Fibrosis metabolic associated steatohepatitis risk assessment system

Abstract

The present application belongs to the technical field of medical risk assessment, and proposes a fibrosis metabolism related fatty hepatitis risk assessment system. It comprises: a data acquisition module configured to determine candidate prediction indicators; a primary screening module configured to perform single-factor logistic regression analysis on the candidate prediction indicators to determine key prediction indicators of fibrosis metabolism related fatty hepatitis; a secondary screening module configured to screen four independent prediction factors from the key prediction indicators of fibrosis metabolism related fatty hepatitis through stepwise multiple regression analysis; and a model training module configured to train a constructed multi-factor logistic regression model based on the four screened independent prediction factors. The present application integrates four indicators of aspartate aminotransferase, alanine aminotransferase, triglyceride and high-density lipoprotein cholesterol to construct a fibrosis MASH evaluation prediction model for obese people, which has better evaluation performance.

Description

Technical Field

[0001] This invention belongs to the field of medical risk assessment technology, and in particular relates to a risk assessment system for fibrosis-related steatohepatitis. Background Technology

[0002] Metabolic dysfunction-associated steatotic liver disease (MASLD) is a common liver disease worldwide. Within the entire spectrum of MASLD, patients with metabolic dysfunction-associated steatohepatitis (MASLD) and significant liver fibrosis, termed "fibrotic MASLD," have the highest risk of developing liver-related complications and are closely associated with a significantly increased all-cause mortality rate.

[0003] The pathophysiology of MASH is strongly associated with obesity. Importantly, obesity-related comorbidities, such as cardiovascular disease and type 2 diabetes, can independently promote the progression of liver fibrosis. Although only about 20% of patients with MASLD will progress to MASH, its prevalence is significantly higher in obese individuals, some of whom already exhibit clinically significant fibrosis.

[0004] Clinically, the lack of routine liver biopsies means that a large number of obese patients with fibrotic MASH go undiagnosed, missing the opportunity for early treatment intervention. This situation can lead to disease progression to advanced liver complications and continuously increase the health burden. Currently, clinical guidelines for MASH with fibrosis are undergoing a shift, moving from focusing on the management of fibrosis in patients diagnosed with MASLD to screening high-risk populations. The key challenge remains accurately identifying fibrotic MASH in high-risk individuals with both MASLD and obesity. However, achieving this goal faces significant difficulties.

[0005] In recent years, several non-invasive diagnostic methods for assessing liver fibrosis have been applied in clinical practice. These methods include:

[0006] (1) Liver stiffness measurement (LSM).

[0007] Liver stiffness is primarily measured using transient elastography (FibroScan), which quantifies tissue stiffness to assess the degree of fibrosis. However, its accuracy in obese individuals is significantly reduced due to signal attenuation caused by excessive abdominal fat.

[0008] (2) Traditional magnetic resonance imaging (MRI).

[0009] Traditional magnetic resonance imaging, as the current mainstream imaging method, has diagnostic limitations because it is difficult to reliably distinguish fibrotic changes and accompanying inflammatory activity in liver tissue.

[0010] (3) Serum biomarkers based on omics.

[0011] Before omics-based serum biomarkers can be widely adopted and applied in highly heterogeneous clinical settings, they still face significant hurdles in terms of standardization and validation.

[0012] For individuals with moderate to severe obesity, existing non-invasive assessment methods for liver fibrosis still have significant limitations in terms of accuracy and reliability. Summary of the Invention

[0013] To overcome the shortcomings of the existing technologies, this invention provides a risk assessment system for fibrosis-related steatohepatitis. It integrates four indicators—aspartate aminotransferase (AST), alanine aminotransferase (ALT), triglycerides, and high-density lipoprotein cholesterol (HDL-C)—to construct a fibrosis MASH assessment and prediction model for obese individuals. This system offers superior assessment performance and provides an accurate and convenient solution for assessing liver disease risk in obese populations. It aims to address the problems of low accuracy and reliance on expensive or invasive techniques in existing non-invasive assessment methods for moderate to severe obesity.

[0014] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions:

[0015] This invention provides a risk assessment system for fibrosis-related steatohepatitis.

[0016] A risk assessment system for fibrosis-related steatohepatitis includes:

[0017] The data acquisition module is configured to: acquire the weight information, clinical examination information and pathological evaluation results of liver tissue samples of patients with fibrosis-related steatohepatitis, determine candidate predictive indicators from them, and construct a dataset;

[0018] The screening module is configured to perform univariate logistic regression analysis on candidate predictive indicators to identify key predictive indicators for fibrotic metabolic-related steatohepatitis.

[0019] The secondary screening module is configured to: screen out four independent predictors from key predictors of fibrotic metabolic steatohepatitis through stepwise multiple regression analysis.

[0020] The model training module is configured to: train the constructed multi-factor logistic regression model based on the four selected independent predictors, and use the trained multi-factor logistic regression model as a risk assessment model.

[0021] The assessment and prediction module is configured to assess the risk of the person being assessed for fibrotic metabolic steatohepatitis based on a risk assessment model.

[0022] As an alternative technical solution, the weight information, clinical examination information, and pathological evaluation results of liver tissue samples of patients with fibrosis-related metabolic steatohepatitis are obtained in the following ways:

[0023] Identify individuals who underwent bariatric surgery to treat morbid obesity within a specified time period in the past;

[0024] Obtain the results of physical, clinical, and laboratory examinations 1-2 days prior to bariatric surgery for individuals undergoing bariatric surgery to treat morbid obesity.

[0025] Weight and clinical examination information of patients with fibrosis-related steatohepatitis were extracted from physical examination, clinical and laboratory examination results.

[0026] The pathological evaluation results of liver tissue samples were obtained through systematic evaluation by liver pathologists.

[0027] As an alternative technical solution, the weight information, clinical examination information, and pathological evaluation results of liver tissue samples of patients with fibrosis-related metabolic steatohepatitis are specifically measured as follows:

[0028] The weight information is measured using the Body Mass Index (BMI).

[0029] The clinical examination information is measured by hematological indicators from fasting blood samples collected from patients 1-2 days before bariatric surgery.

[0030] The pathological evaluation results of the liver tissue samples were measured by the degree of liver fibrosis and histological changes; among which, histological changes were scored separately for steatosis, lobular inflammation and ballooning degeneration, and the sum of the scores for steatosis, lobular inflammation and ballooning degeneration was calculated.

[0031] As alternative technical solutions, the specific candidate predictive indicators include: age, sex, body mass index, waist-to-hip ratio, alanine aminotransferase, aspartate aminotransferase, triglycerides, total cholesterol, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, albumin, platelet count, uric acid, glycated hemoglobin, fasting blood glucose, fasting C-peptide, fasting insulin, insulin resistance index, free triiodothyronine, free thyroxine, and thyroid-stimulating hormone.

[0032] As an alternative technical solution, candidate predictive indicators whose univariate logistic regression analysis results are less than a set value are selected from the candidate predictive indicators and used as key predictive indicators for fibrosis-related steatohepatitis. These key predictive indicators specifically include:

[0033] Alanine aminotransferase (ALT), aspartate aminotransferase (AST), triglycerides (TG), total cholesterol (TC), high-density lipoprotein cholesterol (HDL-c), low-density lipoprotein cholesterol (LDL-c), platelet count (PLT), glycated hemoglobin (HbA1c), fasting plasma glucose (FPG), fasting C-peptide (FCP), and the homeostasis model assessment index (HOMA-IR) of insulin resistance.

[0034] As an alternative technical solution, four independent predictors were selected through stepwise multiple regression analysis using self-sampling. These four independent predictors include alanine aminotransferase (ALT), aspartate aminotransferase (AST), triglycerides (TG), and high-density lipoprotein cholesterol (HDL-c).

[0035] As an optional technical solution, the risk assessment model is specifically as follows:

[0036] ;

[0037] in, , , , These represent the concentrations of ALT, AST, TG, and HDL-c, respectively. The unit is per liter. It represents millimoles per liter.

[0038] As an optional technical solution, in the evaluation and prediction module:

[0039] Based on the risk assessment model, the risk of the person being assessed to have fibrotic metabolic steatohepatitis is assessed, and a risk assessment value is obtained, which is a probability value in the range of 0 to 1.

[0040] As alternative technical solutions, the following also include:

[0041] The threshold determination module is configured to use a dual-threshold strategy to define risk ranges, dividing personnel into low-risk, uncertain, and high-risk groups, specifically:

[0042] Set exclusion and inclusion thresholds;

[0043] Individuals whose risk assessment scores are below the exclusion threshold will be classified into the low-risk group;

[0044] Individuals whose risk assessment scores are higher than the inclusion threshold will be classified into the high-risk group;

[0045] Individuals whose risk assessment values ​​fall between the exclusion and inclusion thresholds are classified into the uncertain group.

[0046] As alternative technical solutions, the following also include:

[0047] The model validation module is configured as follows:

[0048] Internal validation was performed by calculating the average C-index using multiple Bootstrap resampling methods on the training set obtained by splitting the dataset, to assess the model’s internal consistency and potential overfitting.

[0049] External validation assesses the model’s discriminative power, calibration, and net clinical benefit on a validation set derived from the dataset.

[0050] The above one or more technical solutions have the following beneficial effects:

[0051] This invention provides a risk assessment system for fibrosis-related steatohepatitis, integrating four indicators: aspartate aminotransferase, alanine aminotransferase, triglycerides, and high-density lipoprotein cholesterol. It constructs a fibrosis MASH assessment and prediction model for obese individuals, exhibiting superior assessment performance. This provides an accurate and convenient solution for assessing liver disease risk in obese populations, aiming to address the problems of low accuracy and reliance on expensive or invasive techniques in existing non-invasive assessment methods for moderate to severe obese individuals.

[0052] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0053] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0054] Figure 1 This is an overall flowchart of Embodiment 1 of the present invention.

[0055] Figure 2 This is the ROC curve of the FMO model in Embodiment 1 of the present invention on the training set and validation set.

[0056] Figure 3 This is the ROC curve of the training set in Embodiment 1 of the present invention.

[0057] Figure 4 This is the ROC curve of the validation set in Embodiment 1 of the present invention.

[0058] Figure 5 This is a comparison of ROC curves between FMO and control indicators in an independent weight loss surgery cohort according to Embodiment 1 of the present invention.

[0059] Figure 6 This is the ROC curve of the NHANES queue in Embodiment 1 of the present invention.

[0060] Figure 7 This is the ROC curve of the UK biobank cohort in Embodiment 1 of the present invention.

[0061] Figure 8 This is a comparison chart of FMO and the existing non-invasive index C in Embodiment 1 of the present invention. Detailed Implementation

[0062] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0063] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations of the present invention.

[0064] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0065] Example 1

[0066] This invention provides a risk assessment system for fibrosis-associated steatohepatitis (MAH). Based on readily available biomarkers and anthropometric parameters, a model is constructed using easily accessible clinical indicators. A clinically practical scoring tool has been developed and validated to reliably identify MASH with fibrosis in obese individuals, thereby enabling risk assessment of fibrosis-associated steatohepatitis in obese patients.

[0067] like Figure 1 As shown, the fibrosis-related steatohepatitis risk assessment system includes:

[0068] The data acquisition module is configured to: acquire the weight information, clinical examination information and pathological evaluation results of liver tissue samples of patients with fibrosis-related steatohepatitis, determine candidate predictive indicators from them, and construct a dataset;

[0069] The screening module is configured to perform univariate logistic regression analysis on candidate predictive indicators to identify key predictive indicators for fibrotic metabolic-related steatohepatitis.

[0070] The secondary screening module is configured to: screen out four independent predictors from key predictors of fibrotic metabolic steatohepatitis through stepwise multiple regression analysis.

[0071] The model training module is configured to: train the constructed multi-factor logistic regression model based on the four selected independent predictors, and use the trained multi-factor logistic regression model as a risk assessment model.

[0072] The assessment and prediction module is configured to assess the risk of the person being assessed for fibrotic metabolic steatohepatitis based on a risk assessment model.

[0073] The fibrosis-associated steatohepatitis risk assessment system disclosed in this embodiment is a robust and clinically implementable scoring system that combines routinely available biomarkers and anthropometric parameters for the non-invasive assessment of obese patients with fibrosis-associated steatohepatitis. The objective of this embodiment is:

[0074] (1) It can accurately identify MASH with fibrosis in high-risk groups of obesity;

[0075] (2) Facilitates large-scale screening and risk stratification;

[0076] (3) Support disease progression monitoring and treatment decision making.

[0077] Furthermore, in this embodiment, the weight information, clinical examination information, and pathological evaluation results of liver tissue samples of patients with fibrosis-related steatohepatitis are obtained in the following ways:

[0078] Identify individuals who underwent bariatric surgery to treat morbid obesity within a specified time period in the past;

[0079] Obtain the results of physical, clinical, and laboratory examinations 1-2 days prior to bariatric surgery for individuals undergoing bariatric surgery to treat morbid obesity.

[0080] Weight and clinical examination information of patients with fibrosis-related steatohepatitis were extracted from physical examination, clinical and laboratory examination results.

[0081] The pathological evaluation results of liver tissue samples were obtained through systematic evaluation by liver pathologists.

[0082] Going further:

[0083] After obtaining weight information, clinical examination information, and pathological evaluation results of liver tissue samples from multiple obese patients who had undergone bariatric surgery in the past, candidate predictive indicators were identified based on the obtained information. Univariate logistic regression analysis and bidirectional stepwise regression analysis were then performed on the candidate predictive indicators.

[0084] First, each candidate predictive indicator is analyzed independently using univariate logistic regression analysis, and the corresponding p-value is calculated. The p-value is used to test whether there is a significant relationship between the current variable and the dependent variable. In this embodiment, the dependent variable in the univariate logistic regression analysis is MASH with fibrosis. Generally, when the p-value is less than 0.05, it indicates that there is a statistically significant relationship between the current variable and the dependent variable. Therefore, these significant variables are included in the multivariate logistic regression model to construct a preliminary regression model and screen out key predictive indicators.

[0085] Next, a bidirectional stepwise regression procedure is used for further variable selection. This method combines forward selection and backward elimination, optimizing the model by progressively adding and removing variables. At each step, the recalculated p-value determines whether to retain a variable. Specifically, variables with p-values ​​less than 0.05 are retained, while those with p-values ​​greater than 0.05 are removed. Through this iterative process, the most predictive combination of variables is gradually identified. The p-value in bidirectional stepwise regression analysis is a core threshold criterion used to automatically determine whether a particular independent variable is included or removed from the regression model.

[0086] Finally, through bidirectional stepwise regression screening, four independent predictors were identified. These factors were significant in the multivariate logistic regression model and could independently predict the target variable. Based on these factors, the final predictive model was constructed and used for further data analysis and prediction.

[0087] In this embodiment, bariatric surgery for morbid obesity specifically refers to reducing body weight by decreasing stomach volume and / or shortening the effective length of the small intestine to limit food intake and / or reduce nutrient absorption. Indications mainly include:

[0088] (1) Individuals aged 18-70 years with a BMI ≥ 32.5, or a BMI ≥ 27.5 with type 2 diabetes mellitus (T2DM; regardless of the effectiveness of medical treatment for T2DM). BMI refers to body mass index.

[0089] (2) Individuals aged 18-70 with a BMI of 27.5-32.5 who have not responded to medical weight loss or have other obesity-related diseases.

[0090] In this embodiment, the 21 candidate predictive indicators are: age, sex, body mass index, waist-to-hip ratio, alanine aminotransferase, aspartate aminotransferase, triglycerides, total cholesterol, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, albumin, platelet count, uric acid, glycated hemoglobin, fasting blood glucose, fasting C-peptide, fasting insulin, insulin resistance index, free triiodothyronine, free thyroxine, and thyroid-stimulating hormone.

[0091] More specifically, in this embodiment, univariate logistic regression analysis was performed on 21 biochemical, metabolic, and thyroid function parameters to screen out 11 indicators that were significantly associated with MASH associated with fibrosis (P<0.05), including: alanine aminotransferase (ALT), aspartate aminotransferase (AST), triglycerides (TG), total cholesterol (TC), high-density lipoprotein cholesterol (HDL-c), low-density lipoprotein cholesterol (LDL-c), platelet count (PLT), glycated hemoglobin (HbA1c), fasting plasma glucose (FPG), fasting C-peptide (FCP), and insulin resistance homeostasis model assessment index (HOMA-IR).

[0092] Through stepwise multiple regression analysis using bootstrapping, core predictive factors were identified and a model was built. Ultimately, four independent predictive factors were selected: alanine aminotransferase (ALT), aspartate aminotransferase (AST), triglycerides (TG), and high-density lipoprotein cholesterol (HDL-c).

[0093] Based on the regression coefficients of these four independent predictors, a fibrotic MASH prediction model for obese individuals was constructed—named the FMO (Fibrotic MASH in Obesity) model:

[0094] ;

[0095] In the formula, , , , These represent the concentrations of ALT, AST, TG, and HDL-c, respectively.

[0096] The unit for ALT and AST is per liter, which is a unit of enzyme activity concentration. The parameter unit for TG and HDL-c is millimoles per liter.

[0097] The FMO model uses a logistic regression formula to calculate the probability prediction (range 0-1) of an individual having MASH with fibrosis.

[0098] Interpretation of prediction results: The FMO model outputs a continuous probability value between 0 and 1. For example, an output value of 0.30 indicates that the predicted probability of MASH with fibrosis in the currently obese individual is 30%.

[0099] Furthermore, the model performance was comprehensively evaluated using discrimination, calibration, and clinical usability to quantify the net clinical benefit provided by the model compared to the "treat-all" and "treat-none" strategies at a specific threshold probability.

[0100] Furthermore, a dual-threshold strategy is employed to define the risk range:

[0101] Rule-out thresholds are used to exclude low-risk individuals by setting a critical value for high sensitivity (Sensitivity ≥ 0.90, where Sensitivity represents sensitivity).

[0102] The inclusion threshold (Rule-in) sets a critical value for high specificity (Specificity ≥ 0.90, where Specificity represents specificity) to identify high-risk individuals.

[0103] The specific calculations of the above-mentioned sensitivity and specificity are common knowledge in this field and will not be elaborated here.

[0104] Based on this, the target population was divided into low-risk, uncertain, and high-risk groups. The report presents the sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) for each risk group.

[0105] Furthermore, the model was validated:

[0106] Internal Validation: On the training set, the average C-index is calculated using the Bootstrap resampling method with 1000 iterations to assess the internal consistency of the model and potential overfitting.

[0107] External validation: The constructed model is applied to the validation set to comprehensively evaluate the model's discrimination (AUROC), calibration (calibration curve), and net clinical benefit (decision curve analysis).

[0108] Furthermore, the Delong test is used to compare the AUROC of the FMO model in this embodiment with existing models. A head-to-head comparison is performed between the AUROC of the model and several mature non-invasive scoring metrics (including AAR, FIB-4, FNI, NFS, BARD and AP-index) to demonstrate its superiority. AUROC is the area under the ROC curve.

[0109] Among them, the Delong test is a statistical method used to compare whether there is a significant difference in the area under two ROC curves (Receiver Operating Characteristic curves) (AUROC).

[0110] AAR, FIB-4, FNI, NFS, BARD, and AP-index are all non-invasive diagnostic models or indices used to assess liver fibrosis or MASLD. Among them, FIB-4 is the four-factor index for liver fibrosis, and NFS is the fibrosis score for non-alcoholic fatty liver disease. They predict the degree of liver fibrosis or steatosis by combining routine blood test indicators with basic patient information (such as age, height, and weight), thereby reducing reliance on liver biopsy (the gold standard). These refer to: AAR (AST / ALT), the ratio of aspartate aminotransferase (AST) to alanine aminotransferase (ALT); FIB-4, calculated using four readily available indicators: age, aspartate aminotransferase (AST), alanine aminotransferase (ALT), and platelet count; FNI, calculated using aspartate aminotransferase (AST), high-density lipoprotein cholesterol (HDL-c), and glycated hemoglobin (HbA1c); NFS, calculated using indicators such as age, alanine aminotransferase (ALT), fasting insulin, and platelet count; BARD, a weighted calculation based on AAR, BMI, and type 2 diabetes; and AP-index, calculated using age and platelet count.

[0111] Furthermore, to validate its long-term prognostic predictive ability, the FMO score was evaluated using UK Biobank data to assess its predictive power for future serious liver-related outcomes (SLROs). A risk score was calculated based on a Cox proportional hazards model, and its discriminative power in predicting future serious liver disease events was assessed. The Cox proportional hazards model is a statistical model used to analyze the "timing of an event," considering both whether an event occurs and the duration of its occurrence. In this study, it was used to calculate the FMO score to evaluate its predictive power for future serious liver-related outcomes.

[0112] The technical solution of this embodiment will be explained in detail below.

[0113] (a) Data acquisition and processing.

[0114] This embodiment constructed training and validation sets. Using the training set, univariate logistic regression analysis was performed on 21 biochemical, metabolic, and thyroid function parameters, identifying 11 indicators significantly associated with MASH associated with fibrosis (P<0.05). Stepwise multivariate regression analysis using bootstrapping was then used to determine independent predictors and construct the model. Simultaneously, model performance was tested in internal and external trials.

[0115] In this embodiment, the training cohort comprising the training set includes 1095 individuals who underwent bariatric surgery for morbid obesity between June 2018 and October 2024. Of the 1095 patients in the training set, 143 had fibrotic MASH and 952 did not. Univariate and multivariate logistic regression variables were used to screen them, and the model efficacy was validated in multiple validation sets and compared with multiple existing non-invasive indices.

[0116] The validation cohort that constitutes the validation set consists of four bariatric surgery cohorts from tertiary hospitals in different regions, containing a total of 866 individuals who underwent bariatric surgery to treat morbid obesity.

[0117] Physical examinations, clinical tests, and laboratory examinations are all conducted 1-2 days before surgery, including:

[0118] The Body Mass Index (BMI) is calculated as weight (kg) / height (m). 2 kg represents kilogram, m represents meter;

[0119] The formula for calculating the insulin resistance index (HOMA-IR) is: fasting glucose (mmol / L) x fasting insulin (μIU / mL) / 22.5, where mmol / L represents millimoles per liter and μIU / mL represents micro international units per milliliter.

[0120] The diagnostic criteria for diabetes are a fasting plasma glucose (FPG) level ≥7.0 mmol / L or a glycated hemoglobin (HbA1c) level ≥6.5% (48 mmol / mol), where mmol / mol represents millimoles per mole (the international unit for measuring HbA1c).

[0121] Hypertension is defined as a systolic blood pressure ≥140 mmHg and / or a diastolic blood pressure ≥90 mmHg at the time of measurement, or a clear record of hypertension diagnosis and current treatment with antihypertensive drugs. mmHg means millimeters of mercury.

[0122] Liver biopsy samples were systematically evaluated by experienced hepatologists, with the training and test sets analyzed separately by the same experts. Liver fibrosis was graded using a 5-point scale (0 to 4). Histological changes were assessed separately according to the NASH CRN guidelines and the NAFLD Activity Scoring System (NAS), including steatosis (0 to 3 points), lobular inflammation (0 to 3 points), and ballooning degeneration (0 to 2 points). The NAS score is the sum of the scores for steatosis, lobular inflammation, and ballooning degeneration, ranging from 0 to 8.

[0123] The aforementioned NASH CRN guidelines refer to the Nonalcoholic Steatohepatitis Clinical Research Network established by the National Institutes of Health (NIH). The NAFLD Activity Scoring System (NAS) is a semi-quantitative scoring system used in these guidelines to quantify disease activity. It integrates scores for key lesions and is commonly used to assess histological changes in the liver.

[0124] The baseline information for the training and test sets is as follows: The demographic characteristics of each cohort are balanced, with an average age of 29-34 years, and a higher proportion of women (61.23%-79.38%). Biochemical and metabolic indicators show high consistency across the entire study population.

[0125] Histologically confirmed MASH (defined as the presence of hepatic steatosis combined with lobular inflammation and ballooning degeneration) accounted for 45.94% in the training set and 37.11%-48.62% in the four validation cohorts. The prevalence of fibrotic MASH (MASH, NAS score ≥4 and combined with ≥2 stage liver fibrosis) was 13.06% in the training set and 10.69%-18.72% in the four validation sets. The overall MASH prevalence across the five cohorts was 45.18%, and the fibrotic MASH prevalence was 13.82%.

[0126] (II) Development of the MASH prediction model for fibrosis.

[0127] In the training set, potential independent risk factors for fibrotic MASH were systematically evaluated. Initial univariate logistic regression analysis of 21 biochemical, metabolic, and thyroid function parameters showed that 11 of them were significantly associated with fibrotic MASH (P<0.05), referred to as key predictive indicators, including: alanine aminotransferase (ALT), aspartate aminotransferase (AST), triglycerides (TG), total cholesterol (TC), high-density lipoprotein cholesterol (HDL-c), low-density lipoprotein cholesterol (LDL-c), platelet count (PLT), glycated hemoglobin (HbA1c), fasting plasma glucose (FPG), fasting C-peptide (FCP), and the Homeostasis Model for Insulin Resistance Assessment Index (HOMA-IR).

[0128] Through stepwise multiple regression analysis using bootstrapping, the key predictive indicators were refined into four independent predictors: ALT, AST, TG, and HDL-c. These parameters form the basis of the FMO (Fibrosis in Obese Patients MASH) score, and its formula is derived from their regression coefficients.

[0129] ;

[0130] The resulting FMO score provides a continuous probability estimate (0-1 points) for assessing the likelihood of an obese individual having fibrotic MASH, where a value of 0.30 corresponds to a 30% predicted probability for that individual.

[0131] (iii) Comparative performance evaluation of prediction models.

[0132] The FMO score demonstrated robust predictive performance for fibrotic MASH in all study populations. Figure 2 The figure shows the ROC curves of FMO on the training and validation sets. Figure 2 The presentation specifically shows the receiver operating characteristic (ROC) curves of the FMO model (orange curve) predicting fibrotic MASH on the training set (training cohort) and four independent biopsy validation sets (validation cohort 1, validation cohort 2, validation cohort 3, and validation cohort 4). The horizontal axis represents specificity, and the vertical axis represents sensitivity. The curves for each dataset, along with their areas under the curves and 95% confidence intervals, are also indicated.

[0133] The area under the ROC curve (AUC) of the FMO model in the training cohort was 0.874 (95% CI 0.844–0.905).

[0134] The area under the ROC curve (AUC) of the FMO model in the total validation cohort was 0.815 (95% CI 0.775–0.855).

[0135] The area under the ROC curve (AUC) of the FMO model in the validation cohort was 0.827 (95% CI 0.759–0.894).

[0136] The area under the ROC curve (AUC) of the FMO model in the two validation cohorts was 0.829 (95% CI 0.745–0.913).

[0137] The area under the ROC curve (AUC) of the FMO model in the three-cohort validation was 0.874 (95% CI 0.786–0.961).

[0138] The area under the ROC curve (AUC) of the FMO model in the four validation cohorts was 0.803 (95% CI 0.730–0.876).

[0139] Where CI represents the confidence interval.

[0140] The results show that the FMO model exhibits good and stable discrimination ability on both the training set and all validation sets.

[0141] Specifically, the area under the curve (AUC) of the FMO model was 0.874 (95% CI 0.844–0.905), demonstrating good calibrability and excellent cost-effectiveness. This superior performance was consistently reproduced in the five externally validated weight-loss surgery cohorts mentioned above, with AUC values ​​ranging from 0.803 to 0.874, confirming the model's general applicability to weight-loss candidates across diverse populations.

[0142] To enhance clinical applicability, optimal cutoff values ​​were determined. In the training set, an exclusion threshold (rule-out) of 0.05 achieved high sensitivity (0.902) and high negative predictive value (NPV, 0.976); while an inclusion threshold (rule-in) of 0.22 provided excellent specificity (0.900) and positive predictive value (PPV, 0.481). External validation confirmed the reliability of these thresholds: the sensitivity for exclusion diagnoses ranged from 0.737 to 1.000 (NPV 0.907–1.000), and the specificity for inclusion diagnoses ranged from 0.778 to 0.939 (PPV 0.333–0.630).

[0143] Comparative analysis with existing non-invasive indices (AAR, FIB-4, FNI, NFS, BARD, AP-index, FAST, acFibroMASH, and SAFE) shows that the FMO model has better diagnostic accuracy. Figure 3 For performance comparison of the FMO model and non-invasive scoring on the training set, such as Figure 3As shown, the receiver operating characteristic (ROC) curves of FMO (orange curve) and various non-invasive scores predicting fibrotic MASH in the training set are displayed. The horizontal axis represents specificity, the vertical axis represents sensitivity, and the curves of each model, their areas under the curve, and the 95% confidence intervals are also indicated.

[0144] in:

[0145] The area under the ROC curve (AUC) of the FMO model on the training set was 0.874 (95% CI 0.844–0.905).

[0146] The area under the ROC curve (AUC) of AAR on the training set was 0.626 (95% CI 0.573–0.679).

[0147] The area under the ROC curve (AUC) of FNI on the training set was 0.824 (95% CI 0.788–0.861).

[0148] The area under the ROC curve (AUC) of FIB-4 on the training set was 0.696 (95% CI 0.649–0.743).

[0149] The area under the ROC curve (AUC) of NFS on the training set was 0.559 (95% CI 0.509–0.610).

[0150] BARD's area under the ROC curve (AUC) on the training set was 0.569 (95% CI 0.517–0.621).

[0151] The area under the ROC curve (AUC) of the AP-index on the training set was 0.516 (95% CI 0.465–0.567).

[0152] The results showed that in the training cohort, the FMO score was significantly better than all control indicators (all P<0.05), indicating that the FMO model exhibited good and stable discrimination ability compared with other non-invasive scores in the training set.

[0153] Figure 4 For performance comparison of the FMO model and non-invasive scoring on the validation set, such as Figure 4 As shown, the receiver operating characteristic (ROC) curves of FMO (orange curve) and various non-invasive scores predicting fibrotic MASH in the validation set are displayed. The horizontal axis represents specificity, the vertical axis represents sensitivity, and the curves, areas under the curves, and 95% confidence intervals for each model are also indicated.

[0154] in:

[0155] The area under the ROC curve (AUC) of the FMO model on the validation set was 0.815 (95% CI 0.775–0.855).

[0156] The area under the ROC curve (AUC) of AAR on the validation set was 0.637 (95% CI 0.582–0.691).

[0157] The area under the ROC curve (AUC) of FNI on the validation set was 0.806 (95% CI 0.765–0.848).

[0158] The area under the ROC curve (AUC) of FIB-4 on the validation set was 0.608 (95% CI 0.551–0.664).

[0159] The area under the ROC curve (AUC) of NFS on the validation set was 0.571 (95% CI 0.517–0.625).

[0160] BARD's area under the ROC curve on the validation set (AUC) was 0.498 (95% CI 0.449–0.547).

[0161] The area under the ROC curve (AUC) of the AP-index on the validation set was 0.470 (95% CI 0.417–0.523).

[0162] The results showed that in the validation cohort, the FMO score not only outperformed other indicators but also performed comparably to the FNI, indicating that the FMO model demonstrated good and stable discriminative ability compared to other non-invasive scores in the validation set. This establishes its clinical value in identifying MASH patients at risk of fibrosis. Furthermore, the model maintained consistent performance across multiple independent cohorts, confirming its robustness and clinical applicability.

[0163] Figure 5 To compare the performance of the FMO model and non-invasive scoring in an independent bariatric surgery cohort, specifically, the evaluation was conducted in an independent bariatric surgery cohort (285 patients) that had preoperative transient liver elasticity data and was distinct from the training cohort.

[0164] like Figure 5 As shown, the receiver operating characteristic (ROC) curves of FMO (orange curve) and FNI, SAFE, FAST, and acFibroMASH predicting fibrotic MASH in an independent bariatric surgery cohort are presented. The horizontal axis represents specificity, the vertical axis represents sensitivity, and the curves, areas under the curve, and 95% confidence intervals for each model are also indicated.

[0165] in:

[0166] The area under the ROC curve (AUC) of the FMO model in an independent bariatric surgery cohort was 0.873 (95% CI 0.825–0.921).

[0167] The area under the ROC curve (AUC) for FNI in an independent bariatric surgery cohort was 0.816 (95% CI 0.752–0.880).

[0168] SAFE had an area under the ROC curve (AUC) of 0.700 (95% CI 0.623–0.777) in an independent weight loss surgery cohort.

[0169] The area under the ROC curve (AUC) of FAST in an independent bariatric surgery cohort was 0.815 (95% CI 0.758–0.872).

[0170] The area under the ROC curve (AUC) of acFibroMASH in an independent bariatric surgery cohort was 0.595 (95% CI 0.496–0.694).

[0171] Among them, *** represents P <0.001, * indicates P <0.05.

[0172] In this external validation set, the FMO score demonstrated superior discriminative ability for fibrotic MASH compared to four existing models (FAST, acFibroMASH, FNI, and SAFE), indicating that the FMO model exhibited good and stable discrimination ability in independent bariatric surgery cohorts. ROC analysis confirmed that the AUROC obtained by the FMO score was significantly higher than that of all control groups (P<0.05). These findings fully demonstrate the superior discriminative performance and broader applicability of the FMO score, establishing its clinical applicability in the identification of fibrotic MASH. The model's consistent performance across multiple independent cohorts underscores its robustness and clinical applicability.

[0173] To further validate the generalizability of this predictive model, the efficacy of the FMO score and existing indicators was evaluated in two large international cohorts of overweight and obese individuals (BMI ≥ 25 kg / m²).

[0174] Figure 6 For performance comparison of the FMO model and non-invasive scoring in the NHANES cohort, such as Figure 6 As shown, the receiver operating characteristic (ROC) curves for FMO (orange curve) and various non-invasive scores in the US NHANES cohort predicting fibrotic MASH are presented. The horizontal axis represents specificity, the vertical axis represents sensitivity, and the curves for each model, along with their areas under the curve and 95% confidence intervals, are also indicated.

[0175] in:

[0176] The area under the ROC curve (AUC) of the FMO model in the NHANES cohort was 0.866 (95% CI 0.830–0.901).

[0177] The area under the ROC curve (AUC) of AAR in the NHANES cohort was 0.662 (95% CI 0.606–0.717).

[0178] The area under the ROC curve (AUC) of FNI in the NHANES cohort was 0.903 (95% CI 0.876–0.929).

[0179] The area under the ROC curve (AUC) of FIB-4 in the NHANES cohort was 0.652 (95% CI 0.601–0.702).

[0180] The area under the ROC curve (AUC) for NFS in the NHANES cohort was 0.673 (95% CI 0.626–0.721).

[0181] The area under the ROC curve (AUC) of BARD in the NHANES cohort was 0.497 (95% CI 0.437–0.557).

[0182] The area under the ROC curve (AUC) of the AP-index in the NHANES cohort was 0.561 (95% CI 0.509–0.612).

[0183] In the US NHANES cohort (1573 participants), the FMO score demonstrated excellent discriminative ability in predicting fibrotic MASH, with an AUROC of 0.866 (95% CI 0.830–0.901). These results indicate that the FMO model, except for the FNI, exhibits good and stable discriminative ability compared to other non-invasive scores in the NHANES cohort.

[0184] Figure 7 A performance comparison of FMO models and non-invasive scoring in the UK biobank cohort, such as Figure 7 As shown, the receiver operating characteristic (ROC) curves of FMO (orange curve) and the noninvasive scoring index for predicting fibrotic MASH in the UK Biobank cohort are presented. The horizontal axis represents specificity, the vertical axis represents sensitivity, and the curves for each model, along with their areas under the curve and 95% confidence intervals, are also indicated.

[0185] in:

[0186] The area under the ROC curve (AUC) of the FMO model in the UK biobank cohort was 0.753 (95% CI 0.718–0.787).

[0187] The area under the ROC curve (AUC) of AAR in the UK biobank cohort was 0.719 (95% CI 0.681–0.757).

[0188] The area under the ROC curve (AUC) of FNI in the UK biobank cohort was 0.733 (95% CI 0.697–0.769).

[0189] The area under the ROC curve (AUC) of FIB-4 in the UK biobank cohort was 0.618 (95% CI 0.578–0.658).

[0190] The area under the ROC curve (AUC) of NFS in the UK biobank cohort was 0.499 (95% CI 0.455–0.542).

[0191] BARD's area under the ROC curve (AUC) in the UK biobank cohort was 0.437 (95% CI 0.387–0.487).

[0192] The area under the ROC curve (AUC) of the AP-index in the UK biobank cohort was 0.588 (95% CI 0.548–0.628).

[0193] In the UK Biobank cohort (15,745 participants), the FMO score maintained robust performance with an AUROC of 0.753 (95% CI 0.718–0.787). These results indicate that the FMO model demonstrated good and stable discriminative ability compared to other non-invasive scoring methods in the UK Biobank cohort.

[0194] like Figure 6 and Figure 7 As shown, in the NHANES cohort, the FMO score was superior to all other indicators except for FNI; in the UK Biobank cohort, the result was significantly better than all six control indicators.

[0195] This embodiment can also achieve longitudinal prediction of severe liver disease:

[0196] During a median follow-up period of 13.7 years (interquartile range, 12.9–14.4 years), a total of 1955 new cases of serious liver-related outcomes (SLROs) were recorded in overweight and obese individuals. The FMO score showed clinically meaningful predictive value for SLRO events (C-index 0.703).

[0197] The above Figures 2-7 The diagonal gray line in the curve provides a reference for a "random classifier" (area under the ROC curve = 0.5), representing that the model has no discriminative ability. The further the curve is from this diagonal gray line and bulges upwards to the left, the better the model's performance.

[0198] Figure 8 This compares the C-index value of FMO with existing non-invasive indices. For example... Figure 8 As shown, the concordance index (C-index) between FMO (blue column) and noninvasive scoring in predicting serious liver-related outcomes (SLROs) is illustrated. The height of each column corresponds to the C-index value, and each column is labeled with a specific numerical value for accurate comparison.

[0199] in:

[0200] The C-index value of FMO is 0.703;

[0201] The C-index value of AAR is 0.496;

[0202] BARD's C-index value is 0.577;

[0203] The C-index value of AP-index is 0.665;

[0204] The C-index value of NFS is 0.705;

[0205] The C-index value of FIB-4 is 0.719;

[0206] The C-index value of FNI is 0.752.

[0207] ***express P <0.001, * indicates P <0.05.

[0208] Compared with existing indicators, FMO showed higher consistency than AAR (0.496; P<0.001), BARD (0.577; P<0.001), and AP index (0.665; P<0.001), and was comparable to NFS (0.705; P=0.839), but slightly lower than FIB-4 (0.719; P=0.040). Among all control indicators, FNI achieved the highest consistency index (0.752) and was significantly better than FMO (P<0.001).

[0209] The fibrosis-related steatohepatitis risk assessment system provided in this invention can assess the risk of fibrosis-related steatohepatitis in obese individuals. A predictive model was established and validated internally and externally by analyzing liver biopsy data from multiple bariatric surgery cohorts. The model integrates four indicators: aspartate aminotransferase (AST), alanine aminotransferase (ALT), triglycerides, and high-density lipoprotein cholesterol (HDL-C). It demonstrated stable discriminative performance (AUROC 0.753-0.874) in both training and validation cohorts and effectively predicted severe liver-related outcomes. Compared with existing indicators, this method offers superior diagnostic performance, providing an accurate and convenient solution for liver disease risk assessment in obese individuals.

[0210] Those skilled in the art will understand that the various modules of the present invention described above can be implemented using general-purpose computer devices. Optionally, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by a computer device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. The present invention is not limited to any particular combination of hardware and software.

[0211] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A risk assessment system for fibrosis-related steatohepatitis, characterized in that, include: The data acquisition module is configured to: acquire weight information, clinical examination information, and pathological evaluation results of liver tissue samples from patients with fibrosis-related steatohepatitis, identify candidate predictive indicators, and construct a dataset; identify individuals who underwent bariatric surgery for morbid obesity within a specified time period; acquire physical examination, clinical, and laboratory examination results of individuals undergoing bariatric surgery 1-2 days prior to the surgery; and extract weight information and clinical examination information from patients with fibrosis-related steatohepatitis. The pathological evaluation results of liver tissue samples were obtained through systematic evaluation by liver pathologists; The specific candidate predictive indicators include: age, sex, body mass index, waist-to-hip ratio, alanine aminotransferase, aspartate aminotransferase, triglycerides, total cholesterol, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, albumin, platelet count, uric acid, glycated hemoglobin, fasting blood glucose, fasting C-peptide, fasting insulin, insulin resistance index, free triiodothyronine, free thyroxine, and thyroid-stimulating hormone; The primary screening module is configured to perform univariate logistic regression analysis on candidate predictive indicators to identify key predictive indicators for fibrosis-related steatohepatitis; the key predictive indicators specifically include: Alanine aminotransferase (ALT), aspartate aminotransferase (AST), triglycerides (TG), total cholesterol (TC), high-density lipoprotein cholesterol (HDL-c), low-density lipoprotein cholesterol (LDL-c), platelet count (PLT), glycated hemoglobin (HbA1c), fasting plasma glucose (FPG), fasting C-peptide (FCP), and insulin resistance homeostasis model assessment index (HOMA-IR). The secondary screening module is configured to: screen out four independent predictors from key predictive indicators of fibrosis-related steatohepatitis through stepwise multivariate regression analysis; the four independent predictors specifically include: alanine aminotransferase (ALT), aspartate aminotransferase (AST), triglycerides (TG), and high-density lipoprotein cholesterol (HDL-c). The model training module is configured to: train the constructed multi-factor logistic regression model based on the four selected independent predictors, and use the trained multi-factor logistic regression model as a risk assessment model. The risk assessment model is as follows: ； in, , , , These represent the concentrations of ALT, AST, TG, and HDL-c, respectively. The unit is per liter. It represents millimoles per liter; The assessment and prediction module is configured to assess the risk of the person being assessed for fibrotic metabolic steatohepatitis based on a risk assessment model.

2. The fibrosis-related steatohepatitis risk assessment system as described in claim 1, characterized in that, The weight information, clinical examination information, and pathological evaluation results of liver tissue samples of patients with fibrosis-related steatohepatitis were measured using the following methods: The weight information is measured using the Body Mass Index (BMI). The clinical examination information is measured by hematological indicators from fasting blood samples collected from patients 1-2 days before bariatric surgery. The pathological evaluation results of the liver tissue samples were measured by the degree of liver fibrosis and histological changes; among which, histological changes were scored separately for steatosis, lobular inflammation and ballooning degeneration, and the sum of the scores for steatosis, lobular inflammation and ballooning degeneration was calculated.

3. The fibrosis-related steatohepatitis risk assessment system as described in claim 1, characterized in that, In the evaluation and prediction module: Based on the risk assessment model, the risk of the person being assessed to have fibrotic metabolic steatohepatitis is assessed, and a risk assessment value is obtained, which is a probability value in the range of 0 to 1.

4. The risk assessment system for fibrosis-related steatohepatitis as described in claim 3, characterized in that, Also includes: The threshold determination module is configured to use a dual-threshold strategy to define risk ranges, dividing personnel into low-risk, uncertain, and high-risk groups, specifically: Set exclusion and inclusion thresholds; Individuals whose risk assessment scores are below the exclusion threshold will be classified into the low-risk group; Individuals whose risk assessment scores are higher than the inclusion threshold will be classified into the high-risk group; Individuals whose risk assessment values ​​fall between the exclusion and inclusion thresholds are classified into the uncertain group.

5. The risk assessment system for fibrosis-related steatohepatitis as described in claim 1, characterized in that, Also includes: The model validation module is configured as follows: Internal validation was performed by calculating the average C-index using multiple Bootstrap resampling methods on the training set obtained by splitting the dataset, to assess the model’s internal consistency and potential overfitting. External validation assesses the model’s discriminative power, calibration, and net clinical benefit on a validation set derived from the dataset.

Images