A clinical prediction model for the probability of developing hypersplenism in a patient with hepatolenticular degeneration and a design method

Abstract

A clinical prediction model for predicting the probability of hepatosplenomegaly in a patient with hepatolenticular degeneration and a design method thereof belong to the field of medical information technology. The clinical prediction model comprises: an input module receiving patient parameter indexes composed of aspartate aminotransferase to platelet ratio index (APRI), liver fibrosis 4 index (FIB-4), liver stiffness value (LSM) and portal vein diameter (PVD); a processing module connected to the input module, which calculates the prediction probability of the patient according to the model; and an output module connected to the processing module, which outputs the risk probability of hepatosplenomegaly in the patient with hepatolenticular degeneration obtained by the clinical prediction model. The clinical prediction model established by the application has good discrimination, calibration and clinical practicability, and has high accuracy, and can be conveniently used for early identification and risk prediction of hepatosplenomegaly in patients with hepatolenticular degeneration.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of medical information technology, and particularly relates to a clinical prediction model for the probability of splenomegaly in a hepatolenticular degeneration patient and a design method. BACKGROUND

[0002] Hepatolenticular degeneration, also known as Wilson disease (WD), is an autosomal recessive genetic copper metabolism disorder caused by mutations in the ATP7B gene. The mutations in the gene lead to pathological deposition of copper in the liver and other tissues. The copper transport ATPase has the highest expression level in the liver, so liver damage often occurs first. Among patients with hepatolenticular degeneration liver damage, about 34% eventually develop into cirrhosis with splenomegaly. Compared with hepatitis B patients, the probability of splenomegaly in hepatolenticular degeneration patients is 4.4 times higher. Splenomegaly can lead to a decrease in blood cells, thereby affecting the patient's continued copper excretion treatment.

[0003] Therefore, it is helpful to delay the occurrence of splenomegaly to be able to assess the risk of splenomegaly in the early stage of hepatolenticular degeneration and take intervention measures for high-risk patients. At present, the increase in the size of the spleen and the decrease in the blood cell count are generally used as the basis for determining splenomegaly in clinical practice. The assessment of the size of the spleen mainly relies on ultrasonic examination. However, the degree of cooperation of the patient and the skill level of the operator can cause certain errors in the measurement results. Laboratory detection indicators have high sensitivity, but once these indicators significantly decline, it often means that splenomegaly has entered a more serious stage.

[0004] There are studies in clinical practice that use APRI, FIB-4, and 2D-SWE in various liver disease diagnosis models. APRI and FIB-4 have been used to assess liver fibrosis and cirrhosis in hepatitis patients. In addition, studies have shown that 2D-SWE has high diagnostic accuracy for liver fibrosis staging with magnetic resonance elastography as the reference standard. Ultrasonic elastography combined with APRI and FIB-4 has also been used to assess the severity of hepatolenticular degeneration liver manifestations. However, there are relatively few studies on the correlation between the use of these technologies for the assessment of splenomegaly, a complication of hepatolenticular degeneration cirrhosis. This study intends to use APRI, FIB-4, and 2D-SWE to predict the occurrence of splenomegaly in hepatolenticular degeneration patients, in order to provide a basis for early assessment of the occurrence of splenomegaly in hepatolenticular degeneration. SUMMARY

[0005] The purpose of the present application is to provide a clinical prediction model for the probability of splenomegaly in a hepatolenticular degeneration patient and a design method. The established clinical prediction model has high accuracy and can be conveniently used for early identification and risk prediction of splenomegaly in hepatolenticular degeneration patients.

[0006] To achieve the above object, the technical scheme adopted by the present application is:

[0007] Firstly, the present application proposes a clinical prediction model for the probability of splenomegaly in patients with hepatolenticular degeneration, comprising:

[0008] An input module receives patient parameter indicators composed of aspartate aminotransferase to platelet ratio index (APRI), liver fibrosis 4 index (FIB-4), liver stiffness value (LSM), and portal vein diameter (PVD);

[0009] A processing module is connected to the input module and calculates the score and prediction probability of splenomegaly based on the APRI, FIB-4, LSM, and PVD indicator scores according to the following formula:

[0010] Logit(P) = -11.834 + 0.342* APRI + 0.259* FIB-4 + 1.225* LSM + 2.494*PVD

[0011]

[0012] In the formula, Logit(P) is the obtained score, and P is the prediction probability;

[0013] An output module is connected to the processing module and outputs the risk probability of splenomegaly in patients with hepatolenticular degeneration obtained by the clinical prediction model.

[0014] Secondly, the present application also proposes a design method for a clinical prediction model for the probability of splenomegaly in patients with hepatolenticular degeneration, comprising the following steps:

[0015] Step 1, at least the following patient parameter indicators are recorded:

[0016] Basic information: age, gender, body mass index (BMI), and duration of disease in years;

[0017] Serological indicators: red blood cell count (RBC), white blood cell count (WBC), platelet count (PLT), alanine aminotransferase (ALT), and aspartate aminotransferase (AST);

[0018] Calculate the aspartate aminotransferase to platelet ratio index (APRI):

[0019] APRI=[(AST(U / L) / ULN)×100] / PLT(×10 9 / L) (1)

[0020] In formula (1), ULN is the upper limit of normal reference value of AST;

[0021] Calculate the liver fibrosis 4 index (FIB-4):

[0022] FIB-4 = [age (years) x AST (U / L)] / [PLT (x 10 9 / L) x ALT (U / L) 1 / 2 ]; (2)

[0023] 2D-SWE detection indicators: liver stiffness value (LSM), portal vein diameter (PVD), portal vein blood flow velocity, length and thickness of the spleen;

[0024] Step 2: Using inter-group comparison to analyze the variables of age, gender, body mass index (BMI), course of disease, red blood cell count (RBC), white blood cell count (WBC), platelet count (PLT), alanine aminotransferase (ALT), aspartate aminotransferase (AST), aspartate aminotransferase to platelet ratio index (APRI), liver fibrosis 4 index (FIB-4), liver stiffness value (LSM), portal vein diameter (PVD), portal vein blood flow velocity, length and thickness of the spleen, so as to obtain the P<0.05 prediction factors of hepatolenticular degeneration with hypersplenism: aspartate aminotransferase to platelet ratio index (APRI), liver fibrosis 4 index (FIB-4), liver stiffness value (LSM), portal vein diameter (PVD);

[0025] Then, using the multivariate Logistic regression method to analyze the above prediction factors, the results show that aspartate aminotransferase to platelet ratio index (APRI), liver fibrosis 4 index (FIB-4), liver stiffness value (LSM), portal vein diameter (PVD) are independent risk factors for hepatolenticular degeneration patients with hypersplenism;

[0026] Step 3: Construct a calculation formula for the probability of hepatolenticular degeneration patients with hypersplenism composed of the independent risk factors:

[0027] First, according to formula (3), the score value Logit(P) is calculated by the independent risk factors APRI, FIB-4, LSM, and PVD;

[0028] Logit(P) = -11.834 + 0.342* APRI + 0.259* FIB-4 + 1.225* LSM + 2.494*PVD (3)

[0029] Then, the prediction probability P is calculated according to formula (4);

[0030] (4)

[0031] wherein, ln is natural logarithm.

[0032] Step 4, a clinical prediction model composed of an input module, a processing module and an output module is constructed, the input module receives patient parameter indexes composed of aspartate aminotransferase to platelet ratio index (APRI), liver fibrosis 4 index (FIB-4), liver stiffness value (LSM) and portal vein diameter (PVD); the processing module is connected with the input module, and scores and prediction probabilities of hypersplenism are calculated according to the formula obtained in step 3 based on the APRI, FIB-4, LSM and PVD index scores, and the output module is connected with the processing module;

[0033] Step 5, after obtaining the clinical prediction model, the area under the curve (AUC) of the receiver operating characteristic curve (ROC), and the calibration curve, the decision curve and the clinical impact curve are used to verify the prediction model in the modeling population and the verification population respectively to judge the discrimination, the calibration degree and the clinical practicability.

[0034] Compared with the prior art, the beneficial effects of the present application mainly manifest in:

[0035] 1. The present application establishes a clinical prediction model of the probability of hypersplenism in patients with hepatolenticular degeneration. First, the clinical medical record data of 234 patients with hepatolenticular degeneration treated in the Department of Neurology of the First Affiliated Hospital of Anhui University of Chinese Medicine from October 2023 to October 2024 is retrospectively collected. Based on whether the patient has hypersplenism, it is divided into a hypersplenism group and a non-hypersplenism group. Through multivariate Logistics regression, independent risk factors for the occurrence of hypersplenism in patients with hepatolenticular degeneration are screened out, and a clinical prediction model is established. The results show that aspartate aminotransferase to platelet ratio index (APRI), liver fibrosis 4 index (FIB-4), liver stiffness value (LSM) and portal vein diameter (PVD) are independent risk factors for the occurrence of hypersplenism in patients with hepatolenticular degeneration (P<0.05).

[0036] 2. The present application uses the area under the curve (AUC) of the receiver operating characteristic curve (ROC), and the calibration curve, the decision curve and the clinical impact curve to verify the prediction model in the training set and the verification set respectively to judge the discrimination, the calibration degree and the clinical practicability. The results show that the clinical prediction model constructed by the present application has good discrimination, calibration degree and clinical practicability, and has high accuracy, and can be conveniently used for early identification and risk prediction of hepatolenticular degeneration with hypersplenism. BRIEF DESCRIPTION OF DRAWINGS

[0037] Figure 1is a ROC curve of splenomegaly in patients with hepatolenticular degeneration (training set).

[0038] Figure 2 is a ROC curve of splenomegaly in patients with hepatolenticular degeneration (validation set).

[0039] Figure 3 is a ROC curve of splenomegaly in patients with hepatolenticular degeneration (validation set).

[0040] Figure 4 is a calibration curve of splenomegaly in patients with hepatolenticular degeneration (training set).

[0041] Figure 5 is a calibration curve of splenomegaly in patients with hepatolenticular degeneration (validation set).

[0042] Figure 6 is a decision curve of splenomegaly in patients with hepatolenticular degeneration (training set).

[0043] Figure 7 is a decision curve of splenomegaly in patients with hepatolenticular degeneration (validation set).

[0044] Figure 8 is a clinical impact curve of splenomegaly in patients with hepatolenticular degeneration (training set).

[0045] Figure 9 is a clinical impact curve of splenomegaly in patients with hepatolenticular degeneration (validation set). DETAILED DESCRIPTION

[0046] The clinical prediction model for the probability of splenomegaly in patients with hepatolenticular degeneration and the design method are further described below in conjunction with the accompanying drawings and examples.

[0047] Example 1

[0048] 1. Data and methods

[0049] 1.1 Study subjects

[0050] The clinical medical record data of 234 patients with hepatolenticular degeneration hospitalized in the Department of Neurology of the First Affiliated Hospital of Anhui University of Chinese Medicine from October 2023 to October 2024 were retrospectively collected. Based on whether the patients developed splenomegaly, they were divided into the splenomegaly group and the non-splenomegaly group. The research scheme was approved by the Medical Ethics Committee of the First Affiliated Hospital of Anhui University of Chinese Medicine, with the approval number being 2024AH-140, which can ensure that all operation processes comply with relevant ethical regulations.

[0051] 1.2 Inclusion and exclusion criteria

[0052] Inclusion criteria: (1) meet the diagnostic criteria of hepatolenticular degeneration; (2) age between 18 to 55 years old; (3) complete clinical data.

[0053] Exclusion criteria: (1) other liver diseases that can cause splenomegaly or hypersplenism (such as alcoholic, fatty, viral, drug, parasitic or autoimmune liver disease); (2) pregnant and lactating women; (3) combined with malignant tumor.

[0054] 234 patients with hepatolenticular degeneration were randomly divided into training set (163) and validation set (71) according to the ratio of 7:3.

[0055] 1.3 Collection of clinical data and test methods

[0056] Basic information: age, gender, body mass index (BMI) and the length of disease course;

[0057] Serological indicators: red blood cell count (RBC), white blood cell count (WBC), platelet count (PLT), alanine aminotransferase (ALT) and aspartate aminotransferase (AST);

[0058] Calculate the aspartate aminotransferase to platelet ratio index (APRI):

[0059] APRI = [(AST (U / L) / ULN) x 100] / PLT (x 10 9 / L) (1)

[0060] In formula (1), ULN is the upper limit of normal reference value of AST;

[0061] Calculate the liver fibrosis 4 index (FIB-4):

[0062] FIB-4 = [age (years) x AST (U / L)] / [PLT (x 10 9 / L) x ALT (U / L) 1 / 2 ] (2)

[0063] 2D-SWE detection indicators: liver stiffness value (LSM), portal vein diameter (PVD), portal vein blood flow velocity, length and thickness of spleen;

[0064] 1.4 Statistical analysis

[0065] Count data was expressed as frequency (%), and χ 2 test was used for comparison between groups. For measurement data conforming to normal distribution, x ± s was used, and t test was used for comparison within two groups; non-normal distribution measurement data was expressed as M (P25, P75), and Mann-Whitney U test was used for comparison between groups.

[0066] To develop the model, 234 participants were randomly divided into a training set (163) and a validation set (71) in a ratio of 7:3. The training set was used to develop the model, and the validation set was used for internal validation. Multivariate Logistic regression method was used to screen independent risk factors; finally, according to the risk factors with statistical significance in multivariate Logistic regression, nomograms were drawn by the "rms" software package. The variance inflation factor (VIF) was calculated to check whether the final model had multicollinearity. The area under the curve (AUC) of the receiver operating characteristic curve (ROC) was used to evaluate the predictive effectiveness of the model. The calibration curve was used to determine the calibration degree of the model, and the Hosmer-Lemeshow test was used to further determine the accuracy of the model. The decision curve and clinical impact curve were used to evaluate the clinical utility of the model.

[0067] 2 Results

[0068] 2.1 Clinical data of the study subjects

[0069] A total of 234 patients with hepatolenticular degeneration were included, including 61 cases of hypersplenism and 173 cases of non-hypersplenism. There was no significant difference in gender, age, BMI, disease duration, and portal vein blood flow velocity between the two groups; compared with the non-hypersplenism group, the APRI, FIB-4, LSM, and PVD of the hypersplenism group were increased, and the difference was statistically significant (P < 0.05) (see Table 1 for details).

[0070] Table 1 Comparison of clinical data between the two groups

[0071] Variable Splenomegaly group (n=61) Non-splenomegaly group (n=173) [t / z / x 2 ]] P Gender 0.178 0.673 Male 34.00（55.70%） 91.00（52.60%） Female 27.00（44.30%） 82.00（47.40%） Age (years) 27.00（22.00,33.00） 28.00（22.00,33.00） 0.397 0.692 BMI (kg / m 2 )]> 19.35（17.30,21.59） 19.31（17.35,22.89） -0.587 0.557 Course (years) 8.00（4.00,13.50） 9.00（5.00,14.00） -0.441 0.659 Portal vein blood flow velocity (cm / s) 20.47 ± 4.28 21.60 ± 4.26 1.777 0.077 PVD (mm) 12.00（11.00,14.00） 12.00（10.75,13.00） 2.818 0.005 LSM (Kpa) 11.89（10.26,14.54） 8.97（7.26,10.93） 6.812 <0.001 FIB-4 (points) 2.46（1.44,3.88） 0.87（0.60,1.37） -9.232 <0.001 APRI (points) 1.21（0.76,2.04） 0.43（0.31,0.57） 9.784 <0.001

[0072] There was no statistically significant difference in clinical data between the training set and the validation set (all P > 0.05, see Table 2 for details).

[0073] Table 2 Comparison of clinical data between the training set and the validation set

[0074] Variable Training set (n=163) Validation set (n=71) [t / z / x 2 ]] P Group 0.333 0.939 Splenomegaly group 39（23.93%） 22（30.99%） Non-splenomegaly group 124（76.07%） 49（69.01%） Gender 0.027 0.870 Male 86.00（52.76%） 39.00（54.93%） Female 77.00（47.24%） 32.00（45.07%） Age (years) 28.00（22.00,33.00） 27.00（22.00,31.50） 5647 0.770 BMI (kg / m 2 )]> 19.35（17.30,21.59） 19.31（17.35,22.89） -0.587 0.557 Course (years) 9.00（5.00,14.00） 8.00（4.00,13.00） 5362 0.372 Portal vein blood flow velocity (cm / s) 21.41 ± 4.41 21.00 ± 4.03 -0.684 0.495 PVD (mm) 12.00（11.00,13.00） 12.00（11.00,13.00） 6664 0.062 LSM (Kpa) 10.05（7.58,12.07） 9.73（7.70,12.31） 5767 0.968 FIB-4 (points) 1.07（0.67,1.57） 1.21（0.66,1.99） 6191 0.396 APRI (points) 0.54（0.33,1.15） 0.51（0.35,0.76） 6173 0.417

[0075] 2.2 Multivariate Logistic regression analysis of risk factors for hepatolenticular degeneration with hypersplenism

[0076] Multivariate Logistic regression analysis of the above risk factors showed that APRI, FIB-4, LSM, and PVD were independent risk factors for hepatolenticular degeneration with hypersplenism (see Table 3 for details). No significant collinearity was found between variables.

[0077] Table 3 Logistic regression analysis of hepatolenticular degeneration patients with hypersplenism

[0078] Variable B value Wald OR 95% CI P value VIF LSM 0.259 11.838 1.295 1.118～1.501 0.001 1.239 PVD 0.277 4.236 1.319 1.013～1.717 0.040 1.065 FIB-4 1.608 13.135 4.991 2.092～11.906 <0.001 3.015 APRI 1.721 7.042 5.593 1.568～19.942 0.008 2.984

[0079] 2.3 Establishment of nomogram prediction model for hepatosplenomegaly in patients with Wilson's disease

[0080] The four independent risk factors were incorporated, and a clinical prediction model for the probability of hepatosplenomegaly in patients with Wilson's disease was successfully established. By the scale above the model, the single score corresponding to the four independent influencing factors can be obtained, and the total score is obtained by adding each single score. The prediction probability corresponding to the total score is the probability of hepatosplenomegaly in patients with Wilson's disease (see Figure 1 for details). Specifically:

[0081] First, according to formula (3), the score value Logit(P) is calculated by the independent risk factors APRI, FIB-4, LSM, and PVD;

[0082] Logit(P) = -11.834 + 0.342* APRI + 0.259* FIB-4 + 1.225* LSM + 2.494*PVD (3)

[0083] Logit(P) is the score obtained by calculation, which integrates the influence of each risk factor:

[0084] Constant term: -11.834 (intercept, representing the base logit value when all prediction variables are 0)

[0085] Coefficient term: for example, 0.342* APRI, indicating that for every one-unit increase in APRI, Logit(P) increases by 0.342 (i.e., increases the log-odds of hepatosplenomegaly).

[0086] Then, the prediction probability P is calculated according to formula (4);

[0087] (4)

[0088] Where ln is the natural logarithm. Formula (4) is used to convert the log-odds to the probability P, for example, if P = 0.7 is calculated, it means that the patient with Wilson's disease has a 70% probability of developing hepatosplenomegaly.

[0089] 2.4 Validation of the nomogram prediction model

[0090] 2.4.1 Discrimination

[0091] After the model was built using the training set (n = 163), its predictive performance was evaluated on the validation set (n = 71). The AUC of the model in the training set was 0.927 (95% CI: 0.885-0.970) and in the validation set was 0.951 (95% CI: 0.905-0.996), showing a very strong discriminatory power (see Figure 2 , 3 ).

[0092] 2.4.2 Calibration

[0093] Bootstrap corrected calibration analysis showed that the predicted and observed results of the risk of hepatolenticular degeneration with hyperfunctioning spleen were very consistent, with a mean absolute error of 0.027 (training set) and 0.064 (validation set). The Hosmer-Lemeshow test showed no significant deviation (training set: p = 0.404; validation set: p = 0.136), indicating that the deviation was not significant. These results showed that the prediction model had a high calibration degree (see Figure 4 , 5 ).

[0094] 2.4.3 Clinical utility

[0095] The decision curve results showed that in the threshold range of 5-95% (training set) and 5-85% (validation set), the model (red curve) showed statistically superior clinical net benefit compared to the "all treatment" (gray curve) and "no treatment" (blue curve) strategies (see Figure 6 , 7 ).

[0096] The clinical impact curve results showed that as the threshold increased, both the predicted high-risk cases (blue curve) and the true event rate (red curve) showed a monotonous downward trend (see Figure 8 , 9 ).

[0097] By drawing the decision curve and clinical impact curve of the training set and the validation set, it was found that the prediction model had good clinical utility.

[0098] Example 2

[0099] Construction of the clinical prediction model:

[0100] The input module receives patient parameter indicators composed of aspartate aminotransferase to platelet ratio index (APRI), liver fibrosis 4 index (FIB-4), liver stiffness value (LSM), and portal vein diameter (PVD).

[0101] The processing module is connected with the input module, and based on the APRI, FIB-4, LSM and PVD index scores, the score and prediction probability of hypersplenism are calculated according to the formula obtained in step 3.

[0102] The output module is connected with the processing module, and outputs the risk probability of the hepatolenticular degeneration patient to develop hypersplenism obtained by the clinical prediction model.

Claims

1. A method for designing a clinical prediction model of the probability of the development of hypersplenism in a patient with hepatolenticular degeneration, characterized in that, The steps are as follows: Step 1, at least record the following patient parameter indicators: Basic information: age, gender, body mass index (BMI), and duration of disease; Serological indicators: red blood cell count (RBC), white blood cell count (WBC), platelet count (PLT), alanine aminotransferase (ALT), and aspartate aminotransferase (AST); Calculate the aspartate aminotransferase to platelet ratio index (APRI): APRI = [(AST (U / L) / ULN) x 100] / PLT (x 10 9 / L) (1) In formula (1), ULN is the upper limit of the normal reference value of AST; Calculate the liver fibrosis 4 index (FIB-4): FIB-4 = [age (years) x AST (U / L)] / [PLT (x 10 9 / L) x ALT (U / L) 1 / 2 ]; (2) 2D-SWE detection indicators: liver stiffness value (LSM), portal vein diameter (PVD), portal vein blood flow velocity, length and thickness of the spleen; Step 2, use inter-group comparison to analyze the variables of age, gender, body mass index (BMI), duration of disease, red blood cell count (RBC), white blood cell count (WBC), platelet count (PLT), alanine aminotransferase (ALT), aspartate aminotransferase (AST), aspartate aminotransferase to platelet ratio index (APRI), liver fibrosis 4 index (FIB-4), liver stiffness value (LSM), portal vein diameter (PVD), portal vein blood flow velocity, length and thickness of the spleen, and obtain the prediction factors of hepatolenticular degeneration with hypersplenism: aspartate aminotransferase to platelet ratio index (APRI), liver fibrosis 4 index (FIB-4), liver stiffness value (LSM), and portal vein diameter (PVD) with P<0.05; Further analysis of the above prediction factors by multivariate Logistic regression method shows that aspartate aminotransferase to platelet ratio index (APRI), liver fibrosis 4 index (FIB-4), liver stiffness value (LSM), and portal vein diameter (PVD) are independent risk factors for hepatolenticular degeneration patients developing hypersplenism; Step 3, construct a calculation formula for the probability of hepatolenticular degeneration patients developing hypersplenism composed of the independent risk factors: First, calculate the score value Logit(P) according to formula (3) through the independent risk factors APRI, FIB-4, LSM, and PVD; Logit(P) = -11.834 + 0.342* APRI + 0.259* FIB-4 + 1.225* LSM + 2.494* PVD (3) Then calculate the prediction probability P according to formula (4); （4） Wherein, ln is the natural logarithm; Step 4, a clinical prediction model composed of an input module, a processing module and an output module is constructed, the input module receives patient parameter indexes composed of aspartate aminotransferase to platelet ratio index (APRI), liver fibrosis 4 index (FIB-4), liver stiffness value (LSM) and portal vein diameter (PVD); the processing module is connected to the input module, based on the APRI, FIB-4, LSM and PVD index scores, the score and prediction probability of hypersplenism are calculated according to the formula obtained in step 3, and the output module is connected to the processing module; Step 5, after obtaining the clinical prediction model, the area under the curve (AUC) of the receiver operating characteristic curve (ROC), and the calibration curve, the decision curve and the clinical impact curve are used to verify the prediction model in the modeling population and the verification population respectively to judge the discrimination, the calibration degree and the clinical practicability.

Images