Adiponectin, Insulin Sensitivity, β-Cell Function, and Racial/Ethnic Disparity in Treatment Failure Rates in TODAY

Abstract

Objective: The Treatment Options for type 2 Diabetes in Adolescents and Youth (TODAY) study demonstrated that glycemic failure rates in the three treatments combined-metformin plus rosiglitazone, metformin alone, and metformin plus lifestyle-were higher in non-Hispanic blacks (NHB; 52.8%) versus non-Hispanic whites (NHW; 36.6%) and Hispanics (H; 45.0%). Moreover, metformin alone was less effective in NHB versus NHW versus H youth. This study describes treatment-associated changes in adiponectin, insulin sensitivity, and β-cell function over time among the three racial/ethnic groups to understand potential mechanism(s) responsible for this racial/ethnic disparity.

Research design and methods: TODAY participants underwent periodic oral glucose tolerance tests to determine insulin sensitivity, C-peptide index, and oral disposition index (oDI), with measurements of total and high-molecular-weight adiponectin (HMWA).

Results: At baseline NHB had significantly lower HMWA than NHW and H and exhibited a significantly smaller increase (17.3% vs. 33.7% vs. 29.9%, respectively) during the first 6 months overall. Increases in HMWA were associated with reductions in glycemic failure in the three racial/ethnic groups combined (hazard ratio 0.61, P < 0.0001) and in each race/ethnicity separately. Over time, HMWA was significantly lower in those who failed versus did not fail treatment, irrespective of race/ethnicity. There were no differences in treatment-associated temporal changes in insulin sensitivity, C-peptide index, and oDI among the three racial/ethnic groups.

Conclusions: HMWA is a reliable biomarker of treatment response in youth with type 2 diabetes. The diminutive treatment-associated increase in HMWA in NHB (∼50% lower) compared with NHW and H may explain the observed racial/ethnic disparity with higher therapeutic failure rates in NHB in TODAY.

© 2017 by the American Diabetes Association.

Figures

Figure 1

Temporal patterns of insulin sensitivity (A), C-peptide index (B), C-peptide oDI (C), and HMWA (D) in the three racial/ethnic groups with the three treatments combined. Model-adjusted geometric mean ± SE asymmetric limits (obtained as exp[mean ± SE of log values]) of insulin sensitivity (1/IF) (A), C-peptide index (ΔC30/ΔG30) (B), C-peptide oDI (1/IF × ΔC30/ΔG30) (C), and HMWA (D) in the three racial/ethnic groups (NHB, H, NHW) over 36 months of follow-up in TODAY, analyzed using log-transformed values. P values refer to the overall effect of race/ethnicity in the longitudinal models, adjusted for the baseline value of the outcome, sex, baseline BMI, age at randomization, medication adherence, and treatment group. The geometric mean is a good approximation of the median as the log-transformed data are approximately symmetric. NS, not significant (P > 0.05).

Figure 2

Short-term effect of therapy as the mean percent change (± SE) from baseline to 6 months for insulin sensitivity, β-cell function, and HMWA in each race/ethnicity with the three treatments combined. The log-transformed value of insulin sensitivity and other β-cell function biomarkers (HMWA) are modeled and results are presented as the mean percent change from baseline (± SE). Differences by race/ethnicity (NHB, H, NHW) in the 0–6 month percent change was examined in models adjusted for the baseline value of the outcome, sex, baseline BMI, age at randomization, medication adherence, and treatment group.

Figure 3

Temporal patterns of total adiponectin (A) and HMWA (B) by treatment failure in the three racial/ethnic groups with the three treatments combined. Model-adjusted geometric mean ± SE asymmetric limits (obtained as exp[mean ± SE of log values]) of measures total adiponectin (A) and HMWA (B) in the three racial/ethnic groups (NHB, H, NHW) over 36 months of follow-up in TODAY and by treatment failure/loss of glycemic control (failed: dashed lines vs. not fail: solid lines), analyzed using log-transformed values. P values from longitudinal models adjusted for the baseline value of the outcome, sex, baseline BMI, age at randomization, medication adherence, treatment group, race/ethnicity, treatment failure, and the interaction of race/ethnicity with treatment failure. The geometric mean is a good approximation of the median as the log-transformed data are approximately symmetric. NS, not significant (P > 0.05).