The Relationship Between Time-Varying Achieved HbA_1c and Risk of Coronary Events Depends on Haptoglobin Phenotype Among White and Black ACCORD Participants

Despite the well-established positive relationship between glycated hemoglobin (HbA_1c) and cardiovascular disease (CVD), the optimal glycemic target for CVD prevention among people with type 2 diabetes remains unclear. Large randomized clinical trials that have targeted blood glucose to nearly “normal” (<6.5%) levels among participants with type 2 diabetes have not shown a significant reduction in CVD and have reported some inconsistent results for mortality (1–4). The Action to Control Cardiovascular Risk in Diabetes (ACCORD) trial investigated whether intensive blood glucose-lowering therapy (targeting glycated hemoglobin [HbA_1c] <6.0%) compared with standard therapy (targeting HbA_1c of 7.0–7.9%) reduced the risk of CVD in people with type 2 diabetes who were at high CVD risk. The trial reported an increased risk of mortality among participants randomized to receive intensive therapy compared with standard therapy and concluded that an intensive therapy strategy could not be recommended for CVD prevention in high-risk patients with type 2 diabetes (1,2). However, the average HbA_1c throughout the study in the intensive therapy group did not reach the <6.0% target. Further, unmeasured differences between individuals, such as genetics, that affect the role of blood glucose in CVD physiology could help to provide an explanation for the inconclusive results previously reported by trials of intensive glycemic control.

A common variation in the gene that codes for the abundant antioxidant plasma protein haptoglobin (Hp) occurs in ∼40% of people worldwide (5), producing an Hp protein product (the Hp2-2 phenotype) that is associated with risk of coronary artery disease (CAD), the most common form of CVD, from hyperglycemia (6,7). In individuals with the Hp2-2 genotype (who thus have the Hp2-2 phenotype) and hyperglycemia, the antioxidant function of Hp is impaired, which leads to increased susceptibility to atherosclerosis and ultimately CAD (such as fatal and nonfatal myocardial infarction) (8–10). The frequency of the Hp2-2 phenotype differs among race-based and geographic populations (5) and may help to partially explain differences in the efficacy of intensive glycemic control between study populations.

We recently reanalyzed the ACCORD study data stratified by Hp phenotype and observed that the intensive glycemic control treatment of the ACCORD study prevented CAD events only in individuals with the Hp2-2 phenotype, while participants without the Hp2-2 phenotype had no CAD benefit and had an increased risk of mortality from intensive glycemic control (11). As in the original ACCORD study, our analysis was conducted according to the intention-to-treat principle where participants were analyzed based on treatment group assignment, and so it did not account for the differences between the treatment groups’ HbA_1c target and the actual attained HbA_1c over time in response to the treatment (many ACCORD participants in the intensive glycemic control group did not meet the glycemic target of HbA_1c <6.0% [1,2]). Thus, the optimal blood glucose target for CAD prevention for either Hp phenotype remains unclear. Therefore, in the current study, we conducted a time-varying analysis of the HbA_1c concentrations achieved throughout the duration of the ACCORD study to determine the relationship between achieving specific HbA_1c targets (<6.0%, 6.0–6.4%, 6.5–6.9%, and ≥8%) and risk of incident CAD when compared with the standard therapy target (7.0–7.9%), among participants with and without the Hp2-2 phenotype separately.

The ACCORD study (ClinicalTrials.gov Identifier: NCT00000620) is a randomized, multicenter, double-blinded, two-by-two factorial trial that investigated whether intensive glucose-lowering therapy targeting a normal HbA_1c level of <6.0% versus standard glucose-lowering therapy targeting an HbA_1c level of 7.0–7.9% would reduce cardiovascular events in 10,251 middle-aged or older North American adults with long-standing type 2 diabetes (1,2). Participants were recruited within 77 clinical centers in the U.S. and Canada from January through June 2001 during the vanguard phase and then from February 2003 through October 2005 during the main recruitment phase. Participants were men and women aged 40–79 years with a history of previous clinical CVD (defined as a history of myocardial infarction, stroke, angina, coronary revascularization procedures, and/or other revascularization procedures) or aged 55–79 years without a history of previous clinical CVD, but with anatomical evidence of significant atherosclerosis, albuminuria, left ventricular hypertrophy, or at least two risk factors for CVD (e.g., obesity, hypertension, dyslipidemia, current smoking). Participants were excluded if they had a recent history of serious hypoglycemia events, were unwilling to do home glucose monitoring or inject insulin, or had a BMI >45 kg/m², serum creatinine >1.5 mg/dL, or another serious illness. Of the 10,251 participants who were enrolled, 62.4% self-identified as non-Hispanic White, 19.1% identified as Black, 7.2% identified as Hispanic, and 11.4% identified as a race other than White, Black, or Hispanic. In addition to the overarching glycemia trial, 5,518 participants were enrolled in the lipid arm and were randomly assigned fenofibrate or placebo in addition to open-label background simvastatin to raise HDL-cholesterol and lower triglyceride levels. The remaining 4,733 participants were enrolled in the blood pressure arm and were randomized to receive intensive blood pressure lowering (targeting systolic blood pressure of <120 mmHg) or standard blood pressure lowering (targeting systolic blood pressure of <140 mmHg) treatment. Further details on the rationale, design, methods, protocol, and results have been published elsewhere (1,2,12). Institutional review board or ethics committee approval was obtained for the original ACCORD study protocol at all participating institutions, and all participants provided written informed consent, including consent for future research.

Of note, the intensive glucose-lowering therapy regimen was discontinued in February 2008 due to increased rate of death from any cause in this group compared with the standard therapy group. Participants receiving intensive therapy were subsequently switched to standard therapy, and their target HbA_1c level of <6.0% was changed to 7.0–7.9% for a mean period of 1.2 years until the planned end of the trial in June 2009, resulting in a mean study duration of 4.9 years (1,2,12).

A database of the ACCORD study data for all 10,251 participants and frozen serum samples for 9,271 participants were made available to us by the National Institutes of Health Open Biologic Specimen and Data Repository Information Coordinating Center.

The Hp phenotype of ACCORD participants was determined using a validated high-throughput ELISA that can distinguish the Hp2-2 protein from the non–Hp2-2 proteins with a sensitivity of 99% and specificity of 98.1% (13). The ELISA identifies Hp phenotypes based on the differences in Hp protein size/structure (13). There is a 1:1 correspondence between Hp genotype and Hp phenotype (14).

Hp phenotype does not change over time; therefore, a blood sample from baseline or a follow-up visit was used. Of the 10,251 ACCORD participants, Hp phenotype was determined for 9,271 (90.4%). The exclusion of the other 980 participants occurred because serum samples from these participants had previously been depleted by the measurement of other biomarkers.

Participants’ HbA_1c levels were measured from blood samples every 4 months by a central laboratory that was blinded to study group assignment. We report our primary outcome of CAD events according to the original ACCORD study prespecified diagnosis criteria and definition of “major CAD events,” which is defined as the first occurrence of a fatal coronary event, a nonfatal myocardial infarction, or unstable angina. Fatal coronary events included death from myocardial infarction, congestive heart failure, invasive coronary interventions, documented arrhythmia, and death presumed to be due to coronary causes without confirmation of cause. Although the mechanism is not well understood, stroke is an end point that has been associated with the Hp1-1 phenotype rather than the Hp2-2 phenotype (15,16), suggesting that CAD and stroke should be separated from a composite CVD outcome for analyses by Hp phenotype. For this reason, the present analysis studied the primary outcome of CAD events rather than the original ACCORD study primary outcome of major CVD events.

We also report the additional three outcomes of death from any cause, severe hypoglycemia requiring medical assistance (the participant received care at a hospital, emergency department, or from medical personnel), and hypoglycemia requiring any assistance (the participant received care at a hospital, emergency department, from medical personnel, or from nonmedical personnel). Hypoglycemia events were self-reported by the participant at clinic visits. A medical panel unaware of treatment allocations adjudicated all reported cardiovascular and mortality outcomes using predefined criteria.

All statistical analyses were conducted using Stata/SE 17.0 software (StataCorp, College Station, TX) at a two-tailed α level of 0.05. Other than when testing for Hardy-Weinberg equilibrium (HWE), the common approach of dichotomizing the Hp2-2 phenotype variable to represent Hp2-2 phenotype (yes/no) (6,7,17) was used because of the low frequency of the Hp1-1 phenotype (∼15%) and the structure and function of the Hp2-1 protein being more similar to Hp1-1 than to Hp2-2 (17).

We summarized baseline characteristics, comparing the two Hp phenotype groups using t tests or Kruskal-Wallis tests for continuous variables and the χ² test for categorical variables. Less than 3% of data were missing for any baseline variables. For time-dependent variables, there was 13.8% missing for HbA_1c, 5.9% for total cholesterol, 8.8% for systolic blood pressure, and 6.5% for BMI. The last observation carried forward method was used for missing time-dependent variables. Baseline HbA_1c was missing for 18 participants (0.2%) and thus were excluded from the analysis. We further categorized the time-dependent HbA_1c data into five categories of <6.0%, 6.0–6.4%, 6.5–6.9%, 7.0–7.9%, and ≥8%. Our reasoning for choosing the these cut points is as follows: the cut points of <6.0 and 7.0–7.9% were the targets of the ACCORD study (1,2), the relationship between the Hp2-2 phenotype and CAD risk is associated with the HbA_1c cut point of ≥6.5% (7), and current guidelines suggest a less strict HbA_1c goal of <8% may be appropriate for people with a history of microvascular or macrovascular disease and long-standing diabetes (18). Given that an HbA_1c of 7.0–7.9% was the target for the control group in the ACCORD study (1,2), it served as the reference group.

Multivariable adjusted hazard ratios (aHRs) with 95% CIs estimated from Cox proportional hazards regression models with time-varying covariables were used to quantify the relationship between the time-dependent HbA_1c categories and CAD in the two phenotype groups separately. Time-varying covariables included total cholesterol, BMI, and systolic blood pressure. Time-varying covariables were used to relate the most recent measure for each of those variables to incident outcomes at the time of an event to avoid potential bias from using a single baseline measurement. Cluster variance estimates accounting for within-subject correlation of repeated measures were used. Time-independent covariables recorded at baseline included age, sex (for models not stratified by sex only), race (for models not stratified by race only), and the seven clinical center networks assignment to lipid arm; assignment to the intensive blood pressure intervention in the blood pressure trial; assignment to fenofibrate in the lipid trial; and history of previous CVD at baseline, history of heart failure at baseline, smoking at baseline (yes/no), baseline alcohol consumption (yes/no), diabetes duration (>10 years/≤10 years), family history of premature heart disease (no family history/family history of premature heart disease/family history at unknown age), education (less than high school/high school graduate/some college/college degree or higher), insulin use at baseline, and any lipid medication at baseline.

Due to the differing phenotype frequencies by race and the potential for subpopulation differences, we ran our analyses in each phenotype group overall (including participants categorized as White, Black, Hispanic, and unspecified) and then also within the two largest individual race-based groups in the study (the sample size of the other race-based groups was too small), which were White participants (63%) and Black participants (19%). We could not run the analysis in the Hispanic group alone because the sample size was too small for the models to run. We could not run the analysis in the unspecified race group alone because when diverse populations are collapsed into a single group, racial/cultural relevance is lost, the results for this group cannot be interpreted as race-based data and is not consistent with current guidelines on reporting race-based data where specific racial categories are preferred over collective terms (19,20). Further stratified analyses by sex and previous CVD at baseline were performed for the primary CAD outcome in each phenotype group separately. Interactions were tested between the HbA_1c categories as a continuous variable and Hp phenotype, and then when stratified by Hp phenotype, between the HbA_1c categories as a continuous variable and sex. Follow-up time was defined as the time from randomization to date of documented outcome or until they were censored at 7 years after randomization if no event occurred.

The distribution of Hp phenotype frequencies was 18% Hp1-1, 46% Hp2-1, and 36% Hp2-2 (data not shown) and was not in HWE. Baseline characteristics are described according to Hp phenotype (Table 1). Among participants without the Hp2-2 phenotype (n = 5,949), the mean age was 62.7 ± 6.5 years, 41.2% were women, 33.0% had previous CVD, and 61.7% were White and 22.0% were Black. Among participants with the Hp2-2 phenotype (n = 3,322), the mean age was 62.8 ± 6.4 years, 39.2% were women, 34.2% had previous CVD, and 64.2% were White and 14.1% were Black. Participants with and without the Hp2-2 phenotype had different education, history of congestive heart failure, BMI, systolic blood pressure, total cholesterol, LDL cholesterol, HDL cholesterol, triglycerides, insulin, and statin and any lipid medication use at baseline. HbA_1c distribution throughout the study was similar between phenotype groups in all participants and among White and Black participants separately (Fig. 1).

In multivariable adjusted Cox models, having HbA_1c ≥8.0% compared with 7.0–7.9% was associated with a greater risk of CAD among White (aHR 1.43, 95% CI 1.03–1.98) and Black (2.86, CI 1.09–7.51) participants with the Hp2-2 phenotype, but not when all Hp2-2 participants were combined (1.30, 0.99–1.70), and not among participants without the Hp2-2 phenotype (all P interactions >0.05) (Table 2). Having HbA_1c of any cut point value <7.0% when compared with 7.0–7.9% was not associated with a decreased risk of CAD for either phenotype group overall or among White and Black participants separately. When stratified by sex, having HbA_1c ≥8.0% compared with 7.0–7.9% was associated with a greater risk of CAD among White (aHR 1.84, 95% CI 1.01–3.38) and Black (5.19, 1.22–22.07) females but not among White (1.22, 0.82–1.82) and Black (2.35, 0.61–9.10) males with the Hp2-2 phenotype (P interaction = 0.60 and 0.74 for White and Black participants, respectively) and not among participants with the non–Hp2-2 phenotype (Table 2). Having HbA_1c <6.0% compared with 7.0–7.9% was also associated with a higher risk of CAD among White (aHR 2.11, 95% CI 1.01–4.39) females and Black (7.12, 1.04–48.9) males with the Hp2-2 phenotype, and having HbA_1c of 6.0–6.4% was associated with a higher risk of CAD among Black males with the Hp2-2 phenotype (4.40, 1.05–18.41).

For secondary outcomes, having HbA_1c <6.0% compared with 7.0–7.9% was associated with an increased risk of mortality among White participants with the non–Hp2-2 phenotype (aHR 1.56, 95% CI 1.03–2.37; P-interaction = 0.96) and an increased risk of any and severe hypoglycemia events among White and Black participants with and without the Hp2-2 phenotype (Table 3.) When compared with having HbA_1c of 7.0–7.9%, having HbA_1c of 6.0–6.4% was associated with an increased risk of any and severe hypoglycemia events among White and Black participants with the non–Hp2-2 phenotype and with any hypoglycemia events among White participants with the Hp2-2 phenotype. Having HbA_1c of 6.5–6.9% when compared with 7.0–7.9% was associated with an increased risk of any and severe hypoglycemia events among White participants with the Hp2-2 phenotype. Having HbA_1c ≥8.0% compared with 7.0–7.9% was associated with an increased risk of any and severe hypoglycemia events among Black participants with the non–Hp2-2 phenotype and with severe hypoglycemia events among White participants with the Hp2-2 phenotype.

In this study that used longitudinal time-varying achieved HbA_1c data from the ACCORD study of participants with type 2 diabetes, we found that when compared with having HbA_1c 7.0–7.9%, having HbA_1c of ≥8.0 was consistently associated with a higher CAD risk in both White and Black participants with the Hp2-2 phenotype. There was no evidence to support an HbA_1c target of <7.0% for CAD prevention for either phenotype group. When all race-based groups of participants were combined in the analysis, there were no significant findings, suggesting a confounding effect of factors related to race. Achieved HbA_1c concentrations were similar between Hp groups throughout the study, suggesting that our results were not due to differences in HbA_1c between the phenotypes and that Hp function may be the mechanism linking HbA_1c to the risk of CAD.

Our CAD findings among White and Black participants with the Hp2-2 phenotype are supported by the literature (8–10,21–24). In brief, the common Hp polymorphism is a copy number variant that gives rise to three Hp genotypes (Hp1-1, Hp2-1, and Hp2-2) that each produce a physically and functionally distinct Hp protein, thus creating three distinct Hp phenotypes (5,17). People with the Hp2-2 genotype produce an Hp protein (Hp2-2 phenotype) that is larger and less effective at removing oxidative Hb from the blood (a primary function of Hp) compared with the Hp1-1 and Hp2-1 proteins. This difference is magnified in hyperglycemia (HbA_1c ≥6.5%) when Hb is glycated, resulting in more circulating Hb-Hp complexes that are dysfunctional as antioxidants in people with the Hp2-2 phenotype (8,10,21,23,25). The Hp2:HbA_1c complexes oxidize lipoproteins and their related compounds, thus increasing susceptibility to atherosclerosis, deterioration of cardiac function, and ultimately CAD (8,9,22–24). Therefore, glycemic control may be particularly important for CAD prevention among people with the Hp2-2 phenotype to help reduce Hp2:Hb-mediated oxidative damage to blood vessels. This was what we observed in the current study when having HbA_1c ≥8.0% compared with 7.0–7.9% was associated with an increased risk of CAD among White and Black participants with the Hp2-2 phenotype but not among White and Black participants with the non–Hp2-2 phenotype.

In our previous work, we found that intensive (targeting HbA_1c <6.0%) versus standard (targeting HbA_1c of 7.0–7.9%) glycemic control reduced the risk of CAD in White participants with the Hp2-2 phenotype but had no benefit and increased the risk of mortality among those with the non–Hp2-2 phenotypes (11).The present results suggest that our earlier findings among the Hp2-2 phenotype were likely not related to participants achieving strict glycemic control (HbA_1c <6.0%), but rather were related to participants not having high HbA_1c (≥8.0%) and do not support a glycemic target of <7.0% for either phenotype. In alignment with the results from our earlier study, we also found that having HbA_1c <6.0% was associated with an increased risk of mortality among White participants without the Hp2-2 phenotype.

In a post hoc analysis of the ACCORD trial investigating the relationship between average HbA_1c and mortality, Riddle et al. (26) found persistently higher average HbA_1c, rather than low HbA_1c, was a likely contributor to the increased risk of mortality associated with intensive glycemic group in ACCORD. The results of our current study suggest potential effect modification in the original ACCORD study by Hp phenotype. Having HbA_1c <6.0% compared with 7.0–7.9% was associated with an increased risk of mortality among White participants without the Hp2-2 phenotype, which is in agreement with the finding of the ACCORD trial that intensive glycemic control targeting HbA_1c <6.0% may be associated with harm (12). When stratified by sex, the positive association between having HbA_1c ≥8.0% and CAD when compared with having HbA_1c of 7.0–7.9% among participants with the Hp2-2 phenotype was significant among White and Black females, although the interactions between sex and HbA_1c cut points were not significant. A possible explanation for this finding is that females have naturally higher levels of HDL-cholesterol (27), and thus, females with the Hp2-2 phenotype would have theoretically had higher serum concentrations of dysfunctional HDL compared with males. Concordant with this hypothesis, in a reanalysis of the ACCORD lipid arm, we previously found that fenofibrate (HDL cholesterol–raising therapy) compared with placebo increased the risk of CAD among females with the Hp2-2 phenotype (28). The finding that having lower HbA_1c (>6.0% or 6.0–6.4%) compared with having HbA_1c 7.0–7.9% was associated with increased CAD risk with Hp2-2 phenotype was unexpected and warrants further investigation, but may be related to hypoglycemia as severe hypoglycemia can trigger cardiovascular events in vulnerable patients who are at high cardiovascular risk (29).

There are social factors that likely contributed to our findings. Race is a social construct that captures differential access to power and resources and underlies persistent health inequities (30), and self-identified race is an important consideration in optimizing health care. Racial minorities in North America tend to have a higher burden of cardiovascular risk factors and worse outcomes (31). Race-based discriminatory attitudes and behaviors within health care systems contribute to suboptimal diagnosis and management of CVD among patients from minority groups, particularly Black patients, and structural inequities that disproportionally affect racialized communities lead to inequities in access to and quality of care (32,33). Further, racism is a multidimensional chronic environmental stressor that perpetuates racial inequities in CVD (33,34). Racism may have also affected whether someone was enrolled in the ACCORD study, thus inadvertently selecting subpopulations with differing risk and disease progression. In a sensitivity analysis where we compared baseline characteristics, we found that White and Black ACCORD participants differed on several characteristics (Supplementary Table 1). Additionally, Hp phenotype distribution varies according to ethnicity/geography (5), and in the current study, we saw that the two phenotype groups had different race distribution (Table 1). Thus, measured and unmeasured risk factors that differ between study populations likely contributed to our results being different when analyzed in all participants together (which included participants from additional race-based groups as well as White and Black) versus stratified into White and Black groups; therefore, stratification by race is important to consider in future Hp studies to reduce confounding related to systemic differences between populations. Hp phenotype frequencies were not in HWE overall. Hp phenotype frequencies were in HWE among White participants (11) but not among Black participants (P < 0.01), which could indicate bias in this subgroup. However, it should be noted that the assumptions underlying HWE are frequently unmet in human populations (35,36). Another limitation of this study was that most participants were White middle-aged males at high risk of CVD, limiting generalizability of results. We were unable to stratify by and account for potential differences between all the individual race populations. There may be other unmeasured confounders that could have influenced the association between HbA_1c and our outcomes. Additionally, all participants in this study were part of a glycemic control trial, and so these results may not be generalizable across populations with wider ranges of HbA_1c or among people with poor diabetes management/care. Power to detect significant associations in some of the subgroup analyses is of concern due to the small number of events, and future studies with larger sample sizes are warranted.

In summary, this study reports that having HbA_1c >8.0% compared with 7.0–7.9% was consistently associated with increased CAD risk among White and Black ACCORD participants with the Hp2-2 phenotype, while no association was observed among participants without the Hp2-2 phenotype. Further, we report that having HbA_1c <7.0% is not associated with further CAD benefit for either phenotype group and may even be associated with harm. We also report that accounting for potential differences between subpopulations by race stratification is important in studies of the Hp phenotype. Hp phenotype testing could help fine-tune study analyses while also having the clinical application of distinguishing people who would most benefit from glycemic control.

Acknowledgments. The authors sincerely thank the National Institutes of Health Open Biologic Specimen and Data Repository Information Coordinating Center for providing access to the ACCORD database and frozen serum samples for haptoglobin phenotyping, as well as the staff and participants of the ACCORD trial.

Funding. The current study was funded by a Dalhousie University Department of Medicine Ad Hoc Operating Grant (Halifax, Nova Scotia) to L.E.C., a Nova Scotia Health Research Fund Grant (Halifax, Nova Scotia) to L.E.C., and a Canadian Institutes of Health Research Project Grant (PJT173471) to L.E.C.

Duality of Interest. A.P.L. is the author of a patent owned by his university regarding use of haptoglobin genotype to predict susceptibility to cardiovascular disease in individuals with diabetes. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. L.E.C. conceived the study idea and design. L.E.C. and R.A.W. drafted the manuscript. R.A.W. performed the statistical analyses with guidance from A.S.C. rerunning all analyses in duplicate to confirm all findings. A.P.L. and O.L. determined the haptoglobin phenotype in their laboratory while blinded to participant identification and outcome. All authors contributed to additional drafts of the manuscript and approved the submitted version, and each author satisfies the authorship criteria of the International Committee of Medical Journal Editors. L.E.C. 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.