A nomogram incorporating clinical and laboratory indicators for predicting metabolic dysfunction‐associated fatty liver disease in newly diagnosed type 2 diabetes patients

ABSTRACT

Aims

To develop and validate a nomogram model based on clinical and laboratory parameters to predict the risk of metabolic dysfunction‐associated fatty liver disease (MAFLD) in the early stage of type 2 diabetes.

Materials and Methods

We performed this study among 883 inpatients with new‐onset type 2 diabetes, and the data were divided randomly into training and validation groups. The logistic regression method was used to identify the independent risk factors of MAFLD, and a nomogram was established according to the logistic regression analysis and these selected parameters. The discrimination, calibration, and clinical utility of the nomogram were measured by receiver operating characteristic curve analysis, calibration curves, and decision‐curve analysis, respectively.

Results

Eight variables were identified and included in the nomogram (body mass index, alanine aminotransferase, triglyceride, low‐density lipoprotein cholesterol, high‐density lipoprotein cholesterol, fasting plasma glucose, urea nitrogen and serum uric acid). The value of the area under the receiver operating characteristic (ROC) curve was 0.898 for the training group and 0.92 for the validation group. The calibration plots indicated that this model had good accuracy, and the decision‐curve analysis revealed high‐clinical practicability of the nomogram.

Conclusions

This study established a convenient and practical nomogram model, which can be used as an easy‐to‐use tool to evaluate the risk of MAFLD among patients with newly diagnosed T2DM.

It is critical to develop a simple, cost‐effective and practical predictive measurement for screening MAFLD individuals in the early stage of T2DM. We developed a nomogram model based on eight clinical and laboratory indicators to predict MAFLD in patients with newly diagnosed T2DM. We demonstrated that this model had a good predictive ability and clinical application value for predicting MAFLD.

INTRODUCTION

Metabolic dysfunction ‐ associated fatty liver disease (MAFLD) has been chosen to replace the term nonalcoholic fatty liver disease (NAFLD) ¹ . The term MAFLD avoids the limitation of relying on exclusionary confounders and emphasizes more the importance of metabolic disorders in diagnosis ² . To sum up, any patient with steatotic liver disease accompanied by obesity or type 2 diabetes mellitus (T2DM), or meeting two of the seven metabolic risk factors including high‐blood pressure, central obesity, high‐triglyceride (TG) levels, low high‐density cholesterol (HDL‐C) levels, insulin resistance, and impaired glucose metabolism, can be categorized as MAFLD ³ . MAFLD has become the most common chronic liver disease worldwide along with the pandemic of obesity and its complications. Globally, it affects over one‐third of the population ⁴ , and in China, the rate is 29–46% ⁵ . By 2040, more than half of the adults are projected to have MAFLD ⁶ . About 20–30% of MAFLD patients develop metabolic dysfunction‐associated steatohepatitis (MASH) ⁷ and approximately 7–12% of patients progress to liver cirrhosis or end‐stage liver disease ⁸ . Besides liver‐related events, MAFLD was notably related to a rise in the development of cardiovascular disease (CVD), and at the same time, CV events were the most critical factor determining mortality among MAFLD patients ⁷ . Diabetes, another complex metabolic disorder, affected 11 percent of adults worldwide, of whom 90% were T2DM ⁹ . Long‐term high‐glucose toxicity and other damage factors due to T2DM can injure multiple organ systems, such as the pancreas, cardiovascular system, retina, kidneys, liver, gut, muscle, and even the brain ¹⁰ . Extremely high fatality and mutilation rates are the typical labels of T2DM ¹⁰ . Both MAFLD and type 2 diabetes aggravated seriously the medical and economic burden worldwide and became important public health issues.

It is known that MAFLD and T2DM frequently coexist and work together synergistically to raise the risk of adverse clinical outcomes ¹¹ , ¹² . T2DM is one of the most powerful risk factors, not only for the occurrence of MAFLD but also for accelerating the progression of MAFLD to steatohepatitis, liver fibrosis or cirrhosis ¹³ , ¹⁴ , ¹⁵ . On the other hand, compelling evidence showed that MAFLD is related to nearly twice the higher risk of developing T2DM, and the resolution or improvement of MAFLD is linked to a decrease in the risk of T2DM ¹⁶ . In addition, in patients with T2DM, MAFLD promote the development of diabetes‐related complications such as diabetic retinopathy (DR), diabetes nephropathy (DN), and CVD ¹⁷ , ¹⁸ . Therefore, early diagnosis for MAFLD in type 2 diabetic patients are of strong importance to enable the provision of an early intervention, thus avoiding the progression and exacerbation of MAFLD and diabetic complications, especially in the early stage of diabetes.

The diagnosis of MAFLD is based on the evidence of hepatic steatosis. Although liver biopsy is considered as the gold standard in the assessment of steatotic liver disease, its limitations have been extensively discussed ¹⁹ . Liver biopsy is rarely used in clinical practice due to its invasive procedure with some dangerous complications such as notable bleeding, and non‐invasive assessments should be considered in the vast majority of patients ²⁰ . Abdominal ultrasonography, magnetic resonance imaging, imaging‐derived proton density fat fraction (MRI‐PDFF), transient elastography and other imaging techniques are primary noninvasive techniques for diagnosis of hepatic steatosis and can quantitatively assess liver fat ²¹ , ²² , ²³ . However, they are all costly and require high‐professional skills. Given that most subjects with MAFLD are identified in primary care settings ²⁴ , these technologies are unsuitable for large‐scale screening in grassroots medical institutions ²⁵ . Thus a low‐cost, simple and effective screening approach is required.

The prediction models for NAFLD using readily available parameters have been gaining attention. Most common clinical variables for hepatic steatosis include anthropometric indicators such as body mass index (BMI), metabolic risk measures such as total cholesterol (TC), triglyceride (TG), glucose, and high‐density lipoprotein cholesterol (HDL‐C) and liver enzymes such as alanine transferase (ALT) and aspartate aminotransferase (AST) ²⁶ , ²⁷ , ²⁸ . Although several convenient predictive models based on the combination of these indexes to diagnose NAFLD have been constructed ²⁹ , ³⁰ , they are rarely used in the prediction of fatty liver in type 2 diabetes, especially in the early stage of diabetes. In the present research, we tried to construct and validate a nomogram model, which is a graphical representation of the individual's risk classification for predicting MAFLD incidence in patients with newly diagnosed T2DM. This simple model may act as a practical screening tool for doctors in primary, secondary, and tertiary care centers by utilizing easily accessible factors for the early detection of MAFLD in type 2 diabetic patients.

MATERIALS AND METHODS

Research design and participants

This study had a retrospective design, in which 1,063 individuals with newly diagnosed T2DM hospitalized in the Endocrine Department of Qilu Hospital of Shandong University Dezhou Hospital between 2016 and 2024 were selected. Then patients with acute complications of diabetes such as ketoacidosis and hyperosmolar coma, trauma, acute infectious diseases, or serious heart, liver or kidney injuries were excluded. A total of 883 patients were included in the final analysis. All the participants in the research signed written informed consent for this study. This study was conducted conforming to the ethical guidelines outlined in the Declaration of Helsinki and has been approved by the ethics committees of Qilu Hospital of Shandong University Dezhou Hospital.

The diagnosis of type 2 diabetes was based on the WHO and American Diabetes Association (ADA) criteria ³¹ : a fasting plasma glucose concentration ≥ 7.0 mmol/L or a 2‐h plasma glucose (2‐h PG) concentration ≥ 11.1 mmol/L following a 75 g oral glucose tolerance test confirmed on two occasions, and after excluding type 1 diabetes, gestational diabetes, transient hyperglycemia or stress hyperglycemia, gestational diabetes, monogenic diabetes, and secondary diabetes.

The diagnostic criteria for MAFLD require hepatic steatosis with at least one of five cardiometabolic features including impaired glucose regulation, T2DM, overweight or obesity, hypertension, or dyslipidemia ³² . In this study, MAFLD was diagnosed by ultrasound‐confirmed hepatic steatosis ³³ in all participants, who had pre‐existing T2DM.

Physical and laboratory examinations

The physical indices of the study participants, namely systolic blood pressure (SBP), diastolic blood pressure (DBP), height, and weight, were measured during the routine procedures. Subsequently, their Body Mass Index (BMI) was computed by dividing weight by the square of height (kg/m²). The demographics such as age, sex, and medications were recorded carefully. Participants were asked to fast for 8–10 h prior to the collection of early morning fasting venous blood. The biochemical parameters were determined by the methods endorsed by the International Federation of Clinical Chemistry and Laboratory Medicine (IFCC). Various biomarkers were measured including alanine aminotransferase (ALT), aspartate aminotransferase (AST), γ‐glutaryl transferase (γ‐GGT), triglycerides (TG), low‐density lipoprotein cholesterol (LDL), high‐density lipoprotein cholesterol (HDL), fasting blood glucose (FBG), blood urea nitrogen (BUN), creatinine (Cr), and serum uric acid (SUA).

Liver ultrasound examination

Each participant underwent a liver ultrasonography after fasting for over 8 h. All ultrasound examinations were performed by experienced radiologists and interpreted through a standardized double‐reading process involving two independent specialists. The ultrasonic diagnostic benchmarks for fatty liver were as follows: (1) Diffuse punctate hyperecho in the near field of the liver area, with a stronger signal than that of the spleen and kidney; (2) The far‐field echo progressively diminished and the light spots were sparse; (3) The intrahepatic duct structure was not distinctly visualized; (4) The liver was mildly to moderately enlarged, and its edges were rounded and blunt. The existence of item #1 in combination with at least one of items #2, #3, and #4 was regarded as fatty liver ³³ .

Statistical analysis

All data management and statistical analyses were conducted with SPSS 21.0 (IBM, Armonk, NY, USA) and R v. 4.3.2 (R Foundation for Statistical Computing, Vienna, Austria). The dataset (n = 883) was randomly partitioned into a training set (n = 618) and a validation set (n = 265) at a 70:30 ratio using the createDataPartition function from the R caret package. To ensure reproducibility, we set a fixed random seed (seed = 850) during the splitting process. Skewness‐kurtosis tests were carried out to assess the normality of the continuous variables. The quantitative data that were consistent with the normal distribution were expressed as means ± SD and those that did not meet the normal distribution were presented as medians (quartile ranges). Qualitative data were described using frequency and percentage. Group differences were examined by means of the independent sample T test for continuous variables (non‐normally distributed data were subjected to logarithmic transformation prior to analysis.) and the chi‐square test for categorical variables. Bivariate and multivariable logistic regression analysis was conducted to screen the main risk factors of MAFLD. Based on these factors, a nomogram model for predicting the probability of fatty liver was established by rms‐package in the training group. Then the nomogram was validated internally in the training group and externally in the validation group. Firstly, the predictive power of the model was validated by performing the area under the receiver operating characteristic curves (AUROC). With the AUC greater than 0.70, the nomogram was regarded as of high performance. In addition, we calculated the true positive (TP), false negative (FN), false positive (FP), and true negative (TN) values, along with derived diagnostic accuracy metrics including sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy to evaluate comprehensively the diagnostic performance of the nomogram. Furthermore, the discriminating accuracy of the MAFLD nomogram was assessed by calibration curves using the rms‐package. Finally, the decision‐curve analysis (DCA) curves were plotted using the rmda‐package to evaluate the clinical practicability of the model. P < 0.05 (Two‐sided) was recognized to indicate statistical significance.

RESULTS

General characteristics of the patients

The physical and biochemical indicators of the patients are presented in Table 1. After excluding individuals with incomplete data, a total of 883 participants were included in the study, including 566 males and 317 females. 607 patients (68.7%) had MAFLD. The data were randomly divided into the training group (n = 618) and the validation group (n = 265). There were no significant differences in clinical or demographic variables, including age, sex ratio, SBP, DBP, ALT, AST, γ‐GGT, TG, HDL, LDL, FBG, SUN, Cr, SUA between the training and validation groups. The prevalence of MAFLD was 68.3% (422 patients) in the training group and 69.8% (185 patients) in the validation group, respectively, and it also showed no significant difference between these two sets.

Table 1.

Clinical and biochemical characteristics of study participants and the comparisons of factors between training and validation datasets

General indexes	All patients (n = 883)	Training dataset (n = 618)	Validation dataset (n = 265)	P value
Gender, male (%)	566 (64.09%)	404 (65.37%)	162 (61.13)	0.251
Age (year)	47.05 ± 13.6	47.6 ± 13.69	46.79 ± 13.32	0.119
BMI (kg/m2)	26.25 ± 4.21	26.32 ± 4.23	26.08 ± 4.14	0.434
SBP (mmHg)	134.51 ± 18.14	134.92 ± 18.31	133.55 ± 17.73	0.306
DBP (mmHg)	84.28 ± 11.63	84.23 ± 11.53	84.41 ± 11.89	0.827
ALT (IU/L)	25 (17–43)	24.45 (16.9–42)	26 (17.55–44)	0.415
AST (IU/L)	23.4 (17.8–32)	23.4 (17.85–32)	23.2 (17.4–32.9)	0.984
GGT (IU/L)	32.9 (21–54.48)	31.6 (20.3–53.4)	35 (22–56)	0.153
TG (mmol/L)	1.68 (1.11–2.79)	1.68 (1.11–2.8)	1.67 (1.11–2.71)	0.99
HDL (mmol/L)	1.14 ± 0.31	1.13 ± 0.28	1.14 ± 0.36	0.643
LDL (mmol/L)	3.21 ± 0.87	3.23 ± 0.84	3.19 ± 0.93	0.532
FBG (mmol/L)	8.76 (6.9–11.86)	8.79 (6.99–11.74)	8.7 (6.65–12.04)	0.877
BUN (mmol/L)	5.27 ± 1.5	5.29 ± 1.5	5.22 ± 1.51	0.499
Cr (μmol/L)	62.7 (51.6–74)	62.95 (52.03–74)	62 (48.4–74)	0.159
SUA (μmol/L)	318 (252–382.9)	315.3 (250.45–379.25)	324 (252.5–390)	0.237
MAFLD (%)	607 (68.7%)	422 (68.3%)	185 (69.8%)	0.636

ALT, alanine aminotransferase; AST, aspartate aminotransferase; BMI, body mass index; BUN, blood urea nitrogen; Cr, creatinine; DBP, diastolic pressure; FBG, fasting blood glucose; GGT, γ‐glutaryl transferase; HDL, high‐density lipoprotein; LDL, low‐density lipoprotein; MAFLD, metabolic dysfunction‐associated fatty liver disease; SBP, systolic pressure; SUA, serum uric acid; TG, triglycerides.

Independent predictive factors of MAFLD in the training group

In training group, the patients with MAFLD had increased levels of BMI, SBP, DBP, ALT, AST, γ‐GGT, TG, LDL, FBG, Cr, and SUA, and reduced levels of age, HDL, and BUN compared with those without MAFLD (Table 2). No difference was observed in the sex ratio between MAFLD and non‐MAFLD group. The univariable logistic regression showed that BMI, SBP, DBP, ALT, AST, γ‐GGT, TG, LDL, FBG, Cr, and SUA were positively associated with the prevalence of MAFLD, and age, HDL, and BUN were negatively related to the risk of MAFLD (Table 3). Furthermore, the multivariable regression analysis indicated that, after adjusting for confounding variables, BMI, ALT, TG, LDL, HDL, FBG, BUN and SUA were the independent predictive factors for MAFLD (Table 3).

Table 2.

Baseline characteristics and the differences between MAFLD and non‐MAFLD patients in the training dataset

General indexes	Without MAFLD (n = 196)	With MAFLD (n = 422)	P value
Gender, male (%)	118 (59.89%)	268 (67.93%)	0.057
Age (year)	51.26 ± 12.52	45.89 ± 13.89	<0.001*
BMI (kg/m2)	23.4 ± 3.35	27.69 ± 3.9	<0.001*
SBP (mmHg)	128.6 ± 16.77	137.88 ± 18.27	<0.001*
DBP (mmHg)	79.58 ± 10.23	86.39 ± 11.47	<0.001*
ALT (IU/L)	17.4 (13.5–22.1)	31 (20.15–51)	<0.001*
AST (IU/L)	18.65 (15.55–24.88)	25.8 (19.05–35)	<0.001*
GGT (IU/L)	21 (15–31.2)	37 (26–61.05)	<0.001*
TG (mmol/L)	1.03 (0.76–1.64)	2.03 (1.36–3.53)	<0.001*
HDL (mmol/L)	1.21 ± 0.29	1.09 ± 0.26	<0.001*
LDL (mmol/L)	2.95 ± 0.84	3.36 ± 0.82	<0.001*
FBG (mmol/L)	7.53 (6.2–10.5)	9.3 (7.57–12.1)	<0.001*
BUN (mmol/L)	5.56 ± 1.63	5.17 ± 1.41	0.001*
Cr (μmol/L)	59 (51–71.55)	64 (54–74.2)	0.021*
SUA (μmol/L)	261.1 (208–324.7)	336.8 (277–406.35)	<0.001*

ALT, alanine aminotransferase; AST, aspartate aminotransferase; BMI, body mass index; BUN, blood urea nitrogen; Cr, creatinine; DBP, diastolic pressure; FBG, fasting blood glucose; GGT, γ‐glutaryl transferase; HDL, high‐density lipoprotein; LDL, low‐density lipoprotein; MAFLD, metabolic dysfunction‐associated liver disease; SBP, systolic pressure; SUA, serum uric acid; TG, triglycerides.

^* P < 0.05 was considered a statistically significant difference.

Table 3.

Univariate and multivariate analysis for the prediction of MAFLD

Variables	Univariate logistic regression analysis			Multivariate logistic regression analysis
	OR	95% CI	P	OR	95% CI	P
Gender (male/female)	1.418	0.999–2.014	0.051
Age (year)	0.971	0.958–0.983	<0.001*
BMI (kg/m2)	1.487	1.38–1.601	<0.001*	1.309	1.199–1.429	<0.001*
SBP (mmHg)	1.032	1.021–1.044	<0.001*
DBP (mmHg)	1.061	1.042–1.079	<0.001*
ALT (IU/L)	1.057	1.042–1.073	<0.001*	1.037	1.008–1.067	0.013*
AST (IU/L)	1.053	1.034–1.072	<0.001*
GGT (IU/L)	1.014	1.007–1.02	<0.001*
TG (mmol/L)	2.542	2.002–3.228	<0.001*	1.361	1.071–1.728	0.012*
HDL (mmol/L)	0.247	0.133–0.458	<0.001*	0.378	0.149–0.759	0.021*
LDL (mmol/L)	1.868	1.493–2.336	<0.001*	1.821	1.317–2.517	<0.001*
FBG (mmol/L)	1.137	1.077–1.201	<0.001*	1.096	1.018–1.181	0.016*
BUN (mmol/L)	0.844	0.754–0.944	0.003*	0.817	0.696–0.959	0.014*
Cr (μmol/L)	1.012	1.001–1.023	0.038*
SUA (μmol/L)	1.011	1.008–1.013	<0.001*	1.005	1.002–1.008	0.002*

ALT, alanine aminotransferase; AST, aspartate aminotransferase; BMI, body mass index; BUN, blood urea nitrogen; Cr, creatinine; DBP, diastolic pressure; FBG, fasting blood glucose; GGT, γ‐glutaryl transferase; HDL, high‐density lipoprotein; LDL, low‐density lipoprotein; MAFLD, metabolic dysfunction‐associated fatty liver disease; SBP, systolic pressure; SUA, serum uric acid; TG, triglycerides.

^* P < 0.05 was considered a statistically significant difference.

Establishment of predictive model

Using these independent risk variables, the multivariate logistic regression was performed to establish the predictive model, which was presented as a nomogram (Figure 1). The nomogram can quantitatively reflect the risk probability of MAFLD in newly diagnosed type 2 diabetes patients. Each variable in the model is represented on the nomogram by a scale. To make a prediction, a user first locates the values of the relevant variables on their respective scales. Then, the total score was calculated by adding the scores of each factor. The point corresponding to the total score is the probability value of each patient developing MAFLD could be calculated. The higher the score, the greater the risk of developing MAFLD. For example, using this nomogram model, we can estimate that for a patient with T2DM of BMI of 26.4 kg/m², ALT of 16 IU/L, TG of 1.08 mmol/L, HDL of 1.36 mmol/L, LDL of 4.37 mmol/L, FBG of 12.5 mmol/L, BUN of 4.1 mmol/L and SUA of 340 μmol/L, the approximated probability of MAFLD is 90%.

Figure 1.

Nomogram for predicting MAFLD incidence in the training group. MAFLD: metabolic dysfunction‐associated fatty liver disease. The model was formed by eight risk factors of MAFLD including body mass index (BMI), triglycerides (TG), low‐density lipoprotein (LDL), high‐density lipoprotein (HDL), blood urea nitrogen (BUN), alanine aminotransferase (ALT), fasting blood glucose (FBG), and serum uric acid (SUA).

Validation of the nomogram

The predictive ability of the model was verified by receiver operating characteristic (ROC) curve (Figure 2). The AUC of the model was 0.898 (95% CI 0.871–0.924) in the development set and 0.92 (95% CI 0.886–0.953) in the validation set, indicating a great predictability (Figure 2). To further evaluate the diagnostic efficacy of this model, we conducted additional verification in the validation cohort. The results revealed that the TP, FN, FP, and TN were 160, 25, 13, and 67, respectively. Furthermore, the sensitivity, specificity, PPV, NPV, and accuracy were 0.865, 0.838, 0.925, 0.728, and 0.857, respectively. These metrics collectively suggest that our model exhibits high sensitivity and specificity, indicative of strong diagnostic performance (Table 4). The nomogram's prediction accuracy was validated by the calibration curve (Figure 3). The result showed that the calibration curve was close to the diagonal line, revealing good agreement between the predicted probabilities from the nomogram and the actual observed probabilities in both the training set and the validation set (Figure 3). The decision‐curve analysis was employed to evaluate the clinical application of the developed nomogram (Figure 4). It was shown that the threshold probability was ≤98% in the training set and ≤100% in the validation set (Figure 4). This implies that when the predicted probability is no greater than 98%, further diagnosis is beneficial. On the other hand, when the predicted risk value exceeds 98%, diagnosis of MAFLD is of no benefit. In short, briefly, compared with the “all” or “none” strategies within the threshold probability range from 0.0 to 0.98, the MAFLD prediction nomogram was more advantageous in predicting the risk of MAFLD in terms of net benefit.

Figure 2.

Receiver operating characteristic curves (ROC) of nomogram in training set (a) and validation set (b). The x‐axis represents the false positive rate of the prediction model, and the y‐axis represents the true positive rate of the prediction model.

Table 4.

Performance metrics of the nomogram

Metric	Value	Formula
TP	160	–
FN	25	–
FP	13	–
TN	67	–
Sensitivity	0.865	TP/(TP + FN)
Specificity	0.838	TN/(TN + FP)
PPV	0.925	TP/(TP + FP)
NPV	0.728	TN/(TN + FN)
Accuracy	0.857	(TP + TN)/(TP + TN + FP + FN)

FN, false negative; FP, false positive; NPV, negative predictive value; PPV, positive predictive value; TN, true negative; TP, true positive.

Figure 3.

Calibration curves of the predictive model in the training set (a) and validation set (b). The x‐axis represents the predicted risk of MAFLD and the y‐axis represents the actual occurrence rate of MAFLD. The diagonal dotted line serves as a symbol representing a perfect prediction made by an ideal model. The closer the alignment with the diagonal dotted line is, the more precise the prediction will be.

Figure 4.

Decision curves for the nomogram in training set (a) and validation set (b). The x‐axis represents the probability at which a patient is considered to have MAFLD. The y‐axis represents the net benefit. The thick line indicates the additional net benefit provided by the model‐making approach compared to the strategy of not intervening at all. The thin line represents the net benefit when all patients are given the intervention.

DISCUSSION

In this cross‐sectional study, we developed a nomogram based on eight clinical and laboratory indicators including BMI, ALT, TG, LDL, HDL, FBG, BUN and SUA, and demonstrated that it had a good predictive ability and clinical application value for predicting MAFLD in patients with newly diagnosed T2DM. This is the first time that the nomogram model has been applied to predict MAFLD in the early stage of type 2 diabetes and the simple nomogram has the potential to economize more medical resources and decrease the incidence of missed diagnosis.

With increasing global incidence, metabolic dysfunction‐associated fatty liver disease (MAFLD), previously named nonalcoholic fatty liver disease (NAFLD), has become the most common liver disease worldwide. Previous models forecast that China will have the largest increase in NAFLD prevalence compared with the rest of the world in 2030 ³⁴ . Hepatic steatohepatitis and fibrosis are the more severe forms of MAFLD and eventually develop into cirrhosis and even hepatocellular carcinoma ³⁵ . Commonly described as the liver manifestation of metabolic syndrome, MAFLD is also considered a strong determinant for the development of metabolic syndrome and many extra‐hepatic chronic diseases ³⁶ . A strong association between MAFLD and type 2 diabetes has been clarified, with the estimated global prevalence of MAFLD in subjects with T2DM being more than 60% ³⁷ . Our study showed that the incidence of MAFLD in Chinese type 2 diabetes subjects was 68.7%, which was close to the previous studies ³⁷ . T2DM and MAFLD can affect each other due to shared pathogenic mechanisms. The concurrent presence of the two diseases augments the likelihood of liver‐related adverse consequences ³⁸ and induces more significant glucose metabolic dysfunctions and insulin resistance ¹⁸ . Our research also indicated that the presence of MAFLD was associated with more serious metabolic disorders, including worse dyslipidemia, hyperglycemia, and hypertension. In addition, MAFLD was also verified to prompt the emergence of macrovascular and microvascular complications in type 2 diabetic patients ¹⁷ . Therefore, it is critical to develop a simple, cost‐effective, and practical predictive measurement for screening MAFLD individuals in the early stage of T2DM, especially in primary‐level medical institutions where imaging or liver biopsy tests may not be routinely ordered.

Previous studies have developed some screening models combining anthropometric parameters (BMI, waist circumference, neck circumference, etc.), demographics (age, sex, etc.), biochemical indicators (ALT, LDL‐C, TG, SUA, FPG, etc.) and comorbidities (diabetes, hypertension, etc.) to achieve early detection and evaluation of NAFLD ³⁹ , ⁴⁰ . However, these models do have some limitations. For instance, the Fatty Liver Index (FLI), which incorporates waist circumference, BMI, γ‐GGT, and triglyceride, and the Hepatic Steatotic Index (HSI), which is based on gender, BMI, the ALT/AST ratio, and T2DM, are not valid predictors of hepatic steatosis in T2DM patients because the discriminatory function of these tests in patients with T2DM decreases accordingly ⁴¹ . Additionally, these tests involve complex calculations and fail to intuitively reflect the incidence of hepatic steatosis, resulting in few of them being applied in clinical practice. A nomogram is a graphical calculating device. It is a two‐dimensional diagram that uses a coordinate system to represent the relationship between several variables. By inputting specific values of relevant variables for an individual patient, a nomogram provides a visual and intuitive way to estimate the probability of a particular outcome, and the prediction is personalized ⁴² . Nomograms have been widely used in the prediction of other diseases, such as cardiovascular disease and cancer ⁴³ , ⁴⁴ . At present, scarce studies have demonstrated the diagnostic accuracy of the nomogram in the diagnosis of MAFLD in patients with T2DM. Xue et al. ⁴⁵ developed a nomogram based on questionnaire and physical measurement data within a type 2 diabetic population. However, its accuracy is relatively low, with a sensitivity of 69.7% and a specificity of 67.7%.

In our study, we identified eight independent predictive factors of MAFLD using classical regression analysis methods. These factors included anthropometric parameters (BMI) and biochemical indicators (ALT, LDL‐C, HDL‐C, TG, SUA, FPG and BUN) which are all easy to obtain. Overweight and obesity are widely recognized as the primary causes of MAFLD, and elevated BMI is one of the main risk factors for MAFLD ⁴⁶ . ALT is an enzyme found mainly in the liver that is regarded as an important biomarker to assess liver function. High ALT not only indicates the injury of liver cells but can also serve as a reference for screening MAFLD ²⁷ . MAFLD is highly associated with metabolic syndrome characterized by insulin resistance, impaired glycemic metabolism, and dyslipidemia ⁴⁷ . TG, LDL‐C, and FBG are the independent risk factors for MAFLD, while HDL is an independent protective factor for MAFLD ⁴⁸ . Additionally, it is well established that total cholesterol (TC) levels are higher in MAFLD patients compared to non‐MAFLD individuals ⁴⁸ . Elevated TC is associated with an increased risk of MAFLD and may serve as a potential predictor for its development ⁴⁹ . Due to collinearity, TC was excluded from the model. Further investigation is needed to fully explore its predictive value for MAFLD. Consistent with previous research, the present work also showed the roles of BUN and SUA in screening MAFLD. BUN is negatively associated with MAFLD ⁵⁰ and SUA is positively related to the risk of MAFLD ⁵¹ . Therefore, these selected variables in our MAFLD‐predictive model are reliable and accurate. Based on these factors, a practical and easy‐to‐use nomogram was developed. The AUC of the nomogram indicated its excellent discriminative ability in predicting the risk of MAFLD, and it possessed high sensitivity and specificity. The calibration and DCA plot revealed that the model was well calibrated and had remarkable clinical utility. The validations were performed in the training and validation cohorts, and the results in these two groups were similar, indicating the stable predictive ability of the nomogram model. This nomogram enables physicians in primary, secondary, and tertiary care centers to make mass predictions for MAFLD in a quick and intuitive manner without having to calculate the complicated regression formulas. By using some easily available indicators and a simplified statistical model, a broad range of healthcare professionals and researchers who may not be specialized in statistics can evaluate, diagnose, and manage the patients with a high risk of MAFLD early. Additionally, for patients with MAFLD, the nomogram enables self‐management, which facilitates them in better controlling these metabolic indicators and thus reduces the risk of MAFLD.

There are several limitations in this study. Firstly, the current study was conducted using data from a single center, which may limit the generalizability of our findings and potentially lead to overestimation of performance metrics. Future studies should use multicenter or prospective cohorts incorporating diverse populations to externally validate this model. Secondly, this study were cross‐sectional designs and so cannot fully address causality or temporality between predictors and MAFLD incidence. Third, the diagnosis of MAFLD in our study was based on ultrasound, which is relatively insensitive for detecting mild steatosis (when hepatic fat accumulation is less than 20%). This limitation may result in some patients with mild fatty liver being undiagnosed, thereby reducing the model's predictive sensibility for lower degrees of steatosis. Nonetheless, ultrasonography remains the first‐line screening modality for hepatic steatosis recommended by major clinical guidelines due to its feasibility, cost‐effectiveness, and safety ⁵² . We recognize that advanced techniques such as transient elastography (FibroScan), MRI‐PDFF, or liver biopsy provide higher accuracy for grading hepatic steatosis. In our ongoing follow‐up studies, we plan to incorporate these techniques to further strengthen diagnostic accuracy and validate our model more comprehensively. Fourth, due to the limitations of our data source, we were unable to collect sufficient data for insulin levels to include the insulin resistance index such as HOMA‐IR. Additionally, some important confounders such as alcohol consumption, smoking status, menopausal state, and several other anthropometric parameters such as waist circumference (WC) were also excluded. These factors are known to be closely associated with fatty liver and may modify the observed associations. Therefore, future studies should consider incorporating more detailed information on these variables and conducting stratified analyses to better assess the predictive model for MAFLD in T2DM patients.

In conclusion, in this study, a practical and convenient nomogram was developed and validated to detect MAFLD in patients with newly diagnosed T2DM. Eight parameters including BMI, ALT, TG, LDL, HDL, FBG, BUN, and SUA were included in the nomogram. This nomogram can be utilized as an easy‐to‐use and objective tool, which is conducive to screening early T2DM subjects with a high risk of MAFLD and promotes better clinical diagnosis and prevention.

DISCLOSURE

The authors declare no conflict of interest.

Approval of the research protocol: This study was conducted according to the regulations of the Helsinki Declaration, and the study protocol was approved by the ethical committees of Qilu Hospital of Shandong University Dezhou Hospital.

Informed Consent: Written informed consent was provided by each participant.

Registry and the Registration No. of the study/trial: Not applicable.

Animal Studies: Not applicable.

FUNDING

This work was supported by the Natural Science Foundation of Shandong Province [grant numbers ZR2021QH181].