Hemoglobin glycation index among adults with type 1 diabetes: Association with double diabetes features

Abstract

BACKGROUND

The hemoglobin glycation index (HGI) represents the discrepancy between the glucose management indicator (GMI) based on mean blood glucose levels and laboratory values of glycated hemoglobin (HbA1c). The HGI is a promising indicator for identifying individuals with excessive glycosylation, facilitating personalized evaluation and prediction of diabetic complications. However, the factors influencing the HGI in patients with type 1 diabetes (T1D) remain unclear. Autoimmune destruction of pancreatic β cells is central in T1D pathogenesis, yet insulin resistance can also be a feature of patients with T1D and their coexistence is called “double diabetes” (DD). However, knowledge regarding the relationship between DD features and the HGI in T1D is limited.

AIM

To assess the association between the HGI and DD features in adults with T1D.

METHODS

A total of 83 patients with T1D were recruited for this cross-sectional study. Laboratory HbA1c and GMI from continuous glucose monitoring data were collected to calculate the HGI. DD features included a family history of type 2 diabetes, overweight/obesity/central adiposity, hypertension, atherogenic dyslipidemia, an abnormal percentage of body fat (PBF) and/or visceral fat area (VFA) and decreased estimated insulin sensitivity. Skin autofluorescence of advanced glycation end products (SAF-AGEs), diabetic complications, and DD features were assessed, and their association with the HGI was analyzed.

RESULTS

A discrepancy was observed between HbA1c and GMI among patients with T1D and DD. A higher HGI was associated with an increased number of SAF-AGEs and a higher prevalence of diabetic microangiopathy (P = 0.030), particularly retinopathy (P = 0.031). Patients with three or more DD features exhibited an eight-fold increased risk of having a high HGI, compared with those without DD features (adjusted odds ratio = 8.12; 95% confidence interval: 1.52-43.47). Specifically, an elevated PBF and/or VFA and decreased estimated insulin sensitivity were associated with high HGI. Regression analysis identified estimated insulin sensitivity and VFA as factors independently associated with HGI.

CONCLUSION

In patients with T1D, DD features are associated with a higher HGI, which represents a trend toward excessive glycosylation and is associated with a higher prevalence of chronic diabetic complications.

Key Words: Type 1 diabetes; Double diabetes; Insulin resistance; Hemoglobin glycation index; Advanced glycation end products; Diabetic complications

Core Tip: Insights into the discrepancy between the glucose management indicator (GMI) based on mean blood glucose levels and the laboratory estimated glycated hemoglobin (HbA1c) levels in patients with type 1 diabetes (T1D) are limited. This study aimed to quantify this discrepancy by calculating the hemoglobin glycation index (HGI) based on laboratory HbA1c and continuous glucose monitoring-derived GMI levels. The results indicated that double diabetes features in patients with T1D are associated with a higher HGI, which represents a trend toward excessive glycosylation and is associated with a higher prevalence of chronic diabetic complications.

INTRODUCTION

Type 1 diabetes (T1D) is characterized by the progressive loss of β-cell function, resulting in absolute insulin deficiency and hyperglycemia[1]. Glycated hemoglobin (HbA1c) is considered the gold standard for evaluating mean ambulatory blood glucose levels and provides an average value of the past 2-3 months[2]. However, only 60%-80% of the variance in HbA1c levels can be explained by mean blood glucose (MBG) levels[3]. To address this issue, Hempe et al[4] developed and validated the hemoglobin glycation index (HGI) in 2002 to quantify the disparity between laboratory measured HbA1c and the predicted HbA1c derived from the glucose management indicator (GMI) based on MBG levels. The HGI is calculated as the difference between the observed and predicted HbA1c values, estimated by inserting a time-matched blood glucose measurement into a regression equation describing the linear relationship between blood glucose and HbA1c levels. Recently, with the increasing use of continuous glucose monitoring (CGM), a new formula has been derived for converting CGM-derived MBG levels into an estimate of HbA1c (the so-called GMI) levels[5], enabling a more precise determination of the HGI.

The HGI has been increasingly recognized as a valuable biomarker in the diagnosis and prognostication of diabetes, risk assessment for complications, and personalized treatment in type 2 diabetes (T2D). The combination of HGI with HbA1c is beneficial for the diagnosis of diabetes[6,7]. Studies have shown that a higher HGI is associated with an increased risk of retinopathy[8], diabetic kidney disease[9], and non-alcoholic fatty liver disease[10] in patients with T2D. Interestingly, there is a U-shaped correlation between HGI values and the risk of major adverse cardiovascular events (MACE) in patients with T2D. Both low and high HGI are associated with an increased risk of MACE[11]. In terms of treatment, the action to control cardiovascular risk in a diabetes study[12] indicated that intensive glycemic control (aiming for HbA1c < 6.0%) in T2D patients led to higher mortality than the standard control (targeting HbA1c: 7.0%-7.9%), highlighting the potential risks of strict glucose management. Notably, patients with low to moderate HGI levels seemed to benefit from intensive treatment, unlike those with high HGI, suggesting that HGI could guide personalized treatment strategies[12,13]. Further studies have indicated that the HGI could serve as a useful predictive marker for the therapeutic response to dipeptidyl peptidase 4 inhibitors in patients with T2DM[14]. Despite widespread research on the implications of HGI in T2D, its prognostic value in patients with T1D remains underexplored. The findings from the diabetes control and complications trial showed that, in patients with T1D, a high HGI was associated with a three-fold greater risk for retinopathy and a six-fold greater risk for nephropathy[15]. The formation of advanced glycation end products (AGEs) is a major determinant of diabetic complications, and HbA1c is a prototype of AGEs[16]. Recently, Maran et al[17] found that a fast-glycator phenotype, defined by a glucose management indicator (GMI) to an HbA1c ratio of < 0.9, was associated with a higher number of skin AGEs and a greater complication burden.

Despite a growing body of evidence supporting the utility of the HGI in the management of diabetes and prognostication, the factors influencing the HGI in patients with T1D remain to be appropriately evaluated. The HGI is a complex quantitative entity, and laboratory, genetic, and environmental factors are related to its level[18]. Given the epidemic of obesity and T2D, significant attention has been paid to the relationship between insulin resistance (IR) and the HGI. The hyperinsulinemic-euglycemic clamp test is the gold standard for evaluating IR; however, it is invasive and time-consuming. Therefore, more accessible clinical features have been commonly used as substitutes to represent IR, including obesity, atherogenic dyslipidemia, hypertension, metabolic syndrome, and validated equations[19]. In patients without diabetes and with fatty liver disease, higher blood pressure (BP), body mass index (BMI), total cholesterol, low-density lipoprotein cholesterol (LDL-C), and triglyceride (TG) levels are associated with a higher HGI[20,21]. Furthermore, similar findings have been observed in individuals with T2D[13,22,23]. Although IR is a characteristic usually linked to T2D, it can also be a feature of patients with T1D, and their coexistence is called “double diabetes” (DD), especially in those with a disease duration ≥ 10 years[24]. However, no study has comprehensively evaluated the relationship between DD and the HGI in patients with T1D. To date, only a limited number of studies have incorporated certain DD-related parameters when investigating the factors contributing to HGI variation in patients with T1D, and the interpretation of results is influenced by several confounding factors, such as ethnicity, disease duration, and use of lipid-lowering medications[16,17,25]. Additionally, measurements of body composition using dual-energy X-ray absorptiometry (DXA) can provide more specific results for fat deposition and have revealed the varying contributions of distinct adipose and muscle tissue phenotypes to the development of DD[26], introducing precise markers for assessing the role of DD features in HGI variation.

Therefore, this cross-sectional study aimed to employ CGM data to evaluate the discrepancy between predicted HbA1c from MBG and laboratory HbA1c values in deriving the HGI, to investigate the chronic complication features of patients with T1D with different levels of the HGI and to systematically assess the association between the HGI and DD-associated metabolic indicators, including anthropometric measurements of body fat, body composition metrics, lipid profiles, BP, and the calculated insulin sensitivity index.

MATERIALS AND METHODS

Study population

This cross-sectional study included patients with T1D who received clinical treatment at the Department of Metabolism and Endocrinology in the Second Xiangya Hospital of Central South University between March 2022 and March 2024. The inclusion criteria of this study were as follows: Diabetes diagnosed according to the 1999 World Health Organization criteria[27]; Patients clinically diagnosed with T1D, those who satisfied (1) and (2) or those who were negative for all three islet autoantibodies but satisfied (2) and having any condition described in (3)-(5): (1) Those who were positive for at least one of the islet autoantibodies, including glutamic acid decarboxylase antibody (GADA), insulinoma-associated 2 molecule antibody (IA-2A), and zinc transporter 8 antibody (ZnT8A); (2) With insulin dependency from disease onset; (3) With diabetic ketoacidosis/diabetic ketosis at onset; (4) With continuous loss of β-cell function (postprandial C-peptide level, < 300 pmol/L); and (5) With an onset age ≤ 30 years and > 6 months; Those with a duration of diabetes ≥ 10 years; Those aged ≥ 18 years; Those willing and able to wear a CGM device for 14 days. Individuals diagnosed with other forms of diabetes and those with a CGM activation time < 70% were excluded. Whole-exome sequencing was performed to rule out specific types of diabetes mellitus.

A total of 83 eligible patients with T1D were included in this study. This study was approved by the Ethics Committee of the Second Xiangya Hospital of Central South University and conducted in accordance with the Declaration of Helsinki. Written informed consent was obtained from all participants prior to data collection.

Demographic and clinical data collection

At the time of enrollment, we collected demographic and clinical data from all participants, including age, sex, age at onset, duration of diabetes, insulin regimen, medication history, dietary and physical activity patterns (determined by conducting the food frequency questionnaire and calculating the metabolic equivalent based on the results of the international physical activity questionnaire), family history of T2D (defined as any first-degree relative diagnosed with T2D), height, weight, waist circumference (WC), hip circumference, BP, hemoglobin levels, HbA1c levels, lipid profile [TG, and high-density lipoprotein cholesterol (HDL-C) levels], fasting C-peptide levels, islet autoantibody levels (GADA, IA-2A, and ZnT8A), and skin autofluorescence of AGEs (SAF-AGEs) (AGE reader: Hefei Institutes of Physical Science, Chinese Academy of Sciences).

Diabetic complications investigated in this study included diabetic microangiopathy (diabetic nephropathy, diabetic retinopathy, and peripheral neuropathy) and atherosclerosis. Diabetic nephropathy was diagnosed as a urine albumin-to-creatinine ratio ≥ 30 mg/g and/or an estimated glomerular filtration rate < 60 mL/minute/1.73 m²[28] while excluding other causes of chronic kidney disease. Diabetic retinopathy was defined as the presence of a microaneurysm, hemorrhagic spots, neovascularization, vitreous hemorrhage, cotton wool spots, and macular lesions on fundus photographs, combined with the diagnosis made by an ophthalmologist. Diabetic peripheral neuropathy was diagnosed based on the clinical symptoms of neuropathy (pain, numbness, and abnormal sensation) and any one abnormality in the five tests (ankle reflex, acupuncture pain sensation, and vibration sensation using a 128-Hz tuning fork; pressure sensation using a 10-g Semmes Weinstein test; and temperature sensation) and any two abnormalities in the five tests for those without clinical symptoms[29]. Atherosclerosis was defined as a carotid intima-media thickness ≥ 1 mm and/or a carotid-femoral pulse wave velocity ≥ 9 m/second[30].

DD features

For adult patients, overweight was defined as a BMI [BMI = weight in kg/(height in m)²] of 24 to < 28 kg/m² and obesity was defined as a BMI ≥ 28 kg/m². Central obesity was defined as a WC ≥ 90 cm for men and ≥ 85 cm for women[31]. Hypertension was defined as repeated BP measurements ≥ 130/85 mmHg or having a confirmed diagnosis of hypertension and receiving antihypertensive medications. Atherogenic dyslipidemia was defined as a TG level ≥ 1.70 mmol/L and/or an HDL-C level < 1.04 mmol/L[32,33].

The body composition was assessed using a DXA scanner (Lunar iDXA; GE Company). The measurements obtained included the body fat mass (FM) (in kg), body lean mass (in kg), and visceral fat area (VFA) (in cm²). Additionally, the percentage of total body fat (PBF) (FM in kg/whole-body mass in kg) was calculated. A PBF ≥ 20% for men and ≥ 32% for women[34] and a VFA ≥ 100 cm² were defined as abnormal[35].

The equation previously developed and validated against hyperinsulinemic-euglycemic clamp data from the Coronary Artery Calcification in T1D study for estimating IR in adults with T1D was used in this study. The estimated insulin sensitivity (eIS) was calculated as follows: EIS = exp [4.1075 - 0.01299 (WC in cm) - 1.05819 (insulin dose, daily units per kg) - 0.31327 (TGs, mmol/L) - 0.00802 (diastolic BP, mmHg)][36]. A lower eIS level indicates greater IR. The eIS cutoff point of ≤ 4.66 mg/kg/minute was adopted in the current study[37].

The DD features included a positive family history of T2D, overweight/obesity/central obesity, hypertension, atherogenic dyslipidemia, abnormal PBF and/or VFA, and eIS ≤ 4.66 mg/kg/minute.

CGM parameters and definition of the HGI

CGM was performed using the SiJoy GS1 CGM system (GS1 CGM, Shenzhen Sibionics Technology Co., Ltd), which has achieved satisfactory performance and usability (mean absolute relative difference = 8.83% ± 4.03%)[38]. The CGM system was placed on the lateral upper arm of the participants, according to the manufacturer’s instructions. CGM data were collected from the start of study enrollment until discharge during a 14-day sensor wear period for all 83 participants with T1D[1,39]. Additionally, 3-month CGM data were collected from a subset of 28 participants. Parameters obtained included the MBG levels, coefficient of variation (CV), mean amplitude glycemic excursion, time in range (TIR) (3.9-10.0 mmol/L), time above range (percentage of readings and time > 13.9 mmol/L and > 10.0 mmol/L), and time below range (percentage of readings and time < 3.9 mmol/L and < 3.0 mmol/L). The GMI and HGI were calculated as follows: GMI = 3.31 + 0.02392 × mean glucose (mg/dL); HGI = HbA1c - GMI[40]. The HGI groups were categorized based on the HGI values (low HGI: ≤ 0.12%; moderate HGI: 0.13%-0.70%; high HGI: ≥ 0.71%).

Statistical analysis

Continuous variables were expressed as medians and interquartile ranges (IQR) (25^th-75^th percentile), and categorical variables were presented as numbers and percentages. The Shapiro-Wilk test was used to assess the distributions of continuous variables. Inter-group comparisons were performed using the one-way analysis of variance for normally distributed continuous variables, the Kruskal-Wallis H test for non-normally distributed variables, and the χ² test for categorical variables. The discrepancy between the GMI and laboratory HbA1c levels was compared using the Wilcoxon signed-rank test. Spearman’s correlation analysis was conducted to assess the association between the GMI and HbA1c levels and between the HGI and SAF-AGEs. Logistic regression analyses with odds ratios (ORs) and 95% confidence intervals (CIs) were performed to assess the association between DD features and HGI tertiles. Bivariate correlations between DD-related metabolic parameters and the HGI were analyzed using partial correlation analysis, and multiple linear regression models were developed to further evaluate independent associations (forward selection stepwise; P < 0.05, the criterion for variable retention). Prior to correlation and regression analyses, the variables were Box-Cox transformed to achieve normality. Two-sided statistical tests were performed, and P < 0.05 was considered statistically significant. Statistical analyses were performed using statistical product and service solutions version 26.0, and figures were generated using R, version 4.2.2.

RESULTS

General characteristics of patients with T1D

The data of 83 patients with T1D were included in the analysis (Table 1). The median age of the patients was 32.8 (IQR: 26.7-38.0) years, and the median duration of diabetes was 15.1 (IQR: 13.2-19.1) years. Of the 83 patients, 33 (39.8%) were male, and 50 (60.2%) had at least one positive islet cell autoantibody. Overall, 12 (14.4%) participants had diabetic nephropathy, 22 (26.5%) had diabetic retinopathy, 16 (19.3%) had diabetic peripheral neuropathy, 33 (39.8%) had at least one complication of diabetic microangiopathy, and 13 (15.7%) had atherosclerosis. Regarding glycemic control, the median HbA1c level, TIR, and CV were 6.9% (6.3%-7.6%), 73.7% (63.2%-83.5%), and 36.1% (31.3%-41.1%), respectively. A total of 20 (24.1%) patients with T1D exhibited three or more DD features: 16 (19.3%) had a positive family history of T2D, 10 (12.0%) had hypertension, 9 (10.8%) had dyslipidemia, 18 (21.7%) were overweight/obesity/central adiposity, 38 (45.8%) had abnormal PBF and/or VFA, and 29 (34.9%) had decreased eIS.

Table 1 Characteristics of the study participants, n (%).

Characteristics	T1D (n = 83)
Male	33 (39.8)
Age (years), median (25th-75th percentiles)	32.8 (26.7, 38.0)
Age at onset (years), median (25th-75th percentiles)	14.8 (9.9, 23.0)
Duration of diabetes (years), median (25th-75th percentiles)	15.1 (13.2, 19.1)
GADA +, IA-2A +, or ZnT8A +	50 (60.2)
Preserved β-cell function	52 (62.7)
Insulin pump users	37 (44.6)
Daily insulin dose (U/kg/day), median (25th-75th percentiles)	0.60 (0.50, 0.72)
DD features
0	26 (31.3)
1-2	37 (44.6)
≥ 3	20 (24.1)
Positive family history of T2D	16 (19.3)
Hypertension	10 (12.0)
Atherogenic dyslipidemia	9 (10.8)
Overweight/obesity/central obesity	18 (21.7)
Abnormal PBF and/or VFA	38 (45.8)
eIS ≤ 4.66 mg/kg/minute	29 (34.9)
HbA1c, median (25th-75th percentiles)	6.9 (6.3, 7.6)
Mean glucose (mmol/L), median (25th-75th percentiles)	7.4 (6.6, 8.2)
GMI, median (25th-75th percentiles)	6.5 (6.2, 6.8)
HGI, median (25th-75th percentiles)	0.35 (0.02, 1.06)
TBR (< 3.0 mmol/L), median (25th-75th percentiles)	0.80 (0.20, 1.75)
TBR (< 3.9 mmol/L), median (25th-75th percentiles)	7.6 (5.0, 11.5)
TIR, median (25th-75th percentiles)	73.7 (63.2, 83.5)
TAR (> 10.0 mmol/L), median (25th-75th percentiles)	17.5 (7.6, 26.9)
TAR (> 13.9 mmol/L), median (25th-75th percentiles)	2.2 (0.3, 7.2)
Coefficient of variation, median (25th-75th percentiles)	36.1 (31.3, 41.1)
MAGE (mmol/L), median (25th-75th percentiles)	6.0 (4.7, 7.6)
Diabetic nephropathy	12 (14.4)
Diabetic retinopathy	22 (26.5)
Diabetic peripheral neuropathy	16 (19.3)
Diabetic microangiopathy	33 (39.8)
Atherosclerosis	13 (15.7)

T1D: Type 1 diabetes; GADA: Glutamic acid decarboxylase antibody; IA-2A: Insulinoma-associated protein 2 antibody; ZnT8A: Zinc transporter 8 antibody; DD: Double diabetes; T2D: Type 2 diabetes; PBF: Percent total body fat; VFA: Visceral fat area; eIS: Estimated insulin sensitivity; HbA1c: Glycated hemoglobin; GMI: Glucose management indicator; HGI: Hemoglobin glycation index; TBR: Time below range; TIR: Time in range; TAR: Time above range; MAGE: Mean amplitude glycemic excursion.

Relationships among the GMI and laboratory HbA1c and SAF-AGE levels

The GMI and HbA1c level displayed a strong positive correlation (Figure 1A; R = 0.708, P < 0.001); however, the GMI was significantly lower than the laboratory HbA1c levels (Figure 1B). Among the 83 patients with T1D, the median HGI was 0.35% (0.02%-1.06%), and 9 (10.8%) patients had similar GMI and laboratory HbA1c levels (absolute HGI, 0.00%-0.10%). The discrepancy between the GMI and laboratory HbA1c levels was further compared among different HGI groups (Figure 1B). Laboratory HbA1c levels increased from the low- to high-HGI group [low- vs moderate- vs high-HGI groups: 6.2% (5.8%-6.4%) vs 6.8% (6.6%-7.1%) vs 8.3% (7.5%-9.0%), respectively; P < 0.001], whereas the GMI remained similar among the three groups [low- vs moderate- vs high-HGI groups: 6.3% (6.2%-6.6%) vs 6.5% (6.1%-6.8%) vs 6.8% (6.2%-7.0%), respectively; P = 0.102]. The GMI was similar to the HbA1c level in the low-HGI group and lower than the HbA1c level in the moderate- and high-HGI groups (Figure 1B). The HGI was positively correlated with the SAF-AGE levels (R = 0.261, P = 0.017; Figure 1C).

Figure 1 Relationships among glucose management indicator, laboratory glycated hemoglobin, and skin autofluorescence of advanced glycation end products. A: The correlation plot between glucose management indicator (GMI) and laboratory glycated hemoglobin (HbA1c); B: The discrepancy of GMI and laboratory HbA1c among different hemoglobin glycation index (HGI) groups (error bars represent standard error of the mean); C: The correlation plot between HGI and skin autofluorescence of advanced glycation end products. HGI groups were determined by HGI value tertile (low HGI: ≤ 0.12%; moderate HGI: 0.13%-0.70%; high HGI: ≥ 0.71%). ^bP < 0.01. ^cP < 0.001. GMI: Glucose management indicator; HbA1c: Glycated hemoglobin; HGI: Hemoglobin glycation index; SAF-AGEs: Skin autofluorescence of advanced glycation end products.

Characteristics of patients with different levels of the HGI

Compared with patients with T1D and a lower HGI, those with a higher HGI had a higher prevalence of diabetic microangiopathy [low- vs moderate- vs high-HGI groups: 7 (25.0%) vs 10 (35.7%) vs 16 (59.3%), respectively; P = 0.030], especially diabetic retinopathy [low- vs moderate- vs high-HGI groups: 4 (14.3%) vs 6 (21.4%) vs 12 (44.4%), respectively; P = 0.031; Figure 2]. The high-HGI group also had a higher proportion of patients with DD features (Table 2). Furthermore, the CGM profiles of the three HGI groups were similar, and there were no significant differences among the three groups in terms of age, sex distribution, disease duration, dietary and physical activity patterns, medication use, β-cell function, hemoglobin, and red blood cell (RBC) counts, or prevalence of atherosclerosis (Table 2).

Figure 2 Prevalence of diabetic microangiopathy among different hemoglobin glycation index groups. ^aP < 0.05. Significant difference between high and low hemoglobin glycation index group, corrected for multiple comparisons using the Bonferroni method. HGI: Hemoglobin glycation index.

Table 2 Characteristics among different hemoglobin glycation index groups, n (%).

Characteristics	Low HGI (n = 28)	Moderate HGI (n = 28)	High HGI (n = 27)	P value
Male	13 (46.4)	9 (32.1)	11 (40.7)	0.546
Age (years), median (25th-75th percentiles)	33.6 (25.5, 37.7)	32.7 (27.8, 39.4)	32.8 (26.7, 37.0)	0.850
Age at onset (years), median (25th-75th percentiles)	14.6 (9.7, 23.8)	15.5 (10.4, 22.6)	16.7 (9.5, 23.0)	0.940
Duration of diabetes (years), median (25th-75th percentiles)	15.2 (12.2, 20.2)	14.3 (12.4, 18.6)	15.2 (13.9, 20.0)	0.618
GADA +, IA-2A +, or ZnT8A +	17 (60.7)	20 (71.4)	13 (48.1)	0.211
Preserved β-cell function	19 (67.9)	20 (71.4)	13 (48.1)	0.159
Insulin pump users	14 (50.0)	15 (53.6)	8 (29.6)	0.158
Daily insulin dose (U/kg/day), median (25th-75th percentiles)	0.60 (0.51, 0.72)	0.54 (0.48, 0.64)	0.66 (0.53, 0.81)	0.103
Taking lipid-lowering medications	2 (7.1)	3 (10.7)	3 (11.1)	0.904
Taking oral hypoglycemic drugs	3 (10.7)	5 (17.9)	4 (14.8)	0.748
Dietary carbohydrate, median (25th-75th percentiles)	53.3 (37.8, 60.6)	51.8 (46.8, 58.8)	48.6 (46.0, 53.6)	0.755
Dietary fat, median (25th-75th percentiles)	28.2 (19.7, 35.0)	26.9 (23.1, 30.3)	29.3 (24.7, 30.8)	0.747
Dietary protein, median (25th-75th percentiles)	19.7 (17.4, 23.8)	19.2 (16.8, 21.9)	20.4 (18.8, 23.2)	0.476
MET-minute/week, median (25th-75th percentiles)	2144.0 (643.5, 3304.5)	1173.0 (490.0, 2013.0)	1386.0 (990.0, 2270.0)	0.414
DD features				0.025
01	13 (46.4)	10 (35.7)	3 (11.1)
1-2	12 (42.9)	12 (42.9)	13 (48.1)
≥ 31	3 (10.7)	6 (21.4)	11 (40.7)
Hemoglobin (g/L), median (25th-75th percentiles)	143.5 (132.5, 152.5)	135.5 (121.5, 149.0)	131.0 (121.0, 146.0)	0.082
RBC count (1012/L), median (25th-75th percentiles)	4.6 (4.4, 4.9)	4.5 (4.2, 4.8)	4.5 (4.1, 4.9)	0.757
Mean glucose (mmol/L), median (25th-75th percentiles)	6.9 (6.6, 7.6)	7.3 (6.4, 8.1)	8.1 (6.8, 8.5)	0.102
TBR (< 3.0 mmol/L), median (25th-75th percentiles)	0.70 (0.10, 1.29)	0.80 (0.20, 1.51)	0.70 (0.18, 2.39)	0.828
TBR (< 3.9 mmol/L), median (25th-75th percentiles)	7.8 (5.0, 9.6)	6.7 (5.7, 12.0)	7.8 (4.8, 12.0)	0.994
TIR, median (25th-75th percentiles)	78.2 (70.1, 85.9)	73.3 (65.7, 82.1)	65.4 (52.2, 83.5)	0.113
TAR (> 10.0 mmol/L), median (25th-75th percentiles)	11.5 (7.8, 20.5)	18.2 (7.3, 27.4)	24.7 (6.2, 34.5)	0.143
TAR (> 13.9 mmol/L), median (25th-75th percentiles)	1.2 (0.2, 3.0)	2.4 (0.3, 6.6)	3.9 (0.6, 10.7)	0.135
Coefficient of variation, median (25th-75th percentiles)	36.6 (31.7, 40.9)	36.1 (32.6, 40.0)	35.7 (30.0, 43.3)	0.959
MAGE (mmol/L), median (25th-75th percentiles)	5.5 (4.6, 6.6)	6.3 (5.4, 7.3)	7.3 (4.9, 8.5)	0.446
Diabetic microangiopathy	7 (25.0)	10 (35.7)	16 (59.3)	0.030
Atherosclerosis	3 (10.7)	5 (17.9)	5 (18.5)	0.674

¹Significant difference between high and low hemoglobin glycation index group, corrected for multiple comparisons using the Bonferroni method.

HGI: Hemoglobin glycation index; GADA: Glutamic acid decarboxylase antibody; IA-2A: Insulinoma-associated protein 2 antibody; ZnT8A: Zinc transporter 8 antibody; MET: Metabolic equivalent; DD: Double diabetes; RBC: Red blood cell; TBR: Time below range; TIR: Time in range; TAR: Time above range; MAGE: Mean amplitude glycemic excursion.

Association between DD features and the HGI

After adjusting for potential confounders, logistic regression analysis demonstrated that individuals with three or more DD features showed an eight-fold increased risk of having a high HGI (high vs low HGI: Adjusted OR = 8.12, 95%CI: 1.52-43.47), compared with those with no DD features. Among the DD features, elevated PBF and/or VFA (high vs low HGI: Adjusted OR = 4.25, 95%CI: 1.27-14.25) and decreased eIS (high vs low HGI: Adjusted OR = 6.92, 95%CI: 1.59-30.03) were associated with an increased risk of having a high HGI. However, no significant association was observed between the HGI and a positive family history of T2D, hypertension, atherogenic dyslipidemia, or overweight/obesity/central obesity (Figure 3).

Figure 3 Odds ratios and 95% confidence intervals for the association between double diabetes features and hemoglobin glycation index groups. Forest plots display odds ratios and 95% confidence intervals. ¹Moderate hemoglobin glycation index group compared to low hemoglobin glycation index group. ²High hemoglobin glycation index group compared to low hemoglobin glycation index group. Confounding variables adjusted: Age, gender, duration of diabetes and hemoglobin. OR: Odds ratio; 95%CI: 95% confidence interval; HGI: Hemoglobin glycation index; DD: Double diabetes; T2D: Type 2 diabetes; PBF: Percent total body fat; VFA: Visceral fat area; eIS: Estimated insulin sensitivity.

Direct association between DD-related metabolic parameters and the HGI

A bivariate analysis was conducted to clarify the direct association between DD-related metabolic parameters and the HGI (Table 3). According to the partial correlation analysis, the parameters for overweight/obesity/central adiposity, including the BMI (β = 0.268, P = 0.017) and WC (β = 0.292, P = 0.009), were positively associated with the HGI. Regarding the body composition parameters, both the PBF (β = 0.280, P = 0.012) and VFA (β = 0.290, P = 0.010) showed a positive correlation with the HGI. For atherogenic dyslipidemia parameters, a significant positive correlation was found between TG levels and the HGI (β = 0.471, P < 0.001), whereas HDL-C levels demonstrated no significant association with the HGI. Furthermore, the calculated eIS showed a negative correlation with the HGI (β = -0.454, P < 0.001). Neither systolic nor diastolic BP was correlated with the HGI. Subsequently, a hierarchical multivariate regression analysis was conducted, adjusting for age, sex, duration of diabetes, and hemoglobin level. When the BMI, WC, PBF, VFA, TG, and eIS were included as potential risk factors, the analysis identified the eIS and VFA as variables that were independently associated with the HGI (Table 4).

Table 3 Correlation of double diabetes-related metabolic parameters with hemoglobin glycation index.

DD-related metabolic parameters	β	P value
BMI (kg/cm2)	0.268	0.017
WC (cm)	0.292	0.009
SBP (mmHg)	0.136	0.233
DBP (mmHg)	0.152	0.182
HDL-C (mmol/L)	-0.076	0.508
TG (mmol/L)	0.471	< 0.001
PBF (%)	0.280	0.012
VFA (cm2)	0.290	0.010
eIS (mg/kg/minute)	-0.454	< 0.001

Data are beta coefficients and P values from partial correlation analysis. Confounding variables adjusted: Age, gender, duration of diabetes and hemoglobin.

DD: Double diabetes; BMI: Body mass index; WC: Waist circumference; SBP: Systolic blood pressure; DBP: Diastolic blood pressure; HDL-C: High-density lipoprotein-cholesterol; TG: Triglyceride; PBF: Percent total body fat; VFA: Visceral fat area; eIS: Estimated insulin sensitivity.

Table 4 Multivariable regression model for hemoglobin glycation index.

Variables	Standardized β	P value
Age (years)	-0.016	0.918
Male	-0.109	0.355
Duration of diabetes (years)	0.182	0.081
Hemoglobin (g/L)	-0.408	0.004
eIS (mg/kg/minute)	-0.466	< 0.001
VFA (cm2)	0.233	0.042

Data include model performance (R² = 0.313 and P < 0.001), standardized coefficients β and P value, with age, gender, duration of diabetes and hemoglobin entered into the first layer, and possible risk factors incorporated into the second layer (forward selection stepwise; P < 0.05 criterion for variable retention). Variables that did not enter into the final model were not included in the table. VFA: Visceral fat area; eIS: Estimated insulin sensitivity.

A subgroup analysis was conducted to determine whether the effect of DD on the HGI was consistent across the different T1D stages. Participants were divided into two subgroups with different disease durations (≤ 15.1 and > 15.1 years, divided based on the median disease duration). Hierarchical multivariate regression analysis showed that the eIS was independently associated with the HGI in both subgroups (Supplementary Table 1).

The association between DD-related metabolic parameters and the HGI was confirmed in a subset of 28 patients with T1D with 3-month CGM data. These patients showed clinical characteristics similar to those of the other patients (Supplementary Table 2), and the HGI calculated from their 14-day and 3-month mean glucose levels showed a strong correlation (Supplementary Figure 1). After adjusting for age, sex, duration of diabetes, and hemoglobin level, the BMI (β = 0.437, P = 0.033) and TG levels (β = 0.484, P = 0.016) showed a positive correlation with the HGI calculated from the 3-month CGM data, whereas the eIS (β = -0.497, P = 0.013) showed a negative correlation with the HGI calculated from the 3-month CGM data (Supplementary Table 3). The hierarchical multivariate regression analysis, after adjusting for age, sex, duration of diabetes, and hemoglobin level, identified eIS as a variable that was independently associated with the HGI calculated from the 3-month CGM data when including the BMI, TG, and eIS as potential risk factors.

DISCUSSION

This study highlights that DD features are significantly associated with higher HGI, which correlate with an increased risk of diabetic complications in patients with T1D. Patients with T1D with a higher HGI had higher glycation levels, as determined by SAF-AGEs. To our knowledge, this is the first study to comprehensively evaluate the relationship between HGI and various factors representing DD in adult patients with T1D. Compared with patients without DD features, those with three or more DD features had an eight-fold increased risk of having a high HGI. Among the DD-related metabolic parameters, the HGI was positively correlated with body fat measurements (BMI, WC, PBF, and VFA) and TG levels and negatively correlated with the eIS. Additionally, after adjusting for age, sex, duration of diabetes, and hemoglobin level, the eIS and VFA were identified as independent risk factors for a high HGI. These findings suggest a positive association between DD features and HGI in patients with T1D.

A high HGI represents a trend toward excessive glycosylation in the human body, which can increase the risk of chronic diabetic complications[17,41]. Consistent with previous findings, this study found a positive correlation between the HGI and skin AGEs in patients with T1D. Despite having comparable mean glucose levels, patients with high HGI exhibited a higher prevalence of diabetic microangiopathy, especially retinopathy, supporting the future utility of the HGI in diabetic management. The HGI is a complex quantitative trait, and the relationship between the HGI and IR-related metabolic markers has received considerable attention. Although usually linked to T2D, IR can also be a feature of patients with T1D, and their coexistence is called “DD”[42]. The prevalence of DD and related metabolic disorders in patients with T1D varies between 3% and 50%, depending on the study population and diagnostic criteria. However, no study has systematically evaluated the relationship between the HGI and DD features in patients with T1D. Maran et al[17] found that patients with the fast-glycator phenotype had a higher proportion of lipid disorders. However, TG, total cholesterol, and LDL-C levels did not significantly differ between the slow- and fast-glycator phenotypes, which may be attributed to differences in age and statin use between the two groups. Sakane et al[25] found no significant differences in the BMI, BP, or blood lipid profiles among different HGI groups; however, the sex ratio was also unevenly distributed across the groups. In the present study, there were no differences in terms of sex, use of lipid-lowering medications, or disease duration across the groups with different levels of the HGI, and potential confounders were further adjusted for in the correlation and regression analyses. The results indicated a positive correlation between the HGI and typical anthropometric measurements of obesity/central adiposity. Moreover, a more precise metric of visceral fat deposition, VFA[43], was also positively correlated with the HGI, providing additional information on the association between body fat content and excessive glycosylation. In addition to body fat measurements, TG levels showed a positive correlation with the HGI, whereas the calculated eIS showed a negative correlation with the HGI.

The observed association between DD-related parameters and the HGI is multifaceted, and recent studies have shed light on the underlying mechanisms. Visceral fat, recognized as an endocrine and inflammatory organ, secretes a plethora of adipokines and inflammatory cytokines, including interleukin 6 (IL-6), tumor necrosis factor α, and resistin, which play pivotal roles in inflammatory responses and oxidative stress[44]. Elevated TG levels lead to the secretion of inflammatory mediators such as IL-1β, prostaglandin E2, and IL-6 by activating macrophages[45,46], and concurrently increase oxidative stress, promote mitochondrial damage, and activate apoptotic pathways[47]. Inflammation and oxidative stress can accelerate glycosylation, as suggested by recent studies highlighting the interplay between reactive oxygen species and glycosylation enzymes[48]. This acceleration can lead to abnormal glycosylation patterns, which have been implicated in the development of complications associated with diabetes and other metabolic disorders. The glycosylation pathway is complex and involves the modification of proteins that play roles in cell adhesion, signal transduction, and immune evasion. Abnormal glycosylation leads to protein malfunction and disease progression. In the context of DD, increased visceral fat deposition and elevated TG levels may affect the HGI via the aforementioned inflammatory and oxidative stress-mediated glycosylation processes. However, the specific mechanisms linking DD features to elevated HGI require further validation in future studies.

A growing body of evidence has shown that DD features are powerful risk factors for T1D-related major complications and mortality, and might be the missing link for explaining the excess complication burden despite optimal glycemic control[49,50]. Our findings suggest that DD features may increase the risk of complications through excess glycosylation. As DD will possibly become the predominant phenotype in patients with T1D over the next few decades[42], lowering the risk of related features is likely to be as important as glycemic control in the prevention of long-term complications. Evidence indicates that metformin, rosiglitazone, sodium glucose cotransporter 2 inhibitors, and glucagon-like peptide-1 receptor agonists may have beneficial effects on DD features in patients with T1D[51-54]. Understanding the relationship between DD features and the HGI in patients with T1D can offer insights into tailoring individualized treatments and enhancing patient prognoses.

The current study has some limitations. First, surrogate measures were used to assess DD. However, the parameters examined in this study were less variable than fasting glucose or insulin concentrations and are widely used in clinical and research approaches, making these measurements comparable and reliable. Second, the 14-day CGM data were included in the main analysis in the current study. Although the 3-month CGM data can best cover RBC longevity, 10-14 days of CGM data can offer a good estimate of CGM metrics for a 3-month period, as recommended by consensus guidelines. Furthermore, the association between DD-related metabolic parameters and the HGI was confirmed in a subset of patients with T1D with 3-month CGM data. Third, the included participants were relatively younger and had a shorter T1D duration; therefore, these results need to be verified in older patients with T1D with a longer disease duration. Finally, this was a single-center cross-sectional study with a relatively small sample size in China, and it is unclear whether our findings can be generalized to other ethnic groups. Further prospective cohort studies with larger sample sizes are needed to confirm the findings of this study.

CONCLUSION

This study suggests that higher levels of DD-related metabolic indicators in patients with T1D are associated with a higher HGI, which represents a trend toward excessive glycosylation in the human body, as well as an increased risk of chronic diabetic complications. Therefore, treatment strategies should be optimized to alleviate the DD features and reduce the glycosylation rate, which may help improve glycemic control and the long-term prognosis in patients with T1D. Future research could focus on longitudinal studies to elucidate the causal relationship between HGI and DD features, exploring underlying molecular mechanisms, investigating treatment impacts on both, and expanding the study population for generalizability.

ACKNOWLEDGEMENTS

The authors thank all of the patients, nurses, doctors, investigators, and technicians for their efforts in data and sample collection.