Risk prediction model for substandard voriconazole valley concentration and construction method thereof

Abstract

The invention belongs to the technical field of therapeutic drug monitoring, and particularly relates to a voriconazole valley concentration substandard risk prediction model and a construction method thereof, and the method comprises the following steps: collecting clinical data of a patient treated by voriconazole and subjected to TDM; randomly dividing the data set into a training set and a verification set according to a ratio of 7: 3; carrying out single factor analysis and binary Logistic regression analysis on the training set to screen independent risk factors of which the voriconazole valley concentration does not reach the standard; carrying out correlation analysis on the screened independent risk factors, incorporating the independent risk factors with low correlation and constructing a Nomogram model; for the Nomogram model, the model distinction degree is evaluated by drawing an ROC curve, the calibration degree of the model is evaluated by drawing a calibration curve, the clinical net profit of the model is evaluated by drawing a decision analysis curve, and model internal verification is carried out by a Bootstrap method; and performing external verification on the model by using the verification set. The independent risk factors comprise LA, administration dosage, AST, severe pneumonia and tumors.

Description

Technical Field

[0001] The present invention belongs to the technical field of therapeutic drug monitoring, and in particular relates to a risk prediction model for non-standard trough concentration of voriconazole and a construction method thereof. Background Art

[0002] Voriconazole (VCZ) is a broad-spectrum triazole antifungal drug that inhibits the biosynthesis of ergosterol in fungal cells, thereby achieving antifungal effects. It has been recommended as a first-line drug for the treatment and prevention of invasive fungal disease (IFD). VCZ is mainly metabolized by the liver, and its metabolism in the human body shows nonlinear pharmacokinetics. Its dosage, trough concentration of the drug, and area under the concentration-time curve show nonlinear changes. Various factors can seriously affect the in vivo process of the drug, causing large differences in its blood concentration between individuals and within individuals. If the trough concentration of the patient's VCZ blood concentration is lower than or exceeds the safety critical value range, it may lead to clinical treatment failure or an increase in adverse reactions such as abnormal liver function, visual changes or visual impairment. Therefore, therapeutic drug monitoring (TDM) of voriconazole is required.

[0003] Previous studies have shown that the probability of VCZ trough concentration not meeting the standard can be as high as 50%, and many risk factors can affect VCZ trough concentration, including: patient age, weight, albumin level, CYP2C19 gene polymorphism, drug-drug interactions, liver dysfunction, etc. Importantly, with the introduction of artificial intelligence and prediction models, integrating multiple risk factors to establish efficient and accurate prediction models has received widespread attention. However, available prediction models for whether voriconazole trough concentrations meet the standard, especially prediction models with high prediction efficiency, have not been reported. Therefore, combining multiple influencing factors to establish a prediction model for voriconazole trough concentrations that do not meet the standard is of great significance for the implementation of individualized medication regimens for voriconazole, improving drug efficacy and reducing adverse reactions. Summary of the invention

[0004] In view of this, in order to solve the problems existing in the prior art, the first object of the present invention is to provide a method for constructing a risk prediction model for substandard trough concentration of voriconazole.

[0005] In order to achieve the above purpose, the present invention adopts the following technical solution:.

[0006] A method for constructing a risk prediction model for substandard trough concentration of voriconazole, comprising: collecting clinical data of patients treated with voriconazole and undergoing TDM; randomly dividing the data set into a training set and a validation set at a ratio of 7:3; screening independent risk factors for substandard trough concentration of voriconazole in the training set by univariate analysis and binary logistic regression analysis; performing correlation analysis on the screened independent risk factors, incorporating independent risk factors with low correlation into a Nomogram model; evaluating the discrimination of the Nomogram model by drawing an ROC curve, evaluating the calibration of the model by drawing a calibration curve, evaluating the clinical net benefit of the model by drawing a decision analysis curve, and performing internal validation of the model by the Bootstrap method; and performing external validation of the model by using the validation set.

[0007] Furthermore, the independent risk factors include: LA (lactic acid), dosage, AST (aspartate aminotransferase), severe pneumonia, and tumors.

[0008] It is worth noting that in order to comprehensively consider the influence of multiple factors, establish a risk prediction model, and help doctors to judge the risk probability of the occurrence of patients' VCZ trough concentration not meeting the standard at an early stage, so as to formulate a correct treatment plan, the present invention discloses a risk prediction model of voriconazole trough concentration not meeting the standard based on Nomogram, including: collecting clinical data of patients treated with voriconazole and undergoing TDM; randomly dividing the data set into a training set and a validation set at a ratio of 7:3; screening the independent risk factors of voriconazole trough concentration not meeting the standard by univariate analysis and binary Logistic regression analysis in the training set; performing correlation analysis on the screened independent risk factors, and incorporating and constructing the Nomogram model for the independent risk factors with low correlation; evaluating the model discrimination by drawing the ROC curve, evaluating the model calibration by drawing the calibration curve, evaluating the model clinical net benefit by drawing the decision analysis curve, and performing internal validation of the model by the Bootstrap method; and performing external validation of the model by using the validation set. The present invention provides a convenient and effective tool for clinicians to judge the risk of trough concentration not meeting the standard in patients treated with voriconazole, which is conducive to the implementation of individualized medication of voriconazole.

[0009] Furthermore, the specific steps of the method for constructing the risk prediction model for substandard trough concentration of voriconazole include:

[0010] S1: Systematic collection of patient data, including demographics and laboratory test and treatment details;

[0011] S2: According to the established inclusion and exclusion criteria, the collected patient data were screened one by one, and the patients who met the criteria were used to generate a data set. The data set was divided into a target group and a non-target group according to whether the voriconazole blood concentration reached the target. The data set was then randomly divided into a training set and a validation set at a ratio of 7:3;

[0012] S3: Univariate and binary logistic regression analysis was used to screen out independent risk factors affecting the failure of voriconazole trough concentration to reach the target in the target group and non-target group in the training set;

[0013] S4: Based on the information obtained in step S3, a nomogram model is established to predict the risk of substandard trough concentration of voriconazole.

[0014] Furthermore, in step S1, the patient information includes: age, gender, height, weight, BMI (body mass index), smoking history, drinking history, dosage, medication days, drug source, severe pneumonia, infection, hypertension, diabetes, heart disease, sepsis, hypoalbuminemia, tumor, Aspergillus, Candida, mixed infection, infection with unknown pathogens, AST (aspartate aminotransferase), ALT (alanine aminotransferase), ST / LT (aspartate aminotransferase / alanine aminotransferase ratio), ALB (plasma albumin), UREA (urea ), CREA (creatinine), UA (uric acid), EGFR (glomerular filtration rate), WBC (white blood cell count), NEUT% (neutrophil percentage), LYM% (lymphocyte percentage), NEUT# (absolute neutrophil count), LYM# (absolute lymphocyte count), RBC (red blood cell count), HGB (hemoglobin), PLT (platelet count), LA (lactic acid), CRP (C-reactive protein), PCT (procalcitonin), IL-6 (interleukin-6), antimicrobial drugs, proton pump inhibitors, and glucocorticoids.

[0015] It is worth noting that in the present invention, the missing values ​​of the three variables CRP, PCT, and IL-6 were greater than 25%, so these three variables were eliminated and not included in the subsequent analysis.

[0016] Furthermore, in step S2, the standard is defined as: the standard is defined as: 0.5 mg / L ≤ C min ≤5.0mg / L.

[0017] Furthermore, in step S2, the inclusion and exclusion criteria established include:

[0018] Inclusion criteria: treatment with voriconazole; age >18 years old; voriconazole treatment drug monitoring; complete clinical data and examination results;

[0019] Exclusion criteria: patients taking drugs that affect voriconazole concentration; patients diagnosed with liver diseases such as cirrhosis, drug-induced liver injury, viral hepatitis, liver cancer, or liver failure at admission; patients taking hemodialysis or other treatments that affect blood volume while taking voriconazole; non-steady-state trough concentration; pregnant or lactating women; incomplete clinical data and examination results.

[0020] Furthermore, the step S3 includes:

[0021] S31: Data were statistically analyzed using SPSS 27.0 statistical software. Qualitative data were expressed as cases (%) and inter-group comparisons were performed using χ 2 Test; Normally distributed quantitative data are expressed as mean ± standard deviation The t-test was used; the non-normally distributed quantitative data were expressed as median (quartile) [M(P25, P75)], and the Mann-Whitney U test was used; finally, 9 risk factors were screened out, including dosage, severe pneumonia, tumor, Aspergillus, Candida, AST, ALT, ST / LT, and LA (P < 0.10).

[0022] S32: SPSS27.0 was used to perform Pearson correlation analysis on the risk factors screened in step S31; the correlation analysis results showed that the Pearson correlation coefficient of AST and ALT was 0.85. ALT was eliminated and finally the dosage, severe pneumonia, tumor, Aspergillus, Candida, and AST were included in the binary logistic regression analysis.

[0023] S33: SPSS27.0 was used to perform binary logistic regression analysis on the risk factors screened in step S32; the results showed that the independent risk factors included: LA, dosage, AST, severe pneumonia, and tumor (P < 0.05).

[0024] It is worth noting that the present invention proposes for the first time to use LA, dosage, AST, severe pneumonia, and tumors (P < 0.05) to jointly predict the probability of substandard trough concentration of voriconazole, aiming to provide a convenient and effective tool for clinicians to judge the risk of substandard trough concentration in patients treated with voriconazole at an early stage, which is conducive to the implementation of individualized medication of voriconazole.

[0025] Furthermore, in step S4, based on the results of binary logistic regression analysis, the independent risk factors are incorporated into the Nomogram model of voriconazole trough concentration not meeting the standard, and RStudio 4.4.1 is used to construct the Nomogram model; in the constructed Nomogram model, the five independent risk factors are used as variables, and the scores corresponding to each variable at different values ​​are calculated, and the total score after adding up the scores of all variables is calculated to predict the probability of each patient having voriconazole trough concentration not meeting the standard.

[0026] Furthermore, the specific steps of the method for constructing the risk prediction model of voriconazole-related liver injury also include model evaluation, namely:

[0027] S5: performing internal analysis on the Nomogram prediction model established in step S4, including: ROC curve analysis, calibration curve analysis, and DCA curve analysis;

[0028] S6: internally verify the Nomogram prediction model established in step S4 by using the Bootstrap method (n=1000 times);

[0029] S7: Use the validation set to perform external evaluation on the Nomogram prediction model established in step S4.

[0030] It is worth noting that the methods used in constructing the model in the present invention include: using the hospital's electronic medical record system (HIS system) to query the patient's clinical data; using Excel to statistically analyze the patient's clinical data; using SPSS27.0 to perform balance test, univariate analysis and correlation analysis; and using RStudio 4.4.1 to construct, evaluate and verify the model.

[0031] Wherein, the step S5 comprises: the Nomogram model established in step S4 is evaluated by drawing an ROC curve to evaluate the discrimination of the model, drawing a calibration curve to evaluate the calibration of the model, and drawing a decision analysis curve to evaluate the clinical net benefit of the model.

[0032] The step S6 includes: performing internal validation of the Nomogram model established in step S4 by using the Bootstrap method (n=1000 times); calculating the mean ± standard deviation of the AUC of the logistic regression model established by 1000 simulated data sets. The internal validation results of the model were evaluated by comparing the AUC of the original model.

[0033] The step S7 includes: placing the Nomogram model established in step S4 into the validation set data to evaluate the model (ie, ROC curve, calibration curve and DCA curve) to evaluate the performance of the model in the external data set.

[0034] Therefore, the present invention establishes a Nomogram model for voriconazole trough concentration that does not meet the standard, and evaluates the model through ROC curve, calibration curve and DCA curve, performs internal validation of the model through Bootstrap method (n=1000 times); performs external validation of the model through validation set; and finally applies the established Nomogram model to specific clinical patients.

[0035] The second object of the present invention is to provide a risk prediction model for voriconazole trough concentration not meeting the standard obtained by the construction method as described above.

[0036] A risk prediction model for substandard trough concentration of voriconazole, the risk prediction model is a Nomogram model, the independent risk factors used include: LA, dosage, AST, severe pneumonia, tumor, and the calculation formula for the probability of voriconazole-related liver injury predicted by the model is:

[0037] Logit(P)=-5.2753+0.3374×LA+0..6886×Dose+0.0274×AST+1.2312

[0038] ×severe pneumonia+2.0855×tumour

[0039] Among them: P represents the probability of voriconazole trough concentration not reaching the standard; LA represents lactic acid; Dose represents the administered dose; AST represents aspartate aminotransferase; severe pneumonia represents severe pneumonia, with a diagnosis of severe pneumonia recorded as 1 and no diagnosis of severe pneumonia recorded as 0; tumour represents tumor, with a diagnosis of tumor recorded as 1 and no diagnosis of tumor recorded as 0.

[0040] When applied, the Nomogram model disclosed in the present invention is used to input clinical data of clinical patients, obtain the probability of voriconazole-related liver injury in the patient, and draw a case nomogram. Independent risk factors are used as variables, and the scores corresponding to each variable under different values ​​are calculated, and the total score after adding up the scores of all variables is given to predict the probability of voriconazole-related liver injury in each patient.

[0041] Compared with the univariate analysis in the prior art which can only show the influence of a single factor on the trough concentration of voriconazole, the risk prediction model of voriconazole trough concentration not meeting the standard based on Nomogram disclosed in the present invention combines multiple risk factors, and uses LA, dosage, AST, severe pneumonia, and tumor to jointly predict the risk of voriconazole trough concentration not meeting the standard, which provides a convenient and effective tool for clinicians to judge the risk of trough concentration not meeting the standard in patients treated with voriconazole, and is of great significance for the implementation of individualized medication regimens for voriconazole, improving drug efficacy and reducing adverse reactions. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0043] Figure 1The present invention is a flow chart of a method for constructing a risk prediction model for voriconazole trough concentration not meeting the standard.

[0044] Figure 2 This is a Nomogram diagram of the risk prediction model for voriconazole trough concentration not meeting the standard of the present invention (LA represents lactic acid; Dose represents the dosage; AST represents aspartate aminotransferase; severe pneumonia represents severe pneumonia; tumour represents tumor; Points represents the score corresponding to each item; Total Points represents the total score of the patient; Risk represents the risk of voriconazole-related liver injury).

[0045] Figure 3 This is a diagram of the model discrimination evaluation of Example 1 of the present invention (left: training set; right: validation set).

[0046] Figure 4 This is a diagram of the model calibration evaluation of Example 1 of the present invention (left: training set; right: validation set).

[0047] Figure 5 This is a graph of the clinical benefit evaluation of the model in Example 1 of the present invention (left: training set; right: validation set).

[0048] Figure 6 This is an internal verification diagram of the model of Example 1 of the present invention.

[0049] Figure 7 This is a case nomogram of the risk prediction model based on voriconazole trough concentration not meeting the standard in Example 2 of the present invention. DETAILED DESCRIPTION

[0050] The technical solutions in the embodiments of the present invention are described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. 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.

[0051] The word "embodiment" used here as an "exemplary" does not necessarily mean that any embodiment described is superior to or better than other embodiments. Unless otherwise specified, the performance index tests in the embodiments of this application are performed using conventional test methods in the art. It should be understood that the terms described in this application are only used to describe specific implementation methods and are not used to limit the content disclosed in this application.

[0052] Unless otherwise specified, the technical and scientific terms used in this document have the same meanings as commonly understood by ordinary technicians in the technical field to which this application belongs; other experimental methods and technical means not specifically specified in this application refer to experimental methods and technical means commonly used by ordinary technicians in this field.

[0053] In order to better illustrate the content of the present application, numerous specific details are provided in the specific examples below. It should be understood by those skilled in the art that the present application can also be implemented without certain specific details. In the embodiments, some methods, means, instruments, equipment, etc. well known to those skilled in the art are not described in detail in order to highlight the main purpose of the present application.

[0054] Under the premise of no conflict, the technical features disclosed in the embodiments of the present application can be combined arbitrarily, and the resulting technical solutions belong to the contents disclosed in the embodiments of the present application.

[0055] The present invention belongs to the technical field of therapeutic drug monitoring, and specifically relates to a risk prediction model for voriconazole trough concentration not meeting the standard and a construction method thereof, collects clinical data of patients treated with voriconazole and undergoing TDM; randomly divides the data set into a training set and a validation set at a ratio of 7:3; the training set screens independent risk factors for voriconazole trough concentration not meeting the standard by univariate analysis and binary logistic regression analysis; performs correlation analysis on the screened independent risk factors, and incorporates and constructs a Nomogram model; the Nomogram model is evaluated by drawing an ROC curve to evaluate the model discrimination, drawing a calibration curve to evaluate the model calibration, drawing a decision analysis curve to evaluate the model clinical net benefit, and the Bootstrap method is used to perform internal validation of the model; and the validation set is used to perform external validation of the model. The independent risk factors include: LA, dosage, AST, severe pneumonia, and tumors. The present invention provides a convenient and effective tool for clinicians to judge the risk of trough concentration not meeting the standard in patients treated with voriconazole, which is conducive to the implementation of individualized medication of voriconazole.

[0056] In order to better understand the present invention, the present invention is further specifically described below through the following examples, but it should not be understood as a limitation of the present invention. Some non-essential improvements and adjustments made by technicians in this field based on the above invention content are also considered to fall within the protection scope of the present invention.

[0057] Example 1

[0058] A method for constructing a risk prediction model for voriconazole trough concentration not meeting the standard Figure 1 ):

[0059] S1: Systematically collect patient data of patients who used voriconazole and underwent TDM in Ningxia Medical University General Hospital, including demographics, laboratory tests and treatment details; the collected patient data include: In step S1, the patient data include: age, gender, height, weight, BMI (body mass index), smoking history, drinking history, dosage, number of days of medication, drug source, severe pneumonia, infection, hypertension, diabetes, heart disease, sepsis, hypoalbuminemia, tumor, Aspergillus, Candida, mixed infection, infection with unknown pathogens, AST (aspartate aminotransferase), ALT (alanine aminotransferase), ST / LT (aspartate aminotransferase Enzyme / alanine aminotransferase ratio), ALB (plasma albumin), UREA (urea), CREA (creatinine), UA (uric acid), EGFR (glomerular filtration rate), WBC (white blood cell count), NEUT% (neutrophil percentage), LYM% (lymphocyte percentage), NEUT# (neutrophil absolute value), LYM# (lymphocyte absolute value), RBC (red blood cell count), HGB (hemoglobin), PLT (platelet count), LA (lactic acid), CRP (C-reactive protein), PCT (procalcitonin), IL-6 (interleukin-6), antibacterial drugs, proton pump inhibitors, glucocorticoids. It is worth noting that in the present invention, the missing values ​​of the three variables of CRP, PCT, and IL-6 are greater than 25%, so these three variables are eliminated and not included in the subsequent analysis. A total of 176 patients' clinical data were collected.

[0060] S2: According to the inclusion and exclusion criteria established by the present invention, the collected patient data of 176 patients were screened one by one, and a data set was generated for patients who met the criteria. The data set was divided into two groups according to whether the voriconazole blood concentration reached the standard (the standard was defined as: 0.5 mg / L ≤ C min ≤5.0mg / L) were divided into target group and non-target group, and the data set was randomly divided into training set and validation set at a ratio of 7:3; the inclusion and exclusion criteria of the patients were as follows:

[0061] (1) Inclusion criteria:

[0062] a) Treatment with voriconazole;

[0063] b) Age >18 years old;

[0064] c) Conduct voriconazole therapeutic drug monitoring;

[0065] d) The clinical data and examination results are complete;

[0066] (2) Exclusion criteria:

[0067] a) Patients are taking drugs that affect voriconazole concentration;

[0068] b) Patients diagnosed with liver disease at admission, such as cirrhosis, drug-induced liver injury, viral hepatitis, liver cancer, or liver failure;

[0069] c) Patients receiving hemodialysis or other treatments that affect blood volume while taking voriconazole;

[0070] d) non-steady-state trough concentration;

[0071] e) Pregnant or breastfeeding women;

[0072] f) Clinical data and examination results are incomplete.

[0073] According to the inclusion and exclusion criteria, a total of 17 patients were excluded (6 patients underwent CRRT or hemodialysis treatment; 3 patients were diagnosed with liver disease at admission; 4 patients had concurrent drugs that interacted with voriconazole; 3 patients were aged <18 years; 1 patient had non-steady-state trough concentrations). Finally, a data set was generated for 159 patients, 111 patients (69.8%) were used as training sets to build the model, and 48 patients (30.2%) were used as validation sets for external validation. The training set and validation set patients were compared for balance and differences. The results showed that there was no significant difference between the training set and the validation set. The validation set can be used to validate the constructed prediction model. The results are shown in Table 1.

[0074] Table 1 Baseline characteristics and balance test of patients in the training set and validation set

[0075]

[0076]

[0077] Note: BMI: body mass index; AST: aspartate aminotransferase; ALT: alanine aminotransferase; ST / LT: aspartate aminotransferase / alanine aminotransferase ratio; ALB: albumin; UREA: urea; CREA: creatinine; UA: uric acid; eGFR: estimated glomerular filtration rate; WBC: white blood cell count; NEUT%: neutrophil percentage; LYM%: lymphocyte percentage; NEUT#: absolute neutrophil count; LYM#: absolute lymphocyte count; RBC: red blood cell count; HGB: hemoglobin; PLT: platelet count; LA: lactic acid.

[0078] S3: Univariate and binary logistic regression analysis was used to screen out independent risk factors affecting the failure of voriconazole trough concentration to meet the target in patients in the training set.

[0079] Specifically, the S3 steps are as follows:

[0080] S3.1: Univariate analysis: SPSS 27.0 statistical software was used. Qualitative data were expressed as cases (%). Inter-group comparison was performed using χ 2 Normally distributed quantitative data are expressed as mean ± standard deviation. The t test was used; the non-normally distributed quantitative data were expressed as median (quartile) [M(P25, P75)] and the Mann-Whitney U test was used; among them, P < 0.1 was considered to indicate that the independent risk factor had a significant difference between the two groups.

[0081] The results of univariate analysis are shown in Table 2 , among which P < 0.10 for dosage, severe pneumonia, tumor, Aspergillus, Candida, AST, ALT, ST / LT, and LA (P < 0.10) indicated that there were significant differences between the group with trough concentration reaching the target and the group with trough concentration not reaching the target.

[0082] Table 2 Univariate analysis of voriconazole trough concentration not reaching the standard

[0083]

[0084]

[0085] Note: BMI: body mass index; AST: aspartate aminotransferase; ALT: alanine aminotransferase; ST / LT: aspartate aminotransferase / alanine aminotransferase ratio; ALB: albumin; UREA: urea; CREA: creatinine; UA: uric acid; eGFR: estimated glomerular filtration rate; WBC: white blood cell count; NEUT%: neutrophil percentage; LYM%: lymphocyte percentage; NEUT#: absolute neutrophil count; LYM#: absolute lymphocyte count; RBC: red blood cell count; HGB: hemoglobin; PLT: platelet count; LA: lactic acid.

[0086] S3.2: SPSS27.0 statistical software was used for correlation analysis: In order to avoid the influence of collinearity, the risk factors with differences in univariate analysis were subjected to Pearson correlation analysis. The results of the correlation analysis showed that AST and ALT were strongly correlated (Pearson correlation coefficient = 0.85). ALT was excluded, and finally the dosage, severe pneumonia, tumor, Aspergillus, Candida, and AST were included in the binary logistic regression analysis.

[0087] S3.3: SPSS27.0 statistical software was used for binary logistic regression analysis: the results are shown in Table 3. LA, dosage, AST, severe pneumonia, and tumor were independent risk factors for substandard trough concentration of voriconazole (P < 0.05).

[0088] Table 3 Binary Logistic regression analysis of voriconazole trough concentration not reaching the standard

[0089]

[0090] Note: β: regression coefficient; SE(β): standard error of regression coefficient; OR: odds ratio.

[0091] S4: Based on the information obtained in step S3, a Nomogram model was established to predict the risk of voriconazole trough concentration not meeting the standard. The results are shown in Figure 2 ; Specifically: RStudio 4.4.1 was used to construct a Nomogram model; in the constructed Nomogram model, LA, dosage, AST, severe pneumonia, and tumor were used as independent risk factors as variables, and the scores corresponding to each variable at different values ​​were calculated, and the total score after adding up the scores of all variables was calculated to predict the probability of each patient having a voriconazole trough concentration that did not meet the standard.

[0092] S5: Evaluate the Nomogram model established in step S4.

[0093] Specifically, the steps of S5 are as follows:

[0094] S5.1: Use the receiver operating characteristic curve (ROC curve) to evaluate the discrimination of the nomogram model: The ROC curve is used to evaluate the effect of certain indicators in distinguishing two different categories of samples. The curve is drawn with the true positive rate (sensitivity) as the vertical axis and the false positive rate (1-specificity) as the horizontal axis. By drawing the ROC curve and calculating the area under the curve (AUC), the classification or diagnostic effects of different indicators can be compared. The closer the AUC is to 1, the better the classification or diagnostic effect of the indicator. The discrimination evaluation results are shown in Figure 3 The AUC of the established Nomogram model for predicting substandard voriconazole blood concentration was 0.8175, indicating that the Nomogram model had good accuracy and discrimination.

[0095] S5.2: Use the calibration curve to evaluate the calibration of the Nomogram model: The calibration curve is used to evaluate the difference between the model prediction results and the actual results. The curve is drawn with the predicted probability as the horizontal axis and the actual probability as the vertical axis. The gray dotted line (Ideal line) in the figure represents the ideal reference line (the model predicted probability and the actual probability are completely consistent), the dotted line (Apparent line) is the current model curve, and the black solid line (Bias-corrrected line) is the curve calculated by multiple sampling (Bootstrap = 1000 times). Therefore, the higher the consistency between the dotted line and the black solid line, the better the calibration of the Nomogram model. Calibration results are shown in Figure 4, indicating that the model can accurately predict the occurrence of voriconazole trough concentration failure.

[0096] S5.3: Use the decision curve (DCA curve) to evaluate the clinical benefit of the Nomogram model: The DCA curve is a method for evaluating the application value of the prediction model in actual clinical decision-making. The curve is drawn with the threshold probability as the horizontal axis and the net benefit as the vertical axis. The black solid line (None line) in the figure means no intervention for anyone, the gray solid line (All line) means intervention for anyone, and the red solid line (train line) represents the overall net benefit of the Nomogram model within the threshold range. Therefore, the larger the area from the red solid line to the gray solid line and the black solid line, and the wider the range, the better the clinical benefit of the Nomogram model. See the clinical benefit results. Figure 5 , the established Nomogram model showed a larger threshold probability (0.05-1.0), indicating that the Nomogram model has better clinical benefits.

[0097] Thus, a risk prediction model for voriconazole-related liver injury was obtained, and the formula for establishing the model was:

[0098] Logit(P)=-5.2753+0.3374×LA+0..6886×Dose+0.0274×AST+1.2312×severepneumonia+2.0855×tumour

[0099] Among them: P represents the probability of voriconazole trough concentration not reaching the standard; LA represents lactic acid; Dose represents the administered dose; AST represents aspartate aminotransferase; severe pneumonia represents severe pneumonia, with a diagnosis of severe pneumonia recorded as 1 and no diagnosis of severe pneumonia recorded as 0; tumour represents tumor, with a diagnosis of tumor recorded as 1 and no diagnosis of tumor recorded as 0.

[0100] A nomogram model was constructed based on this formula, which has good discrimination, calibration and clinical practicality.

[0101] Furthermore, the present invention also verifies the clinical utility value of the established Nomogram model:

[0102] (1) The Bootstrap method is used to internally validate the Nomogram model. The Bootstrap method generates a large number of virtual “resampled” data sets by repeatedly extracting from the original data set (with replacement), and uses these data sets to simulate possible situations.

[0103] The constructed Nomogram model was bootstraped 1000 times to obtain the AUC of 1000 simulated data sets. The AUC of the ROC curve corresponding to the original model was 0.817±0.040, and that of the original model was 0.8175, indicating that the model had good internal validation results.

[0104] (2) Use the validation set to perform external validation on the Nomogram model. That is, the established Nomogram model is placed in the validation set data to evaluate the model (i.e., ROC curve, calibration curve, and DCA curve) and evaluate the performance of the model in the external data set. The external validation results are shown in Figure 3 , Figure 4 , Figure 5 .

[0105] Therefore, external validation of the constructed Nomogram model showed good discrimination, calibration and clinical practicality.

[0106] Example 2

[0107] A risk prediction model for substandard trough concentrations of voriconazole was applied to the prediction of individual patients.

[0108] Results Figure 7 The patient was admitted to the hospital with a diagnosis of severe pneumonia, LA: 1.1mmol / L, AST: 128.0U / L, dosage: 5.17mg / kg, no tumor diagnosis, and the patient's voriconazole trough concentration was measured to be 6.66mg / L. The established Nomogram model showed a comprehensive score of 88.5 points, and the risk of predicting that the voriconazole blood concentration would not meet the standard was 93.2%. The results showed that the constructed Nomogram model had a high prediction accuracy for patients with voriconazole trough concentrations that did not meet the standard.

[0109] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for constructing a risk prediction model for substandard trough concentration of voriconazole, characterized in that: include: Clinical data of patients treated with voriconazole and undergoing TDM were collected; the data set was randomly divided into a training set and a validation set at a ratio of 7:3; the training set was screened for independent risk factors for substandard voriconazole trough concentrations through univariate analysis and binary logistic regression analysis; the screened independent risk factors were subjected to correlation analysis, and independent risk factors with low correlation were included and a Nomogram model was constructed; the Nomogram model was evaluated for model discrimination by drawing ROC curves, model calibration by drawing calibration curves, and model clinical net benefit by drawing decision analysis curves, and the model was internally validated by the Bootstrap method; the model was externally validated using the validation set.

2. The construction method according to claim 1, characterized in that: The independent risk factors include: LA, dosage, AST, severe pneumonia, and tumor.

3. The construction method according to claim 2, characterized in that: The specific steps include: S1: Systematic collection of patient data, including demographics and laboratory test and treatment details; S2: According to the established inclusion and exclusion criteria, the collected patient data were screened one by one, and a data set was generated for patients who met the criteria. The data set was divided into a target group and a non-target group according to whether the voriconazole blood concentration reached the target. The data set was then randomly divided into a training set and a validation set at a ratio of 7:3; S3: Univariate and binary logistic regression analysis was used to screen out independent risk factors affecting the failure of voriconazole trough concentration to reach the target in the target group and non-target group in the training set; S4: Based on the information obtained in step S3, a nomogram model is established to predict the risk of substandard trough concentration of voriconazole.

4. The construction method according to claim 3, characterized in that: In step S1, the patient information includes: age, gender, height, weight, BMI, smoking history, drinking history, dosage, medication days, drug source, severe pneumonia, infection, hypertension, diabetes, heart disease, sepsis, hypoalbuminemia, tumor, Aspergillus, Candida, mixed infection, unknown pathogens, AST, ALT, ST / LT, ALB, UREA, CREA, UA, EGFR, WBC, NEUT%, LYM%, NEUT#, LYM#, RBC, HGB, PLT, LA, CRP, PCT, IL-6, antibacterial drugs, proton pump inhibitors, and glucocorticoids.

5. The construction method according to claim 3, characterized in that: In step S2, the standard is defined as: 0.5 mg / L ≤ C min ≤5.0mg / L.

6. The construction method according to claim 5, characterized in that: In step S2, the inclusion and exclusion criteria established include: Inclusion criteria: treatment with voriconazole; age >18 years old; voriconazole treatment drug monitoring; complete clinical data and examination results; Exclusion criteria: patients taking drugs that affect voriconazole concentration; patients diagnosed with liver diseases such as cirrhosis, drug-induced liver injury, viral hepatitis, liver cancer, or liver failure at admission; patients taking hemodialysis or other treatments that affect blood volume while taking voriconazole; non-steady-state trough concentration; pregnant or lactating women; incomplete clinical data and examination results.

7. The construction method according to claim 3, characterized in that: The step S3 includes: S31: Data were statistically analyzed using SPSS 27.0 statistical software. Qualitative data were expressed as cases (%) and inter-group comparisons were performed using χ 2 Test; Normally distributed quantitative data are expressed as mean ± standard deviation The t-test was used; the non-normally distributed quantitative data were expressed as median (quartile) [M(P25, P75)], and the Mann-Whitney U test was used; finally, 9 risk factors were screened out, including dosage, severe pneumonia, tumor, Aspergillus, Candida, AST, ALT, ST / LT, and LA (P < 0.10). S32: SPSS27.0 was used to perform Pearson correlation analysis on the risk factors screened in step S31; the correlation analysis results showed that the Pearson correlation coefficient of AST and ALT was 0.

85. ALT was eliminated and finally the dosage, severe pneumonia, tumor, Aspergillus, Candida, and AST were included in the binary logistic regression analysis. S33: SPSS27.0 was used to perform binary logistic regression analysis on the risk factors screened in step S32; the results showed that the independent risk factors included: LA, dosage, AST, severe pneumonia, and tumor (P < 0.05).

8. The construction method according to claim 3, characterized in that: In step S4, the independent risk factors are incorporated into the Nomogram model based on the results of binary logistic regression analysis, and RStudio 4.4.1 is used to construct the Nomogram model; in the constructed Nomogram model, the five independent risk factors are used as variables, and the scores corresponding to each variable at different values ​​are calculated, and the total score after adding up the scores of all variables is calculated to predict the probability of each patient having a trough concentration that does not meet the standard.

9. The construction method according to claim 2, characterized in that: The specific steps also include: S5: internal evaluation of the Nomogram prediction model established in step S4, including: ROC curve analysis, calibration curve analysis, DCA curve analysis; S6: internally verify the Nomogram prediction model established in step S4 by using the Bootstrap method (n=1000 times); S7: Use the validation set to perform external evaluation on the Nomogram prediction model established in step S4.

10. The risk prediction model for voriconazole trough concentration not reaching the standard obtained by the construction method according to any one of claims 1 to 9, characterized in that: The risk prediction model is a Nomogram model, and the independent risk factors used include: LA, dosage, AST, severe pneumonia, and tumors. The calculation formula for the probability of voriconazole-related liver injury predicted by the model is: Logit(P)=-5.2753+0.3374×LA+0..6886×Dose+0.0274×AST+1.2312 ×severe pneumonia+2.0855×tumour Where: P represents the probability of voriconazole trough concentration not reaching the target; LA represents lactic acid; Dose represents the dosage; AST stands for aspartate aminotransferase; severe pneumonia stands for severe pneumonia, with a diagnosis of severe pneumonia recorded as 1 and no diagnosis of severe pneumonia recorded as 0; tumour stands for tumor, with a diagnosis of tumor recorded as 1 and no diagnosis of tumor recorded as 0.

Images