Clinical data and insulin resistance index of patients.
The clinical data and insulin resistance index of the two groups are presented in Table 1. In comparison to the NC group, the T2DM group exhibited significantly higher levels of HbA1c, FPG, and TG. There was a notable decrease in HDL-C; TyG-BMI and METS-IR showed an increase, with statistically significant differences observed (P<0.001).
Table 1 Comparison of relevant clinical information between two groups,M(Q1,Q2)
|
Variable
|
NC
N=70
|
T2DM
N=160
|
t(Z)
|
P
|
General index
|
Age(year)
|
40.59±8.01
|
39.21±10.02
|
1.111
|
0.268
|
|
SBP(mmHg)
|
128.52±14.52
|
129.76±14.01
|
0.117
|
0.907
|
|
DBP(mmHg)
|
83.34±10.52
|
85.62±10.26
|
0.064
|
0.949
|
|
BMI(kg/m2)
|
25.24±3.47
|
25.08±4.79
|
1.121
|
0.264
|
|
WC(cm)
|
87.63±11.57
|
89.92±10.49
|
1.476
|
0.141
|
|
FPG(mmol/l)
|
4.73±0.48
|
13.94±5.08
|
13.528
|
0.000
|
|
HbA1c%
|
8.39±1.09
|
10.67±2.23
|
18.422
|
0.000
|
|
TG(mmol/l)
|
2.39±2.48
|
2.62±2.39
|
18.156
|
0.000
|
|
TC(mmol/l)
|
5.24±0.84
|
5.25±1.40
|
0.154
|
0.878
|
|
LDL-c(mmol/l)
|
3.32±0.83
|
3.36±0.93
|
0.997
|
0.320
|
|
HDL-c(mmol/l)
|
1.69±0.32
|
1.40±0.88
|
5.548
|
0.000
|
|
Cr(umol/l)
|
78.44±22.32
|
73.59±15.04
|
1.661
|
0.100
|
IR index
|
TyG
|
8.02±0.21
|
8.32±0.87
|
0.000
|
1.000
|
|
TyG-BMI
|
168.40±38.39
|
210.29±45.60
|
6.713
|
0.000
|
|
TG/HDL-c
|
1.83(1.15,4.36)
|
2.01(0.90,6.48)
|
0.118
|
0.906
|
|
METS-IR
|
47.25±10.51
|
50.75±11.80
|
2.138
|
0.034
|
Abbreviations:SBP,systolic blood pressure; DBP,Diastolic blood pressure; BMI,body mass index; WC,waist circumference; FPG,fasting plasma glucose; HbA1c,glycosylated hemoglobin; HDL-C,high-densitylipoprotein cholesterol; LDL-C,low-density lipoprotein cholesterol; TyG, triglyceride glucose; TG,triglyceride; TC,Serum total cholesterol; METS-IR, Insulin resistance metabolic score
Ultrasound indices of patients
In comparison to the NC group, the T2DM group exhibited increased left atrium (LA) size and left ventricular end-diastolic diameter (LVEDD), as well as thickened interventricular septum (IVS). Furthermore, GLPS-LAX, GLPS-A4C, GLPS-A2C, and GLPS-AVG decreased significantly (P < 0.001)(Table2).
Table 2 Comparison of relevant ultrasound indexes between two groups
Variable
|
Controls
N=70
|
T2DM
N=160
|
t
|
P
|
LA(mm)
|
30.43±3.25
|
32.96±4.09
|
3.027
|
0.003
|
LVEDD(mm)
|
47.33±3.52
|
43.56±10.93
|
2.536
|
0.012
|
IVST(mm)
|
8.87±1.33
|
9.53±1.18
|
4.492
|
0.000
|
LVPWD(mm)
|
8.94±1.32
|
9.00±1.07
|
0.688
|
0.492
|
LVEF(%)
|
65.87±4.35
|
65.75±9.01
|
0.725
|
0.469
|
GLPS-LAX(%)
|
21.08±2.43
|
18.55±2.92
|
6.416
|
0.000
|
GLPS-A4C(%)
|
18.87±3.33
|
16.73±3.05
|
6.008
|
0.000
|
GLPS-A2C(%)
|
20.45±2.94
|
18.18±2.99
|
6.315
|
0.000
|
GLPS-AVG(%)
|
20.07±2.64
|
17.79±2.52
|
7.648
|
0.000
|
E/e'
|
8.70±2.39
|
9.37±0.37
|
1.335
|
0.183
|
Abbreviations:LA,left atrium;LVEDD,Left ventricular end-diastolic diameter;LVESD,Left ventricular end systolic diameter;IVST,interventricular septal thickness;LVPWD,Left ventricular posterior wall diameter;GLPS-LAX,Longitudinal strain on left ventricular long axis;GLPS-A4C,Left ventricular four-chamber longitudinal strain;GLPS-A2C,Left ventricular two-cavity long axis strain;GLPS-AVG,Left ventricular average longitudinal strain;E,Mitral valve orifice early peak blood flow velocity;e',Early peak mitral ring blood flow velocity
The occurrence and comparison of abnormal GLPS-AVG in the T2DM group
The OGTT derived IR index model showed higher BMI, SBP, and LA, and lower HDL-c, GLPS-LAX, GLPS-A4C, GLPS-A2C, and GLPS-AVG; TyG-BMI, TG/HDL-c and METS-IR were significantly increased in abnormal GLPS-AVG group(P < 0.001)(Table 3).
Table 3 Incidence and comparison of abnormal GLPS-AVG in the T2DM group
Variable
|
All T2DM
|
Abnormal GLPS-AVG
|
Normal GLPS-AVG
|
|
|
N=160
|
N=32
|
N=128
|
t(Z)
|
P
|
BMI(kg/m2)
|
25.08±4.79
|
29.29±4.61
|
23.77±4.07
|
6.679
|
0.000
|
SBP(mmHg)
|
129.76±14.01
|
136.77±13.09
|
127.60±13.79
|
3.398
|
0.001
|
DBP(mmHg)
|
85.62±10.26
|
84.15±15.01
|
86.07±8.47
|
0.696
|
0.491
|
FPG(mmol/L)
|
13.94±5.08
|
15.41±5.10
|
13.49±5.44
|
1.807
|
0.073
|
HDL-c(mmol/L)
|
1.40±0.88
|
1.18±0.37
|
1.46±0.69
|
3.131
|
0.002
|
LA(mm)
|
32.96±4.09
|
35.46±3.91
|
32.19±3.87
|
4.267
|
0.000
|
LVPWD(mm)
|
9.00±1.07
|
8.95±1.05
|
9.06±0.43
|
0.581
|
0.565
|
IVSD(mm)
|
9.53±1.18
|
9.77±1.17
|
9.46±1.19
|
1.322
|
0.188
|
GLPS-LAX(%)
|
18.55±2.92
|
16.03±2.31
|
19.33±2.65
|
6.455
|
0.000
|
GLPS-A4C(%)
|
16.73±3.05
|
12.83±1.83
|
17.93±2.24
|
11.915
|
0.000
|
GLPS-A2C(%)
|
18.18±2.99
|
15.12±1.43
|
19.12±2.69
|
11.526
|
0.000
|
GLPS-AVG(%)
|
17.79±2.52
|
14.59±0.89
|
18.78±1.96
|
17.904
|
0.000
|
TyG
|
8.32±0.87
|
8.53±1.06
|
8.25±0.81
|
1.396
|
0.170
|
TyG-BMI
|
210.29±45.60
|
248.53±37.96
|
198.44±41.34
|
6.227
|
0.000
|
TG/HDL-c
|
2.01(0.90,6.48)
|
3.05(1.82,7.33)
|
1.75(0.81,4.53)
|
(3.032)
|
0.000
|
METS-IR
|
50.75±11.80
|
62.01±10.99
|
47.27±9.78
|
7.436
|
0.000
|
The present study aims to conduct a multivariate logistic regression analysis to identify the influencing factors associated with abnormal GLPS-AVG in individuals with T2DM.
The cut-off value of GLPS-AVG in the NC group was determined to be 17.4%. Based on this threshold, it was found that 20% (32/160) of T2DM patients exhibited a decreased GLPS-AVG, indicating an abnormal GLPS-AVG status.
Regression design: A multivariate unconditional logistic regression model was established. Using the data from this study as the sample, we considered abnormal GLPS-AVG in the T2DM group as the dependent variable, assigning a value of 1 for abnormal GLPS-AVG in the T2DM group and 0 otherwise. Independent variables were selected based on univariate analysis (Table 1) with a significance level of P < 0.01 and adjusted according to clinical input. BMI, IVSD, LA, LVPWD, E/e', TyG, TyG-BMI, TG/HDL-c, METS-IR... along with other indicators were included as independent variables. Among them, BMI may be correlated with some non-insulin-based IR indicators (model II TyG-BMI and model III METS-IR), potentially causing collinearity issues. However, after consulting clinicians and statisticians it was determined that due to fusion and weight calculation of other variables in the model its influence was diluted to some extent resulting in weak collinearity; therefore it was retained temporarily in the sequence of independent variables. The assignment design for dummy variables is shown in Table 4. Stepwise regression method was employed for selection and elimination of independent variables using α elimination =0.10 and α inclusion =0.05.
The regression analysis revealed that BMI, IVSD, SPB, LA, and LVPWD were identified as significant risk factors for the reduction of GLPS-AVG (P < 0.05). Additionally, two IR index models (TyG-BMI and METS-IR) demonstrated a strong association with abnormal GLPS-AVG in the T2DM group (P < 0.001). Please refer to Table 5.
Table 4 Logistic regression analysis of GLPS-AVG reduced variable (dummy variable) assignment design
|
|
|
Assign 1
|
Assign 0 (reference)
|
comment
|
Dependent variable
|
Y
|
parameter
|
Abnormal
|
Normal
|
< 17.4% is abnormal
|
Independent variable
|
X1
|
BMI
|
≥25 kg/m2
|
<25 kg/m2
|
The independent variables were transformed into two categorical variables, known as dimensionality reduction treatment, which enhanced the feasibility of implementing robust regression in cases with limited sample size.
|
|
X2
|
IVSD
|
≥9.4mm
|
<9.4mm
|
|
X3
|
LA
|
≥32mm
|
<32mm
|
|
X4
|
HDL-c
|
≥1.40mmol/L
|
<1.40mmol/L
|
|
X5
|
LVPWD
|
≥9.0 mm
|
<9.0 mm
|
|
X6
|
E/e’
|
≥9.0
|
<9.0
|
|
X7
|
SBP
|
≥130mmHg
|
<130mmHg
|
|
X8
|
TyG
|
≥8.3
|
<8.4
|
|
X9
|
TyG-BMI
|
≥210
|
<210
|
|
X10
|
TG/HDL-c
|
≥2.1
|
<2.1
|
|
X11
|
METS-IR
|
≥50
|
<50
|
Table 5 Logistic regression analysis of risk factors for GLPS-AVG reduction
Variable
|
The assignment of dummy variables in regression analysis
|
B
|
Se
|
Waldc2
|
P
|
OR
|
95% CI
|
constant
|
-
|
-0.115
|
0.046
|
6.260
|
0.012
|
-
|
-
|
BMI
|
≥25 kg/m2=1,No=0
|
0.263
|
0.072
|
13.331
|
0.000
|
1.301
|
1.130~1.498
|
IVSD
|
≥9.4mm=1 ,No=0
|
1.224
|
0.532
|
5.298
|
0.021
|
3.402
|
1.199~9.644
|
SBP
|
≥130mmHg=1,No=0
|
0.373
|
0.158
|
5.590
|
0.018
|
1.452
|
1.066~1.978
|
LA
|
≥32mm=1,No=0
|
0.245
|
0.093
|
6.934
|
0.008
|
1.278
|
1.065~1.533
|
LVPWD
|
≥9.0 mm=1,No=0
|
1.088
|
0.508
|
4.594
|
0.032
|
2.967
|
1.097~8.028
|
TyG-BMI
|
≥210=1,<210=0
|
0.750
|
0.186
|
16.283
|
0.000
|
2.117
|
1.471~3.047
|
METS-IR
|
≥50=1,<50=0
|
0.835
|
0.255
|
10.760
|
0.001
|
2.305
|
1.399~3.796
|
The TyG, TyG-BMI, TG/HDL-c, and METS-IR model/index were employed to assess the predictive value for cardiac function deterioration in individuals with diabetes.
The results of the upper regression analysis revealed a significant association between TyG-BMI, METS-IR, BMI, IVSD, SBP, LA, LVPWD and other indicators of the IR model with impaired cardiac function/GLPS-AVG in individuals with diabetes. To establish the ROC prediction analysis model, the normal GLPS-AVG group (N=128) was utilized as negative samples while the abnormal GLPS-AVG group (N=32) served as positive samples.
(1) Screening and selection of predictors: Initially, the most relevant indicators in this study, namely TyG-BMI, METS-IR, and two other IR index models, will be retained. However, considering that the BMI index is already encompassed within the composition of the aforementioned IR model indicators, it is no longer necessary to include it. Additionally, IVSD and LVPWD exhibit a high degree of correlation as well as being similar types of indicators. Following consultation with both clinicians and statisticians, only IVSD will be retained since it holds utmost clinical significance in assessing left ventricular function among T2DM patients. (2) Individually applied: Five measures (TyG-BMI, METS-IR, IVSD, SBP, LA) were individually employed as described above. All the data consisted of continuous numerical values. The five indicators were categorized into multiple segments based on clinical practice guidelines. The software-fitted ROC curve yielded the maximum Youden index value, from which the theoretical threshold (appropriately rounded) and respective parameters were calculated. Sensitivity, specificity, and accuracy were determined based on the measured samples. (3) The combined application (LogP model) employed the logistic regression method to establish a risk assessment/prediction model for TyG-BMI, METS-IR, IVSD, SBP, and LA (refer to Table 6). The resulting regression risk prediction model Ln (P/1-P) = -0.111 +0.018×TyG-BMI +0.080×METS-IR +0.042×SBP +0.088×IVSD +0.070×LA was derived. This model value served as the virtual probability index for the combined utilization of these five indicators, followed by conducting ROC analysis. (4)The results of the ROC analysis demonstrated that TyG-BMI, METS-IR, IVSD, SBP, LA, and their combined application were all effective in predicting and evaluating cardiac dysfunction in individuals with diabetes. The corresponding ROC-AUC values (95% CI) for these predictors were 0.750 (0.564 ~ 0.934), 0.774 (0.582 ~ 0.944), 0.702 (0.461 ~ 0.948), 0.737 (0.478 ~ 0.983), and 0.726 (0.483 ~ 951) respectively; while the combined application of these five indices using the Log P model exhibited the highest predictive efficiency with an ROC-AUC value of 0.878 (0.770 ~0.987). Notably, its sensitivity, specificity, and accuracy were significantly higher than those achieved by each individual predictor alone as shown in Table7. Please refer to Figure1 for the corresponding ROC analysis curve.
Table 6 The construction of risk prediction models for the combined application of indicators is presented
Indicators/factors
|
Regression dummy variable assignment
|
β
|
Se
|
Wald χ2
|
P
|
OR
|
OR 0.95CI
|
constant
|
-
|
-0.111
|
0.046
|
5.901
|
0.015
|
-
|
-
|
TyG-BMI
|
Continuous numerical prototype input
|
0.018
|
0.005
|
12.518
|
0.000
|
1.018
|
1.008~1.028
|
METS-IR
|
Continuous numerical prototype input
|
0.080
|
0.020
|
15.302
|
0.000
|
1.083
|
1.041~1.128
|
SBP
|
Continuous numerical prototype input
|
0.042
|
0.017
|
6.407
|
0.011
|
1.043
|
1.010~1.077
|
IVSD
|
Continuous numerical prototype input
|
0.088
|
0.041
|
4.574
|
0.032
|
1.092
|
1.007~1.184
|
LA
|
Continuous numerical prototype input
|
0.070
|
0.024
|
8.639
|
0.003
|
1.072
|
1.024~1.124
|
Table 7 ROC analysis results of TyG-BMI, METS-IR, IVSD, SBP, LA and their combined application model
index
|
AUC(0.95CI)
|
threshold value
|
Sensitivity
(n/N)
|
Specificity
(n/N)
|
Youden index
|
accuracy rating
(n/N)
|
TyG-BMI
|
0.750(0.564~0.934)
|
210
|
0.750(24/32)
|
0.742(95/128)
|
0.492
|
0.744(119/160)
|
METS-IR
|
0.774(0.582~0.944)
|
50
|
0.750(24/32)
|
0.766(98/128)
|
0.516
|
0.763(122/160)
|
SBP(mmHg)
|
0.702(0.461~0.948)
|
130
|
0.688(22/32)
|
0.727(93/128)
|
0.415
|
0.719(115/160)
|
IVSD(mm)
|
0.737(0.478~0.983)
|
9.5
|
0.719(23/32)
|
0.734(94/128)
|
0.453
|
0.731(117/160)
|
LA(mm)
|
0.726(0.483~0.951)
|
32
|
0.719(23/32)
|
0.703(90/128)
|
0.422
|
0.706(113/160)
|
Collaborative application model(LogP)
|
0.878(0.770~0.987)
|
16.2
|
0.875(28/32)
|
0.875(112/128)
|
0.750
|
0.875(140/160)
|
Note:The thresholds were appropriately rounded in accordance with clinical practice. The threshold system of the joint application virtual indicators is calculated based on the Log (P/1-P) model, which includes the constant term.