumulation showed a strong association with adiponectin after 24 months of treatment. However, an exact causal relationship remains elusive. Whether adipokine changes follow fat redistribution or adipokine changes mediate fat redistribution is not well understood. With regard to the direct effects of adiponectin on β-cell function, there were associations observed with insulin sensitivity and secretion and HbA1c. However, these associations did not differ by treatment group and do not explain the differential treatment effects on glycemic control observed in the TODAY cohort. Furthermore, the relationship between changes in adiponectin and adiposity remained independent of changes in insulin sensitivity. These observations suggest that adiponectin is a stronger predictor of glycemic control than VAT or SAT, but the underlying mechanism by which adiponectin regulates glycemic control has yet to be determined.

Strengths of this study include the well-characterized TODAY cohort and the longitudinal nature of the study with both detailed imaging and biochemical evaluation. All treatment arms had reasonable sample sizes with similar baseline characteristics, including BMI. Because the study population included adolescents, we adjusted the indices of adiposity for height to account for rapid changes in body composition during this stage of life. There are also limitations in this study. First, our study cohort was overweight or obese, and some of the findings of this study may not be translated to nonobese populations. The availability of a normal-weight control group would have been a desirable comparator and may have strengthened our findings. In addition, because all participants were treated with metformin during run-in, baseline VAT and SAT may have already have been affected by metformin, and thus, our baseline measures may underestimate the true degree of VAT and SAT elevations of this population before treatment. Nevertheless, our findings are still relevant to the majority of youth with obesity and T2D, an increasingly growing population.

In conclusion, despite its superiority in maintaining durable glycemic control, treatment with M + R in youth with T2D, unlike in adults, resulted in substantially more visceral and subcutaneous adiposity compared with other treatment modalities. Although the beneficial effect on glycemic control parallels the increase in adiponectin levels, the mechanism of a glucose-lowering effect of rosiglitazone was not apparent in this study. It is evident that this effect was not mediated through concordant treatment effects on fat redistribution; thus, other biological, environmental, and genetic factors yet to be determined are presumed important. Our observations of a dominant role of adiponectin in the dissociation between fat redistribution and glycemic control suggest that adiponectin may be the more important determinant of metabolic response than either VAT or SAT. Future studies, including long-term follow-up of the TODAY cohort, and the identification of factors explaining the role of adiponectin in glycemic control will be critical to better understand the differential effects of diabetes treatment regimens on glycemic control and adiposity across the life span.

Clinical trial reg. no. NCT00081328, clinicaltrials.gov

Acknowledgments. A complete list of participants in the TODAY Study Group is presented in the Supplementary Data. The TODAY Study Group thanks the participants for their commitment and dedication to the goal of diabetes treatment and the following companies for donations in support of the study’s efforts: Becton Dickinson, Bristol-Myers Squibb, Eli Lilly, GlaxoSmithKline, LifeScan, Pfizer, and Sanofi. The authors also gratefully acknowledge the participation and guidance of the American Indian partners associated with the clinical center located at the University of Oklahoma Health Sciences Center, including members of the Absentee Shawnee Tribe, Cherokee Nation, Chickasaw Nation, Choctaw Nation of Oklahoma, and Oklahoma City Area Indian Health Service; the opinions expressed in this article are those of the authors and do not necessarily reflect the views of the respective tribes and the Indian Health Service. Materials developed and used for the TODAY standard diabetes education program and the intensive lifestyle intervention program are available to the public at https://today.bsc.gwu.edu.

Funding. This work was completed with funding from the National Institute of Diabetes and Digestive and Kidney Diseases and the National Institutes of Health Office of the Director through grants U01-DK-61212, U01-DK-61230, U01-DK-61239, U01-DK-61242, and U01-DK-61254; from National Center for Research Resources General Clinical Research Centers Program grants M01-RR-00036 (Washington University School of Medicine), M01-RR-00043-45 (Children’s Hospital Los Angeles), M01-RR-00069 (University of Colorado Denver), M01-RR-00084 (Children’s Hospital of Pittsburgh), M01-RR-01066 (Massachusetts General Hospital), M01-RR-00125 (Yale University), and M01-RR-14467 (University of Oklahoma Health Sciences Center); and from National Center for Research Resources Clinical and Translational Science Awards grants UL1-RR-024134 (Children’s Hospital of Philadelphia), UL1-RR-024139 (Yale University), UL1-RR-024153 (Children’s Hospital of Pittsburgh), UL1-RR-024989 (Case Western Reserve University), UL1-RR-024992 (Washington University in St. Louis), UL1-RR-025758 (Massachusetts General Hospital), and UL1-RR-025780 (University of Colorado Denver).

The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Duality of Interest. K.C.C. serves on a Novo Nordisk data management committee for a T2D research study. M.E.G. receives a consulting fee for a T2D research study conducted by Daiichi Sankyo. R.S.W. is a consultant to Insulogic and a site principal investigator for industry-sponsored clinical trials sponsored by Medtronic MiniMed, Oramed, Kowa Research Institute, and Mylan. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. R.D. researched data and wrote the manuscript. R.D., J.A.S., L.E.g., K.C.C., M.E.G., J.H., L.L.L., K.J.N., and N.H.W. contributed to the study design, data collection and interpretation, and manuscript edits. L.E.g. contributed to data management, statistical data analysis and interpretation, and manuscript writing. R.S.W. contributed to data interpretation and manuscript edits. L.E.g. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this work were presented in poster form at the 78th Scientific Sessions of the American Diabetes Association, Orlando, FL, 22–26 June 2018.

1.
Mayer-Davis
EJ
,
Lawrence
JM
,
Dabelea
D
, et al.;
SEARCH for Diabetes in Youth Study
.
Incidence trends of type 1 and type 2 diabetes among youths, 2002-2012
.
N Engl J Med
2017
;
376
:
1419
1429
[PubMed]
2.
Cali
AM
,
Caprio
S
.
Ectopic fat deposition and the metabolic syndrome in obese children and adolescents
.
Horm Res
2009
;
71
(
Suppl. 1
):
2
7
[PubMed]
3.
Misra
M
,
Bredella
MA
,
Tsai
P
,
Mendes
N
,
Miller
KK
,
Klibanski
A
.
Lower growth hormone and higher cortisol are associated with greater visceral adiposity, intramyocellular lipids, and insulin resistance in overweight girls
.
Am J Physiol Endocrinol Metab
2008
;
295
:
E385
E392
[PubMed]
4.
Zeitler
P
,
Epstein
L
,
Grey
M
, et al.;
TODAY Study Group
.
Treatment options for type 2 diabetes in adolescents and youth: a study of the comparative efficacy of metformin alone or in combination with rosiglitazone or lifestyle intervention in adolescents with type 2 diabetes
.
Pediatr Diabetes
2007
;
8
:
74
87
[PubMed]
5.
Zeitler
P
,
Hirst
K
,
Pyle
L
, et al.;
TODAY Study Group
.
A clinical trial to maintain glycemic control in youth with type 2 diabetes
.
N Engl J Med
2012
;
366
:
2247
2256
[PubMed]
6.
Spiegelman
BM
.
PPAR-γ: adipogenic regulator and thiazolidinedione receptor
.
Diabetes
1998
;
47
:
507
514
[PubMed]
7.
Mayerson
AB
,
Hundal
RS
,
Dufour
S
, et al
.
The effects of rosiglitazone on insulin sensitivity, lipolysis, and hepatic and skeletal muscle triglyceride content in patients with type 2 diabetes
.
Diabetes
2002
;
51
:
797
802
[PubMed]
8.
TODAY Study Group
.
Treatment effects on measures of body composition in the TODAY clinical trial
.
Diabetes Care
2013
;
36
:
1742
1748
[PubMed]
9.
Sharma
AM
,
Staels
B
.
Review: peroxisome proliferator-activated receptor gamma and adipose tissue--understanding obesity-related changes in regulation of lipid and glucose metabolism
.
J Clin Endocrinol Metab
2007
;
92
:
386
395
[PubMed]
10.
Arslanian
S
,
Pyle
L
,
Payan
M
, et al.;
TODAY Study Group
.
Effects of metformin, metformin plus rosiglitazone, and metformin plus lifestyle on insulin sensitivity and β-cell function in TODAY
.
Diabetes Care
2013
;
36
:
1749
1757
[PubMed]
11.
Arslanian
S
,
El Ghormli
L
,
Bacha
F
, et al.;
TODAY Study Group
.
Adiponectin, insulin sensitivity, β-cell function, and racial/ethnic disparity in treatment failure rates in TODAY
.
Diabetes Care
2017
;
40
:
85
93
[PubMed]
12.
Micklesfield
LK
,
Goedecke
JH
,
Punyanitya
M
,
Wilson
KE
,
Kelly
TL
.
Dual-energy X-ray performs as well as clinical computed tomography for the measurement of visceral fat
.
Obesity (Silver Spring)
2012
;
20
:
1109
1114
[PubMed]
13.
Bosch
TA
,
Dengel
DR
,
Kelly
AS
,
Sinaiko
AR
,
Moran
A
,
Steinberger
J
.
Visceral adipose tissue measured by DXA correlates with measurement by CT and is associated with cardiometabolic risk factors in children
.
Pediatr Obes
2015
;
10
:
172
179
[PubMed]
14.
Fan
B
,
Wilson
J
,
Shepherd
J
,
Wu
X
.
How comparable are Hologic and GE-Lunar visceral fat analyses
?
J Clin Densitom
2013
;
16
:
263
273
15.
Punthakee
Z
,
Alméras
N
,
Després
JP
, et al
.
Impact of rosiglitazone on body composition, hepatic fat, fatty acids, adipokines and glucose in persons with impaired fasting glucose or impaired glucose tolerance: a sub-study of the DREAM trial
.
Diabet Med
2014
;
31
:
1086
1092
[PubMed]
16.
Virtanen
KA
,
Hällsten
K
,
Parkkola
R
, et al
.
Differential effects of rosiglitazone and metformin on adipose tissue distribution and glucose uptake in type 2 diabetic subjects
.
Diabetes
2003
;
52
:
283
290
17.
Iozzo
P
,
Hallsten
K
,
Oikonen
V
, et al
.
Effects of metformin and rosiglitazone monotherapy on insulin-mediated hepatic glucose uptake and their relation to visceral fat in type 2 diabetes
.
Diabetes Care
2003
;
26
:
2069
2074
18.
Carey
DG
,
Cowin
GJ
,
Galloway
GJ
, et al
.
Effect of rosiglitazone on insulin sensitivity and body composition in type 2 diabetic patients
[published correction appears in Obes Res 2002;10:following table of contents].
Obes Res
2002
;
10
:
1008
1015
19.
Amato
MC
,
Giordano
C
,
Galia
M
, et al.;
AlkaMeSy Study Group
.
Visceral Adiposity Index: a reliable indicator of visceral fat function associated with cardiometabolic risk
.
Diabetes Care
2010
;
33
:
920
922
[PubMed]
20.
Preis
SR
,
Massaro
JM
,
Robins
SJ
, et al
.
Abdominal subcutaneous and visceral adipose tissue and insulin resistance in the Framingham heart study
.
Obesity (Silver Spring)
2010
;
18
:
2191
2198
[PubMed]
21.
Katzmarzyk
PT
,
Heymsfield
SB
,
Bouchard
C
.
Clinical utility of visceral adipose tissue for the identification of cardiometabolic risk in white and African American adults
.
Am J Clin Nutr
2013
;
97
:
480
486
[PubMed]
22.
Kaess
BM
,
Pedley
A
,
Massaro
JM
,
Murabito
J
,
Hoffmann
U
,
Fox
CS
.
The ratio of visceral to subcutaneous fat, a metric of body fat distribution, is a unique correlate of cardiometabolic risk
.
Diabetologia
2012
;
55
:
2622
2630
[PubMed]
23.
Yeoh
AJ
,
Pedley
A
,
Rosenquist
KJ
,
Hoffmann
U
,
Fox
CS
.
The association between subcutaneous fat density and the propensity to store fat viscerally
.
J Clin Endocrinol Metab
2015
;
100
:
E1056
E1064
[PubMed]
24.
Suliga
E
.
Visceral adipose tissue in children and adolescents: a review
.
Nutr Res Rev
2009
;
22
:
137
147
[PubMed]
25.
Katzmarzyk
PT
,
Shen
W
,
Baxter-Jones
A
, et al
.
Adiposity in children and adolescents: correlates and clinical consequences of fat stored in specific body depots
.
Pediatr Obes
2012
;
7
:
e42
e61
[PubMed]
26.
Yamauchi
T
,
Kamon
J
,
Ito
Y
, et al
.
Cloning of adiponectin receptors that mediate antidiabetic metabolic effects
.
Nature
2003
;
423
:
762
769
[PubMed]
27.
Kadowaki
T
,
Yamauchi
T
,
Kubota
N
,
Hara
K
,
Ueki
K
,
Tobe
K
.
Adiponectin and adiponectin receptors in insulin resistance, diabetes, and the metabolic syndrome
.
J Clin Invest
2006
;
116
:
1784
1792
[PubMed]
28.
Yatagai
T
,
Nagasaka
S
,
Taniguchi
A
, et al
.
Hypoadiponectinemia is associated with visceral fat accumulation and insulin resistance in Japanese men with type 2 diabetes mellitus
.
Metabolism
2003
;
52
:
1274
1278
[PubMed]
29.
Brown
JE
,
Conner
AC
,
Digby
JE
, et al
.
Regulation of beta-cell viability and gene expression by distinct agonist fragments of adiponectin
.
Peptides
2010
;
31
:
944
949
[PubMed]
30.
Rao
JR
,
Keating
DJ
,
Chen
C
,
Parkington
HC
.
Adiponectin increases insulin content and cell proliferation in MIN6 cells via PPARγ-dependent and PPARγ-independent mechanisms
.
Diabetes Obes Metab
2012
;
14
:
983
989
[PubMed]
31.
Ahima
RS
.
Overcoming insulin resistance with CNTF
.
Nat Med
2006
;
12
:
511
512
[PubMed]
32.
Yu
JG
,
Javorschi
S
,
Hevener
AL
, et al
.
The effect of thiazolidinediones on plasma adiponectin levels in normal, obese, and type 2 diabetic subjects
.
Diabetes
2002
;
51
:
2968
2974
[PubMed]
33.
Yang
WS
,
Jeng
CY
,
Wu
TJ
, et al
.
Synthetic peroxisome proliferator-activated receptor-gamma agonist, rosiglitazone, increases plasma levels of adiponectin in type 2 diabetic patients
.
Diabetes Care
2002
;
25
:
376
380
[PubMed]
34.
Miyazaki
Y
,
DeFronzo
RA
.
Rosiglitazone and pioglitazone similarly improve insulin sensitivity and secretion, glucose tolerance and adipocytokines in type 2 diabetic patients
.
Diabetes Obes Metab
2008
;
10
:
1204
1211
[PubMed]
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 http://www.diabetesjournals.org/content/license.

Supplementary data