Sex-specific in the nonlinear associations of hemoglobin glycation index and all-cause mortality in the general US adult population: results from NHANES 1999–2018

Abstract

Background

The hemoglobin glycation index (HGI) and its association with mortality risk in the United States adults remain insufficiently understood. This research explores potential links between HGI levels and all-cause mortality using nationally representative data.

Methods

>The relationship between HGI and mortality was investigated using data from National Health and Nutrition Examination Survey (NHANES) (1999–2018) covering 19,287 U.S. adults. HGI was calculated via linear regression of glycated hemoglobin A1c (HbA1c) on fasting plasma glucose (FPG). The National Death Index was utilized to link mortality outcomes, with tracking continuing through December 31, 2019. To explore sex-specific associations, we utilized weighted Cox regression models with multivariable adjustments, along with restricted cubic splines and segmented Cox analyses. Robustness was confirmed through stratified analyses and sensitivity tests.

Results

In U.S. males, HGI showed a U-shaped association with mortality (threshold: -0.131). Below this, lower HGI was protective (hazard ratio (HR) 0.52; 95% confidence interval (CI) 0.42–0.66); above it, higher HGI increased risk (HR 1.34; 95% CI 1.16–1.55). In females, an L-shaped pattern emerged, where higher HGI below the threshold correlated with a decreased likelihood of mortality (HR 0.48; 95% CI 0.40–0.59). These sex-specific associations were verified through stratified and sensitivity analyses.

Conclusions

The cohort study observed a U-shaped pattern between HGI and all-cause mortality in U.S. males, whereas an L-shaped association was found in females. These sex-specific patterns warrant further investigation to explore clinical implications.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12986-026-01088-2.

Introduction

Hemoglobin A1c (HbA1c) and fasting plasma glucose (FPG) are essential tools for diagnosing diabetes and prediabetes [1], but HbA1c’s accuracy as a diagnostic standard is not infallible. Mismatches between HbA1c and blood glucose levels were revealed in previous studies [2–5], which were influenced by factors [6, 7] including race, age differences, genetic diversity, red blood cell longevity, and iron deficiency, and can lead to misdiagnoses and inappropriate clinical decisions.

The Hemoglobin Glycation Index (HGI), determined by the discrepancy between actual and expected HbA1c levels derived from a linear regression model, reflects glycation variability beyond blood glucose [2, 8]. Elevated HGI was significantly linked to kidney function decline, as reported in a meta-analysis (hazard ratio (HR) 1.53; 95% confidence interval (CI) 1.05–2.23) [9]. Additionally, HGI has been associated with microvascular and macrovascular complications and mortality, as shown in the ADVANCE trial [10]. Furthermore, previous studies have demonstrated associations between HGI and mortality in various populations, such as those with acute decompensated heart failure [11], critical coronary artery disease [12], diabetes mellitus (DM) and coronary artery disease [13, 14], hypertension [15], sepsis [16], diabetic diabetes-related kidney damage [17], and DM [18]. According to the study’s findings, HGI serves as a valuable biomarker for assessing health risks.

Additionally, mortality risk appears to vary by sex, likely influenced by differences in body composition and lifestyle choices [19, 20]. However, the HGI and its link to mortality across population-based cohorts, particularly when stratified by sex, remain unexplored. Therefore, this study investigates this linkage, with a focus on sex-specific differences.

Materials and methods

Study participants

The National Health and Nutrition Examination Survey (NHANES) 1999–2018 provided the data for this research, focusing on the health and nutritional landscape of the U.S. population. The National Center for Health Statistics (NCHS) Research Ethics Review Board, with written informed consent documented for all participants. Following the Strengthening the Reporting of Observational Studies in Epidemiology guidelines, 101,316 participants were initially considered. Exclusions were applied to those under 20 years old (n = 46,235) and participants with missing data in key variables, including FPG, HbA1c, follow-up data, or fasting sample weight; educational background, marital situation, and the poverty income ratio (PIR); triglycerides (TG), low-density lipoprotein cholesterol (LDL-C), fasting insulin (FINS), serum creatinine (Scr), blood urea nitrogen (BUN), total bilirubin, and uric acid (UA). Post-exclusion, the final sample comprised 19,287 participants. The process of participant selection is detailed in the flowchart presented in Fig. 1.

Fig. 1.

Flowchart of the patient selection process. Abbreviations: NHANES, National Health and Nutrition Examination Survey; FPG, fasting plasma glucose; HbA1c, glycated hemoglobin A1c; PIR, the ratio of poverty income; TG, triglycerides; LDL-C, low-density lipoprotein cholesterol; FINS, fasting insulin; Scr, serum creatinine; BUN, blood urea nitrogen; UA, uric acid

Calculation of hemoglobin glycation index

HbA1c and FPG levels from NHANES participants, each measured once at the baseline examination, were analyzed to establish their linear relationship. The derived regression model for predicting HbA1c is HbA1c = 0.027*FPG + 2.854. Estimated HbA1c levels were calculated by applying participants’ FPG values to the linear regression model. The HGI value was determined by subtracting the predicted HbA1c value from the observed HbA1c value [21]: HGI = observed HbA1c − predicted HbA1c. The cohort was stratified into four HGI quartiles (Q1–Q4), with Q2 serving as the reference group for subsequent analyses.

Outcome definitions

The National Death Index, maintained by the CDC, was used to obtain mortality data up to December 31, 2019. The primary outcome measure was all-cause mortality. Follow-up duration commenced concurrently with the initial interview date (when FPG and HbA1c levels were measured), persisting until either death occurrence or the study endpoint of 31 December 2019. After applying the inclusion criteria, 2,838 deaths were recorded over a median follow-up of 110 months.

Assessment of covariates

Potential covariates were identified from prior research [15, 22–26] and clinical significance, encompassing demographic factors (age, sex, ethnicity, educational background, marital situation, PIR), Metabolic and blood markers (FINS, total cholesterol (TC), TG, LDL-C, high-density lipoprotein cholesterol (HDL-C)), renal function indices (serum creatinine (Scr), estimated glomerular filtration rate (eGFR), blood urea nitrogen (BUN)), and antioxidant indices (total bilirubin, uric acid (UA), albumin). Participants were categorized into four racial groups: non-Hispanic white and black, Mexican American, and others [25]. The study classified education into three levels: incomplete high school, high school graduate, and post-high school education [15]. Marital situation was divided into two categories: married or cohabitating, and single (widowed, divorced, separated, or never married). The PIR was distributed into three levels: low (≤ 1.3), medium (1.3–3.5), and high (> 3.5). eGFR values were derived using the CKD-Epidemiology Creatinine Equation [27].

Statistical analysis

Following NHANES analysis guidelines, our analysis incorporated complex sampling designs and utilized sampling weights [28]. Fasting weights were factored into the weighted analysis. For the integrated analysis of NHANES surveys conducted in 1999–2000 and 2001–2002, a four-year fasting weight (WTSAF4YR) was used. For the 2003–2018 data, the two-year fasting weight (WTSAF2YR) set was used. The formula for calculating sampling weights from 1999 to 2018 is as follows: for 1999–2002, weights were calculated as 1/5 × WTSAF4YR; for 2003–2018, weights were calculated as 1/10 × WTSAF2YR. The description of continuous variables was summarized as means with standard deviations (SD) or medians accompanied by interquartile ranges (IQR), while categorical measures were outlined as unweighted counts alongside weighted proportions. Comparative analyses utilized t-tests or Kruskal-Wallis methods for quantitative data depending on distributional assumptions, with chi-square tests employed for categorical comparisons. Given the low percentage of missing data (0–8.43%), no imputation was performed.

Baseline characteristics were stratified by sex and HGI quartiles, where the lowest was Q1 and the highest was Q4 regarding HGI. After confirming the proportional hazards assumptions (Table S3-S5, Figure S2-S4), we utilized Cox hazard regression models to evaluate the association between HGI and all-cause mortality, with stratification by sex. Four models were constructed: Model 1 accounted for survey cycle, age, ethnicity, educational attainment, marital condition, and PIR. Model 2 further incorporated FINS, TG, TC, LDL-C, and HDL-C. Model 3 then added Scr, BUN, and eGFR. Model 4 finally included total bilirubin, UA, and albumin. Covariates were selected considering prior research, clinical relevance, and their statistical significance (p-value < 0.05 in univariate analysis). Restricted cubic splines, marked by knots at the 5th, 50th, and 95th percentiles, were assessed for non-linear associations, while a two-segment Cox model estimated HRs and 95% CIs around the inflection point. Kaplan-Meier curves visualized mortality differences across HGI quartiles, and stratified analyses by age, PIR, and eGFR confirmed the robustness of the findings.

Additionally, to evaluate result robustness, we performed a sensitivity analysis that handled missing covariates via five-fold multiple imputation, yielding unbiased and reliable estimates.

R version 4.2.2 (http://www.R-project.org; The R Foundation, Vienna, Austria) and Free Statistics software version 2.2 (Beijing Free Clinical Medical Technology Co. Ltd., Beijing, China) were used for the analysis. Statistical significance thresholds were established at two-tailed p-values below 0.05.

Results

Baseline characteristics

Over a median monitoring period of 110 months, a total of 2,838 fatalities from various causes were recorded. Table 1 displays population-adjusted characteristics for the nationally representative sample (n = 19287), with sex-specific stratification of cohort members. Most of them are Non-Hispanic white. In comparison to their male counterparts, female participants were generally older, and demonstrated a greater propensity for higher education, although they reported lower household incomes. Females were also more likely to live alone. Furthermore, female participants had higher average values for TC, HDL-C, eGFR, and HGI. Conversely, they exhibited lower levels of TG, LDL-C, Scr, BUN, Total bilirubin, UA, Albumin, FINS, HbA1c, and FPG.

Table 1.

Weighted baseline characteristics of the study population stratified by sex

Characteristic	Total	Male	Female	p-value
No. of participants	19,287	9254	10,033
Age, years	46.78 (0.23)	46.15 (0.25)	47.36 (0.26)	< 0.001
Race, n (%)				< 0.001
Non-Hispanic white	8,923 (69.36%)	4,363 (69.72%)	4,560 (69.03%)
Non-Hispanic black	3,706 (10.80%)	1,737 (10.04%)	1,969 (11.50%)
Mexican American	3,347 (7.89%)	1,597 (8.51%)	1,750 (7.32%)
Others	3,311 (11.95%)	1,557 (11.73%)	1,754 (12.15%)
Educational attainment, n (%)				< 0.001
Pre-high school	4,942 (16.77%)	2,488 (17.53%)	2,454 (16.07%)
High school graduate	4,413 (23.86%)	2,197 (24.80%)	2,216 (23.01%)
Advanced degrees	9,932 (59.37%)	4,569 (57.67%)	5,363 (60.93%)
Marital status, n (%)				< 0.001
Married or living with partners, n (%)	11,978 (65.02%)	6,255 (68.56%)	5,723 (61.76%)
Living alone	7,309 (34.98%)	2,999 (31.44%)	4,310 (38.24%)
PIR, n (%)				< 0.001
Low (≤ 1.3)	5,827 (21.14%)	2,582 (18.83%)	3,245 (23.26%)
Medium (1.3–3.5)	7,429 (36.68%)	3,603 (36.54%)	3,826 (36.81%)
High (> 3.5)	6,031 (42.18%)	3,069 (44.63%)	2,962 (39.94%)
Laboratory results
TG, mmol/La	1.17 (0.81, 1.72)	1.25 (0.87, 1.81)	1.11 (0.77, 1.63)	< 0.001
TC, mmol/L	5.02 (0.01)	4.93 (0.01)	5.10 (0.02)	< 0.001
LDL-C, mmol/L	2.99 (0.01)	3.01 (0.01)	2.98 (0.01)	0.014
HDL-C, mmol/L	1.40 (0.01)	1.25 (0.01)	1.53 (0.01)	< 0.001
Scr, mg/dLa	0.83 (0.70, 1.00)	0.95 (0.85, 1.07)	0.73 (0.64, 0.83)	< 0.001
eGFR, ml/min/1.73m2	94.24 (0.31)	93.82 (0.32)	94.62 (0.37)	0.002
BUN, mmol/L	4.79 (0.02)	5.12 (0.03)	4.49 (0.03)	< 0.001
Total bilirubin, umol/L	12.05 (0.07)	13.55 (0.09)	10.68 (0.07)	< 0.001
UA, mg/dL	5.45 (0.01)	6.12 (0.02)	4.83 (0.02)	< 0.001
Albumin, g/L	42.43 (0.05)	43.48 (0.06)	41.45 (0.06)	< 0.001
FINS, uU/mL	12.12 (0.12)	12.83 (0.18)	11.46 (0.14)	< 0.001
HbA1c, %	5.55 (0.01)	5.58 (0.01)	5.53 (0.01)	< 0.001
FPG, mg/dl	104.19 (0.28)	107.21 (0.36)	101.41 (0.33)	< 0.001
HGI	−0.11 (0.01)	−0.16 (0.01)	−0.07 (0.01)	< 0.001

a TG and Scr expressed as median (IQR) and others as weighted mean (SD) for continuous variables, while categorical variables are shown as unweighted counts (weighted percentages)

PIR, the ratio of poverty income; TG, triglycerides; TC, total cholesterol; LDL-C, low-density lipoprotein cholesterol; HDL-C, high-density lipoprotein cholesterol; Scr, Serum creatinine; eGFR: estimated glomerular filtration rate; BUN, blood urea nitrogen; UA, uric acid; FINS, fasting insulin; HbA1c, glycated hemoglobin A1c; FPG, fasting plasma glucose; HGI, hemoglobin glycation index; SD, standard deviation; IQR, Interquartile Range

Table S1 provides an overview of participants’ baseline characteristics, categorized by HGI quartile groups. The cohort had an average age of 46.78 years (0.23), comprising 9,254 males (47.88%) and 10,033 females (52.12%). Notably, both the average age of participants and the proportion of females progressively increased from the first quarter (Q1) to the fourth quarter (Q4). Compared to Q1, Q2, and Q3, participants in Q4 were characterized by a lower representation of non-Hispanic white individuals, reduced educational attainment, and lower household income. Additionally, Q4 participants had elevated levels of TG, BUN, FINS, HbA1c, along with reduced levels of eGFR, Total bilirubin, and Albumin.

Sex-specific association between HGI and mortality outcomes

Table 2 presents the sex-stratified connections between HGI and overall mortality, with appropriate adjustment for confounding variables. Compared with the Q3 reference group, male participants in the Q1 and Q4 groups exhibited notably elevated rates of mortality from all causes. For Q1 and Q4, the HRs and 95% CIs were 1.25(1.04 ~ 1.50) and 1.21(1.00 ~ 1.45), respectively. For female participants, HGI, when analyzed continuously, exhibited a statistically significant hazard reduction (HR = 0.77, 95% CI 0.68–0.88; p < 0.001). Additionally, individuals in the Q1 group and Q2 group displayed a markedly increased risk of all-cause mortality relative to those in the Q3 group, with a hazard ratio of 1.43 (95% CI: 1.16–1.76) and 1.24 (95% CI: 1.00–1.54).

Table 2.

Association of HGI and all-cause mortality stratified by sex

HGI	No. of	HR (95% CI), P value
	deaths	Model 1		Model 2		Model 3		Model 4
Males
HGI	1619	1.13(0.95 ~ 1.34)	0.156	1.13(0.95 ~ 1.33)	0.165	1.09(0.93 ~ 1.28)	0.29	1.00(0.85 ~ 1.18)	0.978
HGI Quartiles
Q1(−4.691, −0.402)	359	1.19(0.99 ~ 1.42)	0.067	1.14(0.95 ~ 1.36)	0.164	1.15(0.96 ~ 1.38)	0.123	1.25(1.04 ~ 1.50)	0.02
Q2(−0.4, −0.124)	349	0.95(0.80 ~ 1.12)	0.512	0.93(0.79 ~ 1.11)	0.431	0.94(0.79 ~ 1.11)	0.447	0.97(0.82 ~ 1.14)	0.685
Q3(−0.124, 0.148)	405	1 (Ref)		1 (Ref)		1 (Ref)		1 (Ref)
Q4(0.149, 8.807)	506	1.33(1.11 ~ 1.60)	0.002	1.31(1.09 ~ 1.57)	0.003	1.27(1.06 ~ 1.52)	0.008	1.21(1.00 ~ 1.45)	0.045
Females
HGI	1219	0.79(0.69 ~ 0.91)	0.001	0.79(0.70 ~ 0.91)	< 0.001	0.79(0.70 ~ 0.91)	< 0.001	0.77(0.68 ~ 0.88)	< 0.001
HGI Quartiles
Q1(−4.134, −0.288)	297	1.44(1.18 ~ 1.76)	< 0.001	1.41(1.15 ~ 1.74)	0.001	1.38(1.12 ~ 1.69)	0.002	1.43(1.16 ~ 1.76)	< 0.001
Q2(−0.287, −0.046)	276	1.26(1.02 ~ 1.57)	0.034	1.27(1.02 ~ 1.57)	0.032	1.24(1.00 ~ 1.54)	0.055	1.24(1.00 ~ 1.54)	0.047
Q3(−0.046, 0.205)	276	1 (Ref)		1 (Ref)		1 (Ref)		1 (Ref)
Q4(0.205, 9.313)	370	1.23(1.05 ~ 1.44)	0.012	1.21(1.02 ~ 1.43)	0.026	1.18(0.99 ~ 1.40)	0.057	1.19(1.00 ~ 1.42)	0.055

The number of individuals in the sample is not weighted; the percentages represent the weighted frequency of the risk factor within the relevant strata of the US population

The number of individuals in the sample is not weighted; the percentages represent the weighted frequency of the risk factor within the relevant strata of the US population

Model 2: Adjusted for Model 1 plus FINS, TC, TG, LDL-C, HDL-C

Model 3: Adjusted for Model 2 plus Scr, eGFR and BUN 

Model 4: Adjusted for Model 3 plus Total bilirubin, UA and Albumin

Kaplan-Meier survival curves (Figure S1) demonstrated differential mortality incidence across quartiles; the Q4 group exhibited the greatest all-cause mortality among both male and female participants when compared to the Q1, Q2, and Q3 groups (p < 0.001, log-rank test).

Nonlinear associations between HGI and outcomes

To further elucidate the observed nonlinear relationship, we employed restricted cubic splines (RCS) along with segmented Cox models for hazard analysis, aiming to explore in greater depth the connection between HGI and overall mortality. Figure 2A illustrates that, among male participants, HGI exhibits a U-shaped pattern in relation to all-cause mortality following full covariate adjustment, with the threshold identified at −0.131 (Table 3). As shown in Table 3, the hazard ratio for all-cause mortality was 0.52 (95% CI: 0.42–0.66) below the threshold and 1.34 (95% CI: 1.16–1.55) on the right. For female participants, after controlling for all variables, Fig. 2B revealed an L-shaped relationship, identifying an inflection point at −0.096. A progressive decline in the risk of all-cause mortality was evident to the left of the inflection point as HGI rose, with an HR of 0.48 (95% CI: 0.40–0.59) (Table 3). Beyond this inflection point, further increases in HGI were not significantly related to all-cause mortality (p > 0.05).

Fig. 2.

Weighted RCS curve for the association of HGI and all-cause mortality Notes: Association between (A) HGI and all-cause mortality in the male population, (B) HGI and all-cause mortality in the female population. The solid lines and shaded areas represent the estimated values and their corresponding 95% confidence intervals, respectively. Only 97% of data is showing, adjusted for survey cycle, age, race, educational attainment, marital status, PIR, FINS, TG, TC, LDL-C, HDL-C, Scr, BUN, eGFR, total bilirubin, UA, and albumin. Abbreviations: HR, hazard ratio; CI, confidence interval; restricted cubic spline (RCS); HGI, hemoglobin glycation index; PIR, the ratio of poverty income; FINS, fasting insulin; TG, triglycerides; TC, total cholesterol; LDL-C, low-density lipoprotein cholesterol; HDL-C, high-density lipoprotein cholesterol; Scr, serum creatinine; BUN, blood urea nitrogen; eGFR, estimated glomerular filtration rate; UA, uric acid

Table 3.

Weighted threshold effect analysis of HGI on all-cause mortality stratified by sex

HGI	Adjusted HR (95%CI)	P value
Male
Inflection point	−0.131
HGI < −0.131	0.52(0.42, 0.66)	< 0.001
HGI ≥−0.131	1.34(1.16, 1.55)	< 0.001
Female
Inflection point	−0.096
HGI < −0.096	0.48(0.40, 0.59)	< 0.001
HGI ≥−0.096	0.97(0.79, 1.19)	0.77

Cox proportional hazards models were used to estimate HR and 95% CI. Adjusted for survey cycle, age, race, educational attainment, marital status, PIR, FINS, TC, TG, LDL-C, HDL-C, Scr, eGFR and BUN, Total bilirubin, UA and Albumin

HR, hazard ratio; CI, confidence interval; HGI, hemoglobin glycation index; PIR, the ratio of poverty income; FINS, fasting insulin; TG, triglycerides; TC, total cholesterol; LDL-C, low-density lipoprotein cholesterol; HDL-C, high-density lipoprotein cholesterol; Scr, Serum creatinine; BUN, blood urea nitrogen; eGFR, estimated glomerular filtration rate; UA, uric acid

Subgroup analysis

Analyses were stratified across multiple subgroups to examine potential alteration within HGI-mortality associations, as visualized in Fig. 3. Subgroup analyses based on age, PIR, or eGFR were observed no statistically significant interactions.

Fig. 3.

Weighted subgroup analysis for the association between HGI and all-cause mortality. Notes: They were adjusted for survey cycle, age, race, educational attainment, marital status, PIR, FINS, TG, TC, LDL-C, HDL-C, Scr, BUN, eGFR, total bilirubin, UA, and albumin

Sensitivity analyses

A sensitivity analysis was conducted in which missing data were addressed by multiple imputation (Table S2); the results remained consistent with the primary analysis, further validating the robustness of our conclusions.

Discussion

This retrospective cohort study, utilizing NHANES data from 1999 to 2018, investigates how HGI relates to all-cause mortality differently between men and women in US adults. For males, the analysis revealed a U-shaped relationship with a threshold at −0.131, while females displayed an L-shaped association with a −0.096 inflection point. This relationship remained statistically significant after multiple-imputation handling of missing data. This study is the first to explore sex-specific variation in the HGI-mortality relationship within the U.S. adult population, highlighting the importance of considering sex as a modifier in health-related research.

HbA1c is produced through the nonenzymatic binding of glucose to intracellular HbA1 [29], which can lead to discrepancies between actual and predicted HbA1c levels. The relationship between HbA1c and FPG exhibits considerable interindividual variability, influenced by various factors that affect glucose metabolism. HGI serves as a measure to capture this variability in HbA1c. In line with our findings, a low HGI score indicates reduced glycation relative to what would be anticipated at a higher FPG, while a high HGI score suggests increased glycation [3, 30].

The HGI reflects individual glycemic variability within populations and acts as a key marker for the risk of complications related to both macrovascular and microvascular conditions, potentially contributing to adverse clinical outcomes [10]. A retrospective cohort study [17] based on NHANES data reported that among individuals with diabetic kidney disease, the mortality risk was 1.39-fold higher (95% CI, 1.02–1.88) in the lowest HGI relative to the middle tertile, while the mortality risk in the highest tertile increased to 1.48-fold (95% CI, 1.05–2.08). These results correspond with the study conducted by Zhangyu Lin et al. [14], which investigated individuals diagnosed with DM and coronary artery disease. Furthermore, another retrospective cohort study identified a U-shaped association between HGI and 30-day mortality, using RCS for individuals with severe coronary heart disease [12]. The observed connection between HGI and all-cause mortality within the general population reflects results from prior research. These findings point to HGI’s potential to independently predict unfavorable clinical events within the general population.

Jingqi Yang’s cohort study [22] also observed a U-shaped link in males and an L-shaped association in females between HGI and all-cause mortality in populations with pre-diabetes and diabetes, consistent with our results. Similarly, Qing Shangguan analyzed non-linear correlations using RCS and found consistent gender-specific patterns in 7,607 hypertensive patients [15]. The mechanisms underlying the link of both HGI and all-cause mortality remain poorly understood, but several interlinks may account for these observed sex-based differences in the general population. Prior research has demonstrated that HGI is linked to an increased risk of subclinical myocardial injury (SC-MI) (OR = 1.074, 95% CI: 1.008–1.195) [25] and a higher likelihood of major adverse cardiovascular events (MACE) [14, 31, 32] in general patients. Additionally, non-diabetic individuals with higher HGI faced a 2.7-fold increased risk of vascular atherosclerosis [33]. Furthermore, for every one-unit elevation in HGI, the risk of abdominal obesity increased by 28.7%, hypertension by 34.9%, and hypercholesterolemia by 37.6% [34]. Moreover, HGI was significantly associated with an elevated risk of non-alcoholic fatty liver disease [35]. Furthermore, oxidative stress and impaired antioxidant defense have been implicated in the pathogenesis of virtually all major human diseases, ranging from neuropsychiatric disorders to cardiovascular disease, diabetes, and cancer [36]. Consequently, we additionally adjusted for endogenous antioxidants such as bilirubin, uric acid, and albumin, which have been identified as significant prognostic factors for cardiovascular disease, ischemic stroke, lung cancer, and thyroid cancer in large-scale cohort studies [37–41]. Our analyses further demonstrated that, regardless of sex, both low and high HGI levels were associated with an increased risk of all-cause mortality. Thus, HGI may be linked to the functional status of multiple tissues, potentially influencing clinical outcomes in the general population.

Sex differences have been observed in various factors, including the risk of coronary artery calcification [42] and levels of LDL-C, HDL-C, and TG, spanning from early adulthood to middle age and into older age [43]. These differences were also evident in our study. Hormonal variations between men and women may partially explain the observed sex differences in HGI and mortality. In women, estrogens are known to lower cardiovascular risk by providing protection against atherosclerosis and coronary artery disease [42]. Conversely, in men, an imbalance—either an excess or deficiency—of androgens may elevate cardiovascular risk [44]. Together, these distinctions highlight the differential relationship between HGI and all-cause mortality when analyzed by sex.

This study has limitations and strengths to consider. The findings may not apply to other groups, but weighted analysis improves generalizability. Residual confounding factors might still influence results despite adjustments. We used robust methods, including multivariable adjustments, subgroup analyses, and multiple imputations for missing data. Additionally, as an observational design, it does not establish causality among the variables studied, and HGI was derived from a single baseline FPG and HbA1c measurement, which may cause measurement error and regression dilution during long-term follow-up. Further research is needed for more comprehensive insights.

Conclusion

The cohort study revealed that in males of the general US adult population, HGI exhibited a U-shaped relationship with all-cause mortality, whereas it exhibited an L-shaped relationship in females. Further research is required to provide more comprehensive insights.

Abbreviations

HGI

Hemoglobin glycation index

NHANES

National health and nutrition examination survey

HbA1c

Glycated hemoglobin A1c

FPG

Fasting plasma glucose

HR

Hazard ratio

CI

Confidence interval

DM

Diabetes mellitus

NCHS

National center for health statistics

PIR

Poverty income ratio

TG

Triglycerides

LDL-C

Low-density lipoprotein cholesterol

FINS

Fasting insulin

Q

Quartile

TC

Total cholesterol

HDL-C

High-density lipoprotein cholesterol

Scr

Serum creatinine

BUN

Blood urea nitrogen

eGFR

Estimated glomerular filtration rate

UA

Uric acid

SD

Standard deviations

IQR

Interquartile ranges

RCS

Restricted cubic splines

MACE

Major adverse cardiovascular events

Author contributions

Chaolan Wang: Formal analysis, Methodology, Data curation, Writing – original draft. Ke Lin: Conceptualization, Validation, Resources, Visualization, Supervision. Hong Zhang: Methodology, Data curation, Writing - original draft. Tianbao Liu: Investigation, Writing – review & editing. Zhen Zeng: Formal analysis, Data curation, Writing – review & editing. Wanpei Luo: Methodology, Validation, Writing – review & editing. Yan Jiang: Methodology, Validation, Writing – review & editing. Xiang Zhang: Conceptualization, Methodology, Validation, Formal analysis. Shuang Du: Writing – review & editing, Visualization, Supervision, Project administration.

Funding

This work was conducted without specific funding from any public, commercial, or non-profit sources.

Data availability

NHANES data are publicly accessible at [https://wwwn.cdc.gov/nchs/nhanes](https://wwwn.cdc.gov/nchs/nhanes). For further details, the corresponding author can be contacted.

Declarations

Ethics approval and consent to participate

This study used anonymous data from the National Health and Nutrition Examination Survey and complied with the ethical guidelines and regulations of the Declaration of Helsinki. The study was approved by the National Center for Health Statistics Ethics Review Board, and all participants provided written informed consent before the study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.