Abstract
Objective
To create a nomogram-based model to estimate the Chinese population's 5-year risk of metabolic dysfunction-associated steatotic liver disease (MASLD).
Methods
We randomly divided 7582 participants into two groups in a 7:3 ratio: one group was assigned to work with the training set, which consisted of 5307 cases, and the other group was assigned to validate the model using 2275 cases. The least absolute shrinkage and selection operator model was employed to ascertain the variables with the highest correlation among all potential variables. A logistic model was constructed by incorporating these selected variables, which were subsequently visualized using a nomogram. The discriminatory ability, calibration, and clinical utility of the model were assessed using the receiver operating characteristic (ROC) curve, calibration curve, and decision curve analysis (DCA).
Results
During the 5-year follow-up, 1034 (13.64%) total participants were newly diagnosed with MASLD. Using eight variables (gender, body mass index, waist, hemoglobin, alanine aminotransferase, uric acid, triglycerides, and high-density lipoprotein), we built a 5-year MASLD risk prediction model. The nomogram showed an area under the ROC of 0.795 (95% CI: 0.779–0.811) in the training set and 0.785 (95% CI: 0.760–0.810) in the validation set. The calibration curves revealed a 5-year period of agreement between the observed and predicted MASLD risks. DCA curves illustrated the practicality of this nomogram over threshold probability profiles ranging from 5% to 50%.
Conclusion
We created and tested a nomogram to forecast the risk of MASLD prevalence over the next 5 years.
Introduction
In 2023, international authorities adopted the nomenclature metabolic dysfunction-associated steatotic liver disease (MASLD), which was originally referred to as non-alcoholic fatty liver disease (NAFLD) (1). The estimated prevalence of MASLD in the world is greater than 30% (2). MASLD is a multi-systemic disease affecting not only the liver but also extra-hepatic organs (3). A study conducted on a nationwide cohort found that patients with MASLD have a higher likelihood of developing cardiovascular disease (4). Further, multiple studies have shown that MASLD is associated with an increased risk of chronic kidney disease (CKD) (5, 6).
Therefore, MASLD is a problem that needs to be taken extremely seriously. We need to identify patients at risk of progressing to MASLD in healthy populations in a timely manner and intervene in advance, thereby reducing the physical damage to people and the healthcare burden of the disease. Liver biopsy histopathology is the current diagnostic gold standard for MASLD; however, this invasive procedure has limited applicability due to its limited indications and contraindications (7). Imaging methods such as liver ultrasound or MRI can be used to identify patients with fatty liver but at a higher cost. There is a great need to construct models to predict MASLD.
Previous studies have developed equations to predict the development of NAFLD. In clinical practice, the fatty liver index (FLI) makes use of information that is easily accessible, and it exhibits a moderate level of predictive value (the area under the receiver operating characteristic (AUC): 0.84) (8). With an AUC of 0.812, the hepatic steatosis index (HSI) was also developed as an equation utilizing basic clinical signs (9). FLI and HSI have been demonstrated to be simple and efficient screening tools for NAFLD. In addition, there are a number of graphical models that may be used to forecast NAFLD, such as those developed using data from American and Japanese cohorts (10, 11). In China, a series of multiple visualization models based on data from elderly population, non-obese population, and pre-diabetic populations have been proposed (12, 13, 14).
Numerous machine learning-based models have also surfaced in recent years as a result of the use of machine learning techniques in the medical profession. All of these models showed good NAFLD identification ability, but the data they used were cross-sectional (15, 16, 17, 18). Deng et al. constructed a NAFLD prediction model based on longitudinal physical examination data using the long short-term memory algorithm (19). Huang et al. used six machine learning methods to construct a 5-year NAFLD prediction model and discovered that in both internal and external validation sets, the logistic regression model produced the greatest forecast performance (20). All of these models demonstrated good prediction of NAFLD incidence risk. However, given that NAFLD has been renamed as MASLD, these models have not been further validated in the MASLD population.
Therefore, our aim was to develop a nomogram for the 5-year MASLD risk using data obtained from Chinese health checkup records. We utilized receiver operating characteristic (ROC) curve, calibration curve, and decision curve analysis (DCA) to evaluate the model’s capacity to discriminate, calibrate, and provide clinical utility.
Methods
Study participants
A total of 12,312 participants underwent health checkups at the Nanjing Drum Tower Hospital Health Management Center in both 2018 and 2023. Data on smoking and drinking habits, as well as past medical history, were gathered from the participants. These are the criteria that were used to exclude candidates: participants who had been diagnosed with steatotic liver disease (SLD) at baseline; those who had viral hepatitis or other chronic liver diseases; those who had a history of heavy alcohol consumption (average daily alcohol consumption of more than 30 g for males and more than 20 g for females); and those who lacked the necessary information for this study.
Diagnostic criteria for MASLD were ultrasound diagnosis of steatosis of the liver and having at least one cardio-metabolic risk factor, while excluding a history of heavy alcohol consumption or other chronic liver diseases (1). Cardio-metabolic risk factors included: (1) blood pressure ≥ 130/85 mm Hg or specific antihypertensive drug treatment; (2) fasting plasma glucose (FPG) ≥ 5.6 mmol/L or 2-hour post-load glucose ≥ 7.8 mmol/L, hemoglobin A1c (HbA1C) ≥ 5.7%, or type 2 diabetes or treatment for type 2 diabetes; (3) body mass index (BMI) ≥ 23 kg/m² or waist > 94 cm (male) 80 cm (female); (4) plasma high-density lipoprotein (HDL) ≤ 1.0 mmol/L (male) and ≤1.3 mmol/L (female) or lipid-lowering treatment; (5) triglycerides (TG) ≥ 1.70 mmol/L or lipid-lowering treatment (1).
Hypertension was defined in our study as the following: (1) a previous definitive diagnosis of hypertension or the use of antihypertensive medication, or (2) an average of two blood pressure measurements were taken at the time of physical examination with a systolic blood pressure (SBP) of 140 mm Hg or higher or a diastolic blood pressure (DBP) of 90 mm Hg or higher. Diabetes was defined as the following: (1) a previous definitive diagnosis of diabetes or the use of hypoglycemic medication, or (2) FPG ≥ 7 mmol/L or HbA1C ≥ 6.5%.
A total of 7582 participants were enrolled in the study following the application of the exclusion criteria. Their average age was 39.62 ± 13.42 years, with 2870 males and 4712 females. After 5 years of follow-up, 1034 participants were diagnosed with MASLD and 36 with cryptogenic SLD. Figure 1 includes a detailed reference to the flowchart that depicts the enrolling process for research participants.
Flowchart of this study. SLD, steatotic liver disease; MASLD, metabolic dysfunction-associated steatotic liver disease.
Citation: Endocrine Connections 13, 8; 10.1530/EC-24-0186
Written informed consent was obtained from all participants, and the study followed the ethical criteria of the Declaration of Helsinki. It was approved by the Ethics Committee of Nanjing Drum Tower Hospital (Ethics no. 2022-46-01).
Variable characteristics
We used questionnaires, physical exams, and laboratory analysis as primary factors in this study. The questionnaires were used to collect information about the participants' ages, genders, medical backgrounds, and lifestyle behaviors, such as smoking and alcohol intake. Physical exams included measures of height, body weight, blood pressure, and waist circumference. Each person's BMI was determined by dividing their weight in kilograms by the square of their height in meters. Every participant refrained from eating overnight prior to the examination, and blood samples were collected from their peripheral veins in the early hours of the next day. Participants underwent standardized protocols at the same facility to assess their total blood count (Sysmex XN-1000; Sysmex Corp.) and blood biochemical index (Beckman AU5400; Beckman Coulter Corp.). Subsequently, we identified 19 variables with robust data quality and relevant physical measurements as potential variates.
Ultrasound assessments for hepatic steatosis
Two skilled sonographers, who were not privy to the participants' medical records, used abdominal ultrasonography (Philips ultrasound equipment) to diagnose hepatic steatosis. Ultrasound criteria for identifying hepatic steatosis included the following: (1) parenchymal echoes in the liver are elevated (the intensity of the liver echo is greater than that in the spleen and kidneys); (2) the liver's far-field echoes are muted; (3) the intrahepatic venous conduits are not easily visible. The diagnosis of steatohepatitis can be made by the presence of item 1 of the above three items, plus any of items 2 or 3.
Statistical analysis
The transparent reporting of a multivariable prediction model for individual prognosis or diagnosis recommendations is consistent with this research (21). To check if the data followed a normal distribution, the Kolmogorov–Smirnov test was run. Mean and s.d. were used to characterize normally distributed continuous variables, median and interquartile range were employed to characterize non-normally distributed continuous variables, and frequency and percentage were employed to characterize categorical variables. Depending on whether the continuous variables in the two groups had a normal distribution, we compared them using either an independent samples t-test or a rank sum test (Mann–Whitney U test). For categorical variables, the Chi-square test was employed. Multivariate logistic regression analysis was used to identify independent risk factors.
We performed sample size calculations using the three sample size criteria proposed by Riley et al. (22). The results can be found in Supplementary Table 1 (see section on supplementary materials given at the end of this article). Our sample size and the number of outcome events exceed the calculated sample size. Our study is therefore expected to provide reliable estimates. We used multiple imputations for the missing variables in this study.
Two sets, one for training and one for validation, were randomly assigned to participants in a 7:3 ratio. The least absolute shrinkage and selection operator (LASSO) regression method was used to select the optimal predictors for the study. The predictors selected by the LASSO regression model were then further statistically analyzed using multivariate logistic regression. Nomograms were used to visualize the results of the multiple logistic regression established above. The discriminatory power of the model for MASLD was assessed using the ROC curve to evaluate the AUC. The predicted probabilities were compared with the actual observed probabilities using a calibration curve. To evaluate the model's practicality in healthcare settings, DCA curves were employed.
We used R (version 4.3.1; https://www.R-project.org) to conduct our statistical analyses, and a P value of less than 0.05 was considered significant for two-sided statistical tests.
Results
Participant variables before and after follow-up
This study comprised 7582 participants who underwent a health checkup in 2018, excluding those with SLD or other liver diseases. Following a 5-year period, we collected health checkup data from the same study sample. New diagnoses of MASLD occurred in 1034 participants (13.64%) over the 5-year follow-up period. The comparison of variables between MASLD and non-MASLD groups categorized by outcome in 2018 and 2023 are presented in Table 1. In 2018, there was a higher likelihood of male participants (P < 0.001) and older participants (P < 0.001) who would experience MASLD 5 years later. Hypertension (P < 0.001) and diabetes mellitus (P < 0.001) were also found to have among the highest prevalence rates in this group. Furthermore, it should be mentioned that their BMI was significantly higher (P < 0.001), as was their waist circumference (P < 0.001), blood pressure (P < 0.001), and levels of liver enzymes (alanine aminotransferase (ALT), aspartate aminotransferase (AST), gamma-glutamyl transferase (GGT)), serum creatinine (Scr), uric acid (UA), total cholesterol (TC), triglycerides (TG), fasting plasma glucose (FPG), low-density lipoprotein (LDL), and hemoglobin (HB) were all elevated. Nevertheless, there was a significant decrease in their HDL levels (P < 0.001). Dissimilarities between the two categories persisted in 2023 as well.
Comparison of variables between the MASLD and non-MASLD groups categorized by outcome in 2018 and 2023.
Variables | 2018 | 2023 | ||||
---|---|---|---|---|---|---|
Non-MASLD (n = 6548) | MASLD (n = 1034) | P | Non-MASLD (n = 6548) | MASLD (n = 1034) | P | |
Gender (male%) | 2217 (33.9%) | 653 (63.2%) | <0.001 | 2217 (33.9%) | 653 (63.2%) | <0.001 |
Age (years) | 39.12 ± 13.38 | 42.68 ± 13.21 | <0.001 | 44.12 ± 13.38 | 47.68 ± 13.21 | <0.001 |
Hypertension (%) | 579 (8.8%) | 179 (17.3%) | <0.001 | 751 (11.5%) | 265 (25.6%) | <0.001 |
Diabetes (%) | 195 (3%) | 51 (4.9%) | <0.001 | 249 (3.8%) | 101 (9.8%) | <0.001 |
BMI (kg/m2) | 22.17 ± 2.68 | 24.61 ± 2.59 | <0.001 | 22.54 ± 2.71 | 25.93 ± 2.75 | <0.001 |
Waist (cm) | 75.13 ± 8.71 | 83.26 ± 8.24 | <0.001 | 75.21 ± 9.36 | 86.62 ± 8.53 | <0.001 |
SBP (mm Hg) | 119.53 ± 15.85 | 125.60 ± 16.43 | <0.001 | 123.23 ± 16.94 | 132.48 ± 17.10 | <0.001 |
DBP (mm Hg) | 74.44 ± 10.16 | 78.39 ± 10.93 | <0.001 | 76.49 ± 10.44 | 82.34 ± 10.36 | <0.001 |
HB (g/L) | 141.50 ± 14.29 | 149.76 ± 14.33 | <0.001 | 139.02 ± 14.86 | 149.02 ± 14.84 | <0.001 |
ALT (IU/L) | 15.80 (12.50, 21.20) | 20.60 (16.10, 28.10) | <0.001 | 15.20 (11.50, 21.00) | 24.00 (17.70, 34.10) | <0.001 |
AST (IU/L) | 18.10 (15.70, 21.10) | 19.80 (17.00, 23.00) | <0.001 | 17.90 (15.40, 21.10) | 20.70 (17.60, 24.90) | <0.001 |
GGT (IU/L) | 16.30 (12.90, 22.50) | 22.20 (16.80, 33.10) | <0.001 | 18.30 (14.40, 25.40) | 30.20 (22.00, 43.20) | <0.001 |
FPG (mmol/L) | 4.94 ± 0.85 | 5.10 ± 0.84 | <0.001 | 5.04 ± 0.88 | 5.44 ± 1.09 | <0.001 |
Scr (µmol/L) | 59.30 ± 13.86 | 65.89 ± 14.69 | <0.001 | 61.28 ± 14.85 | 67.60 ± 14.95 | <0.001 |
UA (µmol/L) | 315.25 ± 78.77 | 367.89 ± 85.03 | <0.001 | 315.38 ± 80.34 | 381.63 ± 84.80 | <0.001 |
TG (mmol/L) | 0.81 (0.60, 1.14) | 1.21 (0.86, 1.64) | <0.001 | 0.88 (0.65, 1.27) | 1.55 (1.15, 2.19) | <0.001 |
TC (mmol/L) | 4.38 ± 0.81 | 4.53 ± 0.86 | <0.001 | 4.89 ± 0.89 | 5.13 ± 0.98 | <0.001 |
HDL (mmol/L) | 1.46 ± 0.36 | 1.25 ± 0.31 | <0.001 | 1.57 ± 0.39 | 1.27 ± 0.30 | <0.001 |
LDL (mmol/L) | 2.49 ± 0.68 | 2.71 ± 0.71 | <0.001 | 2.83 ± 0.75 | 3.13 ± 0.82 | <0.001 |
ALT, alanine aminotransferase; AST, aspartate aminotransferase; BMI, body mass index; DBP, diastolic blood pressure; FPG; fasting plasma glucose; GGT, gamma-glutamyl transferase; HB, hemoglobin; HDL, high-density lipoprotein; LDL, low-density lipoprotein; MASLD, metabolic dysfunction-associated steatotic liver disease; SBP, systolic blood pressure; Scr, serum creatinine; TG, triglycerides; TC, total cholesterol; UA, uric acid.
Development of the predictive nomogram
We randomly divided the participants into a training set and a validation set in a 7:3 ratio. Table 2 shows a similar distribution of variables between these two groups.
Comparison of variables between training set and validation set in 2018.
Variables | Training set (n = 5307) | Validation set (n = 2275) | P |
---|---|---|---|
Gender (male%) | 1977 (37.3%) | 893 (39.3%) | 0.105 |
Age (years) | 39.57 ± 13.40 | 39.70 ± 13.45 | 0.693 |
Hypertension (%) | 521 (9.8%) | 237 (10.4%) | 0.449 |
Diabetes (%) | 169 (3.2%) | 77 (3.4%) | 0.704 |
BMI (kg/m2) | 22.50 ± 2.82 | 22.51 ± 2.74 | 0.857 |
Waist (cm) | 76.21 ± 9.12 | 76.29 ± 9.01 | 0.714 |
SBP (mm Hg) | 120.25 ± 16.05 | 120.61 ± 16.09 | 0.373 |
DBP (mm Hg) | 75.03 ± 10.35 | 74.86 ± 10.38 | 0.521 |
HB (g/L) | 142.50 ± 14.42 | 142.92 ± 14.94 | 0.248 |
ALT (IU/L) | 16.40 (12.70, 22.20) | 16.60 (12.80, 22.40) | 0.730 |
AST (IU/L) | 18.30 (15.80, 21.40) | 18.40 (15.90, 21.60) | 0.881 |
GGT (IU/L) | 17.00 (13.20, 23.80) | 16.80 (13.20, 23.65) | 0.709 |
FPG (mmol/L) | 4.96 ± 0.82 | 4.97 ± 0.91 | 0.777 |
Scr (µmol/L) | 59.98 ± 14.13 | 60.73 ± 14.21 | 0.034 |
UA (µmol/L) | 322.00 ± 81.97 | 323.44 ± 80.98 | 0.481 |
TG (mmol/L) | 0.86 (0.62, 1.24) | 0.85 (0.62, 1.22) | 0.656 |
TC (mmol/L) | 4.40 ± 0.83 | 4.39 ± 0.79 | 0.628 |
HDL (mmol/L) | 1.43 ± 0.37 | 1.43 ± 0.35 | 0.929 |
LDL (mmol/L) | 2.52 ± 0.70 | 2.52 ± 0.67 | 0.903 |
ALT, alanine aminotransferase; AST, aspartate aminotransferase; BMI, body mass index; DBP, diastolic blood pressure; FPG, fasting plasma glucose; GGT, gamma-glutamyl transferase; HB, hemoglobin; HDL, high-density lipoprotein; LDL, low-density lipoprotein; MASLD, metabolic dysfunction-associated steatotic liver disease; SBP, systolic blood pressure; Scr, serum creatinine; TG, triglycerides; TC, total cholesterol; UA, uric acid.
The initial univariate analysis of rows within the training set indicated a total of 19 variables. These variables were gender, age, BMI, waist, history of hypertension and diabetes, SBP and DBP, HB, ALT, AST, GGT, Scr, FPG, UA, TG, TC, LDL, and HDL. Subsequently, we conducted LASSO regression analyses using a regularization parameter value (λ) of 0.0149. Figure 2 provides a visual representation of the correlation between the regression coefficients and Log λ. A total of eight variables were culled from the original set of 19 by this analysis. Gender, BMI, waist, ALT, HB, UA, TG, and HDL were used as characteristic variables to construct a 5-year MASLD risk prediction model. The logistic regression model's variables are shown in Table 3, along with their regression coefficients, odds ratios, and 95% CI.
Selection of predictor variables via least absolute shrinkage and selection operator (LASSO) regression. (A) Regression coefficients plotted against Log(λ). (B) Mean square error in LASSO regression depicted as a function of Log(λ).
Citation: Endocrine Connections 13, 8; 10.1530/EC-24-0186
Multivariate logistic regression analysis for the prediction of MASLD in the training set.
Variables | B | OR (95% CI) | P |
---|---|---|---|
Gender (male) | −0.30 | 0.74 (0.54–1.01) | 0.059 |
BMI (kg/m2) | 0.19 | 1.21 (1.14–1.28) | <0.001 |
Waist (cm) | 0.02 | 1.02 (1.00–1.04) | 0.034 |
HB (g/L) | 0.02 | 1.02 (1.01–1.02) | <0.001 |
ALT (IU/L) | 0.01 | 1.01 (1.01–1.02) | <0.001 |
UA (µmol/L) | 0.002 | 1.00 (1.00–1.00) | 0.002 |
TG (mmol/L) | 0.42 | 1.52 (1.34–1.72) | <0.001 |
HDL (mmol/L) | −0.54 | 0.58 (0.43–0.78) | <0.001 |
ALT, alanine aminotransferase; BMI, body mass index; HB, hemoglobin; HDL, high-density lipoprotein; MASLD, metabolic dysfunction-associated steatotic liver disease; OR, odds ratio; TG, triglycerides; UA, uric acid.
A nomogram was used to graphically display the logistic regression model. Figure 3 illustrates the nomogram. For example, when a healthy male has a BMI of 22.5 kg/m2, waist circumference of 85 cm, HB of 120 g/L, ALT of 20 IU/L, UA of 200 μmol/L, TG of 0.5 mmol/L, and HDL of 2.0 mmol/L, after 5 years, his probability of having MASLD is 3.62%, according to our calculations.
Nomogram for predicting 5-year metabolic dysfunction-associated steatotic liver disease (MASLD). BMI, body mass index; HB, hemoglobin; ALT, alanine aminotransferase; UA, uric acid; TG, triglycerides; HDL, high-density lipoprotein.
Citation: Endocrine Connections 13, 8; 10.1530/EC-24-0186
Internal validation of predictive model
Figure 4 illustrates the ROC curves for the nomogram. The AUC of the training set is 0.795 (95% CI: 0.779–0.811). The AUC of the validation set is 0.785 (95% CI: 0.760–0.810). This study proves with its findings that the nomogram has a strong discriminatory performance in predicting the development of MASLD over a period of 5 years. In order to evaluate the degree of congruence between the predicted values derived from the nomogram and the actual incidences of MASLD, calibration curves were utilized. These curves are displayed in Fig. 5. Furthermore, the results of the DCA, which are depicted in Fig. 6, demonstrated that the nomogram's clinical value was superior to the scenarios of ‘none developed MASLD’ or ‘all developed MASLD’. The threshold probability for the predictive model is 5–50%.
Receiver operating characteristic (ROC) curves. (A) From the training set and (B) from the validation set. AUC, area under the ROC.
Citation: Endocrine Connections 13, 8; 10.1530/EC-24-0186
Calibration curves of the nomogram. (A) From the training set and (B) from the validation set.
Citation: Endocrine Connections 13, 8; 10.1530/EC-24-0186
Decision curves of the nomogram. (A) From the training set and (B) from the validation set. MASLD, metabolic dysfunction-associated steatotic liver disease.
Citation: Endocrine Connections 13, 8; 10.1530/EC-24-0186
Discussion
We followed 7582 participants in health checkups for 5 years. During the follow-up, 1034 (13.64%) participants were newly diagnosed with MASLD. Using eight variables (gender, BMI, waist, ALT, HB, UA, TG, and HDL) filtered by LASSO regression, we built a 5-year MASLD risk prediction model. The nomogram showed an AUC of 0.795 (95% CI: 0.779–0.811) in the training set and 0.785 (95% CI: 0.760–0.810) in the validation set. The calibration curves revealed a 5-year period of agreement between the observed and predicted MASLD risks. DCA curves demonstrated the clinical utility of this nomogram over threshold probability profiles ranging from 5% to 50%.
Our study identified eight variables for predicting 5-year MASLD risk as gender, BMI, waist, ALT, HB, UA, TG, and HDL. A growing number of people use the FLI (8), which is based on cross-sectional data and includes variables like waist, BMI, TG, and GGT. BMI, waist, and TG are included as predictors in our longitudinal MASLD prediction model, confirming the strong association between MASLD and these variables. This further suggests that the new name MASLD reflects the pathogenesis better than NAFLD. In our study, both LDL and HDL were correlated with the development of MASLD in univariate analysis. LDL was excluded from further multivariate regression analysis. Nonetheless, HDL levels were still negatively correlated with MASLD risk. A recent study using liquid chromatography–mass spectrometry found that the HDL content of patients with MASLD was decreased, and the lipid composition of HDL differed from that of control subjects (23).
A variety of liver disorders, including fibrosis, cirrhosis, and hepatocellular carcinoma, can progress in MASLD. The severity of these conditions can vary from mild steatosis to life-threatening steatohepatitis (24). When it comes to liver disease in Americans waiting for a transplant, nonalcoholic steatohepatitis ranks second (25). Not only that, but MASLD increases the risk of extrahepatic disease. Mounting evidence has linked MASLD to an elevated risk of cardiovascular disease, including both fatal and non-fatal complications (26, 27). Additionally, the level of seriousness of MASLD and a higher prevalence of cardiovascular problems are correlated (28). Several studies conducted in both hospitals and populations have demonstrated that the incidence of CKD is significantly higher in patients who have MASLD, and that the association between MASLD and CKD is significant even after accounting for hypertension, type 2 diabetes, and other risk factors for CKD (6, 29). There is also an increased likelihood of developing colorectal adenomas and osteoporosis when MASLD is present (30, 31). The amount of money spent annually on the treatment of MASLD in the United States is over one hundred billion dollars, whereas in four European countries (the United Kingdom, France, Germany, and Italy) it is over 35 billion Euros (32).
As the global obesity problem continues to grow, the prevalence of MASLD is gradually increasing (7). Considering the high prevalence and potential harm of MASLD, it is especially crucial to identify high-risk individuals at an early stage in healthy populations. Although several models for predicting NAFLD have been proposed in the past, most of them have not been revalidated in the MASLD population, and most of these models are constructed based on cross-sectional data. Our study, based on a longitudinal follow-up cohort of the population, developed a predictive model that can be applied when individuals have not yet been diagnosed with MASLD. At the same time, our model variables are relatively simple and applicable to mass screening. By presenting the model results through nomogram-based visualizations, our study enables participants to more easily understand their own risk of developing MASLD, thus facilitating individual self-awareness and lifestyle changes. All of this contributes to reducing the risk of developing MASLD and further reducing the harm it causes.
Nonetheless, several limitations must be considered: (1) we used ultrasound results of the liver to diagnose MASLD, rather than the results of a liver biopsy, which was biased because the gold standard method was not used. (2) Logistic regression analysis was utilized in lieu of COX regression analysis due to a lack of follow-up duration information. (3) The nomogram was developed using data from a single center. (4) The results of this study had only been validated with an internal validation set. (5) Our study lacked factors regarding demographic characteristics, dietary preferences, exercise habits, and medication history, which can influence the onset of MASLD. Due to the lack of these factors, this may limit the predictive ability of our prediction model.
Conclusion
A predictive nomogram was constructed incorporating eight risk predictors (gender, BMI, waist, ALT, HB, UA, TG, and HDL) for anticipating MASLD occurrence over a 5-year period. This nomogram serves as a straightforward and economical tool to identify high-risk individuals for MASLD, enabling timely interventions aimed at mitigating the onset of the disease.
Supplementary materials
This is linked to the online version of the paper at https://doi.org/10.1530/EC-24-0186.
Declaration of interest
The authors maintain that the study is devoid of any conflicts of interest that could potentially compromise its objectivity.
Funding
The funding for this research came from the Nanjing Health Science and Technology Development Major Project no. ZDX21001.
Author contribution statement
YH and WZ were responsible for the preliminary design and development of the study. L G gathered information, conducted additional data and statistical analysis, and penned the article. The article was conceptualized and substantially revised by WC. Both NL and DM contributed to the article revision and data acquisition processes. All contributors approved and co-authored the manuscript.
Acknowledgements
All of the members who helped with the manuscript’s development have our deepest gratitude.
References
- 1↑
Rinella ME, Lazarus JV, Ratziu V, Francque SM, Sanyal AJ, Kanwal F, Romero D, Abdelmalek MF, Anstee QM, Arab JP, et al.A multisociety Delphi consensus statement on new fatty liver disease nomenclature. Journal of Hepatology 2023 79 1542–1556. (https://doi.org/10.1016/j.jhep.2023.06.003)
- 2↑
Miao L, Targher G, Byrne CD, Cao YY, & Zheng MH. Current status and future trends of the global burden of MASLD. Trends in Endocrinology and Metabolism 2024 (S1043-2760(24)00036-5). (https://doi.org/10.1016/j.tem.2024.02.007)
- 3↑
Targher G, Byrne CD, & Tilg H. MASLD: a systemic metabolic disorder with cardiovascular and malignant complications. Gut 2024 73 691–702. (https://doi.org/10.1136/gutjnl-2023-330595)
- 4↑
Moon JH, Jeong S, Jang H, Koo BK, & Kim W. Metabolic dysfunction-associated steatotic liver disease increases the risk of incident cardiovascular disease: a nationwide cohort study. EClinicalmedicine 2023 65 102292. (https://doi.org/10.1016/j.eclinm.2023.102292)
- 5↑
Roderburg C, Krieg S, Krieg A, Demir M, Luedde T, Kostev K, & Loosen SH. Non-alcoholic fatty liver disease (NAFLD) is associated with an increased incidence of chronic kidney disease (CKD). European Journal of Medical Research 2023 28 153. (https://doi.org/10.1186/s40001-023-01114-6)
- 6↑
Bilson J, Mantovani A, Byrne CD, & Targher G. Steatotic liver disease, MASLD and risk of chronic kidney disease. Diabetes and Metabolism 2024 50 101506. (https://doi.org/10.1016/j.diabet.2023.101506)
- 7↑
Cusi K, Isaacs S, Barb D, Basu R, Caprio S, Garvey WT, Kashyap S, Mechanick JI, Mouzaki M, Nadolsky K, et al.American association of clinical endocrinology clinical practice guideline for the diagnosis and management of nonalcoholic fatty liver disease in primary care and endocrinology clinical settings. Endocrine Practice 2022 28 528–562. (https://doi.org/10.1016/j.eprac.2022.03.010)
- 8↑
Bedogni G, Bellentani S, Miglioli L, Masutti F, Passalacqua M, Castiglione A, & Tiribelli C. The fatty liver index: a simple and accurate predictor of hepatic steatosis in the general population. BMC Gastroenterology 2006 6 33. (https://doi.org/10.1186/1471-230X-6-33)
- 9↑
Lee JH, Kim D, Kim HJ, Lee CH, Yang JI, Kim W, Kim YJ, Yoon JH, Cho SH, Sung MW, et al.Hepatic steatosis index: a simple screening tool reflecting nonalcoholic fatty liver disease. Digestive and Liver Disease 2010 42 503–508. (https://doi.org/10.1016/j.dld.2009.08.002)
- 10↑
Zou H, Zhao F, Lv X, Ma X, & Xie Y. Development and validation of a new nomogram to screen for MAFLD. Lipids in Health and Disease 2022 21 133. (https://doi.org/10.1186/s12944-022-01748-1)
- 11↑
Zhou Y, Chai X, Guo T, Pu Y, Zeng M, Zhong A, Yang G, & Cai J. A prediction model of the incidence of nonalcoholic fatty liver disease with visceral fatty obesity: a general population-based study. Frontiers in Public Health 2022 10 895045. (https://doi.org/10.3389/fpubh.2022.895045)
- 12↑
Pan H, Liu B, Luo X, Shen X, Sun J, & Zhang A. Non-alcoholic fatty liver disease risk prediction model and health management strategies for older Chinese adults: a cross-sectional study. Lipids in Health and Disease 2023 22 205. (https://doi.org/10.1186/s12944-023-01966-1)
- 13↑
Wang J, Tang Y, Peng K, Liu H, & Xu J. Development and validation of a nomogram for predicting nonalcoholic fatty liver disease in the non-obese Chinese population. American Journal of Translational Research 2020 12 6149–6159. (https://doi.org/10.21203/rs.3.rs-24981/v1)
- 14↑
Zhu W, Shi P, Fu J, Liang A, Zheng T, Wu X, & Yuan S. Development and application of a novel model to predict the risk of non-alcoholic fatty liver disease among lean pre-diabetics with normal blood lipid levels. Lipids in Health and Disease 2022 21 149. (https://doi.org/10.1186/s12944-022-01752-5)
- 15↑
Peng HY, Duan SJ, Pan L, Wang MY, Chen JL, Wang YC, & Yao SK. Development and validation of machine learning models for nonalcoholic fatty liver disease. Hepatobiliary and Pancreatic Diseases International 2023 22 615–621. (https://doi.org/10.1016/j.hbpd.2023.03.009)
- 16↑
Su PY, Chen YY, Lin CY, Su WW, Huang SP, & Yen HH. Comparison of machine learning models and the fatty liver index in predicting lean fatty liver. Diagnostics 2023 13 1407. (https://doi.org/10.3390/diagnostics13081407)
- 17↑
Liu YX, Liu X, Cen C, Li X, Liu JM, Ming ZY, Yu SF, Tang XF, Zhou L, Yu J, et al.Comparison and development of advanced machine learning tools to predict nonalcoholic fatty liver disease: an extended study. Hepatobiliary and Pancreatic Diseases International 2021 20 409–415. (https://doi.org/10.1016/j.hbpd.2021.08.004)
- 18↑
Hsu CL, Wu PC, Wu FZ, & Yu HC. LASSO-derived model for the prediction of lean-non-alcoholic fatty liver disease in examinees attending a routine health check-up. Annals of Medicine 2024 56 2317348. (https://doi.org/10.1080/07853890.2024.2317348)
- 19↑
Deng Y, Ma Y, Fu J, Wang X, Yu C, Lv J, Man S, Wang B, & Li L. A dynamic machine learning model for prediction of NAFLD in a health checkup population: a longitudinal study. Heliyon 2023 9 e18758. (https://doi.org/10.1016/j.heliyon.2023.e18758)
- 20↑
Huang G, Jin Q, & Mao Y. Predicting the 5-year risk of nonalcoholic fatty liver disease using machine learning models: prospective cohort study. Journal of Medical Internet Research 2023 25 e46891. (https://doi.org/10.2196/46891)
- 21↑
Collins GS, Reitsma JB, Altman DG, Moons KGM & TRIPOD Group. Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): the TRIPOD statement. The TRIPOD Group. Circulation 2015 131 211–219. (https://doi.org/10.1161/CIRCULATIONAHA.114.014508)
- 22↑
Riley RD, Ensor J, Snell KIE, Harrell FE, Martin GP, Reitsma JB, Moons KGM, Collins G, & Smeden van M. Calculating the sample size required for developing a clinical prediction model. BMJ 2020 368 m441. (https://doi.org/10.1136/bmj.m441)
- 23↑
Mocciaro G, Allison M, Jenkins B, Azzu V, Huang-Doran I, Herrera-Marcos LV, Hall Z, Murgia A, Susan D, Frontini M, et al.Non-alcoholic fatty liver disease is characterised by a reduced polyunsaturated fatty acid transport via free fatty acids and high-density lipoproteins (HDL). Molecular Metabolism 2023 73 101728. (https://doi.org/10.1016/j.molmet.2023.101728)
- 24↑
Ebrahimi F, Hagström H, Sun J, Bergman D, Shang Y, Yang W, Roelstraete B, & Ludvigsson JF. Familial coaggregation of MASLD with hepatocellular carcinoma and adverse liver outcomes: nationwide multigenerational cohort study. Journal of Hepatology 2023 79 1374–1384. (https://doi.org/10.1016/j.jhep.2023.08.018)
- 25↑
Wong RJ, Aguilar M, Cheung R, Perumpail RB, Harrison SA, Younossi ZM, & Ahmed A. Nonalcoholic steatohepatitis is the second leading etiology of liver disease among adults awaiting liver transplantation in the United States. Gastroenterology 2015 148 547–555. (https://doi.org/10.1053/j.gastro.2014.11.039)
- 26↑
Duell PB, Welty FK, Miller M, Chait A, Hammond G, Ahmad Z, Cohen DE, Horton JD, Pressman GS, Toth PP, et al.Nonalcoholic fatty liver disease and cardiovascular risk: a scientific statement from the American Heart Association. Arteriosclerosis, Thrombosis, and Vascular Biology 2022 42 e168–e185 (https://doi.org/10.1161/ATV.0000000000000153)
- 27↑
Mantovani A, Csermely A, Petracca G, Beatrice G, Corey KE, Simon TG, Byrne CD, & Targher G. Non-alcoholic fatty liver disease and risk of fatal and non-fatal cardiovascular events: an updated systematic review and meta-analysis. Lancet. Gastroenterology and Hepatology 2021 6 903–913. (https://doi.org/10.1016/S2468-1253(2100308-3)
- 28↑
Simon TG, Roelstraete B, Hagström H, Sundström J, & Ludvigsson JF. Non-alcoholic fatty liver disease and incident major adverse cardiovascular events: results from a nationwide histology cohort. Gut 2022 71 1867–1875. (https://doi.org/10.1136/gutjnl-2021-325724)
- 29↑
Adams LA, Anstee QM, Tilg H, & Targher G. Non-alcoholic fatty liver disease and its relationship with cardiovascular disease and other extrahepatic diseases. Gut 2017 66 1138–1153. (https://doi.org/10.1136/gutjnl-2017-313884)
- 30↑
Huang KW, Leu HB, Wang YJ, Luo JC, Lin HC, Lee FY, Chan WL, Lin JK, & Chang FY. Patients with nonalcoholic fatty liver disease have higher risk of colorectal adenoma after negative baseline colonoscopy. Colorectal Disease 2013 15 830–835. (https://doi.org/10.1111/codi.12172)
- 31↑
Targher G, Lonardo A, & Rossini M. Nonalcoholic fatty liver disease and decreased bone mineral density: is there a link? Journal of Endocrinological Investigation 2015 38 817–825. (https://doi.org/10.1007/s40618-015-0315-6)
- 32↑
Younossi ZM, Blissett D, Blissett R, Henry L, Stepanova M, Younossi Y, Racila A, Hunt S, & Beckerman R. The economic and clinical burden of nonalcoholic fatty liver disease in the United States and Europe. Hepatology 2016 64 1577–1586. (https://doi.org/10.1002/hep.28785)