System and method for establishing a prediction model of tacrolimus-induced renal injury after renal transplantation

Abstract

The present invention discloses a system and method for establishing a tacrolimus renal injury prediction model after renal transplantation, comprising a collection module, a statistical analysis module, an extraction module, a model building module, and a verification module, wherein: the collection module is used to collect data on patient demographics, FK506 dosage, combined medication, laboratory tests, and CYP3A5 genotypes; the statistical analysis module is used to divide patients into a renal injury group and a non-renal injury group, and perform statistical analysis on the data; the extraction module is used to extract renal injury influencing factors through multi-factor logistic regression analysis; the model building module is used to input the renal injury influencing factors into a renal injury prediction model to establish a renal injury prediction model; and the verification module is used to verify and evaluate the renal injury prediction model. By adopting the present invention, the occurrence of renal injury can be comprehensively predicted according to the renal injury influencing factors, and medical intervention can be performed earlier.

Description

Technical Field

[0001] The present invention relates to the field of computers, and in particular to a system and method for establishing a tacrolimus renal injury prediction model after renal transplantation. Background Art

[0002] At present, organ transplant patients mainly use the calcineurin inhibitor tacrolimus (FK506), combined with a triple immunosuppressive regimen of mycophenolate mofetil and corticosteroids. However, the therapeutic window of FK506 is narrow, insufficient concentration has the risk of rejection, and excessive concentration may cause irreversible renal damage. FK506 toxicity is drug concentration-dependent within a certain dose range, and therapeutic drug monitoring (TDM) is usually performed to adjust the FK506 dose to control the blood drug concentration. Due to genetic polymorphisms of FK506 metabolic enzymes and transporters, there are large individual differences in blood drug concentrations in patients, and there is a certain time difference between TDM test values ​​and dose adjustments. Therefore, dose adjustments based on TDM results are not enough to meet the clinical needs of organ transplant recipients. Real-world research collects real clinical data, analyzes the medication characteristics and efficacy of drugs, continuously pays attention to factors related to adverse reactions, strengthens adverse event monitoring, and reduces medication risks. Existing studies have used machine learning to predict FK506 concentrations in real-world patients with autoimmune diseases and screen factors that affect concentrations. However, there is currently a lack of a prediction model for renal injury in patients taking FK506. Therefore, a prediction model is needed to predict the risk of renal injury in patients receiving FK506 after renal transplantation. Summary of the invention

[0003] The purpose of the present invention is to provide a system and method for establishing a prediction model for tacrolimus renal injury after renal transplantation, and to perform consistency evaluation of renal injury assessment according to three CysC-based glomerular filtration rate (eGFR) calculation formulas to solve the problems raised in the above-mentioned background technology.

[0004] To achieve the above-mentioned purpose, one aspect of the present invention provides a system for establishing a prediction model of tacrolimus renal injury after renal transplantation, comprising a collection module, a statistical analysis module, an extraction module, a model establishment module, and a verification module, wherein:

[0005] The collection module is used to collect data on patient demographics, FK506 dosage, concomitant medications, laboratory tests, and CYP3A5 genotype;

[0006] A statistical analysis module is used to divide the patients into a renal injury group and a non-renal injury group and perform statistical analysis on the data;

[0007] Extraction module, used to extract the influencing factors of renal injury through multivariate logistic regression analysis;

[0008] A model building module is used to input the renal injury influencing factors into a renal injury prediction model to build a renal injury prediction model;

[0009] The validation module is used to validate and evaluate the renal injury prediction model.

[0010] Furthermore, the patients in the collection module are patients who received kidney transplantation, tacrolimus-based immunosuppression regimen, and had baseline examination results, and the baseline examination was a creatinine test within 90 to 120 (±7) days after the first dose.

[0011] Furthermore, in the statistical analysis module, analysis of variance was used for comparison between groups of continuous variables, chi-square test or Fisher’s exact probability method was used for comparison between groups of categorical variables, and Kruskal-Wallis H test was used for comparison between groups.

[0012] Furthermore, the factors affecting renal injury in step S3 are age, lymphocyte percentage, hematocrit, urea, total protein, and neutrophil percentage.

[0013] Another aspect of the present invention provides a method for establishing a prediction model of tacrolimus renal injury after renal transplantation, comprising the following steps:

[0014] Step S1, collect data on patient demographics, FK506 dosage, concomitant medications, laboratory tests, and CYP3A5 genotype;

[0015] Step S2, dividing the patients into a renal injury group and a non-renal injury group, and performing statistical analysis on the data;

[0016] Step S3, extracting the influencing factors of renal injury through multivariate logistic regression analysis;

[0017] Step S4, inputting the renal injury influencing factors into a renal injury prediction model to establish a renal injury prediction model;

[0018] Step S5, verifying and evaluating the renal injury prediction model.

[0019] Furthermore, the patient in step S1 is a patient who has received a kidney transplant, a tacrolimus-based immunosuppression regimen, and a baseline examination result, wherein the baseline examination is a creatinine test within 90 to 120 (±7) days after the first administration.

[0020] Furthermore, the statistical analysis described in step S2 is to use variance analysis for comparison between continuous variables, chi-square test or Fisher's exact probability method for comparison between categorical variables, Kruskal-Wallis H test for comparison between groups, and the area under the receiver operating characteristic curve is used to evaluate the predictive performance of the model.

[0021] Furthermore, the factors affecting renal injury in step S3 are age, lymphocyte percentage, hematocrit, urea, total protein, and neutrophil percentage.

[0022] Compared with the prior art, the present system and method have the following advantages:

[0023] 1. The present invention can be used to comprehensively predict the occurrence of renal injury based on the factors affecting renal injury, preliminarily screen out high-risk patients, and conduct medical intervention earlier, thereby improving the survival rate of transplanted organs and increasing patient benefits.

[0024] 2. The present invention is based on 10 years of renal transplant patients who received FK506 as the main immunosuppressive regimen, and has a certain data basis, and the results are more credible.

[0025] 3. The present invention focuses on the dosage and combined medication of FK506 in clinical practice, which can provide a certain degree of reference for clinical medication.

[0026] 4. The present invention investigated the effect of CYP3A5 genotype on FK506-induced renal injury.

[0027] 5. The present invention performs consistency test on three CysC-based eGFR calculation formulas and the established renal injury prediction model, thus making up for the defects of low popularity and low applicability of CysC detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 Flowchart of the method for establishing a prediction model for tacrolimus-induced renal injury after renal transplantation.

[0029] Figure 2 A process flow chart is provided for the embodiment.

[0030] Figure 3 This is the ROC curve of the Logistic regression model. DETAILED DESCRIPTION

[0031] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0032] This example was conducted on patients who received kidney transplantation at the Organ Transplant Center of Shanghai Changzheng Hospital from June 2010 to June 2020. Based on the previous candidate biomarkers, the clinical risk factors for FK506 nephrotoxicity in patients after kidney transplantation were determined through real-world evidence, providing guidance for the safe application of FK506 in clinical practice.

[0033] like Figure 1 Shown is a flow chart of the method of the present invention, and the specific steps are as follows:

[0034] Step S1, collect data on patient demographics, FK506 dosage, concomitant medications, laboratory tests, and CYP3A5 genotype.

[0035] The inclusion criteria for patients were patients who received renal transplantation at the Organ Transplant Center of Shanghai Changzheng Hospital between June 2010 and June 2020, were on an FK506-based immunosuppression regimen, and had baseline examination results. The baseline period was defined as 90 to 120 (±7) days after the first dose. The baseline examination was defined as a creatinine test within 90 to 120 (±7) days after the first dose. Exclusion criteria were no record of tacrolimus administration, no record of renal transplantation, no creatinine test at baseline or a blank test result, and too high serum creatinine.

[0036] We collected data on demographics, FK506 dosage, concomitant medication, laboratory tests, and CYP3A5 genotype from the Shanghai Changzheng Hospital Information System. This example was divided into a renal injury group (n=292) and a non-renal injury group (n=966) based on whether renal injury occurred.

[0037] Step S2, dividing the patients into a renal injury group and a non-renal injury group, and performing statistical analysis on the data.

[0038] According to the KDIGO criteria, renal injury is defined as: (1) baseline serum creatinine is higher than 1.6 mg / ml; (2) serum creatinine increases by 0.3 mg / ml compared with the baseline value; (3) serum creatinine increases by 50% compared with the baseline value. Statistical analysis was performed using V9.4 software and Python V3.11.5 software. 2Analysis. Analysis of variance was used for comparison of continuous variables between groups; chi-square test or Fisher's exact probability method was used for comparison of categorical variables between groups; rank variables were expressed as cases (%), and Kruskal-Wallis H test was used for comparison between groups. P < 0.05 was considered statistically significant. Logistic regression analysis was used to analyze the influencing factors of renal injury in patients. The area under the receiver operating characteristic curve (AUROC) was used to evaluate the predictive performance of the model; the Kappa test was used for consistency analysis, and a Kappa coefficient of 0.21-0.40 was considered to have general consistency; 0.41-0.60 was considered to have moderate consistency; 0.61-0.80 was considered to have strong consistency; and 0.81-1.00 was considered to have high consistency.

[0039] A total of 1258 patients were selected in this example, and the entire selection process was as follows: Figure 2 The patients were divided into two groups: the group with renal injury (n=292) and the group without renal injury (n=966). The demographic baseline characteristics of the two groups of patients are shown in Table 1.

[0040] Table 1 Demographic characteristics of the patients at baseline

[0041]

[0042] There was a statistically significant age difference between the group with renal injury and the group without renal injury (p < 0.0001). The average age of the group with renal injury was 41.8 years old, and the average age of the group without renal injury was 45.1 years old.

[0043] With reference to the drug instructions and expert consensus, the oral dose of FK506 is divided into ≤21 mg / day (recommended initial dose), >21 mg / day and ≤42 mg / day (2 times the initial dose), >42 mg / day and ≤63 mg / day (3 times the initial dose), >63 mg / day (greater than 3 times the initial dose) in the present invention. The clinical application of FK506 in patients is shown in Table 2.

[0044] Table 2 Dosage of FK506

[0045]

[0046] Among them, 1,226 recorded doses were within the recommended initial dose range, 9 recorded doses were between the recommended initial dose and twice the recommended initial dose, 51 recorded doses were between two times the recommended initial dose and three times the recommended initial dose, and 120 recorded doses were greater than three times the recommended initial dose.

[0047] The patients’ concomitant medications are shown in Table 3.

[0048] Table 3 Patients' concomitant medications

[0049]

[0050]

[0051] As can be seen from the table above, among the drugs used in combination, there are significant differences in the clinical application of calcium channel blockers and systemic antifungal drugs in the two groups (P < 0.0001). Calcium channel blockers are mainly nifedipine, diltiazem, amlodipine and verapamil. The application rates of the four drugs in the renal injury group and the non-renal injury group are 57.5% vs 45.2%, 46.6% vs 32.1%, 39.0% vs 27.7%, 0.3% vs 0.3%, respectively. Among them, nifedipine (P = 0.0002), diltiazem (P < 0.0001) and amlodipine (P = 0.0002) have statistical differences between the two groups of patients. The application rate of the antifungal drug voriconazole in the two groups is 19.8% vs 8.2%, respectively, with statistical differences (P < 0.0001). In addition, only 3 patients used posaconazole in combination. In addition, there was a certain statistical difference in the use of proton pump inhibitors between the two groups (P=0.0454).

[0052] Step S3, extract the influencing factors of renal injury by univariate logistic regression analysis. The variables of influencing factors of renal injury included in the consideration are the 25 indicators selected by univariate analysis in Table 4: age, mean corpuscular volume, mean corpuscular hemoglobin concentration, mean platelet volume, lymphocyte percentage, hematocrit, hemoglobin, γ-glutamyl transferase, lactate dehydrogenase, urea, uric acid, total protein, globulin, albumin, direct bilirubin, indirect bilirubin, neutrophil percentage, monocyte count, basophil percentage, basophil count, FK506 blood concentration, prealbumin, total bile acid, alkaline phosphatase, and genetic testing.

[0053] Table 4 Factors affecting the occurrence of renal injury

[0054]

[0055]

[0056] As can be seen from the above table, in the univariate logistic regression analysis, nine indicators including age, lymphocyte percentage, hematocrit, lactate dehydrogenase, urea, total protein, albumin, monocyte count and prealbumin were statistically significant. Among them, age (OR=0.9721, 95%CI=0.958-0.986), lymphocyte percentage (OR=0.4214, 95%CI=0.258-0.691), hematocrit (OR=0.6093, 95%CI=0.377-0.974), urea (OR=2.1480, 95%CI=1.392-3.309), total protein (OR=0.4503, 95%CI=0.273-0.743) and neutrophil percentage (OR=0.4213, 95%CI=0.209-0.839) still had a significant impact on renal injury in multivariate Logistic regression analysis, and mean corpuscular volume (OR=0.5402, 95%CI=0.313-0.931) also showed a certain degree of significance. Therefore, the above data show that age, lymphocyte percentage, hematocrit, urea, total protein, and neutrophil percentage are risk factors for renal injury in patients after renal transplantation.

[0057] Step S4, when the model is established, laboratory indicators with a data missing ratio of no more than 30% are selected, and a preliminary Logistic regression model is established with age, gender and genetic testing. The variance inflation factor (VIF), condition number and ridge trace diagram of ridge regression are used to determine whether there is multicollinearity between the model independent variables and variables that may have multicollinearity, and the variables are screened and eliminated in combination with clinical knowledge; to avoid overfitting, the model after multicollinearity is processed by LASSO regression, and the variables therein are screened twice. The model after the secondary screening is regressed by stepwise selection, and the criteria for variable inclusion and exclusion are set to 0.15, so as to obtain the variables that will eventually be included in the statistical model and establish a renal injury prediction model.

[0058] Step S5, the prediction performance of the model is evaluated by the area under the receiver operating characteristic curve (AUC), and the renal injury prediction model is verified and evaluated. The ROC curve of the logistic regression model is shown in Figure 3 , AUC is 0.7006

[0059] (95% CI = 0.661-0.740). From the above data, it can be seen that the renal injury prediction model can predict the risk of renal injury in patients to a certain extent.

[0060] The estimated glomerular filtration rate (eGFRcr) based on creatinine is suitable for most people, but its accuracy is easily affected by diet and muscle mass. The clinical use of cystatin C to evaluate renal function has the characteristic of high accuracy, so this study selected three commonly used clinical eGFR calculation formulas based on CysC: eGFRCr-CysC, FAS-SCysc and FAS-Scr-SCysc. Renal injury was defined with reference to eGFR < 60, and consistency was tested with the previously established model. The results showed that the Kappa coefficient of the FAS-SCysc formula was 0.612, the highest, followed by FAS-Scr-SCysc (0.525) and eGFR Cr-CysC (0.451). Therefore, the Logistic regression analysis model established in this study can accurately predict the occurrence of renal injury, as shown in Table 5.

[0061] Table 5 Comparison of consistency between creatinine and CysC in determining renal injury

[0062]

[0063] Note: 1. eGFR Cr-CysC Formula = 135 × (Scr / 0.9) -0.207 ×(CysC / 0.8) -0.375 ×0.995 Age (Male, Scr≤0.9,

[0064] CysC≤0.8); 135×(Scr / 0.9) -0.207 ×(CysC / 0.) -0.711 ×0.995 Age (Male, Scr≤0.9, CysC>0.8); 135×(Scr / 0.9) -0.601 ×(CysC / 0.8) -0.375 ×0.995 Age (Male, Scr>0.9, CysC≤0.8); 135×(Scr / 0.9) -0.601 ×(CysC / 0.8) -0.711 ×0.995 Age (Male, Scr>0.9, CysC>0.8); 130×(Scr / 0.7) -0.248 ×(CysC / 0.8) -0.375 ×0.995 Age (Female, Scr≤0.7, CysC≤0.8); 130×(Scr / 0.7) -0.248 ×(CysC / 0.8) -0.711 ×0.995 Age(Female, Scr≤0.7, CysC>0.8); 2.FAS-SCysc=107.3 / (Scysc / Q Scysc )[×0.988 (Age-40) when age>40years]; 3.FAS-Scr-SCysc=107.3 / [α×(Scr / Q Scr )]+(1-α)(Scysc / Q Scysc )[×0.988 (Age-40) when age>40years]

[0065] The results of this example show that there is a statistical difference in age between the group with renal injury and the group without renal injury. A retrospective study divided patients into two groups, ≥65 years old and <65 years old. The ≥65-year-old group showed a lower acute rejection rate (10.1%) than the <65-year-old group, but also showed a lower 1-year, 5-year, and 10-year graft survival rate (90.6%, 80.8%, and 70.1%); elderly patients with comorbidities have a high risk of severe acute renal failure; the results of the example also show that there is a statistical difference in age between the two groups, among which elderly patients are more likely to suffer from renal injury, which is consistent with the research results.

[0066] Through factor screening, univariate and multivariate logistic regression analysis, the present invention determined that age, lymphocyte percentage, hematocrit, urea, total protein, and neutrophil percentage were 6 influencing factors closely related to the occurrence of renal injury.

[0067] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A system for establishing a prediction model of tacrolimus renal injury after renal transplantation, characterized in that: It includes collection module, statistical analysis module, extraction module, model building module and verification module, among which: The collection module is used to collect data on patient demographics, FK506 dosage, concomitant medications, laboratory tests, and CYP3A5 genotype; A statistical analysis module is used to divide the patients into a renal injury group and a non-renal injury group and perform statistical analysis on the data; Extraction module, used to extract the influencing factors of renal injury through multivariate logistic regression analysis; A model building module is used to input the renal injury influencing factors into a renal injury prediction model to build a renal injury prediction model; The validation module is used to validate and evaluate the renal injury prediction model.

2. The system for establishing a prediction model of tacrolimus renal injury after renal transplantation according to claim 1, characterized in that: The patients in the collection module were patients who received renal transplants, were on a tacrolimus-based immunosuppression regimen, and had baseline examination results, which were creatinine examinations within 90 to 120 (± 7) days after the first dose.

3. The system for establishing a prediction model of tacrolimus renal injury after renal transplantation according to claim 1, characterized in that: In the statistical analysis module, analysis of variance was used for comparison between groups of continuous variables, chi-square test or Fisher's exact probability method was used for comparison between groups of categorical variables, and Kruskal-Wallis H test was used for comparison between groups.

4. The system for establishing a prediction model of tacrolimus-induced renal injury after renal transplantation according to claim 1, characterized in that: The factors affecting renal injury in step S3 are age, lymphocyte percentage, hematocrit, urea, total protein, and neutrophil percentage.

5. A method for establishing a prediction model for tacrolimus-induced renal injury after renal transplantation, characterized in that: The following steps are involved: Step S1, collect data on patient demographics, FK506 dosage, concomitant medications, laboratory tests, and CYP3A5 genotype; Step S2, dividing the patients into a renal injury group and a non-renal injury group, and performing statistical analysis on the data; Step S3, extracting the influencing factors of renal injury through multivariate logistic regression analysis; Step S4, inputting the renal injury influencing factors into a renal injury prediction model to establish a renal injury prediction model; Step S5, using three CysC-based glomerular filtration rate calculation formulas to verify and evaluate the renal injury prediction model.

6. The method for establishing a prediction model of tacrolimus-induced renal injury after renal transplantation according to claim 5, characterized in that: The patient in step S1 is a patient who has received a kidney transplant, a tacrolimus-based immunosuppression regimen, and a baseline examination result, wherein the baseline examination is a creatinine examination within 90 to 120 (± 7) days after the first administration.

7. The method for establishing a prediction model of tacrolimus renal injury after renal transplantation according to claim 5, characterized in that: The statistical analysis described in step S2 is: analysis of variance for comparison between groups of continuous variables, chi-square test or Fisher's exact probability method for comparison between groups of categorical variables, Kruskal-Wallis H test for comparison between groups, and the area under the receiver characteristic curve is used to evaluate the predictive performance of the model.

8. The method for establishing a prediction model of tacrolimus-induced renal injury after renal transplantation according to claim 5, characterized in that: The factors affecting renal injury in step S3 are age, lymphocyte percentage, hematocrit, urea, total protein, and neutrophil percentage.

Images