A Genome-Wide Association Study Identifies Blood Disorder–Related Variants Influencing Hemoglobin A_1c With Implications for Glycemic Status in U.S. Hispanics/Latinos

Glycated hemoglobin (HbA_1c) results from nonenzymatic and mostly irreversible chemical modification by glucose of hemoglobin molecules carried in erythrocytes. HbA_1c reflects the average concentration of blood glucose over the average life span of an erythrocyte (∼3 months in humans) and indicates glycemic status over a longer period compared with fasting glucose (1). Thus, HbA_1c is used both as a measure of glycemic control and as a diagnostic criterion for diabetes (2).

HbA_1c levels are heritable, with a heritability of ∼50% (3). Previous genome-wide association studies (GWAS) have identified several diabetes- and glucose metabolism–related genetic loci associated with HbA_1c (4–7). In addition, GWAS have identified several erythrocyte-related genetic loci that are associated with HbA_1c (4,7–9). HbA_1c levels attributed to nonglycemic-related genetic variants may not reflect glycemic status, and this implication has been noted in diabetes screening and diagnosis (4,7,9,10). For example, the common G6PD variant rs1050828, which affects red blood cell (RBC) life span, was recently found to be associated with lower HbA_1c in African Americans, and it was estimated that 650,000 African Americans with diabetes would be missed when screened by HbA_1c if this genetic information was not taken into account (10). Another study reported that African Americans with sickle cell trait, who are heterozygous for an abnormal hemoglobin allele and usually have no symptoms of sickle cell disease, had lower HbA_1c levels compared with those without sickle cell trait (11).

Prior GWAS of HbA_1c levels have been largely conducted in populations of European ancestry (4,5,12) and East Asians (6,7,13). A recent transethnic genome-wide meta-analysis included individuals from other ethnic groups (e.g., African American, South Asian) (10), but no GWAS of HbA_1c levels has been conducted in U.S. Hispanics/Latinos. Understanding genetic determinants of HbA₁ in this population is of public health and clinical importance because U.S. Hispanics/Latinos, the largest minority group in the U.S., are disproportionally affected by diabetes (14) and also have poorer diabetes management compared with non-Hispanic whites (15). Moreover, the diverse U.S. Hispanic/Latino population, admixture of African, European, and Amerindian ancestries, may offer opportunities to identify novel ancestry-specific alleles affecting HbA_1c levels (16). Therefore, this study conducted a GWAS of HbA_1c in U.S. Hispanics/Latinos of diverse backgrounds using data from the Hispanic Community Health Study/Study of Latinos (HCHS/SOL) and replicated novel genetic variants associated with HbA_1c levels using data from other Hispanic/Latino cohorts. We also examined the potential implications of nonglycemic-related HbA_1c variants for glycemic status evaluated using HbA_1c.

The HCHS/SOL is a population-based study of 16,415 Hispanic/Latino adults, aged 18–74 years, living in four U.S. metropolitan areas (Bronx, NY; Chicago, IL; Miami, FL; and San Diego, CA) (17,18). A comprehensive battery of interviews relating to personal and family characteristics and health status and behaviors, and a clinical assessment with a blood draw, were conducted at an in-person baseline clinic visit during 2008–2011. The Starr County Health Study (SCHS) is a population-based study of 1,980 Mexican-American adults (aged ≥20 years) in Starr County, TX. Survey collection has been previously described (19). The Boston Puerto Rican Health Study (BPRHS) is a longitudinal cohort of 1,500 Puerto Rican adults (aged 45–75 years) living in the greater Boston, MA, area. Baseline data from this ongoing study have been described elsewhere (20). The BioMe Biobank is an ongoing hospital- and outpatient-based population research study that has enrolled more than 34,000 participants since September 2007. This is a electronic medical records–linked biobank that integrates research data and clinical care information for consented patients at the Mount Sinai Medical Center, which serves diverse local communities of upper Manhattan (21).

The analysis excluded participants with diabetes (self-reported, on antihyperglycemic medications, HbA_1c ≥6.5% [48 mmol/mol], fasting glucose ≥200 mg/dL, or glucose ≥200 mg/dL after oral glucose tolerance test [OGTT]), those with self-reported history of major blood abnormalities (self-reported, if known), and those who had received a blood transfusion 3 months before HbA_1c measures. The GWAS of HbA_1c levels described here included 9,636 U.S. Hispanics/Latinos from the HCHS/SOL (discovery study) and 4,777 U.S. Hispanics/Latinos from three replication studies, the SCHS (n = 395), the BPRHS (n = 832), and the BioMe (n = 3,550). Characteristics of study participants are reported in Supplementary Table 1. The study was approved by the Institutional Review Boards at all participating institutions, and all participants gave written informed consent.

In the HCHS/SOL, HbA_1c was measured in EDTA whole blood using a Tosoh G7 automated high-performance liquid chromatography analyzer (Tosoh Bioscience, San Francisco, CA). In the SCHS, HbA_1c was measured in EDTA whole blood using the DCA Vantage Analyzer point of care device (Siemens, Malvern, PA) following standard protocols. In the BPRHS, a Tosoh G7 automated high-performance liquid chromatography analyzer was used to measure HbA_1c. In the BioMe, data on HbA_1c were extracted from electronic medical records of participants at the time of enrollment. In the HCHS/SOL, plasma glucose (fasting and 2 h OGTT glucose) was measured using a hexokinase enzymatic method (Roche Diagnostics Corporation, Indianapolis, IN). Hemogram and platelet count were measured using a Sysmex XE-2100 instrument (Sysmex America, Mundelein, IL). Serum iron and unsaturated iron binding capacity (UIBC) were measured on a Roche Modular P chemistry analyzer using a Fe reagent kit and a UIBC reagent kit, respectively (Roche Diagnostics), and ferritin was measured in serum with Roche reagents on a Cobas 6000 Analyzer (Roche Diagnostics) using a particle-enhanced immunoturbidimetric assay. Total iron binding capacity was calculated as the sum of serum iron and UIBC, and transferrin saturation was calculated as the percentage of serum iron in total iron binding capacity. Alanine aminotransferase (ALT) and aspartate aminotransferase (AST) were measured in serum on a Roche Modular P Chemistry Analyzer using an α-ketoglutaratic enzymatic method (Roche Diagnostics). Suspected nonalcoholic fatty liver disease was defined by the elevated aminotransferase levels as AST >31 IU/mL or ALT >40 IU/mL for men and AST or ALT >31 IU/mL for women (22).

In the HCHS/SOL, genotyping was performed with an Illumina custom array (15041502 B3), which consists of the Illumina Omni 2.5M array (HumanOmni 2.5-8v1-1) plus ∼150,000 custom single nucleotide polymorphisms (SNPs), with the quality control performed at the HCHS/SOL Genetic Analysis Center (16). Genome-wide imputation was performed with 1000 Genomes Project (1000G) phase 1 worldwide reference panel (v3, released March 2012) using SHAPEIT2 and IMPUTE2 software, as described previously (16). In the SCHS, genome-wide SNPs were genotyped using the Affymetrix Genome-Wide Human SNP Array 6.0 and imputed to the 1000G phase 1 reference data (19). In the BPRHS, genome-wide SNPs were genotyped using the Affymetrix Axiom Genome-Wide LAT Array and imputed to the 1000G phase 1 reference data (23). In the BioMe, genome-wide SNPs were genotyped with the Illumina OmniExpressExome (BioMe-OMNI) (n = 1,867) or the Illumina Multi-Ethnic Global BeadChip (BioMe-MEGA) array (n = 1,683) and imputed to the 1000G phase 3 reference data.

Linear mixed-effects regressions were used to test genome-wide SNP-HbA_1c associations among 9,636 individuals without diabetes using data from HCHS/SOL. Correlations between individuals were accounted for by incorporating covariance matrices corresponding to genetic relatedness (kinship), household, and census block group as random effects. The model also included field center, age, sex, the first five principal components to adjust for ancestry (16), and sampling weights (24). Similar linear regression models were used to test associations between 18 potentially novel SNPs and HbA_1c (P < 1 × 10⁻⁵), adjusting for age, sex, top principal components, center, and relatedness (if appropriate), using data from the three replication cohorts. Inverse variance fixed-effect meta-analyses were performed to combine the results of replication studies and to combine these with results from HCHS/SOL.

We examined the potential implications of two variants, HBB-rs334 and G6PD-rs1050828, on the screening of hyperglycemia (prediabetes and diabetes) using HbA_1c among 10,470 HCHS/SOL participants without diagnosed diabetes (those with self-reported diabetes or antidiabetic medication use were not included in this analysis). We reestimated the effects of HBB-rs334 and G6PD-rs1050828 on HbA_1c levels, respectively, because the additional samples included the natural right-end of HbA_1c distribution, instead of cutoff at 6.5% (48 mmol/mol) in our GWAS analysis, but still without prominent influences by self-awareness or medication. Then, measured HbA_1c levels were adjusted to account for genetic variants using the re-estimated effects in the following equation:

where a hemizygous AO of rs1050828 in the X chromosome for men was coded to have two copies of the A allele. A survey logistic regression was used to compare the prevalence of hyperglycemia defined using fasting glucose, 2-h glucose, or HbA_1c between carriers and noncarriers of HBB-rs334 T allele or G6PD-rs1050828 A allele.

We also constructed an unweighted genetic risk score (GRS) based on five SNPs (ANK1-rs4737010, HK1-rs72805692, TMPRSS6-rs855791, HBM-rs145546625, and TMC6-rs2748424) that might be associated with HbA_1c through the erythrocytic pathway according to a previous GWAS of HbA_1c (10), by summing the HbA_1c-raising alleles. The associations of this GRS with HbA_1c level and prevalence of hyperglycemia were examined, taking into account the complex design of the HCHS/SOL study.

All analyses were performed using R 3.3.2 (R Foundation for Statistical Computing, Vienna, Austria) or SAS 9.4 (SAS Institute, Cary, NC) software.

In the discovery GWAS in the HCHS/SOL, we identified 12 genome-wide significant associations with HbA_1c (P < 5.0 × 10⁻⁸), 6 of which were previously known loci (G6PC2, GCK, ANK1, HK1, FN3K, and TMPRSS6) (Fig. 1 and Table 1). We then sought replication of 18 potentially novel independent SNPs with suggestive associations with HbA_1c (P < 1.0 × 10⁻⁵) in three independent studies of Hispanics/Latinos in the U.S. (Supplementary Tables 2 and 3). Of note, G6PD and HBB variants (10,11) were not known to be associated with HbA_1c when we selected SNPs for replication in this study. Four SNPs, including rs334 at HBB (P = 2.7 × 10⁻⁴⁰ in HCHS/SOL), rs145546625 at HBM (P = 8.2 × 10⁻⁹), rs145546625 at TMC6 (P = 2.0 × 10⁻¹⁵), and rs1050828 at G6PD (P = 9.6 × 10⁻¹³²), were robustly associated with HbA_1c in the combined replication analyses with nominal significance (all P ≤ 0.05) (Table 1 and Supplementary Table 2). No significant heterogeneity was observed across cohorts. Regional plots for these four novel HbA_1c loci are shown in Supplementary Fig. 1. The effect sizes of HBB rs334 (β = −0.31% [−3.4 mmol/mol] per minor allele) and G6PD rs1050828 (β = −0.35% [−3.8 mmol/mol] per minor allele) on HbA_1c were ∼10 times larger compared with other HbA_1c-related variants (Table 1).

In analyses stratified by Hispanic/Latino background, associations between these four SNPs and HbA_1c were consistent across groups with no significant heterogeneity (all P for heterogeneity ≥0.14) (Supplementary Table 4). Given that SNP rs1050828 is located at the X chromosome, we further examined the association by sex, and results were consistent between men and women. In addition, a sensitivity analysis excluding participants with iron deficiency showed similar associations between these SNPs and HbA_1c (Supplementary Table 5).

Although our lead SNP rs145546625 at TMC6 has not been reported before, three SNPs at the same locus (rs2748427, rs761772, and rs2073285) were identified in recent HbA_1c GWAS (10,12,13), reaching genome-wide significance in our study as well (Supplementary Table 6 and Supplementary Fig. 2). In the joint analyses of our lead SNP with rs2748427 or rs761772, which are in moderate-to-high linkage disequilibrium (LD) with the lead SNP (r² = 0.70, 0.53), our lead SNP showed attenuated but still suggestively significant signals (P = 2.8 ×10⁻⁵, P = 3.1 × 10⁻⁵), whereas the signals of the other SNPs became null (P = 0.59, P = 0.78) (Supplementary Table 6). The joint analysis with rs2073285 in weak LD (r² = 0.10) showed that the lead SNP was still at genome-wide significance (P = 1.1 × 10⁻¹¹), and rs2073285 was at near suggestive significance (P = 6.2 × 10⁻⁵). Expanding the conditional analysis to ∼500 kb of our lead SNP, we found a suggestive but not genome-wide significant secondary signal at rs80149164 (P = 5.3 × 10⁻⁷) (Supplementary Fig. 2).

We then examined associations of these newly identified four HbA_1c variants with diabetes-related traits, hematologic traits (25,26), iron traits (27), and liver-related traits in the HCHS/SOL to explore potential mechanisms underlying observed SNP-HbA_1c relationships. These newly identified HbA_1c SNPs were significantly associated with hematologic traits rather than glycemic traits or liver-related traits (Supplementary Table 7).

Specifically, HBB-rs334 (A-to-T, Glu7Val) is a causal mutation for sickle cell trait (heterozygous of the T allele) and sickle cell disease (homozygous of the T allele), producing abnormal β-globin in hemoglobin. In line with this, the minor T allele (HbA_1c-lowering allele) was associated with lower hematocrit (P = 1.3 × 10⁻¹⁰), mean corpuscular volume (MCV) (P = 1.1 × 10⁻²²), and mean corpuscular hemoglobin (MCH) (P = 1.3 × 10⁻⁵), and higher mean corpuscular hemoglobin concentration (P = 3.6 × 10⁻¹⁶), which are hematologic characteristics for sickle cell trait and disease. G6PD-rs1050828 (G-to-A, Val98Met) is a causal mutation for glucose-6-phosphate dehydrogenase (G6PD) deficiency, resulting in the premature breakdown of RBCs. The minor A allele was associated with lower RBC count (P = 1.4 × 10⁻¹⁹), higher MCV (P = 3.7 × 10⁻¹³), higher MCH (P = 4.9 × 10⁻¹¹), lower RBC distribution width (P = 3.7 × 10⁻²⁹), higher iron (P = 3.0 × 10⁻⁵), and higher transferrin saturation (P = 1.1 × 10⁻⁶), which are hematologic characteristics for G6PD deficiency and subsequent iron overload from chronic anemia. In addition, the minor T allele of HBM-rs145546625 was associated with lower MCV and MCH.

Given the much larger effect sizes of HBB-rs334 and G6PD-rs1050828 on HbA_1c compared with other HbA_1c-related variants and their nonglycemic-related features (strong associations with hematologic traits rather than glycemic traits), we then examined the influences of these two variants on hyperglycemia screening using HbA_1c.

In 10,470 participants without diagnosed diabetes (self-reported diabetes or antidiabetic medication use), 224 participants were carriers of the HBB-rs334 T allele, of which 222 individuals were heterozygous (sickle cell trait) and two individuals were homozygous of T allele (sickle cell disease) (Supplementary Table 8). Among 309 carriers of the G6PD-rs1050828 A allele in X chromosome, 91 individuals were homozygous of AA for women and AO for men, and 218 women were heterozygous. With exclusion of the two individuals with sickle cell disease by rs334 hereinafter, the A-to-T mutation of HBB-rs334 lowered HbA_1c by 0.33% (3.6 mmol/mol). The G-to-A mutation of G6PD-rs1050828 lowered HbA_1c by 0.35% (3.8 mmol/mol), with no significant difference by sex.

We then classified individuals into two groups, noncarriers and carriers of the HBB-rs334 T allele or G6PD-rs1050828 A allele. Carriers of HbA_1c-lowering alleles tended to have lower HbA_1c levels (P < 0.001) as expected, but fasting glucose and 2-h glucose levels did not differ significantly between groups (P = 0.14 and P = 0.50, respectively) (Supplementary Table 9). At the same fasting glucose levels, carriers tended to have lower measured HbA_1c than noncarriers, whereas after genetic recalibration by Eq. 1, HbA_1c levels in respect to fasting glucose became comparable between carriers and noncarriers (Fig. 2 and Supplementary Fig. 3).

Carriers and noncarriers had comparable prevalence of hyperglycemia (prediabetes and undiagnosed diabetes) defined by fasting glucose ≥100 mg/dL or 2-h glucose ≥140 mg/dL (carriers vs. noncarriers: 21.2% vs. 25.4%, 18.6% vs. 21.8%; P = 0.10, P = 0.17, respectively), whereas carriers had lower prevalence of hyperglycemia defined by HbA_1c ≥5.7% (39 mmol/mol [12.2% vs. 28.4%], P < 0.001) compared with noncarriers (Table 2). After recalibration, there was no significant difference in the prevalence of hyperglycemia defined by genetically adjusted HbA_1c between carriers and noncarriers (31.3% vs. 28.4%, P = 0.28). Results were similar but not significant when we examined the potential implications of these genetic variants in the diabetes screen using HbA_1c levels, which might be due to the small number of individuals with undiagnosed diabetes among carriers (Supplementary Table 10).

Although other nonglycemic-related genetic variants have much smaller effects on HbA_1c compared with HBB-rs334 and G6PD-rs1050828, a combined effect may have a considerable impact on HbA_1c for the hyperglycemia screening. Hence, we constructed an unweighted GRS of five potentially erythrocytic genetic variants. As expected, the GRS showed an additive effect on HbA_1c levels (Supplementary Fig. 4). We classified individuals into two groups by a cutoff of 5th percentile of the GRS. Thus, the comparison between these two groups was comparable to that between carriers (∼5% of the study population) and noncarriers (∼95% of the study population) of HBB or G6PD variants. Individuals with the bottom 5% of the GRS had lower HbA_1c levels compared with those with higher GRS (mean 5.38% [35 mmol/mol] vs. 5.50% [37 mmol/mol]; P = 1.2 × 10⁻⁷). The prevalence of hyperglycemia defined by fasting glucose or 2-h glucose was comparable between the two groups, whereas the prevalence of hyperglycemia defined by HbA_1c was lower in the lower GRS group than in the higher GRS group (21.9% vs. 28.1%) (Supplementary Table 11).

This study, the first GWAS of HbA_1c to date among Hispanics/Latinos in the U.S., showed multiple previously known as well as novel loci associated with HbA_1c at genome-wide significance level (P < 5.0 × 10⁻⁸). In particular, two blood disorder–related variants, HBB-rs334 and G6PD-rs1050828, showed ∼10-fold larger effect sizes (0.3–0.4% [3.3–4.4 mmol/mol] per allele) on HbA_1c compared with other variants (0.03–0.04% [0.3–0.4 mmol/mol] per allele). The G6PD variant was recently identified in a GWAS of HbA_1c in African Americans (10), and sickle cell trait (determined by the HBB-rs334) was reported to be associated with HbA_1c in African Americans (11).

We found that the HbA_1c level was significantly lower by 0.33% (3.6 mmol/mol; 95% CI 0.24–0.42 [2.6–4.6 mmol/mol]) per A allele of HBB-rs334 in Hispanics/Latinos, comparable to the reported difference in HbA_1c between African Americans with and without sickle cell trait (0.29% [3.2 mmol/mol]; 95% CI 0.23–0.35 [2.5–3.8 mmol/mol]) (11). The missense A allele of rs334 is observed in African ancestry at a frequency of 0.1, is very rare in Amerindian ancestry at a frequency of 0.01, and does not appear in other ancestries from 1000G (25,28), which comaps with malaria risk. The homozygotes of A allele have sickle cell anemia, a lifelong disease by chronic hemolytic anemia, caused by the rigid and sickling form of RBCs formed by polymerization of hemoglobin S especially at low oxygen concentrations (29). Although the heterozygotes (AT; sickle cell trait) are asymptomatic, they have increased complications, such as urinary tract infection, splenic infarction, or sudden death, when exposed to strenuous exercise, high altitudes, dehydration, or low oxygen levels. Lacy et al. (11) hypothesized that lower HbA_1c in people with sickle cell trait might be due to the shorter life span of the RBCs and accordingly less opportunity for glycation of hemoglobin. In line with this, our data in the HCHS/SOL demonstrated that this variant was not associated with fasting glucose but was associated with small RBCs (microcytes), characterized by lower MCV and MCH. Because microcytes may be more susceptible to oxidative stress, the life span of RBCs might be shortened (30). These data suggest that HBB-rs334 may influence HbA_1c through hematologic mechanisms independent of blood glucose. On the contrary, it has also been postulated that lower HbA_1c in individuals with sickle cell trait might be due to assay interference by hemoglobin S compared with those without sickle cell trait (11,31).

The X chromosome–linked G6PD-rs1050828 is another erythrocytic variant with a relatively large effect on HbA_1c in our study of U.S. Hispanics/Latinos (−0.35% [−3.8 mmol/mol] per A allele) as well as in a previous study of African Americans (−0.40% [−4.4 mmol/mol] per A allele for men and −0.34% [−3.7 mmol/mol] for women) (10). The missense A allele in rs1050828 is African-ancestry specific (frequency 0.13), is rarely found in Amerindian ancestry (frequency 0.01), and is not seen in other ancestries (28). G6PD (G6PD) catalyzes the reaction of the pentose phosphate pathway and in turn generates the reduced form of glutathione as an antioxidant, which protects RBCs against oxidative stress (32). G6PD deficiency reduces G6PD activity, results in a premature breakdown of RBCs, and aggravates to hemolytic anemia when triggered by certain foods (fava beans), drugs, or infection. This explains the survival advantage of G6PD deficiency against malaria because it uses RBCs as a host cell (32). However, the shortened life span of RBCs may result in lower HbA_1c levels, regardless of blood glucose levels (10). Further, the regulation via the erythrocytic pathway is supported by the significant associations with RBC count, red cell distribution width, MCH concentration, MCV, and higher iron levels, potentially from chronic hemolysis (33), but no association with other glycemic traits in our study.

TMC6-rs2748424 in the upstream of transmembrane channel-like 6 (TMC6, also known as epidermodysplasia verruciformis 1 [EVER1]) is a common variant observed across all ancestries (0.19–0.46 of minor allele frequency) (28). Three SNPs (rs2748427, rs761772, and rs2073285) near our lead SNP rs2748424 were previously identified in Japanese, European, and non-Hispanic/Latino transethnic populations (10,13). All of these three previously reported SNPs were at genome-wide significance in our Hispanic/Latino populations. SNPs rs2748427 and rs761772 showed moderate-to-high LD with our lead SNP, whereas rs2073285 showed weak LD (r² = 0.1) with our lead SNP. Of note, only our lead SNP rs2748424 remained at genome-wide significance level in the conditional analysis, suggesting a potential causal signal represented by this SNP. In addition, our lead SNP and two other SNPs (rs2748427 and rs761772) in LD were related to hematologic traits, including the immature fraction of reticulocytes (34), RBCs, MCH, and MCV (35) rather than glycemic markers such as fructosamine and glycated albumin (13,36), although there were no significant associations with hematologic traits in our study. It has been hypothesized that TMC6 genetic variants may affect HbA_1c, independent of blood glucose levels, through erythrocyte life span, iron handling, or glucose concentration difference across the erythrocyte membrane (13).

Lastly, to the best of our knowledge, HBM-rs145546625 was discovered and replicated in this study of Hispanics/Latinos for the first time. HBM-rs145546625 originated from Amerindian ancestry (frequency of T allele 0.08) (28). This Amerindian-specific variant is located upstream of hemoglobin-alpha2 (HBA2) and downstream of hemoglobin-mu (HBM), in high LD (r² > 0.99) with a splice-site variant rs148323035 in HBM (25). There is evidence suggesting that these Amerindian ancestral variants may influence HbA_1c levels through blood cell biology (25), but more study is warranted to elucidate how this variant may influence HbA_1c through the erythrocyte pathway.

The findings of the current study provide evidence supporting the significant influences of nonglycemic-related genetic variants on diabetes screening using HbA_1c tests, consistent with two previous studies (10,11). Specifically, individuals in this study carrying the HbA_1c-lowering alleles of HBB-rs334 or G6PD-rs1050828 had a lower prevalence of hyperglycemia (prediabetes and diabetes) defined by HbA_1c levels compared with noncarriers, whereas there were no differences in the prevalence of hyperglycemia defined by fasting glucose or post-OGTT glucose between carriers and noncarriers. In addition to findings in previous studies (10,11), our current analysis using genetically adjusted HbA_1c levels demonstrated that this underestimation could be recalibrated when genetic information was taken into account. Although our study did not test influences of nonglycemic-related genetic variants on the prediction of diabetes and related complications such as stroke and coronary artery disease, these potential implications have been reported previously (37).

Taken together, these findings support the current American Diabetes Association recommendations that hemoglobin variants need to be considered when evaluating the HbA_1c tests (38). However, it should be noted that the clinical interpretation of HbA_1c may need to consider conditions such as shortened erythrocyte life span, hemolytic anemia, microcytosis, iron deficiency, iron overload, vitamin B₁₂, and folate, which may impact the HbA_1c levels from the changes in the opportunity of glycation, hemoglobin levels, and glycation rate (27,33,38,39). On the other hand, the increased glycated proteins from high glucose concentrations may increase the oxidative stress or alter erythrocyte conditions (40). More investigations are needed to evaluate the potential use of genetic adjustment with HbA_1c levels among individuals without symptoms but at high risk of these blood abnormities (e.g., those with African ancestry and family history of sickle cell disease/trait).

In summary, this study found multiple genetic variants associated with HbA_1c levels in U.S. Hispanics/Latinos, particularly two African ancestry–specific variants at HBB and G6PD and an Amerindian ancestry–specific variant at HBM. Further analysis suggested that these variants may influence HbA_1c through erythrocyte-related pathways rather than glycemic-related pathways. Our findings expand the understandings of erythrocyte-related genetic determinants of HbA_1c and their implications for evaluation of HbA_1c tests of glycemic status in U.S. Hispanics/Latinos.

Acknowledgments. The authors thank the participants and staffs of HCHS/SOL, SCHS, Boston Puerto Rican Health Study, and BioMe Biobank for their important contributions.

Funding. HCHS/SOL was performed as a collaborative study supported by contracts from the National Heart, Lung, and Blood Institute (NHLBI) to the University of North Carolina (N01-HC-65233), University of Miami (N01-HC-65234), Albert Einstein College of Medicine (N01-HC-65235), Northwestern University (N01-HC-65236), and San Diego State University (N01-HC-65237). The following institutes/centers/offices contributed to the HCHS/SOL first funding period through a transfer of funds to the NHLBI: the National Institute on Minority Health and Health Disparities, the National Institute on Deafness and Other Communication Disorders, the National Institute of Dental and Craniofacial Research, the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK), the National Institute of Neurological Disorders and Stroke, and the National Institutes of Health Office of Dietary Supplements. The Genetic Analysis Center at the University of Washington was supported by NHLBI contracts (HHSN268201300005C AM03) and National Institute of Dental and Craniofacial Research contracts (HHSN268201300005C MOD03). Genotyping efforts were supported by NHLBI HSN 26220/20054C, National Center for Advancing Translational Sciences Clinical and Translational Science Institute grant UL1-TR-000123, and NIDDK Diabetes Research Center (DRC) grant DK-063491. SCHS was supported in part by Division of Intramural Research, National Institute of Allergy and Infectious Diseases grant AI-085014, NIDDK grants DK-020595, DK-073541, and DK-085501, and NHLBI grant HL-102830 from the National Institutes of Health and funds from the University of Texas Health Science Center at Houston. Genotyping services were provided by the Center for Inherited Disease Research, which is funded through a federal contract from the National Institutes of Health to The Johns Hopkins University, contract number HHSN268200782096C. BRPHS was funded by NHLBI grant P50-HL-105185 and National Institute on Aging grant P01-AG-023394. The Andrea and Charles Bronfman Philanthropies supported the BioMe Biobank program. Q.Q. is supported by NHLBI grants K01-HL-129892, R01-HL-060712, and R01-HL-140976, and by NIDDK grants R01-DK-119268 and R01-DK-120870.

The funding sources had no role in the study design or execution, data analysis, manuscript writing, or manuscript submission.

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

Author Contributions. J.-Y.M. and Q.Q. performed the literature search, analyzed and interpreted data, and drafted the manuscript. T.L.L., D.J., T.S., and T.W. performed data analyses in the HCHS/SOL. C.S., J.E.B., C.-Q.L., C.E.S., L.E.P., E.P.B., K.L.T., J.M.O., C.L.H., and R.J.F.L. contributed to replication data. M.L.A.-S., G.A.T., Y.-D.I.C., K.D.T., M.L.D., J.C., N.S., J.I.R., and R.C.K. contributed to data generation in the HCHS/SOL. Q.Q. designed the study. All authors contributed to data interpretation and editing and reviewing the manuscript. Q.Q. is the guarantor of this work and, as such, has 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 study were presented in abstract and poster form at the 77th Scientific Sessions of the American Diabetes Association, San Diego, CA, 9–13 June 2017.