Baseline characteristics of the study participants
Overall, 229 patients were enrolled in this study. Table 1 lists the baseline characteristics of subjects with and without MAFLD. Individuals with MAFLD had higher levels of BMI, ALT, AST, FPG, UREA, CREA, Cys-C, URIC, TG, TC, ALP, GGT, LDH, HBDH, non-HDL-C, TyG, TyG-BMI, and lower A/A, TP, ALB, GLB, and HDL-C levels (all P < 0.005). As a result, the ratios of the above indicators, such as LDL-C/HDL-C, non-HDL-C/HDL-C, TG/HDL-C, and TC/HDL-C were greater in MAFLD patients than in the control group.
Table 1. The basic characteristics of the Non-MAFLD and MAFLD groups
|
Non-MAFLD ( n=132)
|
MAFLD ( n=97 )
|
P value
|
age
|
36.00 (32.00,46.00)
|
43.00 (33.00,51.00)
|
0.019
|
Male/Female
|
80/52
|
75/22
|
0.008
|
BMI
|
22.56 (20.90,24.91)
|
27.38 (25.25,29.24)
|
<0.001
|
TBIL
|
11.85 (9.50,16.05)
|
12.10 (9.00,15.90)
|
0.918
|
DBIL
|
3.40 (2.70,4.50)
|
3.40 (2.60,4.40)
|
0.597
|
IBIL
|
8.10 (6.50,11.22)
|
8.80 (6.20,11.90)
|
0.606
|
ALT
|
17.50 (14.00,25.00)
|
46.00 (28.00,76.00)
|
<0.001
|
AST
|
19.00 (16.00,23.00)
|
31.00 (24.00,42.00)
|
<0.001
|
A/A
|
1.06 (0.89,1.29)
|
0.67 (0.53,0.86)
|
<0.001
|
TP
|
76.10(3.35)
|
74.74(4.48)
|
0.010
|
ALB
|
48.91(2.40)
|
48.17(2.78)
|
0.033
|
GLB
|
26.85 (25.48,29.20)
|
25.40 (23.60,29.00)
|
0.012
|
A/G
|
1.81 (1.67,1.93)
|
1.90 (1.66,2.04)
|
0.111
|
GLU
|
4.83 (4.63,5.18)
|
5.86 (4.92,7.33)
|
<0.001
|
UREA
|
4.45 (3.80,5.20)
|
4.70 (4.20,5.50)
|
0.015
|
CREA
|
70.00 (60.00,80.00)
|
77.00 (67.00,89.00)
|
0.001
|
Cys-C
|
0.79 (0.73,0.86)
|
0.86 (0.81,0.93)
|
<0.001
|
URIC
|
329.10(65.94)
|
393.14(102.24)
|
<0.001
|
TG
|
1.06 (0.78,1.32)
|
2.47 (1.76,3.32)
|
<0.001
|
TC
|
4.59(0.66)
|
4.87(1.02)
|
0.011
|
HDL-C
|
1.50 (1.28,1.70)
|
1.00 (0.90,1.20)
|
<0.001
|
LDL-C
|
2.77(0.61)
|
2.87(0.84)
|
0.284
|
ALP
|
69.50 (58.75,83.00)
|
82.00 (68.00,99.00)
|
<0.001
|
GGT
|
20.00 (13.00,28.00)
|
50.00 (34.00,74.00)
|
<0.001
|
CK
|
96.50 (74.00,137.75)
|
109.00 (73.00,131.00)
|
0.586
|
LDH
|
169.50 (155.75,188.00)
|
183.00 (171.00,208.00)
|
<0.001
|
HBDH
|
131.00 (120.00,145.00)
|
138.00 (126.00,160.00)
|
0.003
|
nonHDL-C
|
3.07(0.71)
|
3.84(1.00)
|
<0.001
|
LDL-C/HDL-C
|
1.87 (1.52,2.37)
|
3.00 (2.46,3.27)
|
<0.001
|
nonHDL-C/HDL-C
|
2.05 (1.63,2.63)
|
3.94 (3.03,4.45)
|
<0.001
|
TG/HDL-C
|
0.70 (0.48,1.05)
|
2.43 (1.51,3.76)
|
<0.001
|
TC/HDL-C
|
3.05 (2.63,3.63)
|
4.94 (4.03,5.45)
|
<0.001
|
TyG
|
8.30 (7.99,8.54)
|
9.36 (9.01,9.66)
|
<0.001
|
TyG-BMI
|
112.58 (87.85, 153.62)
|
201.52 (150.72, 261.22)
|
<0.001
|
Variable selection and model construction
Among these indexes, 29 variables showed significant differences between the MAFLD and non-MAFLD groups. The differences in 18 variables remained after adjusting for a variety of factors, including age, sex, and BMI (Table 2). A predictive model consisting of gender, BMI, ALT, TG, HDL-C, TyG, and TyG-BMI by binary logistic regression was then constructed. As shown in Figure 1, a dynamic nomogram was used to describe the probability of the MAFLD in the model. For example, a female patient with a BMI of 32.0 kg/m2, ALT of 35 IU/L, TG of 2.0 mmol/L, HDL-C of 1.4 mmol/L, TyG of 9.00, and TyG-BMI of 288 had a likelihood of MAFLD of 0.944 based on the model.
Table 2. logistic regression analysis of risk factors for MAFLD patients after adjusting age, gender and BMI
|
OR
|
95% CI
|
P Value
|
ALT
|
1.123
|
(1.078, 1.170)
|
.000
|
AST
|
1.210
|
(1.130, 1.296)
|
.000
|
A/A
|
.024
|
(0.006, 0.101)
|
.000
|
FPG
|
7.158
|
(3.342, 15.334)
|
.000
|
Cys-C
|
100.408
|
(3.698, 2726.533)
|
.006
|
URIC
|
1.009
|
(1.005, 1.014)
|
.000
|
TG
|
13.550
|
(5.747, 31.944)
|
.000
|
HDLC
|
.007
|
(0.001, 0.040)
|
.000
|
ALP
|
1.020
|
(1.004, 1.037)
|
.016
|
GGT
|
1.050
|
(1.029, 1.071)
|
.000
|
nonHDL-C
|
2.261
|
(1.464, 3.492)
|
.000
|
LDL-C/HDL-C
|
1.905
|
(1.251, 2.902)
|
.003
|
nonHDL-C/HDL-C
|
2.285
|
(1.628, 3.207)
|
.000
|
TG/HDL-C
|
5.387
|
(2.986, 9.718)
|
.000
|
TC/HDL-C
|
2.285
|
(1.628, 3.207)
|
.000
|
TyG
|
107.945
|
(25.824, 451.222)
|
.000
|
TyG-BMI
|
1.191
|
(1.130, 1.256)
|
.000
|
Diagnostic performance of vital indexes and the predictive model in MAFLD
Using the random forest method, patients were randomly divided into a training set and a test set at a ratio of 2:1. When all the variables in the model were adopted, a graph provided an overview of the importance score of each variable in predicting MAFLD. As shown in Figure 2, TyG, TG/HDL-C, and TG are the most important indicators for identifying MAFLD.
ROC curve analyses were conducted to identify the diagnostic value of TG, TG/HDL-C, TyG, and the predictive model (Figure 3). As a result, the area under the receiver operating characteristic curve (AUROC) of the predictive model was 0.989 (95% CI 0.980-0.998) with 0.979 sensitivity and 0.947 specificity when the cut-off value was 0.323, showing the best capacity for assessing MAFLD. TG, TG/HDL-C, and TyG could also detect MAFLD as shown in Table 3.
Table 3 Areas under the curve (AUC) of TG, TG/HDL-C, TyG and the predictive model for MAFLD
Variable
|
AUROC
|
95%CI
|
Cut-off value
|
Sensitivity (%)
|
Specificity (%)
|
TG
|
0.921
|
0.886 to 0.957
|
1.745
|
75.3
|
97.0
|
TG/HDL-C
|
0.925
|
0.892 to 0.959
|
1.260
|
83.5
|
90.9
|
TyG
|
0.943
|
0.913 to 0.973
|
8.804
|
87.6
|
93.9
|
Predictive model
|
0.989
|
0.980 to 0.998
|
0.323
|
97.9
|
94.7
|
As for calibration capability, the P-value of the Hosmer-Lemeshow test of the predictive model was 0.3869. The calibration plot and DCA curves of the model are shown in Figures 4 and 5, respectively.