A model for predicting the risk of purpura nephritis and a method for constructing a nomogram thereof

Abstract

The application discloses a clinical prediction model for predicting the risk of purpura nephritis and a construction method of a nomogram thereof. Four independent risk factors are obtained through mathematical statistical methods, the risk independent factors include age, rash duration, D-dimer and IgG, four corresponding clinical prediction models are established according to the risk independent factors, and the optimal prediction model containing three prediction factors is obtained through mathematical analysis methods, the three prediction factors are the age of the child, D-dimer and IgG, and the occurrence risk of purpura nephritis can be predicted through the established nomogram. The application provides a convenient and effective tool for clinicians to evaluate the risk of HSPN in children with HSP, and the determination of the IgG and D-dimer levels in blood can be realized in community clinics.

Description

Technical Field

[0001] This invention relates to the field of medical technology for Henoch-Schönlein purpura nephritis, specifically to a model for predicting the risk of developing Henoch-Schönlein purpura nephritis and a method for constructing its Nomogram. Background Technology

[0002] Henoch-Schönlein purpura (HSP) is a systemic syndrome primarily characterized by small vessel vasculitis. Clinically, it presents as non-thrombocytopenic purpura, with or without abdominal pain, gastrointestinal bleeding, joint pain, and kidney damage. Approximately 20%–60% of children with HSP develop kidney involvement within 6 months of the disease course, leading to HSPN, characterized by microscopic hematuria and / or proteinuria; severe cases can result in acute renal failure. With timely diagnosis and aggressive treatment, most cases of HSPN have a good prognosis; however, a small percentage of children progress to chronic kidney disease, becoming a significant cause of end-stage renal disease. Although some studies on risk factors for HSPN indicate that age, persistent and recurrent rash, thrombocytopenia, decreased serum albumin, decreased high-density lipoprotein, and elevated fibrinogen are independent risk factors, no effective clinical predictive models, especially those with high predictive efficiency, have been reported. Therefore, finding a predictive method for the development of purpuric nephritis in children with Henoch-Schönlein purpura is of great significance for the timely diagnosis and treatment of purpuric nephritis and for strengthening the follow-up management of children with Henoch-Schönlein purpura. Summary of the Invention

[0003] To address the aforementioned issues, this invention provides a model for predicting the risk of developing Henoch-Schönlein purpura nephritis (HSPN) and a method for constructing its Nomogram. The risk of HSPN in children with Henoch-Schönlein purpura (HSP) is assessed using scores of three predictive factors: age, D-dimer level, and IgG. The model is simple to operate and provides intuitive results. Furthermore, the measurement of D-dimer and IgG, two of the three independent risk factors, can be performed in community clinics, making it simple and convenient.

[0004] The technical solution of the present invention is as follows:

[0005] This invention provides a model for predicting the risk of developing Henoch-Schönlein purpura nephritis and a method for constructing its Nomogram, the construction steps of which include:

[0006] S1: Several HSP and HSPN patients were divided into a training group and an internal validation group, while several HSP and HSPN patients from external hospitals were set up as an external validation group.

[0007] S2: Collect information on HSP and HSPN children from step 1. The information includes general population information (sex, age, weight), duration of rash (Dor), peripheral blood cell count and distribution, C-reactive protein (CRP), immunoglobulin classification and complement, lymphocyte subset analysis, coagulation function, and D-dimer.

[0008] S3: Based on the information obtained in step S2, establish a clinical prediction model, test the model, and obtain the optimal prediction model;

[0009] S4: Construct a Nomogram to predict the risk of developing Henoch-Schönlein purpura nephritis based on the optimal prediction model, including three predictive factors: the child's age, D-dimer, and IgG.

[0010] As described above, in a model for predicting the risk of Henoch-Schönlein purpura nephritis and a method for constructing its Nomogram, step S3 further includes:

[0011] S31: Statistical tests were used to compare the training group of HSP patients and the cohort of HSPN patients to identify indicators with significant differences.

[0012] S32: Apply univariate logistic regression to the indicators with significant differences in step S31 to further screen candidate predictive indicators, which include: age, weight, duration of rash, PLT, CRP, D-dimer, IgG, and C3.

[0013] S33: The predictive indicators in step S32 are subjected to multivariate logistic regression analysis to identify independent risk factors, which include age, duration of rash, D-dimer and IgG.

[0014] S34: Using the independent risk factors in step S33, construct four clinical prediction models to predict the risk of HSPN in children with HSP, namely, the clinical prediction model based on the child's age, IgG, D-dimer and rash duration (AIDD), the clinical prediction model based on the child's age, IgG and D-dimer (AIDi), the clinical prediction model based on the child's age, IgG and rash duration (AIDo), and the clinical prediction model based on the child's age, D-dimer and rash duration (ADD);

[0015] S35: The clinical prediction model obtained in step S34 is used to screen the best prediction model using receiver operating characteristic (ROC), Hosmer-Lemeshow goodness-of-fit test, clinical decision analysis curve (DCA), net reclassification index (NRI), and integrated discriminant index (IDI).

[0016] The final analysis determined that model AIDi was the best clinical prediction model, and R language was used as a statistical tool to obtain the Nomogram of the clinical prediction model AIDi.

[0017] The above describes a model for predicting the risk of Henoch-Schönlein purpura nephritis and the process of constructing its Nomogram, wherein the statistical tests include the T-test, the Mann-Whitney U test, and the chi-square test.

[0018] This technical solution also includes a model for predicting the risk of Henoch-Schönlein purpura nephritis and its Nomogram, established through the above construction steps.

[0019] The beneficial effects of this invention are as follows:

[0020] The AIDi clinical prediction model and its nomogram provide clinicians with a convenient and effective tool for assessing the risk of HSPN in children with HSP. Blood IgG and D-dimer levels in children with allergic purpura can be collected using routine venous blood sampling, allowing for rapid testing and accurate laboratory results. The determination of IgG and D-dimer levels is widely available even in community clinics. The nomogram based on the AIDi clinical prediction model is easy to apply clinically, helping to assess the early risk of HSPN in children with HSP. Its simple operation and intuitive results provide clinicians with an opportunity to optimize clinical treatment. Attached Figure Description

[0021] The solutions and advantages of this application will become clear to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of the invention.

[0022] In the attached diagram:

[0023] Figure 1 This is the Nomogram of the model of this invention;

[0024] Figure 2 This is a roadmap for constructing the model and its Nomogram in this invention;

[0025] Figure 3 ROC analysis for four independent risk factors and four clinical prediction models;

[0026] Figure 4 Probability calibration plots for four clinical prediction models in the training and internal validation groups;

[0027] Figure 5 Plotting the net benefit of four clinical prediction models in the training and internal validation groups;

[0028] Figure 6 The predictive effects of the clinical prediction model AIDi and the clinical prediction model AIDD are plotted.

[0029] Figure 7 The area under the receiver operating curve (AUROC) of the clinical prediction model AIDi in the external validation group, different histological grades, and different genders;

[0030] Figure 8 This invention describes the method of using the Nomogram based on the clinical prediction model AIDi. Detailed Implementation

[0031] Exemplary embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. It should be noted that these embodiments are provided to enable a more thorough understanding of this disclosure and to fully convey the scope of this disclosure to those skilled in the art. This disclosure can be implemented in various forms and should not be limited to the embodiments set forth herein.

[0032] The terms "front and back" and "left and right" mentioned in this invention are only used to express relative positional relationships and are not constrained by any specific directional reference in actual application.

[0033] See Figure 1 This embodiment presents a clinical prediction model for predicting the risk of developing Henoch-Schönlein purpura nephritis (HSPN). The model includes three independent risk factors: age of the child with Henoch-Schönlein purpura, D-dimer level, and IgG level. This clinical prediction model can predict the risk of developing HSPN in children with Henoch-Schönlein purpura.

[0034] See Figure 2 In this embodiment, a model for predicting the risk of Henoch-Schönlein purpura nephritis and its Nomogram construction method are as follows:

[0035] S1: 266 subjects from Shandong Provincial Hospital, including 174 children with Henoch-Schönlein purpura (HSP) and 92 children with Henoch-Schönlein purpura (HSPN), were divided into a training group (138 subjects, including 90 children with HSP and 48 children with HSPN) and an internal validation group (128 subjects, including 84 children with HSP and 44 children with HSPN). At the same time, 72 children with HSP and 50 children with HSPN from the Children's Hospital Affiliated to Soochow University were set up as an external validation group. By setting up two different validation control groups under different environments, the predictive model derived from mathematical statistics can be further validated to ensure the accuracy of the predictive model in predicting the occurrence of Henoch-Schönlein purpura nephritis in children with allergic purpura.

[0036] In this embodiment, the inclusion and exclusion criteria for the children are shown in the table below.

[0037]

[0038] S2: Collect information on HSP and HSPN patients from step 1. This information includes general demographic information (sex, age, weight), duration of rash (Dor), peripheral blood cell count and distribution, C-reactive protein (CRP), immunoglobulin classification and complement, lymphocyte subset analysis, coagulation function, and D-dimer. Missing data were supplemented using statistical methods. The collected patient information is shown in the table below.

[0039]

[0040]

[0041] S3: Based on the information obtained in step S2, establish a clinical prediction model, perform the Hosmer-Lemeshow goodness-of-fit test on the model, and obtain the optimal prediction model.

[0042] In this embodiment, step S3 further includes the following steps:

[0043] S31: Using three test methods—T-test, Mann-Whitney U test, and chi-square test—we compared the training group of HSP patients (n=90) and HSPN patients (n=48) to screen for indicators with significant differences. The specific indicators are shown in the table below.

[0044]

[0045] S32: Apply univariate logistic regression to the indicators with significant differences in step S31 to further screen candidate predictive indicators. The predictive indicators include: age, weight, duration of rash, platelet count (PLT), D-dimer, IgG, and C3.

[0046] S33: The predictive indicators in step S32 are subjected to multivariate logistic regression analysis to identify independent risk factors, including age, duration of rash, D-dimer, and IgG.

[0047] The results of univariate and multivariate logistic regression analyses in steps S32 and S33 are shown in the table below.

[0048]

[0049] S34: Using the independent risk factors in step S33, construct four clinical prediction models to predict the risk of HSPN in children with HSP, namely, the clinical prediction model based on the child's age, IgG, D-dimer and rash duration (AIDD), the clinical prediction model based on the child's age, IgG and D-dimer (AIDi), the clinical prediction model based on the child's age, IgG and rash duration (AIDo), and the clinical prediction model based on the child's age, D-dimer and rash duration (ADD);

[0050] The ROC curve is a curve plotted using a series of different binary classification methods (cutoff values ​​or decision thresholds), with the true positive rate (sensitivity) on the ordinate and the false positive rate (1 - specificity) on the x-axis. The closer the area under the receiver operating curve (AUROC) is to 1, the higher the diagnostic accuracy. See also Figure 3 In the training set, the AUROCs of the four clinical prediction models mentioned above showed good ability to distinguish between children with HSPN and HSP. As can be seen from the data in the figure, the AUROCs of the four prediction models showed good ability to distinguish between children with Henoch-Schönlein purpura nephritis. The AUROCs of the four clinical prediction models were AIDD: 0.931, 95% CI: 0.890-0.972; AIDi: 0.920, 95% CI: 0.876-0.965; AIDo: 0.856, 95% CI: 0.792-0.920; and ADD: 0.907, 95% CI: 0.860-0.954.

[0051] S35: Using the four clinical prediction models obtained in step S34 as research subjects, and taking HSP and HSPN patients in the training group and internal validation group as research subjects, the best prediction model is selected by using ROC analysis, calibration curve and clinical decision curve. By comparing the research subjects in the training group and the internal validation group, the best prediction model can be selected from the four clinical prediction models.

[0052] Specifically, in this embodiment, the net benefit of the four clinical prediction models is first evaluated using Clinical Decision Analysis (DCA). The clinical decision curve is a method used to evaluate the diagnostic accuracy of diagnostic models. Compared to the traditional ROC curve, DCA also takes into account the impact of false positives and false negatives on the prediction model in actual clinical practice. The formula for calculating the net benefit in DCA is as follows:

[0053]

[0054] See Figure 4 and Figure 5 The results showed that the clinical prediction models AIDD and AIDi had the highest net benefit in both the training and internal validation groups.

[0055] We then used the Net Reclassification Index (NRI) and the Integrated Discriminant Index (IDI) to evaluate whether the clinical prediction model AIDi could improve the efficiency of the clinical prediction model AIDD.

[0056] The Net Reclassification Index (NRI) is commonly used to evaluate the quantitative change in how well a new diagnostic indicator correctly classifies subjects compared to an older one in a diagnostic trial. Since the NRI can be used to assess the accuracy of a diagnostic trial's predictions, it can also be used to judge the accuracy of a predictive model. Based on published clinical research literature, the NRI is more widely used to compare the accuracy of two predictive models.

[0057] Simultaneously, the children were reclassified using the net reclassification index (NRI) and the clinical prediction model AIDi and AIDD. As these are two different clinical prediction models, some children will be classified differently in the two models. Therefore, we use this reclassification change to calculate the net reclassification index (NRI) to compare the two models and judge their predictive ability.

[0058] Similarly, the Integrated Discriminant Index (IDI) can also reflect the change in the difference between the predicted probabilities of two models, and thus judge the overall predictive ability of the two models.

[0059] IDI = (P model1,events -P model2,noevents )-(P model1,noevents -P model2,nonevents )

[0060] Among them, P model1,events and P model2,nonevents This represents the average probability of disease occurrence predicted by the two models for each individual in the disease group. Subtracting the two models gives the increase in the predicted probability. For the disease group, a higher predicted probability indicates a more accurate model. Therefore, P... model1,events and P model2,noevents The larger the difference, the better the former, Model 1, is; conversely, the smaller the difference, the better the latter, Model 2 is.

[0061] And P model1,noevents and P model2,nonevents P represents the mean probability of disease occurrence predicted by the two models for each individual in the non-disease group. The difference between the two models represents the reduction in the predicted probability. For the non-disease group, the lower the predicted probability of disease, the more accurate the model. model1,nonevents and P model2,nonevents The smaller the difference, the better the former, Model 1, is; conversely, the better the latter, Model 2 is.

[0062] Finally, subtracting the two parts above gives the IDI. Generally speaking, if IDI > 0, it means that Model 1 has better predictive ability than Model 2. If IDI < 0, it means that Model 2 has better predictive ability than Model 1.

[0063] The clinical prediction model AIDD is used as the standard model for HSPN risk prediction. The clinical prediction model AIDi is obtained by removing the independent risk factor Dor from the standard model AIDD. See [link to relevant documentation]. Figure 6 As shown in the figure, the clinical prediction model AIDi accurately diagnosed more HSPN cases in high-risk children compared to the standard model AIDD. Therefore, the clinical prediction model AIDi has a more significant predictive effect on the high risk of HSPN. Finally, the Nomogram based on the clinical prediction model AIDi was obtained by using R language as a statistical tool.

[0064] See Figure 7 The clinical prediction model AIDi showed good predictive performance for HSPN in internal and external validation groups, different histological grades, and different genders. To optimize the clinical application of AIDi nomogram, ROC analysis was used to perform gender stratification and histological grade stratification analysis of HSPN.

[0065] AUROC was 0.897 (95% CI: 0.840-0.953) in the external validation group and 0.920 (95% CI: 0.886-0.954) in the internal validation group.

[0066] AUROC was 0.933 (95% CI: 0.889-0.977) in HSPN histological grades (I, II) and 0.939 (95% CI: 0.908-0.970) in HSPN histological grades (III, IV).

[0067] The AUROC level was 0.949 (95% CI: 0.907-0.990) in female patients and 0.926 (95% CI: 0.885-0.967) in male patients.

[0068] ROC analysis showed that the AUROC values ​​were all above 0.7. These results indicate that the AIDi clinical prediction model has a high predictive effect on HSPN patients, regardless of gender or histological grade.

[0069] Specifically, the usage of Nomograms based on the AIDi clinical prediction model:

[0070] 1. Clinicians should diagnose HSP according to diagnostic criteria; at the same time, serum IgG and D-dimer levels should be measured.

[0071] 2. Sum the points identified for each predictor on the "index" scale to calculate the total score. The individual risk of the HSPN can be obtained by comparing the "total index" scale with the "risk" scale.

[0072] 3. If the risk of the child is greater than 0.7, close follow-up and monitoring for the occurrence of HSPN are strongly recommended.

[0073] See Figure 8 The following example illustrates the clinical prediction model and its Nomogram in this invention. The Nomogram shows that the patient's age score is 80, the IgG score is 92, and the D-dimer score is 100. By adding these three scores together and comparing them with the total index and risk reference chart, it can be seen that the patient's risk of developing HSPN is 95.8%.

[0074] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations, additions, or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A model for predicting the risk of Henoch-Schönlein purpura nephritis and a method for constructing its Nomogram, characterized in that, Its construction steps include: S1: Several children with allergic purpura and purpuric nephritis were divided into a training group and an internal validation group, while several children with allergic purpura and purpuric nephritis from external hospitals were set up as an external validation group. S2: Collect information on children with allergic purpura and children with purpuric nephritis from step 1; the information includes general population information, duration of rash, peripheral blood cell count and distribution, C-reactive protein, immunoglobulin classification and complement, lymphocyte subset analysis, coagulation function, and D-dimer. S3: Establish a clinical prediction model based on the information obtained in step S2, test the model, and obtain the optimal prediction model; the clinical prediction models include: a clinical prediction model based on the child's age, IgG, D-dimer and rash duration, a clinical prediction model based on the child's age, IgG and D-dimer, a clinical prediction model based on age, IgG and rash duration, and a clinical prediction model based on the child's age, D-dimer and rash duration; The clinical prediction model and its Nomogram were constructed using R language as a statistical tool; S4: Construct a Nomogram to predict the risk of developing Henoch-Schönlein purpura nephritis based on the optimal prediction model; Step S3 includes: S31: Statistical tests were used to compare the training group of HSP patients and the cohort of HSPN patients to identify indicators with significant differences. S32: Apply univariate logistic regression to the indicators with significant differences in step S31 to further screen out candidate predictive indicators. The candidate predictive indicators include: age, weight, duration of rash, platelet count, C-reactive protein, D-dimer, immunoglobulin G, and complement C3. S33: The candidate predictive indicators in step S32 are subjected to multivariate logistic regression analysis to identify independent risk factors, which include age, duration of rash, D-dimer and IgG. S34: Construct four clinical predictive models for the risk of developing HSPN in children with HSP using the independent risk factors in step S33. S35: Select the best prediction model from the four clinical prediction models obtained in step S34 through analysis methods. The analysis methods include any one or more of the following: receiver operating procedure (ROC), Hosmer-Lemeshow goodness-of-fit test, clinical decision analysis curve, net reclassification index, and comprehensive discriminant index.

2. The model for predicting the risk of Henoch-Schönlein purpura nephritis and the method for constructing its Nomogram according to claim 1, characterized in that, In step S31, the statistical test method includes any one or more of the following: T-test, Mann-Whitney U test, and chi-square test.

Images