Demographic and metabolic variables at baseline

A total of 268 adults (145 African American, 123 European American; 193 women, 75 men) who completed baseline assessment of insulin secretion and insulin sensitivity in the POP-ABC study were included in the present analysis (online supplemental figure S1). The participants’ baseline characteristics are shown in table 1. The mean age of participants was 44.6±10.1 years and mean BMI was 30.3±7.37 kg/m2. Compared with European American participants, African American participants were younger, had higher BMI and HbA1c, similar waist circumference and 2hrPG, and lower FPG values (table 1). Approximately 85% of participants had one parent with type 2 diabetes and ~15% had both parents affected. As reported in the primary results of the POP-ABC study,29 African American participants were more likely to report maternal diabetes history, whereas European American participants were more likely to report paternal diabetes history.

Table 1

View inline|View popup

Baseline characteristics of study subjects

Basal and dynamic insulin secretion and clearance

At enrollment, the mean fasting plasma insulin level was similar in African American versus European American offspring of parents with type 2 diabetes, as was HOMA-B (table 2). African American offspring of parents with type 2 diabetes had lower basal insulin clearance compared with European American offspring of parents with type 2 diabetes, before and after adjustment for multiple baseline variables, including sex, age, maternal/paternal diabetes history, BMI, waist circumference, FPG, and HbA1c (table 2).

Table 2

View inline|View popup

Baseline insulin sensitivity, secretion, and clearance under basal and dynamic conditions by race/ethnicity

The mean insulinogenic index was significantly higher in African American versus European American participants, adjusted for baseline variables (table 2). Compared with European American participants, African American participants had higher AUC insulin during IVGT, despite similar plasma glucose excursions (table 2 and online supplemental figure S2). The peak insulin response during IVGT as a percentage of baseline fasting insulin level (AIR%) was higher in African American offspring of parents with type 2 diabetes compared with European American offspring of parents with type 2 diabetes (table 2). Compared with European American participants, African American participants had significantly lower dynamic insulin clearance based on molar ratios of plasma insulin and C peptide during IVGT (table 2). The MCR for insulin during hyperinsulinemic euglycemic clamp also was lower in African American offspring of parents (table 2).

Insulin sensitivity

There was no significant difference between African American versus European American participants in the Matsuda index of insulin sensitivity derived from OGTT plasma insulin and glucose values (table 2). Figure 1 shows plasma glucose (figure 1A) and insulin (figure 1B) excursions during hyperinsulinemic euglycemic clamp in African American and European American participants. The mean steady-state (final 60 min of the clamp) plasma glucose (102±6.00 mg/dL vs 103±4.53 mg/dL, p=0.20) was similar in African American versus European American participants (figure 1C and table 2). The mean steady-state plasma insulin during the hyperinsulinemic euglycemic clamp was numerically higher in African American versus European American participants (figure 1C), but the difference did not reach statistical significance (127±48.2 µU/mL vs 113±39.1 µU/mL, p=0.11). Mean values for M (glucose metabolized during euglycemic clamp) were not significantly different in African American versus European American participants (table 2). However, the mean whole-body insulin sensitivity (Si-clamp) was lower in African American versus European American participants (figure 1D). The ethnic difference in Si-clamp remained significant (p=0.038) after adjusting for sex, baseline variables age, BMI, waist circumference, FPG, 2hrPG, HbA1c, and parental diabetes history.

Request permissionFigure 1

Plasma glucose (A) and insulin (B) levels during hyperinsulinemic euglycemic clamp in African American (red symbols) and European American (blue symbols) offspring of parents with type 2 diabetes at enrollment in the Pathobiology of Prediabetes in a Biracial Cohort (POP-ABC) study. Glucose infusion rate during steady-state plasma glucose (final 60 min of the clamp procedure) was corrected for ambient steady-state plasma insulin levels (C) to derive insulin sensitivity (Si-clamp) (D). (E) The regression of acute insulin response versus Si-clamp in African American (red symbols) and European American (blue symbols) offspring of parents with type 2 diabetes. The disposition index was calculated as the product of acute insulin response and Si-clamp (F). *P=0.017.

Disposition index

Figure 1E shows the regression of AIR versus Si-clamp in African American and European American participants. The regression equations were Y=126.223*eˆ(−6.755*X) in African American offspring of parents with type 2 diabetes and Y=82.776*eˆ(−5.654*X) in European American offspring of parents with type 2 diabetes (p<0.0001). The mean disposition index was higher in African American versus European American offspring of parents with type 2 diabetes (11.0±9.09 vs 8.50±5.37; 0.027) (figure 1F, table 2). The ethnic difference in disposition index remained significant (p=0.03) after adjustment for age, sex, BMI, FPG, and HbA1c at baseline, but not after additional adjustment for baseline waist circumference and maternal/paternal diabetes history (table 2).

As already noted, among study participants who had one parent with type 2 diabetes, maternal diabetes predominated among African American participants, whereas paternal diabetes predominated among European American participants. However, we found no significant interaction between maternal/paternal diabetes status and insulin sensitivity, secretion, or clearance. We also did not observe sex differences in the measures of insulin sensitivity, secretion, or clearance among our study population.

Progression to dysglycemia

Online supplemental table S1 shows the baseline characteristics of participants who developed dysglycemia (progressors) compared with non-progressors. As was observed in the overall POP-ABC study population,29 there were significant baseline differences in age, BMI, waist circumference and blood glucose between progressors and non-progressors (online supplemental table S1) among the 268 participants in the present analysis. During 5.5 years (mean 2.62 years) of follow-up in the POP-ABC study, 91 of 268 participants developed incident dysglycemia (83 pre-diabetes (43 African American, 40 European American) and 8 type 2 diabetes (4 African American, 4 European American)). The remainder 177 participants (98 African American, 79 European American) maintained normoglycemia (non-progressors). The cumulative incidence of pre-diabetes/diabetes was 34% (32.4% in African American participants and 35.8% in European American participants). Consistent with our previous report of the main results from the POP-ABC study, the incidence of dysglycemia did not differ significantly by race/ethnicity, after adjustment for sex, age, BMI, waist circumference and blood glucose at baseline (logrank p=0.30).29

Table 3 shows baseline measures of insulin secretion, sensitivity, and clearance in progressors and non-progressors. All measures of insulin clearance were similar in progressors versus non-progressors; however, there were differences in measures of insulin action and secretion.

Table 3

View inline|View popup

Baseline insulin sensitivity, secretion, and clearance by dysglycemia progression status

Insulin secretion and incident dysglycemia

In adjusted comparisons, participants with incident dysglycemia had higher baseline fasting insulin levels but lower AIR (%) (peak insulin secretion during IVGT), compared with non-progressors (1401±1074% vs 2070±1567%, p=0.0007) (table 3). The progressors versus non-progressors difference in AIR (%) was more marked in African American participants (1544±1118% vs 2784±1784%, p<0.0001) compared with European American participants (1222±1002 vs 1432±985, p=0.29) (figure 2), and lost statistical significance after adjustment for baseline age, sex, BMI, waist circumference, FPG, and HbA1c (table 3).

Request permissionFigure 2

Baseline insulin secretion (A), insulin sensitivity (B) and disposition index (C) in progressors (P) and non-progressors (NP) to dysglycemia among offspring of parents with type 2 diabetes. (D) Kaplan-Meier plot of dysglycemia survival probability stratified by baseline tertiles of disposition index. Higher baseline disposition index was associated with increased dysglycemia survival probability (logrank p=0.015). *P=0.042; **p=0.015; ***p=0.0022; ****p<0.0001. AIR, acute insulin response.

Insulin sensitivity and incident dysglycemia

The Matsuda index of insulin sensitivity was similar in progressors to dysglycemia versus non-progressors (table 2). However, baseline Si-clamp and M values were lower in progressors versus non-progressors in unadjusted but not adjusted comparisons (table 3). The finding of difference in Si-clamp between progressors and non-progressors was less evident in African American participants (0.114±0.074 vs 0.133±0.065, p=0.17) compared with European American participants (0.119±0.059 vs 0.158±0.061, p=0.002) (figure 2).

Disposition index and incident dysglycemia

The baseline disposition index (insulin secretion adjusted for insulin sensitivity) was lower in progressors to dysglycemia (7.89±5.41) versus non-progressors (11.1±8.55, p=0.005) (table 3). The lower disposition index in progressors versus non-progressors was consistent in African American (8.87±6.07 vs 12.7±9.06, p=0.042) and European American (6.82±4.41 vs 9.63±5.69, p=0.015) participants (figure 2C), and persisted after adjustment for sex, age, BMI, waist circumference, FPG, and HbA1c at baseline (table 3). In relative terms, baseline disposition index was significantly lower by 29.2% in European American progressors versus non-progressors and by 30.1% in African American progressors versus non-progressors.

In Cox proportional hazards models, baseline AIR (%) predicted incident dysglycemia with an HR (per 1 SD change) of 0.695 (95% CI 0.540 to 0.896), p=0.005, adjusted for age, sex and race/ethnicity. Specifically, there was no interaction by race/ethnicity in the relationship between AIR (%) and incident dysglycemia (p=0.34). In the same models, insulin sensitivity had an HR (per 1 SD change) of 0.700 (95% CI 0.546 to 0.896), p=0.005 (interaction by race, p=0.34); and baseline disposition index had an HR (per 1 SD change) of 0.692 (95% CI 0.516 to 0.928), p=0.005 (interaction by race, p=0.48) (online supplemental table S2). After additional adjustment for BMI, the HR (per 1 SD change) for insulin sensitivity (0.858 (95% CI 0.651 to 1.131), p=0.28) and insulin secretion (0.783 (95% CI 0.609 to 1.007), p=0.056) became attenuated, indicating that the effects were mediated, at least in part, by adiposity. However, the association of baseline disposition index with incident dysglycemia remained significant after additional adjustment for BMI (HR (per 1 SD change) 0.700 (95% CI 0.517 to 0.947), p=0.02) (interaction by race, p=0.80). Figure 2D shows Kaplan-Meier plot of dysglycemia survival probability stratified by tertiles of baseline disposition index. Higher baseline disposition index was associated with decreased probability of developing dysglycemia, adjusted for age, sex, BMI and race/ethnicity (logrank p=0.0186).

Insulin clearance and incident dysglycemia

Neither the clearance of endogenous insulin (in the fasting state and during IVGT) nor that of exogenous insulin infused during euglycemic clamp (MCR) differed significantly in progressors versus non-progressors (table 3). However, in Cox proportional hazards models, adjusted for age, sex, BMI and race/ethnicity, the HR (per 1 SD change) for incident dysglycemia was 1.372 (95% CI 1.056 to 1.781, p=0.018) for basal insulin clearance and 1.339 (95% CI 1.050 to 1.707, p=0.019) for dynamic insulin clearance (online supplemental table S2). After additional adjustments for FPG, fasting plasma insulin, insulin sensitivity, and insulin secretion, the HR (per 1 SD change) for incident dysglycemia was 1.741 (95% CI 1.231 to 2.464, p=0.0017) for basal insulin clearance and 1.850 (95% CI 1.204 to 2.844, p=0.005) for dynamic insulin clearance (online supplemental table S2). The MCR for exogenous insulin infused during euglycemic clamp was not a significant predictor of dysglycemia in unadjusted (HR 0.861 (95% CI 0.687 to 1.079), p=0.19), minimally adjusted (HR 0.859 (95% CI 0.682 to 1.081), p=0.20), or fully adjusted (HR 0.950 (95% CI 0.747 to 1.209), p=0.68) models.

Multicollinearity diagnostics28 determined that the largest condition index (conditional number) was 17 for baseline AIR (%), 19 for Si-clamp and 17 for disposition index in the regression models predicting incident dysglycemia. A condition number between 10 and 30 indicates the presence of multicollinearity and a number larger than 30 indicates strong multicollinearity in the model. Thus, the condition numbers of 17–19 associated with our regression models indicate moderate multicollinearity among AIR (%), Si-clamp, and disposition index, as predictors of incident dysglycemia.