# Importance of the Hemoglobin Glycation Index for Risk of Cardiovascular and Microvascular Complications and Mortality in Individuals with Type 2 Diabetes

## Article information

## Abstract

### Background

This study investigated the prognostic importance of the hemoglobin glycation index (HGI) for macrovascular and microvascular outcomes, mortality, and hypoglycemia occurrence in a type 2 diabetes cohort and compared it to glycated hemoglobin (HbA1c).

### Methods

Baseline and mean first-year HGI and HbA1c, and the variability thereof, were assessed in 687 individuals with type 2 diabetes (median follow-up, 10.6 years). Multivariable Cox regression was conducted to evaluate the associations of HGI and HbA1c parameters with macrovascular (total and major cardiovascular events) and microvascular outcomes (microalbuminuria, advanced renal failure, retinopathy, and peripheral neuropathy), mortality (all-cause and cardiovascular), and moderate/severe hypoglycemia occurrence.

### Results

During follow-up, there were 215 total cardiovascular events (176 major) and 269 all-cause deaths (131 cardiovascular). Microalbuminuria developed in 126 patients, renal failure in 104, retinopathy in 161, and neuropathy in 177. There were 90 hypoglycemia episodes. Both HGI and HbA1c predicted all adverse outcomes, except microalbuminuria and hypoglycemia. Their adjusted risks were roughly equivalent for all outcomes. For example, the adjusted hazard ratios (HRs) with 95% confidence intervals (CIs), estimated for 1 standard deviation increments, of mean first-year HGI were 1.23 (1.05 to 1.44), 1.20 (1.03 to 1.38), 1.36 (1.11 to 1.67), 1.28 (1.09 to 1.67), and 1.29 (1.09 to 1.54), respectively, for cardiovascular events, all-cause mortality, renal failure, retinopathy, and neuropathy; whereas the respective HRs (95% CIs) of mean HbA1c were 1.31 (1.12 to 1.53), 1.28 (1.11 to 1.48), 1.36 (1.11 to 1.67), 1.33 (1.14 to 1.55), and 1.29 (1.09 to 1.53).

### Conclusion

HGI was no better than HbA1c as a predictor of adverse outcomes in individuals with type 2 diabetes, and its clinical use cannot be currently advised.

**Keywords:**Cardiovascular events; Cohort studies; Hemoglobin glycation index; Microvascular complications; Mortality; Diabetes mellitus, type 2

## INTRODUCTION

Several previous studies have shown that individuals with diabetes can exhibit persistently higher or lower serum glycated hemoglobin (HbA1c) levels than others with similar fasting blood glucose (FPG) levels [1-3]. Research suggests that inter-individual variation exists in glycation tendencies among both healthy individuals and those with diabetes [1-5], which may limit the effectiveness of using HbA1c as the sole measure of glycemic control. Consequently, the hemoglobin glycation index (HGI), which is calculated by determining the difference between an individual’s measured HbA1c and the predicted HbA1c (estimated using a population linear regression equation between HbA1c and FPG), has been proposed as an additional parameter for assessing glycemic control and as a potential risk marker for diabetes-related complications [1].

Nonetheless, the prognostic significance of the HGI remains a topic of debate. Following the pivotal Action to Control Cardiovascular Risk in Diabetes (ACCORD) trial report, which showed that individuals with type 2 diabetes and high HGI faced increased risks of mortality and hypoglycemia and did not benefit from intensive treatment compared to those with low or intermediate HGI values [6], several studies have confirmed its predictive value for cardiovascular, mortality, and renal outcomes, even after adjusting for corresponding HbA1c levels [7,8]. However, numerous earlier studies failed to demonstrate that HGI was superior to HbA1c in predicting these outcomes [9-13]. Recent meta-analyses of observational cohorts have also confirmed that a high HGI is associated with greater risks of cardiovascular and all-cause mortality, though these associations were primarily driven by elevated HbA1c levels [14,15]; thus, it appears that HGI does not offer advantages over the measured HbA1c itself. Furthermore, most previous studies only utilized baseline HbA1c and FPG measurements to calculate HGI, neglecting other measurements during follow-up, especially when antidiabetic treatments were adjusted or intensified, which could influence HGI estimates [13]. Additionally, no study has explored the prognostic significance of HGI for separate microvascular outcomes in type 2 diabetes, with the exception of renal outcomes [8,16]. Moreover, no study has conducted comprehensive interaction/sensitivity subgroup analyses for clinical-laboratory variables to determine if there are specific subgroups where HGI might be a more effective predictor of adverse outcomes than the corresponding measured HbA1c.

Hence, with these data in mind, we aimed to investigate the prognostic significance of baseline and mean first-year HGI, as well as its variability, for macrovascular and microvascular outcomes, mortality, and hypoglycemia occurrence in a type 2 diabetes cohort. Additionally, we sought to compare these findings with the respective measured HbA1c parameters. Furthermore, we explored whether there were subgroups in which the HGI parameters were more effective predictors of adverse outcomes than their corresponding HbA1c parameters by conducting interaction/sensitivity analyses.

## METHODS

### Patients and baseline protocol

This prospective study involved 687 participants from the Rio de Janeiro Type 2 Diabetes (RIO-T2D) Cohort Study, who were enrolled between August 2004 and December 2008. These individuals were reexamined annually until December 2019 at the diabetes outpatient clinic of our tertiary-care university hospital. The study received approval from the Research Ethics Committee of the School of Medicine and University Hospital, Federal University of Rio de Janeiro (number 124/2004), and all participants provided written informed consent. Detailed descriptions of this cohort, the inception methods, and the diagnostic definitions have been previously published [17-20]. In summary, the inclusion criteria for the RIO-T2D cohort consisted of all adults with type 2 diabetes up to 80 years old who had either any microvascular (retinopathy, nephropathy, or neuropathy) or macrovascular (coronary, cerebrovascular, or peripheral artery disease) complication, or at least two other modifiable cardiovascular risk factors (hypertension, dyslipidemia, or smoking). The exclusion criteria included morbid obesity (body mass index [BMI] ≥40 kg/m^{2}), advanced renal failure (serum creatinine >180 μmol/L or estimated glomerular filtration rate [eGFR] <30 mL/min/1.73 m^{2}), or the presence of any serious concomitant disease limiting life expectancy.

All participants underwent a standard baseline protocol that included a thorough clinical laboratory evaluation. The diagnostic criteria for chronic complications of diabetes have been detailed in previous publications [17-20]. Briefly, coronary heart disease was diagnosed using clinical and electrocardiographic criteria, or by positive ischemic stress tests. Cerebrovascular disease was diagnosed based on history and physical examination, and peripheral arterial disease was identified with an anklebrachial index <0.9. The diagnosis of nephropathy required at least two instances of albuminuria exceeding 30 mg/day or a confirmed reduction in glomerular filtration rate (eGFR ≤60 mL/min/1.73 m^{2}, estimated by the Chronic Kidney Disease Epidemiology Colaboration [CKD-EPI] equation, or serum creatinine >130 μmol/L). Peripheral neuropathy was assessed through clinical examination (including knee and ankle reflex activities, feet sensation using the Semmes-Weinstein monofilament, vibration with a 128-Hz tuning fork, pinprick, and temperature sensations) and neuropathic symptoms were evaluated using a standard validated questionnaire [18]. Blood pressure (BP) in the clinic was measured three times using a digital oscillometric BP monitor (HEM-907XL, Omron Healthcare, Kyoto, Japan) with a suitably sized cuff on two separate occasions 2 weeks apart at the start of the study. The first measurement of each visit was discarded, and the BP considered was the average of the last two readings from each visit. Arterial hypertension was diagnosed if the mean systolic BP was ≥140 mm Hg or diastolic BP was ≥90 mm Hg, or if antihypertensive medications had been prescribed. Laboratory evaluations included FPG, HbA1c, serum creatinine, and lipids. Albuminuria was assessed in two non-consecutive sterile 24-hour urine collections.

### HGI calculation

HGI was calculated as previously proposed [1]. We conducted a linear regression analysis to examine the relationship between baseline HbA1c and FPG across the entire cohort. The resulting linear equation was HbA1c (%)=0.015×FPG (mg/dL)+5.6, with a correlation coefficient of 0.55. To estimate the predicted baseline HbA1c, we inserted the baseline FPG values into this equation and then calculated the baseline HGI by subtracting the predicted HbA1c from the observed HbA1c. Consequently, individuals with a high HGI exhibit a higher HbA1c than expected based on their FPG levels, whereas those with a low HGI show a lower HbA1c than predicted. Given that HGI may particularly change during the first year of follow-up due to medication adjustments and the initiation of insulin [12], we also evaluated the mean first-year HGI. This was done by calculating the HGI for each of the four HbA1c-FPG pairs measured during the first year and then averaging these values. The linear equation for the mean first-year HbA1c and FPG was HbA1c (%)=0.019×FPG (mg/dL)+5.0, with a correlation coefficient of 0.59. Additionally, we assessed the variability in HGI and HbA1c during the first year by calculating the standard deviation of these four measurements. Thus, we obtained measures of baseline and mean first-year HGI and HbA1c, along with their respective variabilities during the first year of follow-up.

### Follow-up and outcome assessment

All participants were followed up regularly, at least three to four times a year, until December 2019, receiving standardized treatment. The observation period was defined as the number of months from cohort entry to either the last clinical visit in 2019 or the occurrence of the first endpoint, whichever occurred first. The primary outcomes included the occurrence of any macrovascular or microvascular events and mortality. Macrovascular outcomes encompassed total cardiovascular events (CVEs; fatal or non-fatal myocardial infarctions [MIs], sudden cardiac deaths, new-onset heart failure, death from progressive heart failure, any myocardial revascularization procedure, fatal or non-fatal strokes, any aortic or lower limb revascularization procedure, any amputation above the ankle, and deaths from aortic or peripheral arterial disease), major adverse cardiovascular events (MACEs; non-fatal MIs and strokes plus cardiovascular deaths), and all-cause and cardiovascular mortality [17]. Cardiovascular and mortality outcomes were adjudicated from medical records (most non-fatal and fatal in-hospital events occurred at our hospital), death certificates, and interviews with attending physicians and patient families, using a standard questionnaire reviewed by two independent observers [17]. Microvascular outcomes included the development or worsening of retinopathy [19], renal outcomes [20] (new microalbuminuria development or progression to macroalbuminuria, and renal function deterioration defined as a doubling of serum creatinine or end-stage renal failure requiring dialysis or resulting in death from renal failure), and the development or worsening of peripheral neuropathy [18]. Microvascular outcomes were evaluated annually. Information on hypoglycemia occurrence during follow-up was available for 496 participants (72% of the entire cohort). Symptomatic moderate-severe hypoglycemia was defined as a symptomatic episode of confirmed hypoglycemia (blood glucose <50 mg/dL) where the individual could not self-treat.

### Statistical analysis

In the initial exploratory analysis, participants were divided into tertiles based on their mean first-year HGI, and their baseline clinical-laboratory characteristics were compared using analysis of variance, the Kruskal-Wallis test, or the chi-square test as appropriate. We assessed differences in the cumulative incidences of adverse outcomes during follow-up among tertile subgroups of HGI, HbA1c, and their variabilities using Kaplan-Meier curves with the log-rank test. The adjusted risks associated with baseline and mean first-year HGI, HbA1c, and their variabilities were examined through multivariable Cox survival analyses for each outcome. Initially, each parameter was adjusted for age and sex (model 1); subsequently, adjustments were made for additional potential confounders (model 2: age, sex, BMI [body height in neuropathy analyses], physical activity, smoking status, diabetes duration, pre-existing macrovascular and microvascular complications, serum low-density and high-density lipoprotein cholesterol, the use of insulin, statins, and aspirin, and the number of antihypertensive drugs in use). Renal outcomes were further adjusted for baseline albuminuria and eGFR. Finally, HGI and HbA1c parameters were simultaneously adjusted (model 3). Both HGI and HbA1c parameters were assessed as continuous variables, with hazard ratios (HRs) estimated for standardized 1-standard deviation increments to allow comparisons between them; they were also categorized into tertiles, with HRs estimated for the upper tertile subgroup in relation to the lower tertile. Due to the high correlation between continuous HGI and HbA1c parameters (correlation coefficients between 0.84 to 0.93), to avoid collinearity in model 3, we uncorrelated these parameters by regressing HGI parameters on their corresponding HbA1c parameter and vice versa, using the residual of one parameter as the adjustment covariate for the other [21]. For the hypoglycemia outcome, logistic regressions were used with odds ratios adjusted for the same covariates as in the Cox analyses. In analyses of first-year mean HGI and HbA1c parameters, participants who reached any of the endpoints during the first year of follow-up were excluded. We assessed the predictive performance of each fully-adjusted model—whether with HGI, HbA1c, or both parameters—for each outcome by calculating the C-statistics of each model (analogous to the area under the curve [AUC] applied to time-to-event analysis) and compared them using the method proposed by DeLong [22]. Finally, we conducted interaction-sensitivity subgroup analyses for age (>/≤60 years), sex, diabetes duration (≥/<8 years), presence/absence of macro- and microvascular complications at baseline, and poor/good glycemic control during the first year of follow-up (mean HbA1c ≥/<7.5%), to determine whether there were any subgroups in which HGI parameters were better predictors of adverse outcomes than their corresponding HbA1c parameters. All statistics were performed using SPSS version 19.0 (IBM Corp., Armonk, NY, USA), and a 2-tailed P value <0.05 was considered significant.

## RESULTS

### Baseline characteristics and outcome incidence during follow-up

Table 1 displays the primary baseline clinical-laboratory characteristics of all 687 individuals with type 2 diabetes who were evaluated, as well as those categorized by tertiles of mean HGI during the first year of follow-up. The individuals in the upper tertile HGI subgroup were younger, predominantly female, and had a longer duration of diabetes compared to those in the lower tertile subgroup. Additionally, they exhibited higher prevalence rates of retinopathy and nephropathy, primarily due to increased albuminuria, and were more likely to use insulin and less likely to use sulfonylureas. As anticipated, their glycemic control was poorer than that of individuals in the lower tertile subgroups.

Over a median follow-up period of 10.6 years (interquartile range, 6.3 to 13.2), corresponding to 6,596 person-years, there were 215 total CVEs, including 176 MACEs, and 269 all-cause deaths, 131 of which were due to cardiovascular diseases. Additionally, microalbuminuria developed in 126 participants, advanced renal failure occurred in 104, retinopathy developed or worsened in 161, and peripheral neuropathy developed or worsened in 177. The incidence rates of all adverse outcomes, except for the development of microalbuminuria, were higher among individuals in the upper tertile subgroup of the mean HGI than among those in the lower tertile subgroup (Table 1, bottom). This observation was supported by the Kaplan-Meier estimation of the cumulative incidence of endpoints during the follow-up, as illustrated in Figs. 1, 2. During the follow-up, 90 individuals experienced at least one episode of symptomatic moderate to severe hypoglycemia, which occurred marginally more frequently (23.9% vs. 13.3%, *P*=0.05) in the upper tertile HGI subgroup than in the lower tertile subgroups.

### Adjusted risks associated with HGI parameters and comparisons with their corresponding HbA1c parameters

Figs. 1, 2 illustrate the cumulative incidences of adverse outcomes within tertile subgroups based on average HbA1c levels during the first year of follow-up. It is evident that there was no distinct advantage in using tertile subgroups to differentiate between higher incidences of outcomes when comparing mean first-year HGI and HbA1c, either for cardiovascular/mortality outcomes (Fig. 1) or for microvascular outcomes (Fig. 2). Similar observations were made for baseline HGI and HbA1c (Supplemental Figs. S1, S2), as well as for the variabilities of HGI and HbA1c during the first year of follow-up (Supplemental Figs. S3, S4).

Table 2 displays the adjusted risks linked with baseline and mean first-year HGI and HbA1c parameters for macrovascular and microvascular outcomes, mortality, and hypoglycemia. Generally, in the fully-adjusted analyses (model 2), both HGI and HbA1c parameters were significantly associated with increased risks for cardiovascular, mortality, and microvascular outcomes, with this association being particularly pronounced for mean first-year parameters. Conversely, the adjusted risks for HGI and their corresponding HbA1c parameters were roughly equivalent, showing no distinct advantage of one over the other for any of the outcomes. This observation was reinforced when both parameters were adjusted simultaneously in model 3. Neither HGI nor HbA1c parameters were linked to an increased likelihood of symptomatic hypoglycemia during follow-up. Table 3 presents the predictive performance of each fully-adjusted model for each outcome, whether including HGI, HbA1c, or both parameters, as evaluated by the C-statistic. The AUCs for each model showed no differences across the outcomes, further supporting the predictive equivalence of HGI and HbA1c.

Supplemental Tables S1-S3 present the results of the interaction/subgroup analyses. While there were some interactions noted (for example, the adjusted risks associated with HGI and HbA1c parameters for the retinopathy outcome were higher in younger individuals, in men, and in those without pre-existing macrovascular complications compared to their counterparts), none of the subgroups evaluated showed clear superiority of HGI parameters over their corresponding HbA1c parameters for any of the adverse outcomes.

## DISCUSSION

In a long-term prospective cohort of individuals with type 2 diabetes, we demonstrated that neither baseline nor mean first-year HGI and its variability were superior to their corresponding HbA1c parameters as predictors of macrovascular (total CVEs, MACEs, cardiovascular mortality) and microvascular (microalbuminuria, advanced renal failure, retinopathy, and peripheral neuropathy) complications, all-cause mortality, or severe hypoglycemia occurrence. Furthermore, we showed that there were no differences in their overall predictive performance, as assessed by the C-statistic of each model. Additionally, interaction/subgroup analyses revealed no differences in risk prediction across any of the evaluated subgroups. These findings indicate that there was no clear superiority of HGI over their corresponding HbA1c parameter for any of the adverse outcomes. Therefore, considering the complexity of obtaining HGI and the lack of additional predictive benefit for diabetes-related outcomes compared to HbA1c, its clinical use is currently not recommended.

Several previous cohort studies have explored the relationship between baseline HGI and cardiovascular and mortality outcomes in individuals with type 2 diabetes [6,7,9-11,13]. All of these studies found that a high HGI was linked to increased risks for these outcomes. However, in most cases [9-11,13], as in our study, baseline HGI did not outperform baseline HbA1c in predicting risks. These findings were recently supported by a meta-analysis [15]. Only two earlier studies indicated some advantages of baseline HGI over HbA1c. In a *post hoc* analysis of the ACCORD trial [6], participants in the highest tertile of HGI saw no benefits from intensive treatment and experienced increased risks of mortality and hypoglycemia. In a Korean hospital-based cohort [7], baseline HGI proved to be a better predictor of future cardiovascular disease development than HbA1c. Our study not only confirmed the lack of superiority of baseline HGI over HbA1c in predicting cardiovascular and mortality outcomes but also extended this observation to include the mean first-year HGI and its variability, which have been previously unexplored. Only one previous study [23] evaluated changes in HGI over a 4-year period as a predictor of a positive coronary artery calcium score (CACS), a surrogate marker of cardiovascular disease, in people without diabetes. The subgroup with high HGI had the highest chances of having a positive CACS. However, this association was abolished when the baseline HbA1c was included in the predictive model, suggesting that the association was mainly mediated by the higher HbA1c levels of the subgroup with a persistently high HGI [23]. In that study, only the subgroup with elevated HGI had significantly higher odds of a positive CACS, independent of baseline HbA1c levels [23].

Regarding microvascular outcomes, previous analyses had been conducted solely for renal outcomes [8,16]. Both studies showed that a high HGI independently predicted the risk of renal function deterioration, even after adjusting for HbA1c and other covariates [8,16]. In a *post hoc* analysis of the Preterax and Diamicron Modified Release Controlled Evaluation (ADVANCE) trial [11], HGI predicted a composite microvascular outcome (nephropathy or retinopathy). However, it presented lower risks compared to HbA1c, and these increased risks were not sustained after further adjustment for HbA1c, aligning with findings from the Diabetes Control and Complications Trial (DCCT) report in individuals with type 1 diabetes [12]. In the current study, separate analyses were conducted for each microvascular outcome (retinopathy, renal [microalbuminuria and renal failure], and peripheral neuropathy), assessing not only baseline HGI but also mean first-year HGI and its variability. Our findings advance knowledge by demonstrating that neither baseline nor first-year HGI parameters were superior risk predictors compared to their corresponding HbA1c parameters for any of the individual microvascular outcomes.

Few previous studies have explored the differences in risk prediction associated with the HGI among various subgroups of individuals with type 2 diabetes [6,9,11]. *Post hoc* analyses of the ACCORD [6] and ADVANCE [11] trials assessed whether the effects of HGI varied according to treatment allocation (standard or intensive), yielding conflicting results. In the ACCORD trial, HGI was a predictor of all-cause mortality and hypoglycemia risk solely in the intensive treatment subgroup [6]. Conversely, in the ADVANCE trial, the treatment subgroup did not affect the impact of baseline HGI on macro- and microvascular outcomes or mortality, which was comparable to baseline HbA1c [11]. Another study compared the predictive value of HGI in individuals with and without established cardiovascular diseases at baseline, finding that HGI predicted total CVEs only in those without pre-existing cardiovascular conditions, with risks similar to those associated with HbA1c [9].

We assessed interactions and subgroups among several clinically significant characteristics (older/younger age, male/female, shorter/longer diabetes duration, better/poorer glycemic control, presence/absence of macro- and microvascular complications). Although some significant interactions were observed, they occurred in both the HGI and HbA1c parameters, and we could not identify any specific subgroup of individuals with type 2 diabetes for whom HGI parameters were clearly superior to their corresponding HbA1c parameters in predicting any adverse outcomes. Moreover, we demonstrated that the predictive performances of the HGI parameters and the corresponding HbA1c parameters were equivalent for all outcomes when analyzed by the C-statistic of each model. Overall, similar to most previous studies [9-11,13,15], the risks associated with HGI were very similar to those associated with HbA1c. Therefore, it appears that HGI may simply serve as a substitute for HbA1c. However, given that the clinical use of HGI relies on a linear regression fitting equation between HbA1c and FPG specific to the studied population, and that this equation varies across different populations and studies, there is no clear advantage to its clinical use.

This study has several limitations worth noting. First, our cohort primarily consisted of middle-aged individuals with high cardiovascular risk and long-standing type 2 diabetes, all recruited from a tertiary-care university hospital. Consequently, our findings may not be applicable to other diverse populations of people with diabetes. Second, despite adjustments for numerous potential confounding variables, it is possible that some unmeasured or unidentified confounders were still present. For instance, we did not have data on several factors that could influence HGI, such as glucose-6-phosphate dehydrogenase deficiency, fructosamine-3-kinase activity, and red blood cell glucose transport [24]. Nonetheless, a significant strength of this study is its relatively large cohort of individuals with type 2 diabetes, who were followed over a long term and underwent frequent medical evaluations. This enabled the collection of current and accurate data on clinical variables, laboratory parameters, and outcomes related to adverse events and mortality.

In conclusion, our long-term prospective cohort study of middle-aged, high-risk individuals with type 2 diabetes showed that no HGI parameter—whether assessed at baseline or during the first year of follow-up, in terms of mean values and variability—outperformed its corresponding HbA1c parameter as a predictor of risk for macrovascular and microvascular outcomes, all-cause mortality, and severe hypoglycemia occurrence. Therefore, our findings confirm that HGI does not increase the predictive performance for adverse outcomes compared to HbA1c, and its use is not recommended for improving risk prediction in type 2 diabetes.

## Supplementary Material

## Notes

**CONFLICTS OF INTEREST**

No potential conflict of interest relevant to this article was reported.

**AUTHOR CONTRIBUTIONS**

Conception or design: C.R.L.C., N.C.L., G.F.S. Acquisition, analysis, or interpretation of data: N.C.L., G.F.S. Drafting the work or revising: C.R.L.C., N.C.L., G.F.S. Final approval of the manuscript: C.R.L.C., N.C.L., G.F.S.

## Acknowledgements

This study was supported by grants from the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq, Brazil) and from the Fundação Carlos Chagas Filho de Amparo a Pesquisa do Estado do Rio de Janeiro (FAPERJ, Brazil). The sponsors have no role in study design, data collection and analysis, results interpretation, or in preparation, review and approval of the manuscript.