Risk prediction model for predicting limit exceeding of steady-state blood concentration of linezolid and construction method

Abstract

The invention belongs to the technical field of clinical drug treatment monitoring, and discloses a risk prediction model for predicting the limit exceeding of the steady-state blood concentration of linezolid, and the model is a column diagram model constructed based on multi-factor Logistic regression analysis. The risk probability calculation module is used for calculating the risk probability that the blood concentration is greater than or equal to 8 mg / L according to clinical indexes of a patient; the input variables of the model comprise age, albumin level and serum creatinine level; the albumin level and the serum creatinine level are the albumin level and the serum creatinine level detected last time before blood concentration monitoring of the patient. The method is constructed based on real world data and verified by an independent queue, the model shows good distinguishing ability and calibration degree, and the prediction accuracy is good. The model only contains three clinically conventional easily available indexes of age, albumin and serum creatinine, and is presented in the form of a column diagram without complex calculation, so that clinical doctors can quickly complete risk assessment at the bedside.

Description

Technical Field

[0001] This invention relates to the field of clinical drug therapy monitoring technology, and more specifically, to a risk prediction model and construction method for predicting the exceedance of linezolid steady-state blood drug concentration. Background Technology

[0002] Linezolid is an oxazolidinone antibiotic with good antibacterial activity against a variety of Gram-positive bacteria, including methicillin-resistant Staphylococcus aureus (MRSA), and is widely used in the treatment of complicated skin and soft tissue infections, hospital-acquired pneumonia, and other severe infections. However, its clinical application is limited by dose-related adverse reactions, such as thrombocytopenia, anemia, and peripheral neuropathy. Studies have confirmed that the steady-state trough concentration (C0.05) of linezolid is... min ( ) is a key indicator for assessing its exposure and toxicity risk. When C min When the concentration is consistently above 8 mg / L, the risk of the above-mentioned adverse reactions increases significantly.

[0003] Therefore, therapeutic drug monitoring (TDM) is an important means of optimizing the efficacy and safety of linezolid. However, routine TDM has a time lag, and universal testing for all patients is not cost-effective. Currently, there is a lack of practical predictive tools in clinical practice that can quickly and intuitively identify individuals at high risk of bleeding drug concentration exceeding limits based on readily available baseline characteristics at the start of treatment or early in the treatment period.

[0004] A nomogram is a graphical prediction tool based on a multivariate regression model. It integrates multiple predictive factors into a simple scoring system, visually displaying the probability of an individual experiencing a specific clinical event. This model is well-established in areas such as tumor prognosis and disease diagnosis, but its application in predicting the risk of individualized use of antibacterial drugs, particularly linezolid, has not yet been reported.

[0005] Therefore, a risk prediction model and construction method for predicting linezolid steady-state blood drug concentration exceeding the limit are proposed to provide decision support for clinical practice. Summary of the Invention

[0006] The purpose of this invention is to address the shortcomings of existing technologies by proposing a risk prediction model and construction method for predicting linezolid steady-state blood concentration exceeding the limit.

[0007] To solve the above problems, the present invention adopts the following technical solution: A risk prediction model for linezolid steady-state plasma concentration exceeding the limit is provided. This model is a nomogram model constructed based on multivariate logistic regression analysis, used to calculate the probability of plasma concentration ≥8 mg / L based on the patient's clinical indicators. The input variables for the model include age, albumin level, and serum creatinine level.

[0008] As a further aspect of the present invention: the albumin level and serum creatinine level are the albumin level and serum creatinine level most recently measured before the patient's blood drug concentration monitoring.

[0009] As a further aspect of this invention: the model is presented in the form of a visual nomogram. Different values ​​of each input variable in the nomogram correspond to different scores, and the sum of all variable scores corresponds to a total score. This total score is mapped to the risk probability axis, allowing direct reading of the predicted probability of a blood drug concentration ≥8 mg / L. In clinical use, doctors can find the corresponding points on each variable axis of the nomogram based on the patient's specific age, albumin level, and serum creatinine value, projecting them to obtain individual scores. The three scores are then added together to obtain the total score. Finally, the corresponding predicted probability value is found on the total score table, which represents the estimated risk of the patient's linezolid blood drug concentration ≥8 mg / L.

[0010] This invention also provides a method for constructing the above-mentioned prediction model, comprising the following steps: S1. Data collection: Collect data on patients who used linezolid and underwent steady-state blood drug concentration monitoring. Based on the monitoring results, patients were divided into a control group with blood drug concentration <8 mg / L and an excess group with blood drug concentration ≥8 mg / L, using 8 mg / L as the cutoff value. S2. Cohort establishment: All patient samples that meet the inclusion and exclusion criteria are randomly divided into a training set for model building and a validation set for model validation according to a predetermined ratio. S3. Variable Screening: For the training set data, univariate analysis was performed using statistical test methods applicable to continuous and categorical variables to screen out potential influencing factors related to blood drug concentration exceeding the limit. S4. Model Construction: Variables with a significance level P value ≤ 0.1 in the univariate analysis were included in the multivariate logistic regression analysis to determine the independent risk predictors of blood drug concentration exceeding the limit and their regression coefficients. S5. Nodal plot generation: Using statistical software, based on the results of the multivariate logistic regression analysis, the regression coefficients of each independent predictor are converted into a visual nodal plot, completing the initial construction of the prediction model. S6. Model Validation and Evaluation: Using the data from the validation set, the model's discrimination is evaluated by plotting receiver operating characteristic (ROC) curves and calculating the area under the curves, the model's calibration is evaluated by plotting calibration curves, and the model's clinical applicability and net benefit at different risk thresholds is evaluated by plotting decision curves.

[0011] As a further aspect of the present invention: in step S1, the collected data includes demographic data, medical history, comorbidities, medication information, and laboratory test indicators.

[0012] The collected patient data should include at least: demographic information, clinical diagnosis and comorbidity information, linezolid dosing regimen, and laboratory test results during blood drug concentration monitoring.

[0013] As a further aspect of the present invention: In step S3, for continuous variables, an independent samples t-test or a Mann-Whitney U test is used for univariate analysis based on whether they conform to a normal distribution; for categorical variables, a χ² test is used. 2 Perform a one-way analysis using the Fisher exact test or the test of exactness.

[0014] As a further aspect of the present invention: in step S2, the ratio of the training set to the validation set is 7:3.

[0015] Compared with the prior art, the advantages of this invention are: 1. This invention is built upon real-world data and validated through independent cohorts. The model demonstrates good discriminative power and calibration, with high predictive accuracy. The model includes only three clinically readily available indicators: age, albumin, and serum creatinine, presented in nomogram form. No complex calculations are required, facilitating rapid risk assessment by clinicians at the bedside.

[0016] 2. The model of this invention can identify high-risk patients with excessive bleeding drug concentrations at an early stage, which helps clinicians to strengthen TDM or adjust the dosing regimen in a targeted manner, realizing the transformation from "passive monitoring" to "active prediction", and is expected to reduce the incidence of linezolid-related adverse reactions and improve medication safety.

[0017] 3. This invention is the first to construct and validate a nomogram model for predicting the risk of linezolid blood concentration exceeding the limit, providing a new and easily promoted decision support tool for individualized precision dosing of linezolid. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the screening and queue division of research objects in an embodiment of the present invention; Figure 2 This is a nomogram for predicting the risk of exceeding the steady-state blood concentration limit of linezolid constructed in an embodiment of the present invention; Figure 3 Receiver operating characteristic curves for the training and validation sets; Figure 4 Model calibration curves for the training and validation sets; Figure 5 This is a graph showing the decision curves for the training and validation sets. Detailed Implementation

[0019] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0020] Example 1: Establishment and validation of a risk prediction model for linezolid steady-state plasma concentration ≥8 mg / L This embodiment follows the standard procedure for predictive model research. Through a retrospective study design, a nomogram model for predicting the risk of linezolid exceeding blood concentration limits was constructed and validated. The specific steps are as follows: 1. Research subjects and data sources This was a single-center retrospective study, with data sourced from the electronic medical record system of inpatients at a tertiary teaching hospital from January 2019 to December 2022. The study included all adult patients who received linezolid and underwent therapeutic drug monitoring. According to the Chinese expert consensus on linezolid, when monitoring blood drug concentrations, 8 mg / L was used as the cutoff; blood drug concentrations within the range of 2-8 mg / L were considered normal, while those above this range were considered at risk of exceeding the limit.

[0021] Inclusion criteria for cases: ① Age ≥ 18 years; ② Received intravenous treatment with linezolid for suspected or confirmed Gram-positive bacterial infection; ③ The dosing regimen was the standard dose (600 mg every 12 hours); ④ Blood samples were collected within 30 minutes before the next dose after at least 3 consecutive doses, and the steady-state trough concentration was measured; ⑤ Complete clinical medical records.

[0022] Exclusion criteria for cases: ① patients allergic to linezolid; ② pregnant or lactating women; ③ patients using non-standard dosing regimens; ④ patients with missing key clinical or laboratory data.

[0023] 2. Queue creation and data partitioning Initial screening yielded data from 217 patients. Based on the aforementioned criteria, after excluding patients with incorrect age, sample type, medication regimen, or who were not hospitalized, 172 patients were ultimately included in the analysis. Using computer-generated random numbers, all eligible samples were randomly divided into two independent cohorts at a ratio of 7:3. Training set (n=121): used to select predictor variables, build regression models, and generate nomograms; Validation set (n=51): used to evaluate the predictive performance of the model on independent data; the screening and division process for the two groups of patients is as follows. Figure 1 .

[0024] 3. Data Collection and Variable Definition The following data will be collected through the hospital information system: ① Baseline population data: age, sex, height, weight; ②Past medical history: history of allergies, smoking, and alcohol abuse; ③The patient's primary disease; ④ Accompanying diseases: hypertension, diabetes, coronary heart disease, arthritis, cerebral infarction, heart disease, bronchitis, hepatitis B, chronic digestive tract diseases, tuberculosis, chronic obstructive pulmonary disease, schizophrenia, thyroid disease, tumors; ⑤ Hospitalization details: Department of admission, length of stay, number of surgeries and procedures; ⑥ Medication usage during patient blood drug concentration monitoring: linezolid (dosage, frequency of administration, route of administration, duration of treatment); ⑦ Laboratory test indicators before medication: alanine aminotransferase (ALT), aspartate aminotransferase (AST), alkaline phosphatase (ALP), total bilirubin (TBIL), serum albumin (ALB), uric acid (UA), creatinine (Scr), anion gap (AG), cystatin C (Cys-C), C-reactive protein (CRP), procalcitonin (PCT), white blood cell count (WBC), neutrophil count (ANC), platelet count (PLT); ⑧ Outcome indicator: Monitoring value of linezolid blood concentration.

[0025] 4. Statistical methods Statistical analysis was performed using SPSS 25.0 and R 4.2.0 software. Continuous variables with a normal distribution were expressed as mean ± standard deviation, and independent samples t-tests were used for comparisons between groups. Variables without a normal distribution were expressed as median (interquartile range), and Mann-Whitney U tests were used for comparisons between groups. Categorical variables were expressed as percentages, and chi-square tests were used for comparisons between groups. 2 Test or Fisher's exact test.

[0026] 5. Model Building (Based on Training Set) (1) Univariate analysis: Preliminary screening of variables that may be related to excessive blood drug concentration. For continuous variables (using t-test or U-test based on distribution) and categorical variables (χ² test), further analysis was conducted. 2 The analysis was performed using the Fisher test (or p-value ≤ 0.1), and variables with p-values ​​≤ 0.1 were considered potential predictors and included in subsequent multivariate analyses.

[0027] (2) Multivariate Logistic Regression Analysis: The forward stepwise method (likelihood ratio test) was used to include the variables selected by the univariate analysis into the binary logistic regression model to control for confounding among variables. Finally, variables with a p-value ≤ 0.05 in the multivariate analysis were retained as independent predictors, and their regression coefficients (β values) and odds ratios (OR) were calculated.

[0028] 6. Generating Column Charts Nodal plot generation: Using the "rms" package in R, the results of the final multivariate logistic regression model are visualized, and a prediction nomal plot is constructed. Each predictor is assigned a different score range based on its β value, with the total score corresponding to the risk probability axis at the bottom.

[0029] 7. Model Validation and Evaluation (Based on Validation Set) The validation set patient data was fed into the constructed nomogram model to calculate the predicted risk probability for each patient, and the model performance was evaluated from the following three aspects: ① Discrimination: Plot the receiver operating characteristic curve (ROC) and calculate the area under the curve (AUC) of the 95% confidence interval (CI) of the ROC curve to assess the discrimination ability of the nomogram. The higher the AUC value (closer to 1), the stronger the model's ability to distinguish between high-risk and low-risk patients.

[0030] ② Calibration accuracy: Plot a Logistic calibration curve and use the Hosmer-Lemeshow goodness-of-fit test to evaluate the consistency between the model's predicted probabilities and the actual observed probabilities. The closer the curve is to the diagonal, the better the calibration accuracy.

[0031] ③ Clinical applicability: Decision curve analysis (DCA) was performed to calculate the standardized net benefit of applying this predictive model to guide clinical decision-making (such as enhanced monitoring of high-risk patients) compared to the strategies of "intervention for all patients" or "no intervention for all patients" at different risk threshold probabilities.

[0032] 8. Research Results 8.1 Baseline Characteristics The baseline characteristics of the training set and the validation set are compared in Table 1. There were no statistically significant differences between the two groups of patients in terms of age, gender, weight, comorbidities and key laboratory indicators (such as albumin and serum creatinine) (all P ≥ 0.05), indicating that the randomization was balanced and the validation set could be used for objective evaluation of the model.

[0033]

[0034] 8.2 Univariate analysis of factors influencing blood drug concentration exceeding limits The univariate analysis results in the training set (as shown in Table 2) showed that 50 patients had linezolid plasma concentrations exceeding the limit (≥8 mg / L) (30 in the training set and 20 in the validation set), while 122 patients had plasma concentrations below 8 mg / L (91 in the training set and 31 in the validation set), representing 29.07% of the total. The mean ages of patients in the training and validation sets were 59.09 ± 19.46 and 62.12 ± 19.2, respectively. Univariate analysis results showed that age, comorbid cerebral infarction, comorbid hepatitis B, albumin level, cystatin C level, and serum creatinine level were significantly associated with linezolid plasma concentration exceeding the limit (P≤0.05), and coronary heart disease also showed a potential association (P=0.063).

[0035]

[0036] 8.3 Multivariate Logistic Analysis of Factors Affecting Exceeding Blood Drug Concentration Limits The seven variables with P≤0.1 were included in the multivariate logistic regression analysis. The results (Table 3) showed that age and serum creatinine were independent risk factors for excessive blood drug concentration (P≤0.01). Although albumin did not reach conventional statistical significance in the multivariate analysis (P=0.172), given that numerous pharmacokinetic studies have confirmed it as a key physiological factor affecting the distribution and clearance of linezolid, it was included in the final model along with age and serum creatinine to enhance the clinical rationality and biological interpretability of the model.

[0037]

[0038] 8.3 Establishment of Nodal Chart Model Based on a logistic regression model that includes three predictive factors—age, albumin, and serum creatinine—a nomogram predicting the risk of linezolid exceeding blood concentration limits was generated. Figure 2 In clinical use, doctors can find the corresponding points in the graph based on the patient's age, albumin, and serum creatinine values, project them vertically upwards onto the "Score" axis to obtain individual scores, add the three scores together to get the total score, and finally project the total score vertically downwards onto the "Risk Probability" axis at the bottom to directly read the patient's individualized predicted risk probability.

[0039] 8.4 Model Validation Results The model performance was comprehensively evaluated on an independent validation set: Discrimination: such as Figure 3As shown, the area under the ROC curve (AUC) of the training set is 0.837 (95% CI: 0.762-0.912), and the AUC of the validation set is 0.723 (95% CI: 0.583-0.863), indicating that the model has good discriminative ability.

[0040] Calibration degree: such as Figure 4 As shown, the calibration curves of both the training and validation sets closely follow the ideal diagonal, and the Hosmer-Lemeshow test shows no statistical significance (P≥0.05), indicating that the model's predicted probability is highly consistent with the actual observed out-of-caliber frequency, demonstrating excellent calibration.

[0041] Clinical applicability: such as Figure 5 As shown, the decision curve analysis reveals that, within a threshold probability range of approximately 0.03 to 0.6, the net benefit of using this model to guide clinical decision-making (i.e., strengthening monitoring of those predicted to be at high risk) is higher than the strategies of "intervening in all patients" or "not intervening in all patients," demonstrating the model's clinical application value.

[0042] In summary, this invention successfully constructed and validated a nomogram model incorporating three indicators—age, albumin, and serum creatinine—for predicting the risk of steady-state linezolid blood concentrations ≥8 mg / L. This model is reliable and easy to use, providing clinicians with an effective tool for rapidly and intuitively identifying high-risk patients with elevated linezolid concentrations in the early stages of linezolid treatment. This facilitates prospective management of individualized therapeutic drug monitoring and optimizes treatment safety.

[0043] The above description is merely a preferred embodiment of the present invention; however, the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and its improved concepts, should be covered within the scope of protection of the present invention.

Claims

1. A risk prediction model for linezolid steady-state plasma concentration exceeding the limit, characterized in that: The model described is a nomogram model constructed based on multivariate logistic regression analysis, used to calculate the risk probability of blood drug concentration ≥8 mg / L based on the patient's clinical indicators. The input variables for the model include age, albumin level, and serum creatinine level.

2. The risk prediction model for predicting linezolid steady-state blood concentration exceeding the limit according to claim 1, characterized in that: The albumin level and serum creatinine level mentioned are the albumin level and serum creatinine level most recently measured before the patient's blood drug concentration monitoring.

3. The risk prediction model for predicting linezolid steady-state blood concentration exceeding the limit according to claim 1, characterized in that: The model is presented in the form of a visual nomogram. Different values ​​of each input variable in the graph correspond to different scores. The sum of all variable scores corresponds to a total score. This total score is mapped to the risk probability axis, and the predicted probability of the corresponding blood drug concentration ≥8 mg / L can be read directly.

4. A method for constructing a risk prediction model for linezolid steady-state blood concentration exceeding the limit as described in any one of claims 1-3, characterized in that, Includes the following steps: S1. Data collection: Collect data on patients who used linezolid and underwent steady-state blood drug concentration monitoring. Based on the monitoring results, patients were divided into a control group with blood drug concentration <8 mg / L and an excess group with blood drug concentration ≥8 mg / L. S2. Queue establishment: The overall sample is randomly divided into a training set for model building and a validation set for model validation according to a predetermined ratio; S3. Variable Screening: For the training set data, univariate analysis was performed using statistical test methods applicable to continuous and categorical variables to screen out potential influencing factors related to blood drug concentration exceeding the limit. S4. Model Construction: Variables with a significance level P value ≤ 0.1 in the univariate analysis were included in the multivariate logistic regression analysis to determine the independent risk predictors of blood drug concentration exceeding the limit and their regression coefficients. S5. Nodal plot generation: Based on the results of the multivariate logistic regression analysis, the regression coefficients of each independent predictor are converted into a visual nodal plot, completing the initial construction of the prediction model; S6. Model Validation and Evaluation: Using the data from the validation set, the model's discrimination is evaluated by plotting receiver operating characteristic (ROC) curves and calculating the area under the curves, the model's calibration is evaluated by plotting calibration curves, and the model's clinical applicability and net benefit at different risk thresholds is evaluated by plotting decision curves.

5. The method for constructing a risk prediction model for linezolid steady-state blood concentration exceeding the limit according to claim 4, characterized in that: In step S1, the collected data includes demographic information, medical history, comorbidities, medication information, and laboratory test results.

6. The collected patient data shall include at least: Demographic data, clinical diagnosis and comorbidity information, linezolid dosing regimen, and laboratory test results during blood drug concentration monitoring.

7. The method for constructing a predictive model for the risk of exceeding the steady-state blood concentration limit of linezolid according to claim 4, characterized in that: In step S3, for continuous variables, univariate analysis is performed using either the independent samples t-test or the Mann-Whitney U test, depending on whether the variable conforms to a normal distribution; for categorical variables, the χ² test is used. 2 Perform a one-way analysis using the Fisher exact test or the test of exactness.

8. The method for constructing a risk prediction model for linezolid steady-state blood concentration exceeding the limit according to claim 4, characterized in that: In step S2, the ratio of the training set to the validation set is 7:3.

Images