Correlation between key indicators of continuous glucose monitoring and the risk of diabetic foot

Abstract

BACKGROUND

Continuous glucose monitoring (CGM) metrics, such as time in range (TIR) and glycemic risk index (GRI), have been linked to various diabetes-related complications, including diabetic foot (DF).

AIM

To investigate the association between CGM-derived indicators and the risk of DF in individuals with type 2 diabetes mellitus (T2DM).

METHODS

A total of 591 individuals with T2DM (297 with DF and 294 without DF) were enrolled. Relevant clinical data, complications, comorbidities, hematological parameters, and 72-hour CGM data were collected. Logistic regression analysis was employed to examine the relationship between these measurements and the risk of DF.

RESULTS

Individuals with DF exhibited higher mean blood glucose (MBG) levels and increased proportions of time above range (TAR), TAR level 1, and TAR level 2, but lower TIR (all P < 0.001). Patients with DF had significantly lower rates of achieving target ranges for TIR, TAR, and TAR level 2 than those without DF (all P < 0.05). Logistic regression analysis revealed that GRI, MBG, and TAR level 1 were positively associated with DF risk, while TIR was inversely correlated (all P < 0.05). Achieving TIR and TAR was inversely correlated with white blood cell count and glycated hemoglobin A1c levels (P < 0.05). Additionally, achieving TAR was influenced by fasting plasma glucose, body mass index, diabetes duration, and antidiabetic medication use.

CONCLUSION

CGM metrics, particularly TIR and GRI, are significantly associated with the risk of DF in T2DM, emphasizing the importance of improved glucose control.

Key Words: Continuous glucose monitoring; Time in range; Glycemia risk index; Diabetic foot; Continuous glucose monitoring target achievement

Core Tip: The study investigated the relationship between continuous glucose monitoring (CGM) metrics, specifically time in range (TIR), and the risk of diabetic foot (DF) in individuals with type 2 diabetes mellitus. Our findings indicate that DF risk was inversely associated with TIR but positively correlated with glycemic risk index, mean blood glucose, and time above range (TAR) level 1. Achieving optimal TIR and TAR targets was negatively influenced by elevated white blood cell count and glycated hemoglobin A1c levels. These insights emphasize the importance of achieving optimal CGM targets for improved management of DF complications.

INTRODUCTION

Diabetic foot (DF) is a serious and common complication of diabetes[1]. It involves an infected, ulcerated, or destructive lesion of the foot tissue below the ankle in diabetic patients, often associated with neuropathy and/or peripheral arterial disease of the lower extremities[2]. DF affects approximately 6.3% of the global population and 5.7% of individuals with diabetes in China, leading to impaired bodily function, reduced quality of life, and increased healthcare costs[3]. Moreover, DF ulcers (DFUs) are associated with high rates of disability and mortality. Between 15% and 20% of individuals with DFUs require lower limb amputation, and the five-year mortality rate for DFU patients ranges from 30% to 50%[4-6]. Unfortunately, patient awareness and attention to DF, particularly among those with low health literacy, remains relatively low[7,8]. Furthermore, the management of DF presents significant challenges. Therefore, early recognition and appropriate intervention for DF in diabetic patients are crucial.

Several studies have demonstrated that inadequate blood glucose management is a major contributor to the development and progression of DF[3,9]. Poor glycated hemoglobin A1c (HbA1c) control is linked to an increased risk of DF, while intensive glycemic control therapy (HbA1c: 6%-7.5%) reduces the risk of amputation in DF patients by 35%[10,11]. HbA1c is considered the "gold standard" for assessing glycemic control, as established by the American Diabetes Association (ADA) and the European Association for the Study of Diabetes (EASD)[12]. Substantial evidence supports a strong correlation between HbA1c levels and diabetic complications[13,14]. However, HbA1c has limitations. Firstly, it provides an average glucose level over the past 8-12 weeks, failing to capture individual glycemic patterns, such as hypoglycemia, hyperglycemia, and daily glucose variability[14,15]. Secondly, conditions like anemia, pregnancy, iron deficiency, hemoglobinopathy, and ethnicity can influence HbA1c accuracy[15,16]. Additionally, significant discrepancies can exist between measured HbA1c and daily mean glucose levels, even within the same individual[17,18]. Therefore, relying solely on HbA1c to assess glycemic control is insufficient, necessitating the exploration of alternative reliable indicators.

Indeed, emerging continuous glucose monitoring (CGM) technology has been recognized as the most effective method for obtaining comprehensive glycemic data. CGM-derived glycemic profiles offer valuable supplementary information to HbA1c, including metrics such as mean blood glucose (MBG), coefficient of variation (CV), time in range (TIR), time above range (TAR), time below range (TBR), and glycemic risk index (GRI)[19]. Among these, TIR (3.9-10.0 mmol/L) has gained prominence as a preferred glycemic metric and is recommended by the ADA and EASD as a useful adjunct to HbA1c for assessing glycemic control[12,20]. Recent studies have linked TIR to various diabetes-related clinical outcomes, including diabetic retinopathy (DR), diabetic sensorimotor polyneuropathy, microalbuminuria, and cardiovascular mortality[15,21-23]. Additionally, researchers have found that TIR is negatively correlated with the risk of amputation in patients with DF[10], and decreased TIR has been associated with an increased risk of major amputation[24] and adverse effects on post-amputation wound healing[25]. However, the specific impact of CGM indexes, including TIR, on the incidence of DF remains unclear. Furthermore, the factors influencing the achievement of CGM indicators in individuals with DF are not fully understood. Therefore, this study aimed to investigate the relationships between CGM indexes and the risk of DF in individuals with type 2 diabetes mellitus (T2DM), as well as to identify the factors influencing the attainment of CGM indicators in individuals with DF.

MATERIALS AND METHODS

Study population

In our study, a sample size calculation was performed prior to data collection using the Events per Variable (EPV) method, a widely accepted approach in the field. The EPV value was set to 10, consistent with the literature[26]. A total of 591 individuals with T2DM were recruited from hospitalized patients in our department between January 2018 and September 2022, including 297 with DF and 294 without DF. The diagnosis of T2DM was based on the World Health Organization's 1999 criteria[27], and DF was diagnosed according to the 1981 Wagner standard[28]. Among the DF patients, 125 had positive wound secretion cultures. These infected cases received appropriate antibiotic treatment based on antibiotic sensitivity testing. Inclusion criteria were: (1) Diagnosis of T2DM; (2) Valid 3-day CGM data; and (3) Complete clinical data. Exclusion criteria were: (1) Absence of T2DM; (2) Skin lesions above the ankle in individuals with T2DM; (3) Acute diabetic complications (diabetic ketoacidosis, hyperglycemic hyperosmolar state, or severe or recurrent hypoglycemic events); (4) Pregnancy or lactation; (5) History of mental illness, autoimmune disease, biochemical abnormalities, or severe kidney or liver dysfunction; (6) Incomplete clinical or CGM data; and (7) Infected patients in the non-DF (NDF) group. Informed written consent was obtained from each participant prior to enrollment. The study protocol was approved by the Ethics Committees of the Affiliated Hospital of Yunnan University and adhered to the Declaration of Helsinki.

Clinical information and biochemical assays

Demographic data and information regarding age, sex, diabetes duration, family history of diabetes, smoking and drinking status (current or not), diabetic complications, and comorbidities were collected during the admission. All participants underwent a routine examination, including waist circumference, hip circumference, height, and weight measurements. We calculated the body mass index (BMI) by dividing the weight (kg) by the squared height (m²). Waist-to-hip ratio (WHR) was calculated as waist circumference divided by hip circumference. After at least 8 hours of overnight fasting, we collected a venous blood sample for biochemical assays during the admission. The measurements of fasting blood glucose (FPG), HbA1c, blood routine examination, total cholesterol (TC), triglycerides (TG), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), albumin (ALB), serum creatinine, and blood urea nitrogen were performed as previously described[29]. The Chronic Kidney Disease Epidemiology Collaboration equation was utilized to calculate the estimated glomerular filtration rate (eGFR)[23].

CGM parameters

During the hospitalization period, a retrospective CGM system (Medtronic Inc., Northridge, CA, United States) was used to monitor subcutaneous interstitial glucose levels for three consecutive days. CGM sensors were inserted on day 0 and removed after 72 hours, recording glucose values every 3 minutes, resulting in 480 consecutive sensor readings per day. Capillary blood glucose measurements were obtained at least six times daily using a Sure Step blood glucose meter (LifeScan, Milpitas, CA, United States) to calibrate the CGM system. Glycemic metrics calculated from the CGM data included TIR, GRI, MBG, CV, TAR level 1, TAR level 2, TBR level 1, TBR level 2, and maximum glucose fluctuation range (LAGE). The methodology used to calculate these blood glucose indicators was consistent with standardized CGM metrics for clinical care, as described previously[15,19,30].

Statistical analysis

We performed a sample size calculation prior to data collection using the EPV method. This method assumes a minimum of 10 events per variable in a logistic regression model. Given a maximum of 20 variables in our logistic regression model, a minimum sample size of 200 DF patients (20 variables * 10 patients/variable) was determined. Continuous variables were presented as the mean ± SD for normally distributed data or as the median (25^th-75^th percentiles) for skewed distributions. Categorical variables were expressed as frequencies and percentages. Differences between the two groups were compared using a student’s t-test or Mann-Whitney U test for continuous variables and the χ² test for categorical variables. A restricted cubic spline model was employed to assess nonlinear associations between CGM parameters (MBG, TIR, GRI, TAR level 1, and TAR level 2) as continuous variables and the risk of DF. Binary logistic regression analyses were conducted to explore the effects of a 5% increase in MBG, a 10% increase in TIR, TAR level 1, and TAR level 2, and a 10-unit increase in GRI on the risk of DF occurrence after adjusting for confounding factors. In addition, in patients with DF, we further investigated potential factors influencing the performance of TIR, TAR level 1, and TAR level 2. Receiver operating characteristic (ROC) curve analyses were used to determine cutoff points and assess the performance of the aforementioned CGM indicators in predicting the occurrence of DF in T2DM patients. Statistical analyses were performed using SPSS version 27.0 software (SPSS, Inc., Chicago, IL, United States) and R statistical version 4.0.2 software (R Foundation for Statistical Computing). A P value < 0.05 (two-sided) was considered statistically significant.

RESULTS

Basic characteristics of patients with and without DF

A total of 591 patients with T2DM were enrolled in this study, comprising 294 patients with NDF and 297 patients with DF. Among the DF patients, ten underwent major or minor amputations (3.37%). The normality of the data for both the DF and NDF groups was evaluated using the Kolmogorov-Smirnov test. The results indicated that most variables in both groups significantly deviated from normality (P < 0.05) (Supplementary Table 1). Table 1 presents the demographic and clinical characteristics of the included participants. Compared to the control group, patients in the DF group were more likely to be male, older, have a longer duration of diabetes, and have a smaller waist-hip ratio (all P < 0.05). Fewer current smokers [107 (36.39%) vs 82 (27.61%), P = 0.022] and alcohol drinkers [98 (33.33%) vs 67 (22.56%), P = 0.004] were observed in the DF group. Compared to NDF patients, DF patients tended to have higher rates of hypertension [165 (56.12%) vs 196 (65.99%), P = 0.014], diabetic autonomic neuropathy (DAN) [49 (16.67%) vs 88 (29.63%), P = 0.001], and diabetic nephropathy (DN) [126 (42.86%) vs 167 (56.23%), P = 0.001], while lower rates of DR [143 (48.64%) vs 119 (40.07%), P = 0.036]. No significant differences were observed in BMI, history of diabetes, diabetic peripheral neuropathy (DPN), or diabetic peripheral vascular disease (DPVD). DF patients exhibited higher white blood cell (WBC) count and HbA1c levels but lower levels of hemoglobin, TC, TG, and ALB compared to NDF patients (all P < 0.05). No differences were observed in levels of HDL-C, LDL-C, aspartate aminotransferase, alanine aminotransferase, eGFR, or fasting plasma glucose between the DF and NDF groups (all P > 0.05). Patients in the NDF group exhibited significantly lower left lower limb ankle-brachial index (ABI) values [1.03 (1.01-1.06) vs 1.08 (1.01-1.15), P < 0.001] and right lower limb ABI values [1.06 (1.03-1.10) vs 1.08 (1.02-1.15), P = 0.026] compared to the DF group. Regarding antidiabetic treatment, the proportion of patients solely using insulin was higher in the DF group [133 (44.78%) vs 49 (16.67%), P < 0.001]. Conversely, the proportions of patients using non-insulin medications [24 (8.08%) vs 78 (26.53%), P < 0.001] and those using both insulin and non-insulin medications [140 (47.14%) vs 167 (56.80%), P = 0.019] were lower in the DF group. Fasting insulin levels were comparable between the two groups, with DF patients showing a value of 10.55 (5.52-19.05) µIU/mL compared to 9.60 (5.69-15.60) µIU/mL in the NDF group (P = 0.350). However, C-peptide levels were significantly different, with DF patients exhibiting lower levels at 1.59 (0.86-2.59) ng/mL compared to 1.87 (1.24-2.59) ng/mL in the NDF group (P = 0.045). Islet function parameters were assessed between the two groups. No significant difference was observed in the homeostasis model assessment of beta cell function between the two groups, with DF patients at 47.72 (22.38-125.24) and NDF patients at 44.97 (24.39-97.91) (P = 0.546). Similarly, homeostasis model assessment of insulin resistance values were also comparable, showing 3.42 (2.10-7.11) in the DF group and 3.55 (1.74-6.01) in the NDF group (P = 0.376).

Table 1 Comparison of general clinical data between the non-diabetic foot and diabetic foot groups, n (%).

Variables	NDF (n = 294)	DF (n = 297)	P value
Age, years	60.00 (52.00-69.00)	64.00 (54.00-71.00)	0.045
Male	160 (54.42)	189 (63.64)	0.023
BMI, kg/m2	24.23 (22.50-26.54)	24.34 (22.22-26.25)	0.672
WHR	0.93 (0.89-0.96)	0.90 (0.90-1.00)	< 0.001
Diabetes duration, months	72.00 (12.00-144.00)	120.00 (60.00-216.00)	< 0.001
Family history of diabetes	111 (37.76)	124 (41.75)	0.321
Smoking (current)	107 (36.39)	82 (27.61)	0.022
Drinking (current)	98 (33.33)	67 (22.56)	0.004
Presence of hypertension	165 (56.12)	196 (65.99)	0.014
Amputation	0	10 (3.37)	< 0.001
Left lower limb ABI	1.03 (1.01-1.06)	1.08 (1.01-1.15)	< 0.001
Right lower limb ABI	1.06 (1.03-1.10)	1.08 (1.02-1.15)	0.026
Antidiabetic drugs
Non insulin	78 (26.53)	24 (8.08)	< 0.001
Insulin	49 (16.67)	133 (44.78)	< 0.001
Both	167 (56.80)	140 (47.14)	0.019
Diabetic complications
DPN	286 (97.28)	282 (94.95)	0.143
DPVD	230 (78.23)	245 (82.49)	0.192
DAN	49 (16.67)	88 (29.63)	< 0.001
DN	126 (42.86)	167 (56.23)	0.001
DR	143 (48.64)	119 (40.07)	0.036
Biochemical indexes
WBC, × 109/L	6.68 (5.58-7.96)	7.30 (6.05-9.00)	0.000
NE, × 109/L	4.13 (3.21-4.99)	4.60 (3.70-6.40)	0.000
Hb, g/L	144.00 (132.00-155.00)	134.00 (118.00-148.50)	0.000
TC, mmol/L	4.66 (3.73-5.44)	4.20 (3.40-4.90)	0.000
TG, mmol/L	1.72 (1.12-2.55)	1.40 (0.90-2.00)	0.000
H-DLC, mmol/L	0.99 (0.83-1.18)	1.00 (0.80-1.10)	0.366
L-DLC, mmol/L	2.58 (1.89-3.18)	2.40 (1.80-3.10)	0.085
ALB, g/L	39.20 (36.55-41.70)	36.30 (32.40-39.50)	0.000
AST, U/L	19.00 (15.00-23.00)	19.000 (15.00-25.00)	0.812
ALT, U/L	19.00 (14.00-27.00)	18.00 (13.00-26.00)	0.072
eGFR, mL/min/1.73 m2	83.66 (65.15-109.49)	79.14 (57.83-102.70)	0.086
FPG, mmol/L	7.44 (5.54-10.31)	7.70 (5.70-10.45)	0.342
HbA1c, %	8.80 (7.20-10.53)	9.50 (7.75-11.00)	0.002
Islet function
FINS, μIU/mL	9.60 (5.69-15.60)	10.55 (5.52-19.05)	0.350
C-P, ng/mL	1.87 (1.24-2.59)	1.59 (0.86-2.59)	0.045
HOMA-β	44.97 (24.39-97.91)	47.72 (22.38-125.24)	0.546
HOMA-IR	3.55 (1.74-6.01)	3.42 (2.10-7.11)	0.376

Bold values indicate P < 0.05. Data were expressed as medians (25^th-75^th percentiles) for skewed distribution variables. Categorical variables were expressed as frequency (percentage). NDF: Non-diabetic foot; DF: Diabetic foot; BMI: Body mass index; WHR: Waist-to-hip ratio; ABI: Ankle-brachial index; DPN: Diabetic peripheral neuropathy; DPVD: Diabetic peripheral vascular disease; DAN: Diabetic autonomic neuropathy; DN: Diabetic nephropathy; DR: Diabetes retinopathy; WBC: White blood cell; NE: Neutrophils; Hb: Hemoglobin; TC: Total cholesterol; TG: Triglyceride; HDL-C: High-density lipoprotein cholesterol; LDL-C: Low-density lipoprotein cholesterol; ALB: Albumin; AST: Aspartate aminotransferase; ALT: Alanine aminotransferase; eGFR: Estimated glomerular filtration rate; FPG: Fasting plasma glucose; HbA1c: Hemoglobin A1c; FINS: Fasting insulin; C-P: C-Peptide; HOMA-β: Homeostasis model assessment of β cell function; HOMA-IR: Homeostasis model assessment of insulin resistance.

CGM-derived metrics

Compared to individuals with NDF, individuals with DF exhibited higher MBG levels and increased proportions of TAR, TAR level 1, TAR level 2, and GRI. Conversely, the proportions of TIR and TBR level 2 were lower in the DF group (Table 2).

Table 2 Comparison of continuous glucose monitoring-derived metrics between the non-diabetic foot and diabetic foot groups.

Variables	NDF (n = 294)	DF (n = 297)	P value
CGM-derived metrics
MBG, mmol/L	8.23 (7.33-9.81)	9.20 (8.00-10.70)	< 0.001
SD, mmol/L	2.34 (1.80-2.92)	2.40 (1.80-3.25)	0.129
CV, %	26.80 (22.11-32.80)	26.32 (20.56-33.78)	0.496
LAGE	11.82 (8.98-14.71)	11.50 (8.80-15.25)	0.929
GRI	24.52 (10.21-42.77)	35.98 (18.87-59.29)	< 0.001
TIR, %	76.90 (58.00-88.73)	65.91 (43.68-81.95)	< 0.001
TAR, %	20.26 (9.37-41.59)	32.57 (16.26-55.43)	< 0.001
TAR level 1, %	17.03 (8.44-31.71)	25.35 (13.56-38.90)	< 0.001
TAR level 2, %	2.81 (0.00-8.60)	5.04 (0.00-14.97)	< 0.001
TBR, %	0.00 (0.00-1.24)	0.00 (0.00-0.69)	0.114
TBR level 1, %	0.00 (0.00-1.08)	0.00 (0.00-0.62)	0.197
TBR level 2, %	0.00 (0.00-0.00)	0.00 (0.00-0.00)	0.031
CGM-derived indicator achievement, n (%)
CV target	250 (85.03)	240 (80.81)	0.172
TIR target	180 (61.22)	128 (43.10)	< 0.001
TAR level 1 target	196 (66.67)	146 (49.16)	< 0.001
TAR level 2 target	182 (61.90)	148 (49.83)	0.003
TBR level 1 target	262 (89.12)	269 (90.57)	0.558
TBR level 2 target	279 (94.90)	284 (95.62)	0.678

Bold values indicate P < 0.05. Data were expressed as medians (25^th-75^th percentiles) for skewed distribution variables. Categorical variables were expressed as frequency (percentage). NDF: Non-diabetic foot; DF: Diabetic foot; CGM: Continuous glucose monitoring; MBG: Mean blood glucose; CV: Coefficient of variation; LAGE: Maximum glucose fluctuation range; GRI: Glycemia risk index; TIR: Time in range; TAR: Time above range; TBR: Time below range.

The achievement rates of CGM targets based on DF status are presented in Table 2. The targets were selected in accordance with the latest ADA recommendations[30]. Compared to patients with NDF, patients with DF exhibited significantly lower rates of achieving the targets for TIR > 70%, TAR level 1 < 25%, and TAR level 2 < 5%. However, no significant differences were observed in the achievement rates of CV ≤ 36%, TBR level 1 < 4%, and TBR level 2 < 1% between the two groups.

Associations of CGM metrics with the risk of DF

Figure 1 illustrates the distribution of DF patients classified using the Wagner grading system. The chart shows that the majority of patients are classified as Grade 2 (27.27%), indicating mild infection. Grade 1 (21.55%) and Grade 0 (16.16%) follow, suggesting no infection. Grade 3 (24.58%) and Grade 4 (10.44%) also have certain proportions, representing deep ulcers with osteomyelitis and localized gangrene, respectively. It is noteworthy that there are no patients in Grade 5. As shown in Figure 2, the restricted cubic spline model indicated a linear relationship between MBG, TIR, GRI, TAR level 1, and TAR level 2, and the risk of DF (P for nonlinearity = 0.887, 0.822, 0.823, 0.197, and 0.376, respectively). Compared to the reference points, patients with MBG > 8 mmol/L, TAR level 1 > 25%, TAR level 2 > 5%, GRI > 30, and TIR < 70% had a significantly higher risk of DF.

Figure 1 Grading of diabetic foot patients (according to the Wagner grading system).

Figure 2 Restricted cubic spline model demonstrating the linear relationship of mean blood glucose, glycemia risk index, time above range level 1, time above range level 2, and time in range with the risk of diabetic foot. A: Mean blood glucose; B: Glycemia risk index; C: Time above range (TAR) level 1; D: TAR level 2; E: Time in range. MBG: Mean blood glucose; GRI: Glycemia risk index; TAR: Time above range; TIR: Time in range.

Logistic regression analyses revealed that MBG, GRI, and TAR level 1 were positively correlated with the risk of DF, while TIR showed a negative correlation in both crude and adjusted models (adjusted for age, sex, BMI, WHR, diabetes duration, smoking, drinking, presence of hypertension, DAN, DN, DR, WBC, Hb, TG, TC, ALB, HbA1c, ABI, and antidiabetic drug use; all P < 0.05, Table 3). A 10% increase in TIR was associated with a nearly 13.3% decreased risk of DF (OR = 0.867, 95%CI: 0.784-0.957, P = 0.005) (Table 3, Model 5). Additionally, a 5% increase in MBG and a 10% increase in TAR level 1 were associated with a 125.5% and 22.5% increase in the incidence of DF, respectively (Model 5). Although TAR level 2 was positively associated with DF, this association was not statistically significant after adjusting for HbA1c in model 4 (P = 0.250, Table 3). After adjusting for the aforementioned multivariable factors, a 10-unit increase in GRI was associated with an 11.5% increased risk of developing DF (P = 0.012). Multivariate logistic regression analyses examined the association between CGM-derived metrics and the risk of DF. Results (Supplementary Table 2) showed that in both individuals with Wagner grade < 2 DF and those with Wagner grade ≥ 2 DF, higher MBG, GRI, and TAR level 1 and 2 were significantly associated with increased DF risk in unadjusted models (Model 1). After adjusting for potential confounders (Model 2), MBG, GRI, and TAR level 1 remained significant predictors of DF risk in both groups; while the association between TAR level 2 and DF risk was not statistically significant. Conversely, TIR was inversely associated with DF risk in both adjusted and unadjusted models for both groups.

Table 3 Logistic regression analysis of the correlations between continuous glucose monitoring-derived metrics and the risk of diabetic foot.

Variable	Multivariate logistic regression
	β	OR	95%CI	P value
MBG
Model 1	1.056	2.876	1.876-4.409	< 0.001
Model 2	0.948	2.582	1.642-4.058	< 0.001
Model 3	0.998	2.712	1.697-4.334	< 0.001
Model 4	0.654	1.923	1.103-3.353	0.021
Model 5	0.813	2.255	1.222-4.163	0.009
TIR
Model 1	-0.182	0.834	0.776-0.895	< 0.001
Model 2	-0.167	0.846	0.784-0.913	< 0.001
Model 3	-0.180	0.836	0.772-0.904	< 0.001
Model 4	-0.114	0.892	0.814-0.978	0.015
Model 5	-0.143	0.867	0.784-0.957	0.005
TAR level 1
Model 1	0.249	1.283	1.156-1.424	< 0.001
Model 2	0.230	1.258	1.126-1.406	< 0.001
Model 3	0.243	1.275	1.138-1.429	< 0.001
Model 4	0.173	1.189	1.044-1.353	0.009
Model 5	0.203	1.225	1.065-1.407	0.004
TAR level 2
Model 1	0.258	1.294	1.121-1.493	< 0.001
Model 2	0.236	1.266	1.088-1.473	0.002
Model 3	0.253	1.287	1.099-1.508	0.002
Model 4	0.104	1.109	0.930-1.324	0.250
Model 5	0.168	1.183	0.970-1.442	0.098
GRI
Model 1	0.141	1.151	1.084-1.223	< 0.001
Model 2	0.131	1.140	1.068-1.216	< 0.001
Model 3	0.141	1.151	1.077-1.231	< 0.001
Model 4	0.084	1.087	1.007-1.175	0.033
Model 5	0.109	1.115	1.024-1.213	0.012

Odds ratios represent per 5% increase in mean blood glucose, per 10% increase in time in range, time above range (TAR) level 1, TAR level 2, and per 10-unit increase in glycemia risk index. Model 1: Non-adjusted; Model 2: Adjusted for age, sex, body mass index, waist-to-hip ratio, diabetes duration, smoking, drinking; Model 3: Model 2+ presence of hypertension, diabetic autonomic neuropathy, diabetic nephropathy, diabetes retinopathy; Model 4: Model 3+ white blood cell, hemoglobin, triglyceride, total cholesterol, albumin, and hemoglobin A1c; Model 5: Model 4+ left lower limb ankle-brachial index (ABI), right lower limb ABI, and antidiabetic drugs. MBG: Mean blood glucose; GRI: Glycemia risk index; TAR: Time above range; TIR: Time in range.

ROC analysis for the risk of DF

ROC curve analysis was next conducted to predict the risks of DF using the CGM indexes (including MBG, TIR, GRI, and TAR level 1) and HbA1c (Figure 3). The area under curve of the MBG, TIR, GRI, TAR level 1, and HbA1c was 0.619 (95%CI: 0.578-0.658), 0.617 (95%CI: 0.576-0.656), 0.609 (95%CI: 0.569-0.649), 0.607 (95%CI: 0.566-0.646), and 0.574 (95%CI: 0.533-0.614), respectively. The optimal cutoff points of MBG, TIR, TAR level 1, TAR level 2, and HbA1c for predicting the DF were 8.69 (Youden index = 0.201), 77.78 (Youden index = 0.208), 19.23 (Youden index = 0.186), 20.63 (Youden index = 0.215), and 9.03 (Youden index = 0.151), respectively (all P < 0.05).

Figure 3 Receiver-operating characteristic analysis for identification of patients with the risk of diabetic foot. ROC: Receiver operating characteristic; TIR: Time in range; GRI: Glycemia risk index; MBG: Mean blood glucose; TAR: Time above range; HbA1c: Hemoglobin A1c.

Factors affecting CGM metric performance

To further investigate the factors influencing the achievement of recommended CGM targets in patients with DF, we performed univariate and multivariate regression analyses. As shown in Figure 4, after adjusting for sex, age, BMI, duration of diabetes, duration of DF, DPN, FPG, Wagner grade, and antidiabetic drug use, WBC and HbA1c were negatively correlated with achieving recommended TIR targets. Similarly, WBC, FPG, and HbA1c inversely influenced the performance of TAR level 1 after adjusting for sex, age, BMI, duration of diabetes, duration of DF, FPG, and antidiabetic drug use. The performance of the TAR level 2 target was negatively influenced by the duration of diabetes, WBC, HbA1c, and antidiabetic drug use but positively influenced by BMI after adjusting for the same confounders.

Figure 4 Logistic regression analysis of the proportion of continuous glucose monitoring metrics meeting the standards and their influencing factors. A: The attainment of time in range > 70%; B: The attainment of time above range (TAR) level 1 < 25%; C: The attainment of TAR level 2 < 5%. ¹Adjusted for sex, age, body mass index (BMI), duration of diabetes, duration of diabetic foot, diabetic peripheral neuropathy, white blood cell (WBC), fasting plasma glucose (FPG), hemoglobin A1c (HbA1c), Wagner grade, antidiabetic drugs. ²Adjusted for sex, age, BMI, duration of diabetes, duration of diabetic foot, WBC, FPG, HbA1c, antidiabetic drugs. TAR: Time above range; TIR: Time in range. WBC: White blood cell; HbA1c: Hemoglobin A1c; BMI: Body mass index; FPG: Fasting plasma glucose.

DISCUSSION

In this study, we assessed the relationships between CGM-derived TIR, GRI, and other key metrics and the risk of DF in patients with T2DM. An inverse relationship between TIR and the presence of DF remained statistically significant after adjusting for confounders. However, MBG, TAR level 1, TAR level 2, and GRI were positively associated with an increased risk of DF after adjusting for clinical risk factors. Significantly, this relationship was not observed for TBR. Notably, the correction for certain covariates, particularly HbA1c, attenuated the association between TAR level 2 and the risk of DF (Table 3, Model 4).

Early intensive glycemic control can reduce the risk of DF and lessen the severity of adverse DF outcomes, such as osteomyelitis and amputation[24,31-33]. HbA1c, a metric for average glycemic levels, is significantly associated with an increased risk of both macrovascular and microvascular complications in individuals with type 1 diabetes mellitus (T1DM) and T2DM[14,33]. However, it provides no information on glycemic variability (GV) or the daily pattern of glycemic levels (e.g., hypoglycemic or hyperglycemic events). Furthermore, physiological or pathological factors can influence HbA1c detection, making it an imprecise representation of true blood glucose levels. With the increasing use of CGM, new indices like TIR, TAR, TBR, MBG, and LAGE have been identified to provide a more detailed view of blood glucose control, addressing the limitations of HbA1c. Numerous studies have demonstrated a strong link between CGM indices, especially TIR, and diabetic complications like retinopathy, nephropathy, and neuropathy[15,21,34]. Decreased TIR is associated with a higher risk of DF due to peripheral neuropathy and lower extremity arterial disease[35,36]. Elevated blood sugar can exacerbate inflammation and vascular damage, leading to arterial disease[37]. A 10% decrease in TIR has been linked to a 7% higher risk of lower limb atherosclerosis in T2DM patients[38].

In recent years, few studies have explored the link between CGM indices and DF risk. Our study found that as TIR quartiles increased, the incidence of DF decreased (P for trend < 0.05). Multinomial logistic regression revealed that a 10% increase in TIR corresponded to a 13.3% lower DF risk, even after adjusting for HbA1c and other factors. Additionally, the present study suggests varying degrees of DF risk are associated with increases in MBG, TAR level 1, and TAR level 2. According to ADA guidelines, the recommended target TIR for most patients with T2DM is above 70%[30]. Our study found that patients with DF presented significantly lower TIR compliance than those without DF. DF patients exhibited a higher percentage of TAR level 1 and level 2, with no significant difference in TBR. This suggests that the decrease in TIR was primarily due to increased TAR, indicating that DF patients generally experience high blood glucose levels with few hypoglycemic events. Previous studies have demonstrated a linear relationship between TIR and HbA1c, consistent with the results of our study (r = -0.470, P < 0.001)[10,39]. Our results, adjusted for HbA1c, suggest that TIR, MBG, and TAR level 1 are independently associated with DF risk in T2DM patients. We propose that MBG, along with TAR level 1 and level 2, could complement TIR for a more comprehensive evaluation of glycemic control.

In addition to TIR, GV metrics obtained from CGM, such as SD and CV, have gained significant attention as independent predictors of diabetes complications, including retinopathy and autonomic neuropathy[15,40]. Both in vitro and in vivo studies have demonstrated an association between GV and endothelial dysfunction and oxidative stress[41,42]. However, our study did not find a statistically significant difference in either GV (including LAGE, CV, and SD) or TBR between T2DM patients with and without DF.

Considering that TIR does not fully account for out-of-range glycemic profiles, particularly hypoglycemic episodes, we also calculated the GRI, a novel composite CGM metric. Recently introduced and validated by 330 experienced clinicians in CGM analysis and interpretation[43], the GRI has proven useful in assessing the overall risk of hypoglycemia and hyperglycemia in both pediatric and adult T1DM patients[44,45]. Additionally, research has linked a higher GRI to an increased risk of developing DR, albuminuria, and arterial stiffness[46-48]. To our knowledge, this study is the first to explore the relationship between the GRI and DF using CGM recordings. Our findings indicate a significant association between a higher GRI and a greater incidence of DF. Logistic regression analysis revealed that a 10-unit increase in GRI was correlated with an approximately 15% increased risk of DF (Table 3, Model 1). This significant association persisted after adjusting for confounding factors. These results suggest that GRI could be a valuable clinical tool for assessing diabetes-related complications. Further research is warranted to explore this potential.

As previously mentioned, numerous studies have established a strong link between CGM-derived parameters and diabetes complications. Our study further supports this connection, identifying correlations between CGM indicators and DF occurrence. However, few studies have explored the factors influencing the attainment of CGM measures in patients with DF. Recent research has demonstrated that the frequency of blood glucose monitoring and the use of treatment modalities, such as insulin pumps, can significantly impact TIR achievement in individuals with T1DM[49-51]. To our knowledge, this study is the first to investigate potential factors affecting CGM attainment in patients with DF. After adjusting for confounding factors, we found that higher levels of HbA1c and WBC were associated with lower target rates of TIR (> 70%), TAR level 1 (< 25%), and TAR level 2 (< 5%). According to Beck et al[39], a TIR of 70% corresponds to an HbA1c level of 7%, while a TIR of 50% corresponds to an HbA1c level of 8%. Additionally, each 10% increase in TIR led to a 0.6% decrease in HbA1c. A recent comprehensive analysis of 18 studies revealed a strong association between HbA1c and TIR (r = -0.84), indicating that a 0.8% change in HbA1c corresponds to a 10% change in TIR[52]. These results confirm a significant correlation between HbA1c and the attainment of the recommended TIR target. Therefore, considering both HbA1c and TIR is advisable for assessing the blood glucose management of diabetic patients with DF. Additionally, FPG, diabetes duration, and antidiabetic drug use were negatively correlated with the achievement of level 1 and level 2 targets, respectively. However, the severity of DF did not affect the attainment of TIR and TAR. It is well-established that DF is driven by hyperglycemia, inflammation, neuropathy, and vascular disease, with poor glycemic control and long diabetes duration being key risk factors[53,54]. Controlling blood sugar is crucial for reducing DF risk and achieving CGM targets. These findings emphasize the importance of incorporating CGM-derived metrics like TIR and GRI alongside HbA1c for optimal diabetes management. ROC curve analysis demonstrated that TIR, GRI, MBG, and TAR level 1 may be more sensitive than HbA1c in predicting the risk of developing DF. This suggests that further research is warranted to determine the threshold values of changes in GRI and MBG for predicting DF.

In the present study, DF patients were predominantly male, older, and had a longer duration of diabetes but a lower waist-hip ratio than NDF patients. These findings suggest that male sex and diabetes duration may be risk factors for the development of DF[55,56]. In our study, we adopted a broad definition of DPVD, encompassing lower limb involvement and conditions affecting other vascular regions, such as the upper limb[57,58]. Additionally, we observed a higher number of current smokers in the NDF group compared to the DF group. Smoking is known to increase the risk of peripheral vascular disease, which may help explain the lack of statistically significant differences in DPVD prevalence between patients with DF and those without[58]. Elevated WBC counts in individuals with DF may indicate the presence of infection. Our study observed that elevated WBC counts negatively impacted the performance of TAR level 2. It is well-established that inflammatory states increase insulin resistance, making glycemic control more challenging for patients[59]. Additionally, clinical research supports the notion that high WBC counts are a risk factor for the development of T2DM[60,61]. DF patients with low levels of ALB and hemoglobin were found to correlate with chronic illness consumption and malnutrition.

This study has several limitations that should be acknowledged. Firstly, the sample size was limited, particularly regarding toe amputation cases, which consisted of only 10 cases. Consequently, we were unable to examine the relationship between CGM parameters and adverse outcomes, such as DF amputation, as previous research studies have done. Future studies will require a larger sample size and the inclusion of amputation patients with more severe outcomes. Secondly, analyzing the relationship between CGM metrics and specific diabetes foot outcomes, such as amputation rates, ulcer healing rates, hospitalization duration, ulcer healing speed, and infection control, is challenging. Thirdly, this study was a retrospective analysis of patients with T2DM in the Department of Endocrinology. Future multicenter and prospective studies with larger sample sizes are needed to further investigate these relationships. Fourth, the three-day CGM data, while insufficient to fully evaluate long-term glycemic control, offers a more accurate assessment of real-time glucose fluctuations compared to HbA1c, which reflects a longer-term average. This real-time monitoring of glucose variability is crucial because such variability contributes to various complications in type 2 diabetes patients[15,62]. Furthermore, short-term glucose variability impacts patient quality of life and treatment adherence, thus influencing long-term outcomes[63].

CONCLUSION

In summary, this study demonstrated that TIR was negatively associated with DF risk, while GRI, MBG, and TAR level 1 were positively associated with DF risk. These associations remained significant even after adjusting for other risk factors, including HbA1c. Our findings suggest that these CGM-derived indices, which provide intuitive measures of glycemic control, may be useful for identifying the occurrence of DF in patients with T2DM. Further prospective studies are needed to validate these findings and gain a comprehensive understanding of the role of CGM-derived metrics in the onset and progression of DF. Additionally, future research should explore the underlying pathogenic mechanisms involved in these associations.

ACKNOWLEDGEMENTS

We would like to express our sincere gratitude to Professor Meng Qiong for her invaluable statistical support throughout this study.