Comparative prognostic value of hemoglobin glycation index and glycation gap for mortality risk in adults: A population-based study

Abstract

Objective

This study aimed to compare hemoglobin glycation index (HGI) and glycation gap (GG), markers of the discordance between measured and predicted hemoglobin A1c levels, for predicting all-cause and cardiovascular disease (CVD) mortality in a nationally representative population.

Methods

Data of 3468 adults were retrieved from the 1999–2004 US National Health and Nutrition Examination Survey. Participants were stratified into four groups based on the median of absolute values of HGI and GG. Associations between HGI, GG and mortality risk were evaluated with Cox proportional hazard model and time-dependent receiver-operating characteristic curves.

Results

Over a median follow-up of 101 months, 733 (12.48%) participants died, of which 210 were from CVD causes. Compared to the HGI-/GG-group, the HGI+/GG+ group had an hazard ratio (95% confidence interval) of 1.36 (1.02−1.82) for all-cause mortality and 1.91 (1.00−3.64) for CVD mortality. Restricted cubic spline curves demonstrated a positively linear relationship between absolute values of HGI, GG and mortality risk. Time-dependent receiver-operating characteristic curves revealed comparable predictive accuracy for HGI and GG, with area under the curve ranged between 0.50 and 0.60 across follow-up periods.

Conclusions

Combined HGI and GG assessment may provide guidance on risk stratification and identification of high-risk individuals.

Keywords

all-cause mortality glycation gap hemoglobin glycation index NHANES receiver-operating characteristic curve

Introduction

The glycated hemoglobin A1c (HbA1c) is a main subcomponent of total glycated hemoglobin and represents the most widely used indicator of glycemic control over two to 3 months in patients with diabetes. The American Diabetes Association recommends HbA1c ≥ 6.5% as a diagnostic threshold for diabetes.¹ However, discordance between HbA1c levels and other markers of glycemic control are very common in clinical practice, which might be related to factors like ethnic background, erythrocyte lifespan, intracellular glucose transport, enzymatic activity, and hemoglobinopathies.²

The hemoglobin glycation index (HGI) and the glycation gap (GG) have thus been proposed to quantify the variability in the relationship between HbA1c and other glucose indicators. The HGI specifically measures interindividual variations in hemoglobin glycation that account for disparities between fasting plasma glucose (FPG) and HbA1c levels.³ Similarly, the GG is an empirical measure of the disagreement between HbA1c and glycated albumin, a reliable biomarker of short-term glycemic control.⁴ Notably, glycated albumin offers distinct advantages as it remains unaffected by hemoglobin turnover, making it a viable alternative to HbA1c in some settings. Previous studies have demonstrated that the GG is consistent over time and showed no significant within-subject variability despite significant changes in HbA1c.⁵

Accumulating evidence has demonstrated that elevated HGI or GG values are associated with both microvascular and macrovascular complication of diabetes, such as diabetic nephropathy, retinopathy, and coronary artery calcification.^6,7 Furthermore, elevated HGI or GG levels have been independently linked to increased cardiovascular disease (CVD) incidence and mortality risk, particularly in individuals with type 2 diabetes.^8,9 Notably, while these associations have been established, a direct comparison of the prognostic value between HGI and GG for mortality risk prediction has not been conducted. Moreover, whether the coexistence of high HGI and high GG confers synergistic mortality risk remains to be elucidated. To address these knowledge gaps, this cohort study was designed to: (1) directly compare the prognostic performance of HGI and GG for predicting all-cause and CVD mortality; and (2) investigate potential synergistic effects of combined high HGI and high GG on mortality risk.

Methods

Data source and participant selection

The present study utilized data from the 1999-2000 & 2001-2002 & 2003-2004 cycles of the National Health and Nutrition Examination Survey (NHANES), a biennial nationwide survey assessing health and nutritional status of non-institutionalized US population. These specific cycles were chosen based on the availability of glycated albumin measurements. Detailed survey methodology has been described elsewhere.^10,11

To ensure data quality, we restricted our analysis to fasting participants to meet the requirements for valid FPG measurements. The exclusion criteria were: (1) age <20 years; (2) missing data for FPG, or HbA1c, or glycated albumin; (3) pre-existing CVD, including self-reported coronary artery disease, congestive heart failure, stroke, angina, and heart attack; (4) current pregnancy; (5) history of cancer; and (6) missing covariate data. Figure 1 illustrates the participant selection and exclusion process.

Figure 1.

Study flowchart showing criteria and corresponding number of participant inclusion and exclusion. FPG: fasting plasma glucose; GG: glycation gap; NHANES: National Health and Nutrition Examination Survey.

Calculation of HGI and GG

Laboratory measurements of HbA1c, FPG, and glycated albumin were obtained from the NHANES database. In brief, HbA1c was measured using the Boronate Affinity High Performance Liquid Chromatography system, with inter-assay coefficient of variation <3.0%. FPG was determined using the glucose hexokinase method, while glycated albumin was measured using the Siemens Dimension Vista 1500 (Siemens Healthcare Diagnostics).

The HGI was calculated using a two-stage approach.¹² First, we established the population-level relationship between HbA1c and FPG through linear regression analysis, yielding the equation: HbA1c = 0.02,748 × FPG (mg/dL) + 2.71,368 (R² = 0.725, p < 0.001, Supplemental Figure 1). For each participant, the predicted HbA1c value was computed by applying their measured FPG to this regression model. The HGI was then derived as the difference between measured HbA1c and model-predicted HbA1c. The restricted cubic spline curves demonstrated U-shaped associations of HGI with all-cause and CVD mortality (Supplemental Figure 2A, B). Therefore, we used absolute HGI values in this study, as both extreme negative and positive values were associated with an increased risk of mortality.

For glycated albumin percentage calculation, we applied the standard formula: [glycated albumin (g/dL)/serum total albumin (g/dL)] × (100/1.14) + 2.9. Similarly, we developed a linear regression model between HbA1c and glycated albumin: HbA1c (%) = 0.21,602 × glycated albumin (%) + 2.54,797 (R² = 0.525, p < 0.001, Supplemental Figure 3). The GG was calculated as the difference between measured HbA1c and glycated albumin-predicted HbA1c values.¹³ Mirroring the relationship observed for HGI, GG also exhibited U-shaped associations with all-cause and CVD mortality in restricted cubic spline models. Consequently, absolute values of GG were employed for subsequent analysis (Supplemental Figure 4A, B).

Mortality assessment

The primary and secondary outcome was all-cause and CVD mortality, respectively. Mortality status was ascertained through linkage with the National Death Index public-use files, with follow-up continuing through December 31, 2019. CVD mortality was identified using International Classification of Diseases, Tenth Revision codes I00-I09, I11, I13, I20-I51, and I60-I69.¹⁴

Covariates

Potential confounders were selected a priori based on established biological mechanisms and prior literature evidence.^15–17 The covariates included age (years), sex (men/women), race (Mexican American/non-Hispanic white/non-Hispanic black/other Hispanics/other), marital status (single/non-single), education level (< high school/high school/> high school), household poverty-income ratio (continuous), smoking (never/former/current), drinking (never/former/mild/moderate/heavy), body mass index (continuous), physical activity (none/moderate/vigorous), hypertension (yes/no), diabetes (yes/no), statin use (yes/no), and laboratory tests of white blood cell count (continuous), hemoglobin (continuous), aspartate aminotransferase (AST, continuous), alanine transaminase (ALT, continuous), total cholesterol (continuous), triglyceride (continuous), low-density lipoprotein cholesterol (continuous), urinary albumin-to-creatinine ratio (continuous), and estimated glomerular filtration rate (continuous) calculated from the 2009 serum creatinine-based formula.¹⁸ Definitions for smoking, drinking, and physical activity were consistent with previous reports.¹⁹ Hypertension was diagnosed based on self-report, or blood pressure measurement ≥140/90 mmHg at the Mobile Examination Center, or taking anti-hypertensive medications. Diabetes was diagnosed by self-report, or taking anti-diabetic medications, or blood measurements (FPG ≥7 mmol/L, or HbA1c > 6.5%, or 2-h post-load glucose ≥11.1 mmol/L), as appropriate.²⁰

Statistical analysis

Participants were arbitrarily stratified into four mutually exclusive groups using median splits of HGI and GG values: (1) HGI+/GG+ (both indices above median); (2) HGI+/GG− (HGI above and GG below median); (3) HGI−/GG+ (HGI below and GG above median); and (4) HGI−/GG− (both indices below median). This classification scheme allowed examination of mortality risk across distinct glycemic variability patterns while ensuring sufficient statistical power. The HGI−/GG− group served as the reference category in all analyses.

Baseline characteristics were compared across the four groups using one-way analysis of variance or chi-squared test, as appropriate, with all analyses incorporating appropriate sampling weights to account for the complex design. We applied the Cox proportional hazard regression model to assess relationships between HGI, GG and mortality risk, reporting hazard ratios (HRs) with corresponding 95% confidence intervals (CIs). Three sequential statistical models were established: the unadjusted crude model, the Model 1 (adjusted for age, sex, race, poverty-income ratio, marital status, and educational attainment), and the final Model 2 that adjusted for all covariates. To examine dose-response relationships, we implemented restricted cubic spline models with knots positioned according to the minimal Akaike Information Criterion. We also conducted subgroup analysis stratified by participant’s age (<60 vs ≥ 60 years) and sex (men vs women). Predictive performance of HGI and GG for mortality was assessed by time-dependent receiver-operating characteristic curves. Sensitivity analysis was carried out by excluding participants who died within 24 months to minimize the risk of reverse causation.²¹ Finally, we constructed nomograms for all-cause and CVD mortality, using significant predictors from the Cox regression. Model performance was evaluated by calculating the area under the curve from the receiver-operating characteristic curves. A two-sided p value <0.05 denoted statistical significance.

Results

Baseline participant characteristics

Following application of the inclusion and exclusion criteria, our analytic cohort comprised 3468 participants (mean age 43.63 years, 49.03% men, 7.67% had diabetes). Presentation and comparison of characteristics stratified by the median values of absolute HGI (0.224) and GG (0.284) were summarized in Table 1. Compared to the reference HGI−/GG− group, participants in the HGI+/GG + group were associated with significantly higher body mass index, more likely to have comorbid diabetes, lower total cholesterol, lower low-density lipoprotein cholesterol, and higher urinary albumin-to-creatinine ratio.

Table 1.

Comparison of baseline characteristics among US adults with different categories of hemoglobin glycation index and glycation gap.

	Total (n = 3468)	HGI−/GG− (n = 954)	HGI−/GG+ (n = 779)	HGI+/GG− (n = 780)	HGI+/GG+ (n = 955)	P value
Age, years	43.63 ± 0.42	43.14 ± 0.36	43.25 ± 0.73	43.84 ± 0.53	44.34 ± 0.80	0.18
Sex, n (%)						0.003
Men	1748 (49.03)	484 (48.30)	342 (40.46)	443 (57.00)	479 (51.02)
Women	1720 (50.97)	470 (51.70)	437 (59.54)	337 (43.00)	476 (48.98)
Race/ethnicity, n (%)						0.05
Non-Hispanic White	1744 (71.80)	486 (70.82)	423 (75.93)	367 (69.69)	468 (70.86)
Non-Hispanic Black	613 (10.86)	144 (10.06)	121 (9.51)	152 (11.74)	196 (12.22)
Mexican American	852 (8.33)	258 (10.74)	179 (5.71)	204 (8.20)	211 (8.22)
Other Hispanic	147 (4.38)	45 (5.09)	31 (4.59)	26 (3.23)	45 (4.39)
Other	112 (4.63)	21 (3.28)	25 (4.26)	31 (7.15)	35 (4.32)
Single marital status, n (%)	1252 (34.46)	322 (31.64)	294 (38.40)	300 (35.51)	336 (33.02)	0.20
Education, n (%)						0.37
<High school	491 (5.65)	129 (6.37)	96 (4.28)	129 (5.90)	137 (5.93)
High school	1339 (37.55)	380 (36.58)	305 (37.17)	298 (41.75)	356 (35.44)
>High school	1638 (56.79)	445 (57.05)	378 (58.55)	353 (52.35)	462 (58.63)
Poverty-income ratio	3.07 ± 0.05	3.12 ± 0.07	2.96 ± 0.09	3.07 ± 0.10	3.12 ± 0.08	0.26
Body mass index, kg/m²	28.26 ± 0.18	27.88 ± 0.29	27.96 ± 0.58	29.05 ± 0.34	28.27 ± 0.40	0.01
Smoking, n (%)						0.34
Never	1819 (51.56)	495 (49.96)	408 (52.10)	410 (52.61)	506 (51.93)
Former	864 (23.27)	241 (24.53)	179 (20.31)	189 (21.42)	255 (26.18)
Current	785 (25.17)	218 (25.51)	192 (27.59)	181 (25.97)	194 (21.89)
Drinking, n (%)						0.23
Never	469 (11.32)	125 (12.21)	119 (11.58)	96 (9.90)	129 (11.32)
Former	616 (15.21)	160 (16.34)	122 (12.31)	155 (17.02)	179 (15.13)
Mild	1166 (34.67)	340 (33.45)	245 (32.92)	258 (36.21)	323 (36.31)
Moderate	494 (16.96)	133 (15.60)	133 (22.20)	89 (13.99)	139 (16.11)
Heavy	723 (21.84)	196 (22.39)	160 (20.99)	182 (22.88)	185 (21.13)
Vigorous activity, n (%)	1119 (36.38)	315 (37.86)	255 (37.89)	251 (33.07)	298 (36.18)	0.51
Moderate activity, n (%)	1632 (55.99)	438 (56.31)	394 (56.84)	359 (57.07)	441 (53.94)	0.82
Hypertension, n (%)	1303 (31.52)	314 (26.58)	299 (32.83)	304 (33.82)	386 (33.75)	0.13
Diabetes, n (%)	329 (7.67)	26 (1.82)	62 (6.71)	60 (7.75)	181 (14.88)	<0.001
Statin use, n (%)	261 (8.19)	56 (5.47)	60 (9.42)	59 (7.89)	86 (10.28)	0.11
WBC, ×10⁹/L	6.69 ± 0.06	6.57 ± 0.08	6.63 ± 0.09	6.86 ± 0.11	6.71 ± 0.13	0.16
Hemoglobin, g/dL	14.63 ± 0.05	14.65 ± 0.08	14.46 ± 0.10	14.85 ± 0.09	14.57 ± 0.10	0.02
ALT, U/L	26.73 ± 0.68	24.88± 0.86	26.34 ± 1.60	27.55 ± 1.10	28.43 ± 1.53	0.06
AST, U/L	24.96 ± 0.43	23.73 ± 0.43	24.19 ± 0.42	25.20 ± 0.89	26.83 ± 1.19	0.04
Triglyceride, mg/dL	127.70 ± 1.83	121.01 ± 3.06	127.31 ± 3.77	133.30 ± 4.22	130.64 ± 3.48	0.08
Total cholesterol, mg/dL	197.93 ± 1.22	200.53 ± 2.15	195.64 ± 1.29	201.42 ± 2.19	194.27 ± 2.03	0.01
LDL-c, mg/dL	119.11 ± 1.03	123.45 ± 1.84	115.53 ± 1.57	122.50 ± 1.82	114.82 ± 1.64	<0.001
eGFR, ml/min/1.73 m²	96.46 ± 0.76	96.35 ± 0.71	95.71 ± 1.33	95.96 ± 0.95	97.68 ± 1.02	0.35
uACR, mg/g	16.03 ± 1.61	9.79 ± 0.83	17.01 ± 3.75	18.84 ± 4.59	19.57 ± 2.92	0.01
FPG, mg/dL	98.60 ± 0.50	94.84 ± 0.64	96.93 ± 1.07	99.40 ± 0.67	103.55 ± 1.36	<0.001
HbA1c, %	5.40 ± 0.02	5.32 ± 0.02	5.37 ± 0.03	5.33 ± 0.02	5.58 ± 0.05	<0.001
Glycated albumin, %	13.31 ± 0.09	12.89 ± 0.06	13.41 ± 0.17	12.95 ± 0.10	13.97 ± 0.18	<0.001
HGI	0.283 ± 0.005	0.109 ± 0.003	0.108 ± 0.003	0.421 ± 0.009	0.520 ± 0.015	<0.001
Glycation gap	0.372 ± 0.009	0.140 ± 0.005	0.551 ± 0.015	0.141 ± 0.005	0.656 ± 0.016	<0.001
All-cause death	733 (12.477)	177 (9.967)	167 (11.536)	171 (13.603)	218 (15.133)	0.004
CVD-cause death	210 (3.380)	40 (2.056)	58 (3.728)	47 (3.456)	65 (4.440)	0.006

ALT: alanine transaminase; AST: aspartate aminotransferase; eGFR: estimated glomerular filtration rate; FPG: fasting plasma glucose; GG: glycation gap; HGI: hemoglobin glycation index; LDL-c: low-density lipoprotein cholesterol; uACR: urinary albumin-to-creatinine ratio; WBC: white blood cell.

P denotes the comparison between the four groups.

Associations of HGI and GG with mortality outcomes

The cohort was followed for a median of 207 months (interquartile range 187–225), during which 733 deaths (12.48%) occurred, including 210 CVD deaths. As shown in Table 2, in unadjusted analyses, both HGI and GG as continuous variables showed significant positive associations with all-cause and CVD mortality. These associations persisted after full adjustment for potential confounders in Model 2, with each 1-unit increase in HGI and GG associated with a 56% and 30% higher risk of all-cause mortality, and 119% and 58% higher risk of CVD mortality, respectively. Participants with HGI+/GG + demonstrated particularly heightened mortality risk compared to the reference group, with adjusted HRs of 1.36 (95% CI 1.02–1.82) for all-cause mortality and 1.91 (95% CI 1.00–3.64) for CVD mortality.

Table 2.

Associations between HGI/GG status with mortality outcomes in US adults.

	Crude model		Model 1		Model 2
	HR (95%CI)	p	HR (95%CI)	p	HR (95%CI)	p
All-cause mortality
HGI (continuous)	1.85 (1.46–2.35)	<0.001	1.55 (1.14–2.11)	0.005	1.56 (1.15–2.11)	0.004
GG (continuous)	1.26 (1.02–1.57)	0.04	1.31 (1.08–1.59)	0.006	1.30 (1.08–1.57)	0.006
HGI/GG status categories
HGI−/GG−	1 (reference)	/	1 (reference)	/	1 (reference)	/
HGI−/GG+	1.17 (0.85–1.60)	0.34	1.08 (0.79–1.47)	0.65	1.06 (0.78–1.45)	0.70
HGI+/GG−	1.39 (0.94–2.05)	0.10	1.09 (0.71–1.65)	0.70	1.08 (0.71–1.63)	0.73
HGI+/GG+	1.53 (1.11–2.11)	0.009	1.38 (1.03–1.85)	0.03	1.36 (1.02–1.82)	0.04
P for trend	0.01		0.05		0.06
Cardiovascular disease mortality
HGI (continuous)	2.36 (1.74–3.19)	<0.001	2.10 (1.44–3.07)	<0.001	2.19 (1.52–3.15)	<0.001
GG (continuous)	1.52 (1.00–2.30)	0.05	1.59 (1.05–2.42)	0.03	1.58 (1.04–2.40)	0.03
HGI/GG status categories
HGI−/GG−	1 (reference)	/	1 (reference)	/	1 (reference)	/
HGI−/GG+	1.83 (1.13–2.98)	0.02	1.66 (1.02–2.70)	0.04	1.65 (1.01–2.69)	0.05
HGI+/GG−	1.71 (0.90–3.25)	0.10	1.31 (0.67–2.55)	0.43	1.31 (0.69–2.47)	0.41
HGI+/GG+	2.16 (1.08–4.31)	0.03	1.96 (1.05–3.69)	0.04	1.91 (1.00–3.64)	0.05
P for trend	0.06		0.10		0.12

CI: confidence interval; GG: glycation gap; HGI: hemoglobin glycation index; HR: hazard ratio.

The crude model was unadjusted. Model 1 was adjusted for age, sex, race, poverty-income ratio, marital status, and educational attainment. Model 2 was adjusted for covariates in Model 1, plus smoking, drinking, body mass index, physical activity, hypertension, diabetes, statin use, white blood cell count, hemoglobin, aspartate aminotransferase, alanine transaminase, total cholesterol, triglyceride, low-density lipoprotein cholesterol, urinary albumin-to-creatinine ratio, and estimated glomerular filtration rate.

Dose-response relationships

The restricted cubic spline curves (Figure 2) indicated that both HGI and GG were positively and significantly correlated with both all-cause and CVD mortality.

Figure 2.

Restricted cubic spline curves depicting the dose-response relationship between absolute hemoglobin glycation index (a, b) and glycation gap (c, d) with risk of all-cause and cardiovascular mortality, respectively. The solid line represents the estimated hazard ratio (HR) for all-cause or cardiovascular mortality, and the shaded areas indicate the 95% confidence intervals. The reference point (dashed line) is set at log HR = 0, which indicates a HR = 1. So for log HR >0, the corresponding HR is >1. P for nonlinearity <0.05 indicated non-linear relationships, whereas P for nonlinearity >0.05 indicated linear relationships.

Subgroup analysis

Stratified analysis indicated that participant’s age and sex did not significantly modify the associations between HGI/GG status with all-cause and CVD mortality (Supplemental Table 1).

Predictive performance of HGI and GG

Time-dependent receiver-operating characteristic analysis demonstrated modest but statistically significant predictive value of HGI and GG for all-cause mortality, with area under the curve values ranging from 0.50 to 0.60 during follow-up (Figure 3). The HGI showed area under the curve values for predicting all-cause and CVD mortality that were comparable to the GG across the entire observation period. The receiver-operating characteristic curves and areas under the curve for 5-years intervals are provided in Supplemental Figure 5.

Figure 3.

Predictive performance of hemoglobin glycation index (HGI) and glycation gap (GG) for mortality over follow-up time. Time-dependent receiver-operating characteristic curves show the area under the curve for all-cause mortality (a) and cardiovascular mortality (b). The shaded areas represent the 95% confidence intervals. An area under the curve of 0.5 indicates prediction no better than chance and 1.0 indicates perfect prediction.

Sensitivity analysis

As shown in the Supplemental Table 2, the HRs (95% CIs) of HGI and GG for all-cause and CVD mortality were all significant. Similar to the primary findings, the HGI+/GG+ group had significantly higher risk of all-cause and CVD mortality as compared to the reference group.

Nomograms

We constructed nomograms for individualized prediction of all-cause and CVD mortality (Figure 4). Each variable was assigned a score proportional to its contribution to overall risk. The sum of these scores yielded an individual’s predicted mortality probability. Notably, HGI was incorporated into both nomograms, while GG was not. The models demonstrated excellent discriminative ability, with area under the curve values at 5-years of follow-up of 0.84 (95% CI 0.80–0.87) for all-cause mortality and 0.87 (95% CI 0.81–0.94) for CVD mortality (Supplemental Figure 6).

Figure 4.

Nomogram for individualized prediction of 5-years and 10-years risk of all-cause (a) and cardiovascular (b) mortality. For a given patient, find their value on each variable axis. Draw a line straight up to the “Points” axis to assign a score for that variable. Sum the points from each variable and locate this value on the “Total Points” axis. Draw a line straight down from “Total Points” to the “5-years survival probability” axis to read the predicted probability of 5- and 10-years survival probability. By knowing the survival probability, then the risk of mortality was 1-survival probability. Note: The HGI represents the absolute value of hemoglobin glycation index. The “Linear predictor” in the nomogram is the statistical model’s underlying risk score, calculated as the weighted sum of all predictor variables. This value is mathematically transformed to produce the final probability displayed on the nomogram. The “Total Points” scale is a linear representation of this predictor. Abbreviations: ALT: alanine transaminase; CVD: cardiovascular disease; eGFR: estimated glomerular filtration rate; FBG: fasting blood glucose; HbA1C: glycated hemoglobin A1c; HGI: hemoglobin glycation index; uACR: urinary albumin-to-creatinine ratio.

Discussion

A main finding of this nationally representative study is that the absolute values of both HGI and GG are independent predictors of all-cause and CVD mortality in US adults with comparable predictive capacity. Nonetheless, the Cox proportional hazard regression analysis indicated the individuals within the HGI+/GG+ group had the highest risk of all-cause and CVD mortality risk, suggesting a synergistic effect of HGI and GG for outcome prediction. Notably, restricted cubic spline curve models showed that the absolute values of HGI and GG exhibited linear dose-response relationships with all-cause and CVD mortality. Subgroup analysis revealed that age and sex did not significantly modify the association between HGI/GG status with mortality risk. Furthermore, sensitivity analyses confirmed the robustness of these findings. The nomograms constructed could potentially aid in the individualized prediction of mortality risk. Collectively, these results demonstrate that discordance between measured and predicted glycemic levels, as indicated by HGI or GG, not only serve as standalone mortality risk indicators but also manifest synergistic predictive effects.

Although the predictive value of both HGI and GG for mortality risk has been previously established across various populations with specific medical conditions, our study provides novel insights through direct comparison of these two measures. Prior research has demonstrated U- or J-shaped associations between HGI and all-cause mortality in patients with congestive heart failure,²² ischemic stroke,²³ metabolic dysfunction-associated steatotic liver disease,²⁴ metabolic syndrome,²⁵ and hypertension,²⁶ with the inflection points identified at −0.419, 0.302, −0.056, 0.83, −0.271, respectively. These inflection points are all remarkably close to the value identified in the current study. Parallel findings have also emerged for GG, with similar U-shaped mortality patterns observed in both diabetic and general populations.^27,28 The unique contribution of our study lies in its head-to-head comparison of the predictive capabilities of HGI and GG, an approach that, to our knowledge, has not been previously undertaken.

Slightly deviating from previous studies, we employed the absolute values of HGI and GG in the current study due to their established U-shaped associations with mortality risk. The observation that the restricted cubic spline curves demonstrating a positively linear relationship between absolute values of HGI and GG with all-cause and CVD mortality rate is largely compatible with this analytic approach. The pathophysiological implications of these findings merit careful consideration. Negative HGI or GG values, denoting predicted HbA1c exceeding actual values, may indicate either prolonged undetected glycemic dysregulation, or significant glycemic variability (e.g., brittle diabetes with postprandial hyperglycemic spikes). After excluding confounders of altered hemoglobin turnover and assay limitations, this pattern predominantly represents true glycemic variability, which have been linked to increased microvascular and macrovascular complications through endothelial dysfunction, oxidative stress, and chronic low-grade inflammation.^29,30 Conversely, positive HGI or GG values, in which actual HbA1c exceeds predicted values, may reflect either persistent glucose fluctuations, or therapeutic over-treatment manifesting as frequent hypoglycemia with compensatory hyperglycemia. These bidirectional pathophysiological mechanisms collectively underscore the mortality risk associated with both positive and negative deviations in glycemic control metrics.

Another interesting finding of our study is the synergistic effect of HGI and GG in mortality prediction, despite the time-dependent receiver-operating characteristic curves indicating only moderate predictive capability for each metric individually. We are aware that there are several reports suggesting a high concordance between HGI and GG in patients with type 2 diabetes.³¹ For instance, Kim et al. in a sample of 105 patients with type 2 diabetes demonstrated that the both HGI and GG were similar in terms of direction and magnitude with a correlation coefficient of 0.722.³² Using data from continuous glucose monitoring, Joung and associates showed that the GG is highly correlated with HGI.³³ This observation is congruent with our report that the HGI-/GG+ and HGI+/GG-group had similar HRs for predicting all-cause mortality. This study significantly advances the field by extending these observations to a general population cohort, where only 7.67% participants included have diabetes. The demonstrated synergy between HGI and GG underscores the clinical utility of considering both metrics concurrently, as their combined use appears to capture distinct aspects of glycemic dysregulation that collectively enhance mortality risk stratification beyond what either measure can achieve alone.

The primary findings of this study may have several practical implications for clinical practice. First, our findings establish the absolute values of both HGI and GG as valid predictive tools when HbA1c discordance is clinically suspected. For example, when a patient’s HbA1c appears discordant with their clinical presentation (e.g., high HbA1c but normal fasting glucose, or vice versa), clinicians can calculate HGI or GG. An elevated absolute value of either metric flags a higher-risk individual, independent of the direction of the discordance. These metrics may prove particularly valuable in cases where standard HbA1c interpretation may be confounded by factors such as glycemic variability or altered red blood cell turnover. Second, the synergistic effect of HGI and GG for mortality prediction persists in the general population, suggesting combined evaluation captures complementary aspects of glycemic dysregulation. Third, our results suggest that HGI may serve as early warning indicators, potentially identifying at-risk patients before conventional markers become abnormal, enabling earlier intervention in patients with normal HbA1c but significant glycemic variability. The developed nomograms provide a tool for calculating risk and stratifying patients based on readily available clinical variables. It is noteworthy that only HGI, and not GG, was retained in the models, indicating its superior prognostic value. The identification of high-risk patients via these nomograms can guide timely clinical interventions, such as intensive lifestyle modifications or pharmacotherapy, to mitigate mortality risk.

Main strength of this study include use of nationally representative data, and adjustment for multiple relevant confounders. Moreover, direct comparison of HGI and GG provide practical evidence for the choice of glycemic control markers for the management of patients. The application of time-dependent receiver-operating characteristic analysis offers robust characterization of the predictive performance of both metrics over the study duration. In addition, the developed nomograms also offer a means for clinically assessing individualized mortality risk. Nonetheless, limitations of this study should also be acknowledged. As an observational study, our findings demonstrate association rather than causation between HGI/GG and mortality risk. Second, both HGI and GG were calculated based on a single measurement of FPG and glycated albumin, respectively, which may not fully capture long-term glycemic patterns and could be susceptible to biological variability. Third, the study subjects was drawn from the US population, which may limit the external generalizability to other populations.

Conclusions

This mortality follow-up study of a nationally representative survey demonstrated potential utility of HGI and GG for predicting all-cause mortality in US adults. In addition, patients with HGI+/GG+ had the highest risk of all-cause and CVD mortality, indicating combined HGI and GG assessment may provide superior risk stratification compared to individual metric. The clinical translation of these findings could lead to more precise identification of high-risk individuals (e.g., the HGI+/GG+ group) who might benefit from intensified monitoring and preventive interventions.

Supplemental Material

Supplemental Material - Comparative prognostic value of hemoglobin glycation index and glycation gap for mortality risk in adults: A population-based study

Supplemental Material for Comparative prognostic value of hemoglobin glycation index and glycation gap for mortality risk in adults: A population-based study by Xiaoying Sun, Ping Yu, Chun Zhang in Diabetes & Vascular Disease Research

Footnotes

ORCID iD

Chun Zhang

Ethical considerations

The NHANES protocol was approved by the National Center for Health Statistics.

Consent to participate

All participants provided informed consent.

Author contributions

XYS and PY contributed to data analysis and writing of the original manuscript. CZ contributed to critical revision. All authors have read and approved the manuscript.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data Availability Statement

The dataset for this study is publicly available at CDC website at .

Supplemental Material

Supplemental material for this article is available online.

References

American Diabetes Association . 2. Classification and diagnosis of diabetes: standards of medical care in Diabetes-2021. Diabetes Care 2021; 44(Suppl 1): S15–S33.

Cohen

Lindsell

. When the blood glucose and the HbA(1c) don’t match: turning uncertainty into opportunity. Diabetes Care 2012; 35(12): 2421–2423.

van Steen

Schrieks

Hoekstra

, et al. The haemoglobin glycation index as predictor of diabetes-related complications in the AleCardio trial. Eur J Prev Cardiol 2017; 24(8): 858–866.

Kim

Yun

Kwon

, et al. Discordance in the levels of hemoglobin A1C and glycated albumin: calculation of the glycation gap based on glycated albumin level. J Diabet Complicat 2016; 30(3): 477–481.

Nayak

Singh

Dunmore

. Potential clinical error arising from use of HbA1c in diabetes: effects of the glycation gap. Endocr Rev 2019; 40(4): 988–999.

Zhang

Wang

Zhang

, et al. Hemoglobin glycation index in relationship to the risk of cardiovascular complication in patients with type 2 diabetes: a systematic review and meta-analysis. J Diabet Complicat 2020; 34(10): 107673.

Thanh

Tran

, et al. The correlation between glycation gap and renal complications in patients with type 2 diabetes mellitus. Diabetes Metab Syndr Obes 2024; 17: 333–341.

Liang

Xie

, et al. Association between hemoglobin glycation index and risk of cardiovascular disease and all cause mortality in type 2 diabetic patients: a meta-analysis. Front Cardiovasc Med 2021; 8: 690689.

Huang

Zhong

, et al. Glycated haemoglobin index is a new predictor for all-cause mortality and cardiovascular mortality in the adults. Sci Rep 2024; 14(1): 19629.

10.

Fei

, et al. Association between glycation gap and impaired cardiorespiratory fitness: evidence from American adults. BMC Cardiovasc Disord 2025; 25(1): 119.

11.

Wang

Chen

, et al. Association of hemoglobin glycation index and glycation gap with cardiovascular disease among US adults. Diabetes Res Clin Pract 2022; 190: 109990.

12.

Hsia

Rasouli

Pittas

, et al. Implications of the hemoglobin glycation index on the diagnosis of prediabetes and diabetes. J Clin Endocrinol Metab 2020; 105(3): e130–e138.

13.

Paleari

Strollo

Guerra

, et al. Glycation gap: an additional tool for glycometabolic monitoring. Clin Chim Acta 2016; 463: 27–31.

14.

Kong

Wang

. Prediabetes phenotypes and all-cause or cardiovascular mortality: evidence from a population-based study. Endocr Pract 2025; 31(4): 486–493.

15.

Wang

Liu

, et al. Association between hemoglobin glycation index and 5-year major adverse cardiovascular events: the REACTION cohort study. Chin Med J 2023; 136(20): 2468–2475.

16.

Yang

Shangguan

Xie

, et al. Sex-specific associations between haemoglobin glycation index and the risk of cardiovascular and all-cause mortality in individuals with pre-diabetes and diabetes: a large prospective cohort study. Diabetes Obes Metabol 2024; 26(6): 2275–2283.

17.

Huang

Luo

, et al. Association of haemoglobin glycation index with all-cause and cardiovascular disease mortality in diabetic kidney disease: a cohort study. Diabetol Metab Syndr 2024; 16(1): 221.

18.

Levey

Stevens

Schmid

, et al. A new equation to estimate glomerular filtration rate. Ann Intern Med 2009; 150(9): 604–612.

19.

Min

Kong

Wang

. L-shaped associations between composite dietary antioxidant index and hearing loss: a cross-sectional study from the national health and nutrition examination survey. Biol Res Nurs 2025; 27(1): 28–36.

20.

Kong

Wang

. Prevalence of individuals with diabetes and chronic kidney disease indicated for finerenone treatment in the United States: a national health and nutrition examination Survey-based estimation. Diabetes Obes Metabol 2025; 27(1): 435–437.

21.

Strain

Wijndaele

Sharp

, et al. Impact of follow-up time and analytical approaches to account for reverse causality on the association between physical activity and health outcomes in UK biobank. Int J Epidemiol 2020; 49(1): 162–172.

22.

Sun

Yuan

Wang

, et al. Association between hemoglobin glycation index and In-Hospital all-cause mortality of patients with congestive heart failure: a retrospective study utilizing the MIMIC-IV database. Front Endocrinol 2025; 16: 1475063.

23.

Huang

Tao

, et al. Nonlinear association between hemoglobin glycation index and mortality in ischemic stroke patients: insights from the MIMIC-IV database. Diabetes Res Clin Pract 2025; 224: 112105.

24.

Zhao

Sun

Cheng

, et al. Hemoglobin glycation index and mortality risk in metabolic dysfunction-associated steatotic liver disease patients: a novel U-shaped association. Sci Rep 2025; 15(1): 1465.

25.

Zhao

, et al. Association of hemoglobin glycation index with all-cause and cardio-cerebrovascular mortality among people with metabolic syndrome. Front Endocrinol 2024; 15: 1447184.

26.

Shangguan

Yang

, et al. Association of the hemoglobin glycation index with cardiovascular and all-cause mortality in individuals with hypertension: findings from NHANES 1999–2018. Front Endocrinol 2024; 15: 1401317.

27.

Wang

, et al. Association of glycation gap with all-cause and cardiovascular mortality in US adults: a nationwide cohort study. Diabetes Obes Metabol 2023; 25(8): 2073–2083.

28.

Nayak

Nevill

Bassett

, et al. Association of glycation gap with mortality and vascular complications in diabetes. Diabetes Care 2013; 36(10): 3247–3253.

29.

Sun

Luo

Zhou

. Comprehensive elaboration of glycemic variability in diabetic macrovascular and microvascular complications. Cardiovasc Diabetol 2021; 20(1): 9.

30.

Huang

Pan

Zhou

, et al. Correlation between glycemic variability and diabetic complications: a narrative review. Int J Gen Med 2023; 16: 3083–3094.

31.

Chalew

McCarter

Thomas

, et al. A comparison of the glycosylation gap and hemoglobin glycation index in patients with diabetes. J Diabet Complicat 2005; 19(4): 218–222.

32.

Kim

Jeong

Kwon

, et al. Concordance the hemoglobin glycation index with glycation gap using glycated albumin in patients with type 2 diabetes. J Diabet Complicat 2017; 31(7): 1127–1131.

33.

Joung

Kwon

Baek

, et al. Consistency of the glycation gap with the hemoglobin glycation index derived from a continuous glucose monitoring system. Endocrinol Metab (Seoul) 2020; 35(2): 377–383.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.31 MB