ANCA-related vasculitis kidney injury risk prediction method
By establishing a kidney injury risk prediction model based on glomerular filtration rate and percentage of glomerular sclerosis, the difficulty of predicting kidney injury risk for ANCA-related vasculitis in the prior art is solved, and accurate risk assessment and personality of pediatric patients are achieved Chemical treatment reduces the risk of end-stage renal disease.
Patent Information
- Application Number
- CN202310897673.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-07-20
- Publication Date
- 2024-12-20
AI Technical Summary
The prior art is difficult to effectively predict the risk of kidney damage in ANCA-related vasculitis, especially in pediatric patients, resulting in misdiagnosis and waste of medical resources.
By collecting the data on glomerular filtration rate and glomerular sclerosis percentage of patients, a kidney injury risk prediction model was established, and the logistic regression analysis and verification set verification was used to select prediction models of accuracy and stability, and divide high-risk and low-risk patient.
Accurate prediction of the risk of kidney damage in ANCA-related vasculitis is achieved, helping to prevent early and personalized treatment, and reducing the probability of end-stage renal disease.
Smart Images
Figure CN119170261A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of disease risk prediction, and in particular to a method for predicting the risk of kidney damage in ANCA-associated vasculitis. Background Art
[0002] Antineutrophil cytoplasmic antibody (ANCA)-associated vasculitis (AAV) is a rare autoimmune disease characterized by small vessel inflammation and fibrin necrosis. As a life-threatening systemic disease, AAV can affect multiple organic injuries, and ANCA-associated glomerulonephritis (AAGN) is a key factor in determining the final prognosis of AAV patients, which may lead to continuous deterioration of renal function. Ultimately, approximately 20% of patients develop end-stage renal disease (ESRD). Although childhood-onset AAV is rare, renal damage worsens faster than in adults. 55-58% of children with AAGN develop chronic kidney disease (CKD), and 29-32% of patients develop ESRD. Therefore, early diagnosis and precise individualized treatment are crucial to improving renal function and reducing mortality in children.
[0003] In the prior art, the Berden classification classifies the pathology of AAGN into focal, crescent-shaped, sclerotic, or mixed lesions to predict the renal outcomes of pathological lesions. Previous studies have shown that the clinical manifestations of AAV vary by region, race, and age, which makes it more difficult to implement these categories in pediatric patients. For clinicians, AAV has many pathological features and few real cases, making it difficult for clinicians to access and observe real cases; it is difficult to directly judge the patient's risk of illness from the actual parameters and patient status, which can easily lead to misdiagnosis, waste of medical resources, or untimely medical treatment. So far, due to the rarity of AAV, there is a lack of predictive models to predict the progression of pediatric AAV and AAGN. Summary of the invention
[0004] In view of the above-mentioned deficiencies in the prior art, the present invention provides a method for predicting the risk of renal injury in ANCA-associated vasculitis.
[0005] In order to achieve the above-mentioned object of the invention, the technical solution adopted by the present invention is:
[0006] A method for predicting the risk of renal injury in ANCA-associated vasculitis is provided, comprising the following steps:
[0007] S1: Collect the glomerular filtration rate eGFR and glomerular sclerosis percentage GGS of the patient to be predicted;
[0008] S2: Establish a kidney injury risk prediction model, collect disease characterization data of several patients with different degrees of pathology to obtain a data set, and train the kidney injury risk prediction model through the data set to obtain a kidney injury risk prediction model;
[0009] S3: inputting the collected characterization data of the patient to be predicted into the kidney injury risk prediction model to obtain the disease prediction value P of the patient to be predicted;
[0010] S4: Patients are divided into risk levels according to the disease prediction value P, and patients are supervised according to the risk level.
[0011] Furthermore, step S2 includes the following specific steps:
[0012] A1: Establish a data set, collect disease characterization data of several patients with different degrees of pathology to obtain a data set, the elements in the data set are disease characterization data of a single patient; randomly extract elements in the data set to obtain a training set and a validation set, and the number of elements in the training set is 2-3 times the number of elements in the validation set;
[0013] Disease characterization data include: immunofluorescence of perinuclear anti-neutrophil cytoplasmic antibodies, immunofluorescence of cytoplasmic anti-neutrophil cytoplasmic antibodies, ELISA values of myeloperoxidase-anti-neutrophil cytoplasmic antibodies MPO-ANCA, ELISA values of proteinase 3-anti-neutrophil cytoplasmic antibodies PR3-ANCA, white blood cell content, hemoglobin content, platelet content, C-reactive protein content, albumin content, immunoglobulin G content and complement C3 content, serum creatinine content Scr, blood urea nitrogen Bun content, urine composition, glomerular filtration rate eGFR, glomerular sclerosis percentage GGS, and childhood vasculitis activity score;
[0014] A2: Patients with chronic kidney disease stage 5 in the training set were divided into the end-stage renal disease group, and the remaining patients in the training set were divided into the non-end-stage renal disease group;
[0015] A3: Logistic regression analysis was used to analyze the disease characteristics of the ESRD group and the non-ESRD group to obtain two groups of independent risk factors, one of which was the independent risk factors at baseline, including serum creatinine content (Scr); the other was the independent risk factors at renal biopsy, including glomerular filtration rate (eGFR) and glomerular sclerosis percentage (GGS);
[0016] A4: According to the two groups of independent risk factors, the prediction model was established to obtain the baseline prediction model:
[0017]
[0018] Prediction model for renal biopsy:
[0019]
[0020] A5: Use the validation set to compare and verify the baseline prediction model and the renal biopsy prediction model. Judge the accuracy and stability of the baseline prediction model and the renal biopsy prediction model according to the prognostic value, and select the prediction model with better accuracy and stability between the baseline prediction model and the renal biopsy prediction model as the renal injury risk prediction model.
[0021] Furthermore, the model verification method in step A5 includes the following steps:
[0022] B1: The prognostic value p1 of the prediction model for the training set data at baseline; the prognostic value p2 of the prediction model for the validation set data at baseline; the prognostic value p3 of the prediction model for the training set data at the time of renal biopsy; and the prognostic value p4 of the prediction model for the validation set data at the time of renal biopsy were calculated by the product limit method.
[0023] B2: The prognostic value p5 of the prediction model for the training set data at baseline; the prognostic value p6 of the prediction model for the validation set data at baseline; the prognostic value p7 of the prediction model for the training set data at the time of renal biopsy; and the prognostic value p8 of the prediction model for the validation set data at the time of renal biopsy were calculated by random sampling method.
[0024] B3: Based on the prognostic values p1, p2, p3, p4, p5, p6, p7 and p8, the renal injury risk prediction model was obtained by one-way analysis of variance.
[0025] Furthermore, the specific steps of dividing the risk level in step S4 include: comparing the disease prediction value P output by the kidney injury risk prediction model with the disease threshold P max For comparison, when the output disease prediction value P is greater than or equal to the disease threshold P max If the patient has a positive test result, the patient is classified as high risk; otherwise, the patient is classified as low risk.
[0026] Furthermore, the disease threshold P max =0.585.
[0027] Furthermore, the optimal cutoff value of serum creatinine content Scr at baseline was 416.2 μmol / L, and the AUC of the ROC curve was 0.885.
[0028] Furthermore, the critical value of eGFR at the time of renal biopsy was 19.42 mL / min / 1.73 m2, and the AUC of the ROC curve was 0.842; the critical value of glomerular sclerosis percentage GGS was 30.8%, and the AUC of the ROC curve was 0.734.
[0029] The beneficial effects of the present invention are:
[0030] The present invention constructs a kidney injury risk prediction model, inputs the patient's test results into the kidney injury risk prediction model, and can obtain the risk probability of the patient suffering from ANCA-associated vasculitis, thereby helping to identify ANCA-associated vasculitis in clinical practice; after classifying and marking high-risk patients, ANCA-associated vasculitis can be prevented in the early stage, ensuring personalized and precise treatment, and reducing the probability of ANCA-associated vasculitis turning into end-stage renal disease. BRIEF DESCRIPTION OF THE DRAWINGS
[0031] Figure 1 It is a schematic diagram of the process of the present invention.
[0032] Figure 2 is the comparison histogram of prognostic values p1, p2 and p5;
[0033] Figure 3 is the comparison histogram of prognostic values p3, p4, and p7;
[0034] Figure 4 It is the comparison histogram of prognostic values p5, p6, p7 and p8. DETAILED DESCRIPTION
[0035] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.
[0036] like Figure 1 As shown, a method for predicting the risk of renal injury in ANCA-associated vasculitis comprises the following steps:
[0037] S1: Collect the glomerular filtration rate eGFR and glomerular sclerosis percentage GGS of the patient to be predicted;
[0038] S2: Establish a kidney injury risk prediction model, collect disease characterization data of several patients with different degrees of pathology to obtain a data set, and train the kidney injury risk prediction model through the data set to obtain a kidney injury risk prediction model;
[0039] S3: inputting the collected characterization data of the patient to be predicted into the kidney injury risk prediction model to obtain the disease prediction value P of the patient to be predicted;
[0040] S4: Patients are divided into risk levels according to the disease prediction value P, and patients are supervised according to the risk level.
[0041] Step S2 includes the following specific steps:
[0042] A1: Establish a data set, collect disease characterization data of several patients with different degrees of pathology to obtain a data set, and the elements in the data set are disease characterization data of a single patient; randomly extract elements in the data set to obtain a training set and a validation set, and the number of elements in the training set is 2-3 times the number of elements in the validation set;
[0043] Disease characterization data include: immunofluorescence of perinuclear anti-neutrophil cytoplasmic antibodies, immunofluorescence of cytoplasmic anti-neutrophil cytoplasmic antibodies, ELISA values of myeloperoxidase-anti-neutrophil cytoplasmic antibodies MPO-ANCA, ELISA values of proteinase 3-anti-neutrophil cytoplasmic antibodies PR3-ANCA, white blood cell content, hemoglobin content, platelet content, C-reactive protein content, albumin content, immunoglobulin G content and complement C3 content, serum creatinine content Scr, blood urea nitrogen Bun content, urine composition, glomerular filtration rate eGFR, glomerular sclerosis percentage GGS, and childhood vasculitis activity score;
[0044] A2: Patients with chronic kidney disease stage 5 in the training set were divided into the end-stage renal disease group, and the remaining patients in the training set were divided into the non-end-stage renal disease group;
[0045] A3: Logistic regression analysis was used to analyze the disease characteristics of the ESRD group and the non-ESRD group to obtain two groups of independent risk factors, one of which was the independent risk factors at baseline, including serum creatinine content (Scr); the other was the independent risk factors at renal biopsy, including glomerular filtration rate (eGFR) and glomerular sclerosis percentage (GGS);
[0046] A4: According to the two groups of independent risk factors, the prediction model was established to obtain the baseline prediction model:
[0047]
[0048] Prediction model for renal biopsy:
[0049]
[0050] A5: Use the validation set to compare and verify the baseline prediction model and the renal biopsy prediction model. Judge the accuracy and stability of the baseline prediction model and the renal biopsy prediction model according to the prognostic value, and select the prediction model with better accuracy and stability between the baseline prediction model and the renal biopsy prediction model as the renal injury risk prediction model.
[0051] In specific implementation, the prediction model at baseline has the same stability as the prediction model at renal biopsy; and both the prediction model at renal biopsy and the prediction model at baseline have good accuracy. Through verification, the prediction model at renal biopsy is better than the prediction model at baseline, but it does not mean that the prediction model at baseline cannot make predictions. In specific implementation, the prediction model at renal biopsy or the prediction model at baseline is used flexibly for prediction according to the difficulty of collecting the patient's serum creatinine content Scr, glomerular filtration rate eGFR and glomerular sclerosis percentage GGS. If necessary, the prediction model at renal biopsy and the prediction model at baseline are used at the same time to predict the patient's disease risk as a reference for the final disease risk classification.
[0052] The model verification method in step A5 includes the following steps:
[0053] B1: The prognostic value p1 of the prediction model for the training set data at baseline; the prognostic value p2 of the prediction model for the validation set data at baseline; the prognostic value p3 of the prediction model for the training set data at the time of renal biopsy; and the prognostic value p4 of the prediction model for the validation set data at the time of renal biopsy were calculated by the product limit method.
[0054] B2: The prognostic value p5 of the prediction model for the training set data at baseline; the prognostic value p6 of the prediction model for the validation set data at baseline; the prognostic value p7 of the prediction model for the training set data at the time of renal biopsy; and the prognostic value p8 of the prediction model for the validation set data at the time of renal biopsy were calculated by random sampling method.
[0055] B3: Based on the prognostic values p1, p2, p3, p4, p5, p6, p7 and p8, the renal injury risk prediction model was obtained by one-way analysis of variance.
[0056] In specific implementation, the prognostic values p1, p2 and p5 are calculated, and the following are made according to the prognostic values: Figure 2 The comparison histogram shown is composed of Figure 2 It can be seen that the prognostic value p1 is significantly greater than the prognostic value p2, and there is no obvious difference between the prognostic value p1 and the prognostic value p5. The accuracy of the prediction model using the validation data set at baseline is lower than the accuracy of the prediction model using the training set data at baseline.
[0057] The prognostic values p3, p4 and p7 are calculated, and the following are made based on the prognostic values: Figure 3 The comparison histogram shown is composed of Figure 3 It can be seen that there is no significant difference between the prognostic values p3, p4 and p7, and the prediction model for renal biopsy has good prediction accuracy in both the training data set and the validation data set.
[0058] The prognostic values p5, p6, p7 and p8 are calculated, and the following are made based on the prognostic values: Figure 4 The comparison histogram shown is composed of Figure 4It can be seen that there is no significant difference between the prognostic value p5 and the prognostic value p6, and the prognostic value p7 is significantly greater than the prognostic value p8. The stability of the prediction model at renal biopsy is higher than that at baseline.
[0059] Therefore, the renal biopsy prediction model was selected as the renal injury risk prediction model.
[0060] Step S4 includes the following steps: comparing the predicted value P output by the kidney injury risk prediction model with the threshold value P max For comparison, when the output disease prediction value P is greater than or equal to the disease threshold P max If the patient is at high risk, the patient is classified as high risk; otherwise, the patient is classified as medium risk.
[0061] Disease threshold P max =0.585.
[0062] The optimal cutoff value of serum creatinine level Scr at baseline was 416.2 μmol / L, and the AUC of the ROC curve was 0.885.
[0063] The critical value of eGFR during renal biopsy is 19.42 mL / min / 1.73 m2, and the AUC of the ROC curve is 0.842; the critical value of glomerular sclerosis percentage GGS is 30.8%, and the AUC of the ROC curve is 0.734. The calculation method of the AUC of the ROC curve is an existing technology, so it will not be described in detail.
Claims
1. A method for predicting the risk of renal injury in ANCA-associated vasculitis, characterized in that: The steps include: S1: Collect the glomerular filtration rate eGFR and glomerular sclerosis percentage GGS of the patient to be predicted; S2: Establish a kidney injury risk prediction model, collect disease characterization data of several patients with different degrees of pathology to obtain a data set, and train the kidney injury risk prediction model through the data set to obtain a kidney injury risk prediction model; S3: inputting the collected characterization data of the patient to be predicted into the kidney injury risk prediction model to obtain the disease prediction value P of the patient to be predicted; S4: Patients are divided into risk levels according to the disease prediction value P, and patients are supervised according to the risk level.
2. The method for predicting the risk of renal injury in ANCA-associated vasculitis according to claim 1, characterized in that: The step S2 comprises the following specific steps: A1: Establish a data set, collect disease characterization data of several patients with different degrees of pathology to obtain a data set, the elements in the data set are disease characterization data of a single patient; randomly extract elements in the data set to obtain a training set and a validation set, and the number of elements in the training set is 2-3 times the number of elements in the validation set; The disease characterization data include: immunofluorescence of perinuclear anti-neutrophil cytoplasmic antibodies, immunofluorescence of cytoplasmic anti-neutrophil cytoplasmic antibodies, ELISA values of myeloperoxidase-anti-neutrophil cytoplasmic antibodies MPO-ANCA, ELISA values of proteinase 3-anti-neutrophil cytoplasmic antibodies PR3-ANCA, white blood cell content values, hemoglobin content values, platelet content values, C-reactive protein content values, albumin content values, immunoglobulin G content values and complement C3 content values, serum creatinine content values Scr, blood urea nitrogen Bun content values, urine components, glomerular filtration rate eGFR, glomerular sclerosis percentage GGS, and childhood vasculitis activity score; A2: Patients with chronic kidney disease stage 5 in the training set were divided into the end-stage renal disease group, and the remaining patients in the training set were divided into the non-end-stage renal disease group; A3: Logistic regression analysis was used to analyze the disease characteristics of the ESRD group and the non-ESRD group to obtain two groups of independent risk factors, one of which was the independent risk factors at baseline, including serum creatinine content (Scr); the other was the independent risk factors at renal biopsy, including glomerular filtration rate (eGFR) and glomerular sclerosis percentage (GGS); A4: According to the two groups of independent risk factors, the prediction model was established to obtain the baseline prediction model: Prediction model for renal biopsy: A5: Use the validation set to compare and verify the baseline prediction model and the renal biopsy prediction model. Judge the accuracy and stability of the baseline prediction model and the renal biopsy prediction model according to the prognostic value, and select the prediction model with better accuracy and stability between the baseline prediction model and the renal biopsy prediction model as the renal injury risk prediction model.
3. The method for predicting the risk of renal injury in ANCA-associated vasculitis according to claim 2, characterized in that: The model verification method in step A5 comprises the following steps: B1: The prognostic value p1 of the prediction model for the training set data at baseline; the prognostic value p2 of the prediction model for the validation set data at baseline; the prognostic value p3 of the prediction model for the training set data at the time of renal biopsy; and the prognostic value p4 of the prediction model for the validation set data at the time of renal biopsy were calculated by the product limit method. B2: The prognostic value p5 of the prediction model for the training set data at baseline; the prognostic value p6 of the prediction model for the validation set data at baseline; the prognostic value p7 of the prediction model for the training set data at the time of renal biopsy; and the prognostic value p8 of the prediction model for the validation set data at the time of renal biopsy were calculated by random sampling method. B3: Based on the prognostic values p1, p2, p3, p4, p5, p6, p7 and p8, the renal injury risk prediction model was obtained by one-way analysis of variance.
4. The method for predicting the risk of renal injury in ANCA-associated vasculitis according to claim 1, characterized in that: The specific steps of dividing the risk level in step S4 include: comparing the disease prediction value P output by the kidney injury risk prediction model with the disease threshold P max For comparison, when the output disease prediction value P is greater than or equal to the disease threshold P max If the patient has a positive test result, the patient is classified as high risk; otherwise, the patient is classified as low risk.
5. The method for predicting the risk of renal injury in ANCA-associated vasculitis according to claim 4, characterized in that: The disease threshold P max =0.
585.
6. The method for predicting the risk of renal injury in ANCA-associated vasculitis according to claim 2, characterized in that: The optimal cutoff value of serum creatinine content Scr at the baseline was 416.2 μmol / L, and the AUC of the ROC curve was 0.
885.
7. The method for predicting the risk of renal injury in ANCA-associated vasculitis according to claim 2, characterized in that: The critical value of eGFR during renal biopsy was 19.42 mL / min / 1.73 m2, and the AUC of the ROC curve was 0.842; the critical value of glomerular sclerosis percentage GGS was 30.8%, and the AUC of the ROC curve was 0.734.