Sickle Cell Trait, European Ancestry, and Longitudinal Tracking of HbA1c Among African Americans: The Jackson Heart Study.

Studies of differences in glycosylated hemoglobin (HbA1c) have defined race/ethnicity as a social construct (1,2), not using objective biological parameters, and thus comparisons may have been confounded (3). Among blacks without diabetes, we investigated whether the proportion of European genetic ancestry (PEA) mediates the sickle cell trait (SCT) and HbA1c relation and whether PEA tracks HbA1c levels over time accounting for SCT.

We used data from three Jackson Heart Study (JHS) visits (2000–2004, 2005–2008, and 2009–2013), occurring at 4-year intervals. We excluded participants with diabetes (use of glucose-lowering medication or fasting blood glucose [FBG] ≥126 mg/dL) or prediabetes (FBG ≥100 mg/dL) at visit 1. We estimated PEA using 1,444 ancestry-informative markers. The rs334 variant defined SCT, and rs1050828 indicated glucose-6-phosphate dehydrogenase (G6PD) deficiency. HbA1c was assessed by high-performance liquid chromatography (Tosoh analyzer; assays coefficient of variation 1.4–1.9%).

Using visit 1 data, we estimated three relations (Fig. 1) corresponding to regressions of 1) SCT on HbA1c (c), 2) SCT on PEA (a), and 3) both SCT and PEA on HbA1c levels (c′). The indirect effect of PEA on HbA1c was estimated using the difference between c and c′. Complete mediation is observed when c′ = 0 and partial mediation when c − c′ ≠ 0 (Fig. 1).

Figure 1

A: Schematic conceptual framework for the mediation analysis investigating the relations among SCT, PEA, and HbA1c. B: Adjusted mean of HbA1c across Jackson Heart Study visits by SCT and PEA status across the three study visits. C: Adjusted mean of HbA1c by SCT and PEA status among Jackson Heart Study participants remaining free of diabetes across the three study visits.

We used mixed models to test longitudinal changes in HbA1c and their differences between groups (defined by PEA and SCT). We also conducted analyses restricted to 1) participants free of diabetes across the three visits or 2) individuals without G6PD deficiency. In all models, we adjusted for baseline age, sex, educational attainment, current smoking, FBG, HOMA of insulin resistance, and ferritin levels.

At enrollment (visit 1, n = 3,569), mean ± SD age was 53 ± 13 years (63% women), mean HbA1c was 5.45% ± 0.45%, and mean FPG was 87.8 ± 6.2 mg/dL.

Using visit 1 data, we observed no association between PEA and SCT (a in Fig. 1) (βPEA [SE] = −0.009 [0.007], P = 0.23). There was a significant association between SCT and HbA1c (c) (βSCT [SE] = −0.18 [0.03], P < 0.001), which was not affected by PEA (c′) (βPEA [SE] = −0.19 [0.10], P = 0.089). Because c − c′ = 0, PEA has no mediating effect on the SCT and HbA1c relation.

Without accounting for SCT, HbA1c did not differ between the low PEA and high PEA groups (difference 0.055 [SE 0.036], P = 0.120) across visits. Differences in HbA1c between high-PEA and low-PEA individuals became apparent after stratification by SCT (Fig. 1). Across the three study visits, individuals with low PEA had significantly higher HbA1c compared with individuals with high PEA in the non-SCT group (difference 0.06 [SE 0.01], P = 0.005). In the SCT group, there was no significant HbA1c difference between the high-PEA and low-PEA groups (difference 0.02 [SE 0.04], P = 0.750). Similar results were observed among participants free of diabetes across the three visits in the non-SCT and SCT groups (Fig. 1). Among non-SCT individuals, HbA1c was higher in the low-PEA group versus the high-PEA group (difference 0.06 [SE 0.01], P = 0.008). After excluding individuals with G6PD deficiency, the results remained unchanged in the non-SCT and SCT groups. In the non-SCT group, low-PEA individuals had higher HbA1c compared with high-PEA individuals (difference 0.05 [SE 0.01], P = 0.011).

Across visits and PEA status, HbA1c was significantly lower in the SCT group versus the non-SCT group (difference 0.23 [SE 0.04], P < 0.001). Similar results were observed among individuals without diabetes at all three visits (difference 0.22 [SE 0.04], P < 0.001) or without G6DP deficiency (difference 0.22 [SE 0.04], P < 0.001).

Among blacks without diabetes, we observed higher HbA1c levels among individuals with low PEA versus those with high PEA in the non-SCT group, but not in the SCT group. This suggests independent PEA and SCT effects, with a predominance of the SCT effect. Our findings are consistent with prior evidence of differences in HbA1c between blacks and whites (1), including the effect of SCT status (4). However, our investigation differs from prior studies, which examined SCT only (5) or genetic ancestry only (3), seldom accounted for G6PD (3,5), included individuals with prediabetes or diabetes, and did not track HbA1c over time (3–5).

The inclusion of individuals from a single racial/ethnic group but with a heterogeneous ancestry allowed a better disaggregation of ancestry-specific HbA1c-related genetic determinants from the social factors that correlate with race/ethnic groupings. The observed differences among individuals without diabetes or prediabetes suggest that PEA captures nonglycemic factors. The PEA effect was explained by neither SCT nor G6PD.