Prediction method for dosage of anti-tuberculosis drug isoniazide

Abstract

The invention provides a method for predicting the dosage of an antituberculous drug, which comprises the following steps of: establishing a group pharmacokinetic model of the antituberculous drug by adopting a nonlinear mixed effect model, a two-atrioventricular model and a mixed residual model through plasma concentration analysis, and calculating the pharmacokinetic parameters of the corresponding model; and the established model is verified based on a graphical method, a visual prediction test method and a nonparametric bootstrap method, and the steps of stability and prediction capability evaluation and the like are completed. According to the anti-tuberculosis drug population pharmacokinetic model established by the invention, individual pharmacokinetic parameters can be estimated, a specific population administration scheme is optimized, the drug curative effect is improved, the adverse reaction risk is reduced, and individualized administration of the anti-tuberculosis drug is realized.

Description

Technical Field

[0001] The present invention relates to the field of pharmaceutical information technology, and particularly relates to a method for predicting the dosage of isoniazid. Background Art

[0002] Tuberculosis (TB) poses a serious threat to global public health. In 2021, there were approximately 10.6 million new TB infection cases globally, with an incidence rate of 134 per 100,000 people. Isoniazide (INH), as an important part of first-line anti-tuberculosis drugs, has high selectivity and strong antibacterial effects against Mycobacterium tuberculosis and is applicable to the treatment of various types of tuberculosis.

[0003] Isoniazide is rapidly and completely absorbed (>90%) after oral administration, and the peak plasma concentration is reached 1-2 hours after administration. It is mainly metabolized by liver N-acetyltransferase 2 (NAT2). Although the World Health Organization has issued tuberculosis treatment guidelines and dosage recommendations based on body weight (WT), due to the large pharmacokinetic variability of isoniazid, there are significant differences in drug exposure among individuals. Studies have found that among patients given standard doses, the difference in isoniazid plasma concentration can reach 3-7 times, which is closely related to adverse outcomes such as delayed sputum conversion, treatment failure, and even drug resistance in tuberculosis treatment. Pharmacokinetic studies have found that NAT2 gene polymorphism, age, gender, food, drug-drug interaction, nutritional status, and comorbidities are significant factors affecting the pharmacokinetic variability of isoniazid, but there is currently no consensus.

[0004] Population pharmacokinetic modeling is a method for quantifying drug-related PK parameters and determining potential covariates that may affect PK variability. Combined with Bayesian prediction, it can be used for drug dosage prediction and optimization of dosing regimens for specific populations. Therefore, the present invention uses the nonlinear mixed effects model method to establish a population pharmacokinetic model of isoniazid in tuberculosis patients, determine significant factors affecting the pharmacokinetic variability of isoniazid, and develop individualized dosing regimens for patients with different characteristics to ensure the effectiveness and safety of isoniazid clinical use. Summary of the Invention

[0005] The first aspect of the present invention provides a method for predicting the dosage of an anti-tuberculosis drug, which includes the following steps:

[0006] (1) Data collection: Collect the clinical information of patients and collect plasma samples. The plasma samples are detected by the LC-MS / MS method to obtain the plasma concentration data of the target drug;

[0007] (2) Basic model: Using the nonlinear mixed-effect modeling software NONMEM, a pharmacokinetic structural model of anti-tuberculosis drugs was established through a one-compartment or two-compartment configuration model with first-order absorption and elimination, and the pharmacokinetic parameters of the corresponding model were calculated;

[0008] The NONMEM program is ADVAN2-TRANS2 or ADVAN4-TRANS4;

[0009] The random effects in population pharmacokinetics include inter-individual variation and intra-individual variation (residual variation). The inter-individual variation of pharmacokinetic parameters is expressed by formula (1):

[0010] P i =P pop *e ηi Formula (1)

[0011] Among them, P i represents the pharmacokinetic parameters of the i-th patient, P pop is the typical value of the pharmacokinetic parameter, ηi is the inter-individual variation of the i-th patient, ηi has a mean of 0 and a variance of ω 2 Normal distribution;

[0012] Formula (2), formula (3) or formula (4) is used to express the intra-individual variation (residual variation) of pharmacokinetic parameters:

[0013] Additive type: C obs,ij =C pred,ij +ε ij,add Formula (2)

[0014] Proportional type: C obs,ij =C pred,ij *(1+ε ij,prop ) Formula (3)

[0015] Hybrid: C obs,ij =C pred,ij *(1+ε ij,prop )+ε ij,add Formula (4)

[0016] Among them C obs,ij and C pred,ij are the observed and predicted values of the plasma concentration of the i-th subject at the j-th time point, ε ij,prop and ε ij,add Represents proportional or additive intra-individual variation, with mean 0 and variance σ1, respectively. 2 and σ2 2 Normal distribution;

[0017] (3) Covariate model: Screening potential covariates based on physiological and clinical rationality, investigating the effects of different covariates on pharmacokinetic parameters using the likelihood ratio, and constructing a population pharmacokinetic model;

[0018] (4) Model evaluation: Evaluating the predictive ability and stability of the population pharmacokinetic model through goodness-of-fit plots, visual predictive checks, and non-parametric bootstrap methods;

[0019] (5) Model-guided prediction of anti-tuberculosis drug dosage: Based on the established population pharmacokinetic model of the drug, combined with Monte Carlo simulation, formulating individualized dosage recommendation plans for tuberculosis patients with different characteristics.

[0020] Preferably, in the step (1), the patients are tuberculosis patients who are over 18 years old, have no serious underlying diseases, are not pregnant, and are not allergic to or intolerant to isoniazid.

[0021] Preferably, in the step (1), the clinical information of the patients includes demographic characteristics, disease status indicators, and laboratory test results.

[0022] Preferably, in the step (1), the method for collecting plasma samples is as follows: After the patients have taken the medicine for more than one week and the drug concentration has reached a steady state, plasma samples are collected half an hour before dosing and at 1, 2, 4, 6, 12, and 23 h after dosing.

[0023] Preferably, in the step (1), the method for processing plasma samples is as follows: Within 2 hours after blood collection, centrifuge at 3000 rmp / min at 4°C for 10 min, separate the upper serum, and store it at -80°C for analysis.

[0024] Preferably, in the step (1), the chromatographic conditions of the LC-MS / MS method are as follows: The chromatographic column specifications are: ZORBAX PRHDSB-Aq (3×50 mm, 1.8 μm, Agilent, USA), the flow rate: 0.3 mL / min, the injection volume: 10 μL, the column temperature: 40°C, using a binary mobile phase system with mobile phase A being 0.1% formic acid aqueous solution and mobile phase B being methanol. The gradient elution of the mobile phase is shown in the following table:

[0025] Time (min) ％A ％B 0 90 10 1.50 90 10 3.00 10 90 4.50 10 90 4.51 90 10 6.00 90 10

[0026] Preferably, in the step (1), the mass spectrometry conditions of the LC-MS / MS method are as follows: The mass spectrometer is a triple quadrupole Xevo TQ-S IVD, operating in the positive ionization mode, and the detection of INH is based on multiple reaction monitoring of m / z 138.1→121.1.

[0027] Preferably, in the step (1), the detection limit of the LC-MS / MS method is 0.2 μg / mL, the linear range is 0.2 - 40 μg / mL, the recovery rate is 97.2% - 101.2%, and the RSD within day and between days ≤ 10%.

[0028] Preferably, in the step (2), the estimation method of pharmacokinetic parameters is the first-order conditional estimation method with ε-η interaction.

[0029] Preferably, in the step (2), the selection of the basic model is determined by comparing the differences in the objective function value OFV, the accuracy of parameter estimation, and the visual inspection of the goodness-of-fit plot.

[0030] Preferably, the covariates in the step (3) include age, gender, weight, creatinine clearance rate, total bilirubin, direct bilirubin, indirect bilirubin, alanine aminotransferase, aspartate aminotransferase, serum urea nitrogen, albumin, and white blood cell level, etc., as well as concomitant medications and complications.

[0031] Preferably, in the step (3), the method for constructing the covariate model is the stepwise method, including the forward inclusion and backward elimination processes; in the forward inclusion process, covariates are added to the basic model one by one. If the introduction of a covariate causes the objective function value OFV of the model to decrease by more than 3.84 (P < 0.05), then this covariate is retained in the model; this process is repeated until no more covariates can be included; in the backward elimination process, one covariate is removed each time. When the removal of one covariate causes the objective function value OFV of the model to increase by more than 6.64 (P < 0.01), then this factor is considered to have significant significance and is retained in the model.

[0032] Preferably, in the step (4), the goodness-of-fit plot includes the fitting scatter plot between the observed values and the individual predicted values and the population predicted values, and the fitting scatter plot between the conditional weighted residuals and time and the population predicted values.

[0033] Preferably, in the step (4), the visual predictive check is to simulate 1000 data sets using the parameters of the final model, and compare the 5th and 95th percentiles and the median of the observed data with the corresponding percentiles and the median of the simulated data.

[0034] Preferably, in the step (4), the non-parametric bootstrap test is to generate 1000 repeated bootstrap data sets from the original data through random resampling, and apply the final model to the regenerated data sets to repeatedly calculate the PK parameters, and compare the median, 2.5 and 97.5 percentiles of the PK parameters estimated from the resampled data sets with the corresponding values estimated by the final model.

[0035] Preferably, in step (5), the Monte Carlo simulation uses the established population pharmacokinetic model to set up virtual patient populations with different characteristics, simulate the concentration-time distribution curves of the drug at different dosing doses, and obtain the peak concentration C of the drug max , and determine the dosing dose that makes the proportion of drug C max lower than the standard treatment range of 3-6 μg / ml less than 20% and the proportion within the treatment range greater than 50%.

[0036] Preferably, the tuberculosis patients with different characteristics in step (5) are: classified into fast metabolizers and slow metabolizers according to the gene polymorphism of the drug, and the weight ranges are set to 35-57 kg, 58-70 kg, and 71-100 kg.

[0037] Preferably, the anti-tuberculosis drug is isoniazid.

[0038] Preferably, the prediction method is for non-diagnostic and non-treatment purposes.

[0039] The second aspect of the present invention provides an anti-tuberculosis drug dosing prediction device, and the prediction device includes:

[0040] (1) A data collection module for collecting the clinical information and plasma samples of patients, and detecting by the LC-MS / MS method to obtain the blood drug concentration data of the target drug;

[0041] (2) A basic model construction module for using the NONMEM program of the non-linear mixed effect model modeling software to establish a pharmacokinetic structure model of the anti-tuberculosis drug through a one-compartment or two-compartment configuration model with first-order absorption and elimination, and calculating the pharmacokinetic parameters of the corresponding model;

[0042] The NONMEM program is ADVAN2-TRANS2 or ADVAN4-TRANS4;

[0043] The random effects in population pharmacokinetics include inter-individual variability and intra-individual variability (residual residual variability), and the inter-individual variability of pharmacokinetic parameters is represented by formula (1):

[0044] P i =P pop *e ηi Formula (1)

[0045] Where P i represents the pharmacokinetic parameter of the i-th patient, P pop is the typical value of the pharmacokinetic parameter, ηi is the inter-individual variability of the i-th patient, and ηi conforms to a normal distribution with a mean of 0 and a variance of ω 2 ;

[0046] The within-subject variability (residual variability) of pharmacokinetic parameters is expressed using formula (2), formula (3), or formula (4):

[0047] Additive type: C obs,ij = C pred,ij + ε ij,add Formula (2)

[0048] Proportional type: C obs,ij = C pred,ij *(1 + ε ij,prop ) Formula (3)

[0049] Mixed type: C obs,ij = C pred,ij *(1 + ε ij,prop ) + ε ij,add Formula (4)

[0050] Where C obs,ij and C pred,ij are the observed value and predicted value of the plasma concentration at the jth time point for the ith subject, respectively, and ε ij,prop and ε ij,add represent the within-subject variability values of the proportional type or additive type, respectively, and both conform to a normal distribution with a mean of 0 and variances of σ1 2 and σ2 2 ;

[0051] (3) Covariate model module: Used to screen potential covariates based on physiological and clinical rationality, and to examine the influence of different covariates on pharmacokinetic parameters using the likelihood ratio to construct a population pharmacokinetic model;

[0052] (4) Model evaluation module: Used to evaluate the predictive ability and stability of the population pharmacokinetic model through goodness-of-fit plots, visual predictive checks, and non-parametric bootstrap methods;

[0053] (5) Model-guided anti-tuberculosis drug dosage prediction module: Used to develop individualized dosage recommendation plans for tuberculosis patients with different characteristics by using the established population pharmacokinetic model of the drug in combination with Monte Carlo simulation.

[0054] The third aspect of the present invention provides an electronic device for predicting anti-tuberculosis drug dosages, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the steps of the above anti-tuberculosis drug dosage prediction method are implemented.

[0055] The fourth aspect of the present invention provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above anti-tuberculosis drug dosage prediction method are implemented.

[0056] The prediction method for the dosage of isoniazid, an anti-tuberculosis drug, constructed in the present invention has a two-compartment configuration model with first-order absorption and elimination, which can better describe the pharmacokinetic characteristics of isoniazid. A proportional and additive mixed model is used to represent the remaining residual variability. After covariate screening, the gene polymorphism of N-acetyltransferase and body weight are significant covariates affecting the pharmacokinetics of isoniazid. Adding the effects of genotype and body weight to the clearance rate (CL) and the effect of body weight to the volume of distribution (V) has better fitting performance than the basic model.

[0057] The prediction method for the dosage of isoniazid, an anti-tuberculosis drug, constructed in the present invention can be used to establish an individualized medication plan: using the finally established population pharmacokinetic model and combining with Monte Carlo simulation, a dosing plan with a peak isoniazid blood concentration of 3 - 6 mg / L can be obtained. Simulations are carried out on virtual patients with different N-acetylase metabolic types and different body weights, and the expected efficacy and adverse reaction risks of different dosing plans for individual patients are analyzed and estimated, and further judged whether the candidate plan matches the patient to determine the most suitable individualized medication plan for the patient.

[0058] Compared with the prior art, the present invention has the following beneficial effects:

[0059] 1. The prediction method for the dosage of isoniazid, an anti-tuberculosis drug, established in the present invention is based on Chinese tuberculosis patients and is more in line with the pharmacokinetic characteristics of Chinese tuberculosis patients, which can provide a scientific basis for the adjustment of isoniazid dosage for domestic tuberculosis patients.

[0060] 2. The prediction method for the dosage of isoniazid, an anti-tuberculosis drug, obtained in the present invention, combined with the Bayesian maximum a posteriori method, can predict the blood drug concentration at different time points after dosing, calculate the overall drug exposure, and evaluate the efficacy of anti-tuberculosis drugs. Combined with Monte Carlo simulation, the recommended dose to reach the target value can be obtained, and the dosing plan for specific populations can be optimized, thereby improving the drug efficacy and reducing the risk of adverse reactions, and realizing individualized administration of anti-tuberculosis drugs. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 is a schematic structural diagram of the final model for predicting the dosage of isoniazid in the present invention;

[0062] Figure 2 is a goodness-of-fit curve graph of the final model for predicting the dosage of isoniazid in the present invention,

[0063] Figure 3 is a visual predictive check graph of the final model for predicting the dosage of isoniazid in the present invention;

[0064] Figure 4 is a simulation graph of dosage recommendations for patients with different N-acetylase metabolic types and different body weights. Detailed implementation mode

[0065] The present invention will be further described in detail below in conjunction with embodiments and the accompanying drawings, but the implementation modes of the present invention are not limited thereto.

[0066] 1. Test method

[0067] 1.1 Research design and patient population

[0068] The research subjects were Chinese tuberculosis patients who received anti-tuberculosis treatment in the Department of Tuberculosis, Hangzhou Red Cross Hospital from February 2022 to August 2023. This research was approved by the hospital's medical research ethics committee, complied with the Helsinki Declaration, and all selected patients obtained written informed consent.

[0069] Patients were included according to the following criteria: (i) diagnosed with pulmonary tuberculosis, (ii) aged ≥ 18 years, (iii) the anti-tuberculosis treatment regimen included isoniazid, (iv) no severe underlying diseases, (v) not pregnant. If a patient (i) was allergic or intolerant to isoniazid, (ii) had severe underlying diseases, they were excluded.

[0070] 1.2 Administration plan and blood sampling

[0071] According to the doctor's prescription, the administration dose of isoniazid was 300 mg / day, administered orally. After the patients took the medicine regularly for more than one week to reach a steady state, samples were taken half an hour before administration and at 1, 2, 4, 6, 12, and 23 h after administration and stored in tubes containing EDTA. Centrifuged at 3000 rpm / min for 10 minutes at 4°C, plasma was collected and stored at -80°C for analysis.

[0072] 1.3 Detection of isoniazid blood drug concentration

[0073] The blood drug concentration of isoniazid was determined using high performance liquid chromatography tandem mass spectrometry (LC-MS / MS).

[0074] The chromatographic conditions of the LC-MS / MS method were as follows: the chromatographic column specification was: ZORBAX PRHD SB-Aq (3 × 50 mm, 1.8 μm, Agilent, USA), the flow rate was 0.3 mL / min, the injection volume was 10 μL, the column temperature was 40°C, a binary mobile phase system with mobile phase A being 0.1% formic acid aqueous solution and mobile phase B being methanol was used, and the gradient elution of the mobile phase is shown in the following table:

[0075] Time (min) ％A ％B 0 90 10 1.50 90 10 3.00 10 90 4.50 10 90 4.51 90 10 6.00 90 10

[0076] The mass spectrometry conditions of the LC-MS / MS method were as follows: The mass spectrometer was a triple quadrupole Xevo TQ-S IVD, operating in positive ionization mode. The detection of INH was based on multiple reaction monitoring of m / z 138.1→121.1.

[0077] The detection limit of this method was 0.2 μg / mL, the linear range was 0.2 - 40 μg / mL, the recovery rate was 97.2% - 101.2%, and the RSD within and between days was ≤10%. On the day of blood collection, indicators such as the patient's liver and kidney functions were detected.

[0078] 1.4 Population pharmacokinetic modeling

[0079] 1.4.1 Basic model

[0080] The NONMEM software (version 7.5, ICON Development Solutions, Ellicott City, MD, USA) was used to analyze data and build models. The R software (version 3.5.1; www.R-project.org) was used for data collation and plotting. The first-order conditional estimation with interaction (FOCE-I) method was used to estimate the pharmacokinetic parameters of isoniazid and their variability. The overall advantages and disadvantages of alternative models were evaluated by comparing the changes in the objective function value (OFV) and plotting the goodness-of-fit (GOF) graph. In this experiment, one-compartment and two-compartment models with first-order absorption and linear elimination were attempted to fit the blood concentration data of isoniazid. The inter-individual variability of pharmacokinetic parameters in the basic model was estimated using an exponential statistical model, as shown in Equation 1:

[0081] P i = P tot *e ηi (1)

[0082] where P i represents the individual pharmacokinetic parameter, P tot is the typical value of the population pharmacokinetic parameter, ηi is the inter-individual variability, and ηi follows a normal distribution with a mean of 0 and a variance of ω 2 .

[0083] The intra-individual variability (residual variability) of isoniazid drug concentration was attempted to be estimated using an additive model, a proportional model, and a mixed statistical model. The calculation formulas are as follows:

[0084] Additive type: C obs,ij = C pred,ij + ε ij,add (2)

[0085] Proportional type: C obs,ij = C pred,ij *(1 + ε ij,prop ) (3)

[0086] Mixed type: C obs,ij = C pred,ij *(1 + ε ij,prop ) + ε ij,add (4)

[0087] where C obs,ij and C pred,ij are the observed and predicted plasma concentrations of the i-th subject at the j-th time point, respectively, and ε ij,prop and ε ij,add represent the proportional and additive within-subject variability values, respectively, and conform to the normal distributions with means of 0 and variances of σ1 2 and σ2 2 respectively.

[0088] 1.4.2 Covariate model

[0089] Potential covariates were screened based on physiological and clinical rationality. Age, gender, CrCL, WT, TBIL, IBIL, DBIL, ALT, AST, BUN, ALB, and WBC levels, as well as co-administered drugs and diseases, were studied as potential variables affecting PK parameters. The likelihood ratio test was used to examine the influence of each variable on the model parameters.

[0090] The method for constructing the covariate model was the stepwise method: including the forward inclusion and backward elimination processes. In the forward inclusion process, covariates were added to the basic model one by one. If the introduction of a covariate caused the objective function value (OFV) of the model to decrease by more than 3.84 (P < 0.05), the covariate was retained in the model; this process was repeated until no more covariates could be included. In the backward elimination process, one covariate was removed each time. When the removal of one covariate caused the OFV of the model to increase by more than 6.64 (P < 0.01), the factor was considered to be significant and was retained in the model.

[0091] 1.4.3 Model evaluation

[0092] The predictive ability and reliability of the final model were evaluated by the goodness-of-fit curve (GOF) and visual predictive check (VPC). The GOF plot includes a scatter plot of observed values against individual or population predicted values and a scatter plot of conditional weighted residuals (CWRES) against time or population predicted values. The VPC method involves simulating 1000 datasets and comparing the 5th and 95th percentiles and median of the observed data with the corresponding percentiles and median of the simulated data. The non-parametric bootstrap method was used to validate the robustness of the final model. 1000 replicated bootstrap datasets were regenerated from the original data by random resampling, and the final model was applied to the regenerated datasets to recalculate the PK parameters. The median, 2.5, and 97.5 percentile values of the PK parameters estimated from the resampled datasets were compared with the corresponding values estimated by the final model.

[0093] 1.4.4 Model-guided prediction of anti-tuberculosis drug dosage

[0094] Based on the established population pharmacokinetic model of isoniazid, combined with Monte Carlo simulation, individualized dosage recommendation schemes were developed for tuberculosis patients with different characteristics. The PK profiles of typical patients and patients at different covariate levels were compared. By simulating the concentration-time distribution curve of isoniazid, the peak concentration (C max ) of isoniazid was obtained, and the dosage was determined such that the proportion of isoniazid C max below the standard therapeutic range (3 - 6 μg / ml) was less than 20% and the proportion within the therapeutic range was greater than 50%.

[0095] 2. Trial results

[0096] 2.1 Characteristics of the subjects

[0097] A total of 1012 blood drug concentration data from 736 tuberculosis patients were included for population pharmacokinetic analysis. Male patients accounted for 65.6% of all patients. The median age and weight were 51 years (18 - 93 years) and 57.0 kg (27 - 100 kg), respectively. 20.9% of the patients had diabetes mellitus. More than 72.6% of the patients received combination therapy with rifampicin, pyrazinamide, and ethambutol simultaneously. Table 1 summarizes the clinical characteristics of the patients included in the analysis.

[0098] Table 1 Demographic data of tuberculosis patients

[0099] Variable Median (Range) Number of patients (n) 736 Plasma drug concentration (n) 1012 Gender (Male) (n, %) 483(65.6％) Age (years) 51(18～93) Height (cm) 165(138～187) Weight (kg) 57(27～100) Concurrent diabetes (n, %) 154(20.9％) Albumin, Alb (g / L) 37.4(15.1～68.9) Total bilirubin, TBIL (μmol / L) 11.8(2.6～78.0) Direct bilirubin, DBIL (μmol / L) 4.78(0.4～27.1) Indirect bilirubin, IBIL (μmol / L) 6.89(1.3～21.2) Alanine aminotransferase, ALT (U / L) 19.3(1～175) Aspartate aminotransferase, AST (U / L) 25.3(6～175) Alkaline phosphatase, ALP (U / L) 88.2(18～484) Gamma-glutamyl transferase, GGT (U / L) 46.2(7～384) <![CDATA[White blood cells, WBC (10* 9 / L)]]> 5.25(0.1～21.9) Hemoglobin, Hgb (g / L) 125.5(15～1443) <![CDATA[Platelets, PLT (10* 9 / L)]]> 229.3(11～760) Blood urea nitrogen, BUN (mmol / L) 4.77(1.14～32.12) Creatinine clearance rate, Ccr (ml / min) 89.2(6.27～252) Concurrent use of rifampicin (n, %) 210(86.1％) Concurrent use of pyrazinamide (n, %) 151(72.6％) Concurrent use of ethambutol (n, %) 251(78.4％)

[0100] 2.2 Population pharmacokinetic model of INH

[0101] The present invention found that a two-compartment model with first-order absorption and elimination could better describe the population pharmacokinetic characteristics of isoniazid (seeFigure 1 )。The final model includes PK parameters such as the absorption rate constant (ka), apparent clearance (CL / F), apparent volume of distribution in the central compartment (Vc / F), clearance between compartments (Q / F), and apparent volume of distribution in the peripheral compartment (Vp / F). Inter-individual variability was incorporated into the CL / F of isoniazid. A mixed model was used to describe the remaining residual variability. After covariate screening, the inclusion of the NAT2 genotype polymorphism in the CL / F of isoniazid led to a significant decrease in the OFV by 11.2 (P < 0.001). Body weight was incorporated into CL / F and Vc / F in an allometric growth form, which also significantly improved the model fit (ΔOFV = -47.9). The final established model is as follows:

[0102]

[0103] where WT is body weight, and CL / F (L / h) fast is the apparent clearance of patients with a fast metabolism genotype of NAT2, and CL / F (L / h) slow is the apparent clearance of patients with a slow metabolism genotype of NAT2. Table 2 summarizes the estimated population pharmacokinetic parameter values and relative precision of the final model.

[0104] Table 2 Estimated population pharmacokinetic parameter values of isoniazid in the final model

[0105]

[0106]

[0107] Note: Ka: Absorption rate constant; CL / F: Apparent clearance; Prop: Proportion of fast metabolizers among tuberculosis patients; V c / F: Apparent volume of distribution in the central compartment; Q / F: Clearance between compartments; V p / F: Apparent volume of distribution in the peripheral compartment; RSE(%): Relative standard error (percentage).

[0108] 2.3 Model evaluation

[0109] The GOF plot shows that the deviation between the dependent variable (observed value) and the individual or population predicted values is small, and the precision is acceptable (see Figure 2 ). The VPC analysis also shows the good predictive performance of the final model, and the observed values are within the 5%, 50%, and 95% confidence intervals of the model predicted values (see Figure 3 ). The non-parametric bootstrap analysis indicates that the resampled dataset is consistent with the estimated PK parameter values of the final model, indicating that the final model is stable and reproducible.

[0110] 2.4 Model-guided dose prediction

[0111] The present invention discovers that the polymorphism of the NAT2 genotype and body weight significantly affect the clearance rate of isoniazid, and body weight affects the apparent volume of distribution. Therefore, based on the established population pharmacokinetic model of isoniazid, the present invention formulates an individualized dosage recommendation scheme according to the NAT2 genotype and body weight of different tuberculosis patients.

[0112] Optimize the isoniazid dosing regimen for patients with different NAT2 genotypes and body weights through Monte Carlo simulation. The goal is to determine the dosing dose that makes the proportion of isoniazid C max lower than the standard treatment range (3 - 6 μg / ml) less than 20% and the proportion within the treatment range greater than 50%. Virtual patients are divided into fast metabolizers and slow metabolizers according to the genetic polymorphism of isoniazid, and the body weight ranges are set as 35 - 57 kg, 58 - 70 kg, and 71 - 100 kg. A patient with a body weight of 57 kg taking 300 mg of isoniazid per day is defined as the typical reference population. Simulate the concentration-time distribution curves of isoniazid at different dosing doses 1000 times for patients with different metabolic types and body weights respectively, and obtain the C max of isoniazid. According to the standard treatment range of C max , determine the optimal dose regimen for patients with different characteristics. The simulation results of the dosing doses for patients with different NAT2 genotypes and body weights are shown in Table 3 and Figure 4 as follows.

[0113] According to the simulation results, for slow-metabolizer tuberculosis patients, those with a body weight of 35 - 57 kg need to reduce the isoniazid dose to 250 mg / day, while for those with a body weight of 58 - 70 kg and 71 - 100 kg, the daily doses need to be 350 mg and 400 mg respectively to reach the standard exposure level. For fast-metabolizer patients, those with a body weight of 35 - 57 kg need 450 mg of isoniazid per day, while for those with a body weight of 58 - 70 kg and 71 - 100 kg, they need to take 600 mg and 650 mg of isoniazid per day respectively to make the C max of isoniazid reach the standard treatment range.

[0114] Table 3 Simulation results of dosing doses for patients with different NAT2 genotypes and body weights

[0115]

[0116]

[0117] To sum up, the present invention is based on large clinical sample data, establishes a prediction method for the dosage of the anti-tuberculosis drug isoniazid, and develops a model-guided optimized dosing regimen for patients with different NAT2 genotypes and body weights, providing a reference for formulating specific dosage recommendation schemes in the process of individualized treatment.

[0118] The above specific embodiments are preferred embodiments described for the convenience of understanding the present invention, but the present invention is not limited to the above embodiments, that is, it does not mean that the present invention must rely on the above embodiments to be implemented. Those skilled in the art should understand that any equivalent changes or modifications completed under the technical spirit prompted by the present invention should fall within the scope of the patent covered by the present invention.

Claims

1. A method for predicting the dosage of an anti-tuberculosis drug, characterized in that, The prediction method includes the following steps: (1) Data collection: Collect the clinical information of patients and collect plasma samples. The plasma samples are detected by the LC-MS / MS method to obtain the blood drug concentration data of the target drug; (2) Basic model: Use the NONMEM program, a non-linear mixed effect model modeling software, to establish a pharmacokinetic structural model of anti-tuberculosis drugs through a one-compartment or two-compartment configuration model with first-order absorption and elimination, and calculate the pharmacokinetic parameters of the corresponding model; The NONMEM program is ADVAN2-TRANS2 or ADVAN4-TRANS4; The random effects in population pharmacokinetics include inter-individual variability and intra-individual variability (residual variability). The inter-individual variability of pharmacokinetic parameters is expressed by formula (1): P i = P pop * e ηi Equation (1) where $P$ i represents the pharmacokinetic parameter of the $i$-th patient, and $P$ pop is the typical value of the pharmacokinetic parameter. $\eta_i$ is the inter-individual variability of the $i$-th patient, and $\eta_i$ follows a normal distribution with a mean of 0 and a variance of $\omega$ 2 ; The intra-individual variability (residual variability) of pharmacokinetic parameters is expressed by formula (2), formula (3) or formula (4): Additive type: C obs,ij = C pred,ij + ε ij,add Formula (2) Proportional type: C obs,ij = C pred,ij *(1 + ε ij,prop ) Formula (3) Hybrid type: C obs,ij = C pred,ij *(1 + ε ij,prop ) + ε ij,add Equation (4) where C obs,ij and C pred,ij are the observed value and predicted value of the plasma concentration of the i-th subject at the j-th time point, respectively, and ε ij,prop and ε ij,add represent the within-subject variability values of the proportional type or additive type, respectively, and conform to the normal distributions with a mean of 0 and variances of σ1 2 and σ2 2 respectively; (3) Covariate model: Screen potential covariates based on physiological and clinical rationality, and use the likelihood ratio to examine the influence of different covariates on pharmacokinetic parameters to construct a population pharmacokinetic model; (4) Model evaluation: Evaluate the predictive ability and stability of the population pharmacokinetic model through goodness-of-fit plots, visual predictive checks and non-parametric bootstrap methods; (5) Model-guided prediction of anti-tuberculosis drug dosage: Based on the established population pharmacokinetic model of anti-tuberculosis drugs, combined with Monte Carlo simulation, develop individualized dosage recommendation plans for tuberculosis patients with different characteristics.

2. The prediction method according to claim 1, wherein The chromatographic conditions of the LC-MS / MS method in step (1) are as follows: The chromatographic column specification is: ZORBAX PRHD SB-Aq, flow rate: 0.3 mL / min, injection volume: 10 μL, column temperature: 40 °C. A binary mobile phase system with mobile phase A being 0.1% formic acid aqueous solution and mobile phase B being methanol is used. The gradient elution of the mobile phase is shown in the following table:

3. The prediction method according to claim 1, characterized in that The estimation method of the pharmacokinetic parameters in step (2) is the first-order conditional estimation method with ε-η interaction.

4. The prediction method according to claim 1, wherein The selection of the basic model in step (2) is determined by comparing the differences in the objective function value OFV, the accuracy of parameter estimation, and the visual inspection of the goodness-of-fit plot.

5. The prediction method according to claim 1, characterized in that The covariates in step (3) include age, gender, weight, creatinine clearance rate, total bilirubin, direct bilirubin, indirect bilirubin, alanine aminotransferase, aspartate aminotransferase, serum urea nitrogen, albumin, and white blood cell levels, etc., as well as concomitant medications and complications.

6. The prediction method according to claim 1, wherein The method for constructing the covariate model in step (3) is the stepwise method, including the forward inclusion and backward elimination processes; in the forward inclusion process, covariates are added to the basic model one by one. If introducing a covariate causes the objective function value OFV of the model to decrease by more than 3.84 (P < 0.05), then the covariate is retained in the model; repeat this process until no more covariates can be included; in the backward elimination process, one covariate is removed each time. When removing 1 covariate causes the objective function value OFV of the model to increase by more than 6.64 (P < 0.01), then it is considered that the factor is significant and is retained in the model.

7. The prediction method according to claim 1, wherein In step (5), the Monte Carlo simulation uses the established population pharmacokinetic model to set up virtual patient populations with different characteristics, simulate the concentration-time distribution curves of the drug at different dosing doses, and obtain the peak concentration C of the drug. max , and determine the dosing dose that makes the proportion of the drug C max below the standard treatment range of 3-6 μg / ml less than 20% and the proportion within the treatment range greater than 50%.

8. An anti-tuberculosis drug dosage prediction device, characterized in that, The prediction device includes: (1) A data collection module for collecting the clinical information and plasma samples of patients. The plasma samples are detected by the LC-MS / MS method to obtain the blood drug concentration data of the target drug; (2) A basic model construction module for using the NONMEM program of the nonlinear mixed effect model modeling software to establish a pharmacokinetic structure model of anti-tuberculosis drugs through a one-compartment or two-compartment configuration model with first-order absorption and elimination, and calculating the pharmacokinetic parameters of the corresponding model; The NONMEM program is ADVAN2-TRANS2 or ADVAN4-TRANS4; The random effects in population pharmacokinetics include inter-individual variability and intra-individual variability (residual variability). The inter-individual variability of pharmacokinetic parameters is represented by formula (1): P i = P pop * e ηi Equation (1) where $P_i$ i represents the pharmacokinetic parameter of the $i$-th patient, and $P$ pop is the typical value of the pharmacokinetic parameter. $\eta_i$ is the inter-individual variability of the $i$-th patient, and $\eta_i$ follows a normal distribution with a mean of 0 and a variance of $\omega$ 2 . The intra-individual variability (residual variability) of pharmacokinetic parameters is represented by formula (2), formula (3) or formula (4): Additive type: C obs,ij = C pred,ij + ε ij,add Formula (2) Proportional type: C obs,ij = C pred,ij *(1 + ε ij,prop ) Formula (3) Hybrid type: C obs,ij = C pred,ij *(1 + ε ij,prop ) + ε ij,add Equation (4) where C obs,ij and C pred,ij are the observed value and predicted value of the plasma concentration of the i-th subject at the j-th time point, respectively, and ε ij,prop and ε ij,add represent the within-subject variability values of the proportional or additive type, respectively, and conform to the normal distributions with a mean of 0 and variances of σ1 2 and σ2 2 respectively; (3) A covariate model module: for screening potential covariates according to physiological and clinical rationality, using the likelihood ratio to examine the influence of different covariates on pharmacokinetic parameters, and constructing a population pharmacokinetic model; (4) A model evaluation module: for evaluating the reliability and stability of the population pharmacokinetic model through diagnostic goodness-of-fit plots, visual predictive checks and non-parametric bootstrap methods; (5) A model-guided anti-tuberculosis drug dosage prediction module: for using the established population pharmacokinetic model of the drug, combined with Monte Carlo simulation, to develop an individualized dosage recommendation plan for tuberculosis patients with different characteristics.

9. An electronic device for predicting the dosage of anti-tuberculosis drugs, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the anti-tuberculosis drug dosage prediction method according to any one of claims 1-7.

10. A non-transitory computer-readable storage medium, on which a computer program is stored, characterized in that, When the computer program is executed by the processor, it implements the steps of the anti-tuberculosis drug dosage prediction method according to any one of claims 1-7.

Images