Construction method of PPK model of r-ATG used by children and determination method of administration scheme

Abstract

The invention provides a construction method of a PPK model of r-ATG used by children and a determination method of a drug administration scheme. The construction method comprises the following steps: acquiring statistical characteristics of a plurality of child patients; r-ATG instillation is conducted on the multiple child patients according to a preset administration plan, blood sample collection is conducted on the multiple child patients according to a preset sampling plan, and blood concentration data of the multiple child patients are obtained; dividing the blood concentration data of the child patients into model construction group data and model verification group data, and constructing an initial PPK model based on the statistical characteristics and the model construction group data; performing internal verification on the initial PPK model based on the model verification group data to obtain an internal verification result, and performing external verification on the initial PPK model based on the model verification group data to obtain an external verification result; and correcting the initial PPK model according to the internal verification result and the external verification result to obtain a target PPK model. The method can improve the safety of r-ATG administration for children.

Description

Technical Field

[0001] The present disclosure relates to the field of medical technology, and in particular to a method for constructing a PPK model of children using r-ATG and a method for determining a dosing regimen. Background Art

[0002] Allogeneic hematopoietic stem cell transplantation is a common method for treating severe β-thalassemia, but after transplantation, complications of graft-versus-host disease are very likely to occur, which can lead to graft failure. At present, rabbit anti-human thymocyte immunoglobulin (Thymoglobulin, r-ATG) is often used before and after transplantation to prevent and treat graft-versus-host disease. However, excessive doses of r-ATG will affect immune reconstruction after allogeneic hematopoietic stem cell transplantation, and low-dose infusion of r-ATG will result in low absolute lymphocyte counts, leading to abnormal immune function. Therefore, it is very important to study the pharmacokinetics of r-ATG. By quantitatively studying the dynamic changes of r-ATG in the body, it provides a basis for the rational use and dosage adjustment of r-ATG.

[0003] However, there is no population pharmacokinetic model for the use of r-ATG in children in the related art, which leads to lower safety of r-ATG administration in children. Summary of the invention

[0004] The embodiments of the present disclosure provide a method for constructing a PPK model for the use of r-ATG in children and a method for determining a dosing regimen, which can improve the safety of r-ATG administration in children.

[0005] According to one aspect of the present disclosure, a method for constructing a PPK model for children using r-ATG is provided, comprising:

[0006] Obtaining statistical characteristics of multiple children, wherein the statistical characteristics include age information and allo-HSCT transplant type information;

[0007] Performing r-ATG drip infusion on the multiple children according to a preset dosing plan, and collecting blood samples from the multiple children according to a preset sampling plan to obtain blood drug concentration data of the multiple children;

[0008] Dividing the blood drug concentration data of the plurality of children into model building group data and model verification group data, and constructing an initial PPK model based on the statistical characteristics and the model building group data;

[0009] Performing internal verification on the initial PPK model based on the model verification group data to obtain an internal verification result, and performing external verification on the initial PPK model based on the model verification group data to obtain an external verification result;

[0010] The initial PPK model is modified according to the internal verification result and the external verification result to obtain a target PPK model.

[0011] Optionally, in one implementation, the externally validating the initial PPK model based on the model validation group data to obtain an external validation result includes:

[0012] Calculate a plurality of relative prediction error values ​​according to the model validation group data and the initial PPK model;

[0013] Calculating a first probability value that the relative prediction error value is within a first error range, and calculating a second probability value that the relative prediction error value is within a second error range;

[0014] An external verification result is determined based on the first probability value and the second probability value.

[0015] Optionally, in one implementation, before determining the external verification result based on the first probability value and the second probability value, the method further includes:

[0016] Calculate the average prediction error and the average absolute prediction error according to the model validation group data and the initial PPK model;

[0017] The determining the external verification result based on the first probability value and the second probability value includes:

[0018] An external validation result is determined based on the first probability value, the second probability value, the average prediction error, and the average absolute prediction error.

[0019] Optionally, in one implementation, determining the external verification result based on the first probability value, the second probability value, the average prediction error, and the average absolute prediction error includes:

[0020] Determine, based on the initial PPK model, a first weight corresponding to the first probability value, a second weight corresponding to the second probability value, a third weight corresponding to the mean prediction error, and a fourth weight corresponding to the mean absolute error;

[0021] Performing weighted calculation on the first probability value, the second probability value, the average prediction error, and the average absolute prediction error based on the first weight, the second weight, the third weight, and the fourth weight to obtain a verification value;

[0022] An external verification result is determined according to the verification value and a preset reference value.

[0023] Optionally, in one implementation, the performing internal verification on the initial PPK model based on the model verification group data to obtain an internal verification result includes:

[0024] Obtain a preset number of bootstrap data, and obtain a reference PPK model based on the bootstrap data by fitting, and obtain reference model parameters corresponding to a preset quantile interval of the reference PPK model;

[0025] The initial PPK model is internally verified based on the reference model parameters and the parameters of the initial PPK model to obtain an internal verification result.

[0026] Optionally, in one implementation, before performing internal verification on the initial PPK model based on the reference model parameters and the parameters of the initial PPK model to obtain an internal verification result, the method further includes:

[0027] Draw a model fit diagnostic graph based on the model validation group data and the initial PPK model;

[0028] The internal verification of the initial PPK model based on the reference model parameters and the parameters of the initial PPK model to obtain an internal verification result includes:

[0029] The initial PPK model is internally verified based on the model fitting diagnostic graph and the comparison result between the reference model parameters and the parameters of the initial PPK model to obtain an internal verification result.

[0030] Optionally, in one embodiment, constructing an initial PPK model based on the statistical features and the model building group data includes:

[0031] Obtain multiple candidate PPK models;

[0032] Fitting each of the candidate PPK models based on the model building group data to obtain multiple fitting models;

[0033] Calculating an objective function value corresponding to each fitting model, and determining a basic model among the multiple candidate PPK models according to the objective function value;

[0034] A plurality of model covariates are determined from the statistical features, and an initial PPK model is determined according to the plurality of model covariates and the basic model.

[0035] Optionally, in one embodiment, determining a plurality of model covariates in the statistical features comprises:

[0036] Incorporating each of the statistical features into the basic model one by one, and evaluating the reduction value of the objective function value of the basic model when each feature is incorporated into the basic model;

[0037] determining a plurality of candidate covariates among the plurality of features based on the reduction values;

[0038] Eliminating the candidate covariates one by one from the basic model incorporating the statistical features, and evaluating the increase in the objective function when each candidate covariate is eliminated;

[0039] A plurality of model covariates are determined among the plurality of candidate covariates based on the elevated values.

[0040] Optionally, in one embodiment, dividing the blood drug concentration data of the plurality of children into model building group data and model verification group data comprises:

[0041] Determining a first amount of model building group data and a second amount of model validation group data based on the number of the plurality of children and a preset division ratio;

[0042] A first number of blood drug concentration data are randomly determined from the blood drug concentration data of the multiple children as model building group data, and a second number of blood drug concentration data are determined from the blood drug concentration data of the multiple children excluding the model building group data as model verification group data.

[0043] According to another aspect of the present disclosure, a method for determining a dosing regimen is provided, comprising:

[0044] Obtain statistical characteristic data of target children;

[0045] Determine the metabolic data of r-ATG in the target child based on the statistical characteristic data and the target PPK model, wherein the target PPK model is a PPK model constructed by the method for constructing a PPK model of r-ATG for children according to any one of claims 1 to 9;

[0046] A target dosing regimen for the target child is determined based on the metabolic data.

[0047] In the method for constructing a PPK model of r-ATG for children in the disclosed embodiment, r-ATG is dripped to multiple children according to a preset dosing plan, and blood samples are collected from multiple children according to a preset sampling plan, thereby obtaining blood drug concentration data for multiple children. An initial PPK model is constructed based on the statistical characteristics of the children and the blood drug concentration data corresponding to the model construction group data, thereby obtaining the influence of statistical characteristics on the pharmacokinetic parameters of r-ATG. The initial PPK model and retrograde internal and external validation are performed based on the blood drug concentration data corresponding to the model validation data. The fitting effect and validation ability of the initial PPK model are evaluated through internal validation, and the precision and accuracy of the initial PPK model are evaluated through external validation. In this way, the target PPK model obtained by correcting the initial PPK model based on internal validation and external validation can accurately describe the influence of children's statistical characteristics on the pharmacokinetic parameters of r-ATG, thereby providing a reliable basis for r-ATG administration and improving the safety of r-ATG administration for children.

[0048] Other features and advantages of the present disclosure will be described in the following description, and partly become apparent from the description, or understood by practicing the present disclosure. The purpose and other advantages of the present disclosure can be realized and obtained by the structures particularly pointed out in the description, claims and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The accompanying drawings are used to provide further understanding of the technical solution of the present disclosure and constitute a part of the specification. Together with the embodiments of the present disclosure, they are used to explain the technical solution of the present disclosure and do not constitute a limitation on the technical solution of the present disclosure.

[0050] Figure 1 is a flow chart of a method for constructing a PPK model for children using r-ATG according to an embodiment of the present disclosure;

[0051] Figure 2 is a graph showing the relationship between blood drug concentration and time according to an embodiment of the present disclosure;

[0052] Figure 3 is a flow chart of a method for determining a dosing regimen according to an embodiment of the present disclosure;

[0053] Figure 4 is a schematic diagram of the correlation between various statistical features and pharmacokinetic parameters according to an embodiment of the present disclosure;

[0054] Figure 5A is a model fit diagnostic plot corresponding to a base model according to one embodiment of the present disclosure;

[0055] Figure 5Bis a model fitting diagnostic graph corresponding to an initial PPK model according to one embodiment of the present disclosure;

[0056] Figure 6 It is a schematic diagram of a predicted value calibration visualization prediction distribution according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solution and advantages of the present disclosure more clear, the present disclosure is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present disclosure and are not used to limit the present disclosure.

[0058] Before further describing the embodiments of the present disclosure in detail, the nouns and terms involved in the embodiments of the present disclosure are described. The nouns and terms involved in the embodiments of the present disclosure are subject to the following interpretations:

[0059] Allogeneic hematopoietic stem cell transplantation (allo-HSCT): An effective method for treating malignant blood diseases and certain genetic diseases. It involves transplanting hematopoietic stem cells from healthy donors into the patient's body to replace the child's damaged or abnormal hematopoietic system.

[0060] β-thalassemia major (β-TM) is an inherited hemolytic disease caused by a defect in the β-globin gene, which leads to abnormal hemoglobin synthesis.

[0061] Graft versus-host disease (GVHD): A serious complication following allogeneic hematopoietic stem cell transplantation (HSCT) in which the donor's immune cells attack the recipient's body, causing multi-organ damage.

[0062] Rabbit anti-human thymocyte immunoglobulin (r-ATG): a drug used for immunosuppressive therapy, mainly used to prevent and treat organ rejection, prevent acute and chronic graft-versus-host disease (GVHD) after hematopoietic stem cell transplantation, and treat hormone-resistant graft-versus-host disease. The basic principle of r-ATG in producing immunosuppression is to deplete lymphocytes by recognizing the active substances on the surface of most T cells that appear during organ rejection, so as to overcome the immune barrier between recipient and donor T cells.

[0063] Absolute lymphocyte count (ALC): The number of lymphocytes in the blood, usually expressed as the number of lymphocytes per liter of blood. It is one of the important indicators for evaluating the function of the immune system.

[0064] Population pharmacokinetic (PPK): A scientific method that studies the absorption, distribution, metabolism and excretion of drugs in the population. It focuses on the variability of drugs between different individuals and the impact of this variability on drug efficacy and safety. The calculation methods of PPK mainly include modeling of drug concentration-time curves and estimation of pharmacokinetic parameters. By monitoring drug concentrations and analyzing related data in a large population, we can obtain information on the dynamic changes of drugs in the population. At the same time, combined with individual characteristics and genetic information, the differences in drug responses can be further explained. In the process of drug development, understanding the dynamic changes of drugs in the population can help optimize drug dosage and administration regimens, improve efficacy and reduce adverse reactions. In personalized treatment, based on the individual's pharmacokinetic characteristics, an individualized treatment plan can be formulated to improve the treatment effect.

[0065] General description of the disclosed embodiments

[0066] According to one embodiment of the present disclosure, a method for constructing a PPK model for children using r-ATG is provided.

[0067] like Figure 1 As shown, it is a flow chart of the method for constructing a PPK model of children using r-ATG provided by the present disclosure. The method for constructing a PPK model of children using r-ATG may include:

[0068] Step 110, obtaining statistical characteristics of multiple children;

[0069] Step 120: administer r-ATG drip to multiple children according to a preset dosing plan, and collect blood samples from multiple children according to a preset sampling plan to obtain blood drug concentration data of multiple children;

[0070] Step 130: dividing the blood drug concentration data of multiple children into model building group data and model verification group data, and constructing an initial PPK model based on the statistical characteristics and the model building group data;

[0071] Step 140: internally verify the initial PPK model based on the model verification group data to obtain an internal verification result, and externally verify the initial PPK model based on the model verification group data to obtain an external verification result;

[0072] Step 150: Modify the initial PPK model according to the internal verification result and the external verification result to obtain a target PPK model.

[0073] In step 110, statistical characteristics of a plurality of children are obtained.

[0074] In one embodiment, the statistical features may include age information and allo-HSCT transplant type information. Since children of different ages have differences in height, weight, body surface area, and drug metabolism levels, the age information of each child can be obtained to conduct stratified studies on children of different ages according to age groups, thereby exploring the individual differences in the pharmacokinetic characteristics of r-ATG in children of different age groups.

[0075] Allo-HSCT transplant type information can indicate the results of classification of allogeneic hematopoietic stem cell transplantation based on factors such as the source and nature of the transplant. The drug metabolism level of r-ATG may be different for different transplant types. Therefore, the allo-HSCT transplant type information corresponding to each child can be obtained to clarify the differences in the pharmacokinetic characteristics of r-ATG corresponding to different transplant types.

[0076] In another embodiment, the statistical features may also include demographic features such as gender, age, height, weight, body surface area, and transplant-related covariates such as the patient's diagnosis, transplant type, number of various lymphocytes, and laboratory test results (alanine aminotransferase, albumin, serum creatinine, etc.). By quantitatively analyzing more abundant statistical features, the influence of each statistical feature on the pharmacokinetic characteristics of r-ATG can be clarified.

[0077] In one embodiment, the statistical characteristics of multiple children can be collected and obtained based on the hospital medical record system and the nursing system.

[0078] In step 120, r-ATG is dripped to multiple children according to a preset dosing plan, and blood samples are collected from multiple children according to a preset sampling plan to obtain blood drug concentration data of the multiple children.

[0079] The preset dosing plan can indicate the method of r-ATG drip infusion for multiple children, which may include preset indicators such as r-ATG dosage, drug concentration, drug source, and administration time.

[0080] The preset sampling plan may indicate the method of collecting blood samples from multiple children after r-ATG infusion is performed on the multiple children, which may include preset indicators such as blood sample collection cycle, blood sample collection volume, and blood sample collection method.

[0081] After blood sample collection, the blood sample can be subjected to blood drug concentration measurement to obtain the blood drug concentration data of the child. The blood drug concentration data can indicate the total concentration of r-ATG in the plasma or whole blood after the child absorbs the infused r-ATG, including r-ATG drugs bound to plasma proteins and r-ATG drugs free in the plasma. In one embodiment, the blood drug concentration data can be the blood drug concentration data of active r-ATG measured by quantitative flow cytometry. Quantitative flow cytometry is a technology for rapidly detecting cell characteristics in a liquid flow. It combines single cells with specific antibodies to perform multi-parameter quantitative analysis of cells, including cell surface markers, intracellular protein expression, cell cycle, DNA content, etc.

[0082] In step 130, the blood drug concentration data of a plurality of children are first divided into model building group data and model verification group data.

[0083] The model building group data can be used to fit the statistical features to build an initial PPK model. The model validation group data can be used to validate the initial PPK model, thereby modifying the initial PPK model to obtain a target PPK model.

[0084] In one embodiment, the blood drug concentration data of multiple children are divided into model building group data and model verification group data, including:

[0085] Determining a first amount of model building group data and a second amount of model validation group data based on the number of multiple children and a preset division ratio;

[0086] A first number of blood drug concentration data are randomly determined from the blood drug concentration data of multiple children as model building group data, and a second number of blood drug concentration data are determined from the data other than the model building group data in the blood drug concentration data of the multiple children as model verification group data.

[0087] The division ratio of the model building group data and the model verification group data can be pre-set, for example, the division ratio is 4:1, that is, the blood drug concentration data of multiple children are evenly divided into 5 parts, 4 of which are used as the model building group data, and the remaining 1 is used as the model verification group data.

[0088] After determining the number of children and the preset division ratio, the first number of model building group data and the second number of model verification group data can be determined. For example, if the number of children is 265 and the preset division ratio is 4:1, the first number of model building group data is 265*4 / 5=212, and the second number of model verification group data is 265*1 / 5=53.

[0089] After determining the first quantity and the second quantity, the first quantity of blood drug concentration data can be determined from the blood drug concentration data of multiple children in a random sampling manner as the model building group data, and the second quantity of blood drug concentration data can be determined from the remaining blood drug concentration data as the model verification group data.

[0090] By dividing the blood drug concentration data of multiple children into model building group data and model verification group data according to a preset division ratio, the data volume of the model building group data and the data volume of the model verification group data can be determined according to the needs of actual application, thereby improving the flexibility of PPK model construction and verification.

[0091] In one embodiment, before dividing the blood drug concentration data of multiple children into model building group data and model verification group data, it also includes: screening the blood drug concentration data of multiple children based on a control relationship diagram between blood drug concentration and time.

[0092] The plot of the relationship between blood drug concentration and time can indicate the changes in blood drug concentration in each child over time, for example Figure 2 As shown, the horizontal axis represents time and the vertical axis represents blood drug concentration. The comparison relationship diagram between blood drug concentration and time can be generated based on the blood drug concentration of blood samples collected at different times and the corresponding time. For different children, the degree of change in blood drug concentration may vary, but the trend of blood drug concentration change will conform to the general pharmacokinetic curve trend of r-ATG. Therefore, based on the comparison relationship diagram between blood drug concentration and time, the data that does not conform to the pharmacokinetic curve trend of r-ATG in the blood drug concentration data of multiple children can be eliminated, and the initial PPK model can be constructed and verified based on the remaining blood drug concentration data.

[0093] Screening the blood drug concentration data of multiple children can eliminate abnormal variables and improve the accuracy of the initial PPK model construction.

[0094] After dividing the model building group data and the model validation group data, an initial PPK model can be constructed based on the statistical characteristics and the model building group data.

[0095] The PPK model can be used to describe the population pharmacokinetics of r-ATG in the pediatric population, which may include: the association between individual child difference variables and r-ATG clearance, the association between individual child difference variables and r-ATG central compartment apparent distribution volume, the association between individual child difference variables and r-ATG inter-compartment clearance, and the association between individual child difference variables and r-ATG peripheral compartment apparent distribution volume. Individual child difference variables can indicate differences in statistical characteristics of different individual children.

[0096] Clearance, central compartment apparent distribution volume, inter-compartmental clearance, and peripheral compartment apparent distribution volume are important parameters in population pharmacokinetics, which can indicate the dynamic changes of drugs in the human body. Clearance refers to the apparent distribution volume of drugs cleared from the body per unit time, that is, how much volume of drugs in body fluids is cleared per unit time. It is used to describe the metabolism and excretion rate of drugs in the body. The central compartment apparent distribution volume refers to the volume of drugs that are instantly distributed to tissues and organs representing the central compartment (such as plasma, liver, kidneys, spleen and other tissues and organs with high blood perfusion) after a certain amount of drugs is injected intravenously, but has not yet been transferred to tissues and organs representing the peripheral compartment. Inter-compartmental clearance refers to the rate at which drugs are transferred from one compartment to another, and is usually used to describe the distribution and elimination process of drugs in the body. The peripheral compartment apparent distribution volume refers to the ratio of the amount of drugs in the body to the blood drug concentration when the drug reaches dynamic equilibrium in the body.

[0097] In one embodiment, constructing an initial PPK model based on statistical features and model building set data includes:

[0098] Obtain multiple candidate PPK models;

[0099] Fitting each candidate PPK model based on the model building group data to obtain multiple fitting models;

[0100] Calculate the objective function value corresponding to each fitting model, and determine the basic model among multiple candidate PPK models according to the objective function value;

[0101] A plurality of model covariates are determined in the statistical characteristics, and an initial PPK model is determined according to the plurality of model covariates and the basic model.

[0102] The multiple candidate PPK models can be multiple base models used to construct the initial PPK model. The candidate PPK models can include compartmental models and random effects models. The compartmental model is a mathematical model commonly used in pharmacokinetic studies to describe the absorption, distribution, metabolism and excretion of drugs in the body. It assumes that the drug can be quickly and evenly distributed to various tissues throughout the body after entering the body and eliminated from the body at a certain rate. The random effects model is a statistical model used to analyze data with hierarchical or nested structures, especially when dealing with individual differences. It assumes that the observed value is composed of group effects and individual effects, where the individual effects are regarded as random variables.

[0103] After obtaining multiple candidate PPK models, the model building group data can be used to fit each candidate PPK model and calculate the objective function value (OFV) corresponding to each fitted model. OFV is an important statistic used to evaluate the quality of model fitting data. It can reflect the degree of match between model parameters and observed data. The smaller the OFV, the more stable the model and the better the model fitting effect.

[0104] In one embodiment, the first-order conditional estimation method of the interaction between inter-individual variation and residual variation can be used to calculate the objective function value. Inter-individual variation refers to the difference in a certain characteristic or feature between different individuals in the same species. This difference may be caused by different genotypes, differences in environmental conditions or other unknown factors. Residual variation refers to the difference between the actual observed value and the model predicted value in statistical modeling, that is, the part of the variation that the model fails to explain. It is a key indicator for evaluating the effectiveness of the statistical model. By analyzing the residual variation, it can be determined whether the model is reasonable and whether there are systematic deviations or other problems. In pharmacokinetic studies, there is a certain correlation between inter-individual variation and residual variation. For example, there may be significant differences in the metabolic rates of different individuals, and this inter-individual variation may lead to an increase in residual variation. Therefore, the objective function value can be calculated based on the interaction between inter-individual variation and residual variation.

[0105] After calculating the objective function values ​​of the respective fitting models, the fitting model with the best objective function value may be determined as the basic model.

[0106] In another embodiment, in addition to the objective function value corresponding to the fitting model, parameters such as model drawing diagnostic graphs, statistical tests, and the precision of parameter estimates can also be obtained, and the basic model is determined among multiple candidate PPK models based on the size of the objective function value, model drawing diagnostic graphs, statistical tests, and the precision of parameter estimates.

[0107] After determining the base model, a number of model covariates can be determined in the statistical features.

[0108] In one embodiment, a plurality of model covariates are determined in the statistical features, including:

[0109] Incorporate each feature of the statistical features into the basic model one by one, and evaluate the reduction in the objective function value of the basic model when each feature is incorporated into the basic model;

[0110] identifying a plurality of candidate covariates among the plurality of features based on the reduction values;

[0111] Eliminate candidate covariates one by one from the basic model that incorporates statistical characteristics, and evaluate the increase in the objective function when each candidate covariate is eliminated;

[0112] A plurality of model covariates are identified among the plurality of candidate covariates based on the elevated values.

[0113] Since not every feature is significant for the construction of the initial PPK model, multiple model covariates can be screened from the statistical features.

[0114] First, each feature in the statistical feature can be incorporated into the basic model one by one, and the objective function value corresponding to the basic model after the feature is incorporated can be calculated. After the feature is incorporated into the basic model, the objective function value corresponding to the basic model should decrease. Therefore, the reduction value between the objective function value after the feature is incorporated into the basic model and the objective function value before the feature is incorporated into the basic model can be obtained. For example, the objective function value after the feature is incorporated into the basic model is 10, and the objective function value before the feature is incorporated into the basic model is 18, then the reduction value corresponding to the feature is 8.

[0115] Determining multiple candidate covariates from multiple features based on the reduction value can be determined based on a predetermined condition. When the reduction value meets the predetermined condition, the corresponding feature can be determined as a candidate covariate. For example, the predetermined condition is: the reduction value is greater than 3.84.

[0116] The full model can be obtained by incorporating all statistical features that meet the preset conditions into the basic model. After determining multiple candidate covariates, each candidate covariate can be reversely eliminated from the full model. When the candidate covariate is eliminated from the full model, the objective function value corresponding to the model should increase. Therefore, the increase value between the objective function value after the candidate covariate is eliminated from the full model and the objective function value before the candidate covariate is eliminated from the full model can be obtained. For example, the objective function value after the candidate covariate is eliminated from the full model is 28, and the objective function value before the candidate covariate is eliminated from the full model is 15, then the increase value corresponding to the candidate covariate is 13.

[0117] Determining the model covariate from among the plurality of candidate covariates based on the elevated value may be determined based on a predetermined condition. When the elevated value satisfies the predetermined condition, the corresponding candidate covariate may be determined as the model covariate. For example, the predetermined condition is: the elevated value is greater than 10.83.

[0118] Screening multiple model covariates from statistical characteristics through the process of forward inclusion and backward elimination can ensure that the model covariates have significant significance for the function value changes of the basic model, thereby improving the accuracy of screening model covariates.

[0119] After determining the multiple model covariates, an initial PPK model can be determined based on the multiple model covariates and the basic model. Specifically, the multiple model covariates can be incorporated into the basic model for fitting, thereby obtaining the initial PPK model.

[0120] Based on the objective function value, a basic model is determined from multiple candidate PPK models, and an initial PPK model is constructed based on the basic model. The initial PPK model can be constructed by selecting the model with the best fitting effect from multiple candidate PPK models, thereby improving the accuracy of the initial PPK model construction.

[0121] In step 140, first, the initial PPK model is internally verified based on the model verification group data to obtain an internal verification result. The internal verification can evaluate the fitting effect and verification ability of the initial PPK model, so the internal verification result can indicate the degree of fit of the initial PPK model to the data.

[0122] In one embodiment, the initial PPK model is internally verified based on the model verification group data to obtain an internal verification result, including:

[0123] Obtain a preset number of bootstrap data, and obtain a reference PPK model based on the bootstrap data, and obtain reference model parameters corresponding to a preset quantile interval of the reference PPK model;

[0124] The initial PPK model is internally verified based on the reference model parameters and the parameters of the initial PPK model to obtain an internal verification result.

[0125] The bootstrap data can be obtained by performing random sampling with replacement on the blood drug concentration data of multiple children to generate a bootstrap data set. The preset number can be equal to the sample size corresponding to the blood drug concentration data of multiple children. Repeat the above operation multiple times to obtain multiple bootstrap data sets. For each bootstrap data set, multiple bootstrap data in the bootstrap data set can be used to fit and obtain a reference PPK model.

[0126] Obtaining the reference model parameters corresponding to the preset quantile interval of the reference PPK model can be performed by obtaining the values ​​corresponding to the reference model parameters in the preset quantile interval in the reference PPK model corresponding to multiple bootstrapped data sets. For example, the median value. The reference model parameter can be any parameter in the reference PPK model, for example, r-ATG clearance, r-ATG central compartment apparent distribution volume, and r-ATG inter-compartmental clearance, etc.

[0127] After obtaining the reference model parameters, the initial PPK model can be internally verified based on the reference model parameters and the corresponding parameters in the initial PPK model. The internal verification result can be determined by judging whether the parameters in the initial PPK model and the reference model parameters meet the preset conditions. For example, whether the difference between the reference model parameters and the parameters of the initial PPK model is less than a predetermined value, if so, it means that the initial PPK model has good stability and prediction performance.

[0128] Internal validation of the initial PPK model based on the reference PPK model obtained by fitting the bootstrap data is conducive to a more accurate estimation of the model parameters and statistics, thereby improving the reliability and validity of the initial PPK model.

[0129] In another embodiment, the initial PPK model is internally verified based on the reference model parameters and the parameters of the initial PPK model. Before obtaining the internal verification result, the method further includes:

[0130] Draw model fit diagnostic graphs based on the model validation group data and the initial PPK model;

[0131] The initial PPK model is internally verified based on the reference model parameters and the parameters of the initial PPK model to obtain internal verification results, including:

[0132] Based on the model fitting diagnostic diagram and the comparison results between the reference model parameters and the parameters of the initial PPK model, the initial PPK model was internally verified to obtain the internal verification results.

[0133] The model fit diagnostic plot can be used to evaluate the goodness of fit of the statistical model. It helps to understand whether the statistical assumptions such as residual distribution, homoscedasticity, and normality of the initial PPK model are met through visualization.

[0134] Model fitting diagnostic graphs may include: dependent variable-group predicted value scatter plot, dependent variable-individual predicted value scatter plot, conditional weighted residual-group predicted value diagnostic graph, and conditional weighted residual-time graph. The dependent variable-group predicted value scatter plot can be used to evaluate the prediction error between the actual value and the predicted value corresponding to the patient group of the model validation group data. The dependent variable-individual predicted value scatter plot can be used to evaluate the prediction error between the predicted value and the actual value obtained after prediction for each patient in the model validation group data. The conditional weighted residual-group predicted value diagnostic graph can intuitively understand the model fitting effect and potential problems by comparing the conditional weighted residual with the predicted value corresponding to multiple blood drug concentration data in the model validation group data of the initial PPK model. The conditional weighted residual-time graph can intuitively obtain the change of the prediction error of the initial PPK model over time by comparing the conditional weighted residual with time.

[0135] The model fitting diagnostic graph can be drawn by inputting the statistical characteristics of the children corresponding to the model validation group data and the blood drug concentration data into the initial PPK model, and then drawing the predicted values ​​and actual values ​​based on the initial PPK model.

[0136] After the model fitting diagnostic graph is obtained, the initial PPK model can be internally verified based on the model fitting diagnostic graph and the comparison result between the reference model parameters and the parameters of the initial PPK model.

[0137] The process of internally verifying the initial PPK model based on the comparison results of the reference model parameters and the parameters of the initial PPK model will not be described here. When the initial PPK model is internally verified based on the model fit diagnostic graph, the internal verification result can be obtained by comparing the fit between the model fit diagnostic graph and the initial PPK model. For example, if the initial PPK model is compared with the dependent variable-group prediction value scatter plot and the overlap is high, it means that the initial PPK model has good prediction performance for the group.

[0138] Double internal validation of the initial PPK model based on the model fitting diagnostic diagram and the reference PPK model is conducive to further improving the reliability and effectiveness of the internal validation of the initial PPK model.

[0139] After obtaining the internal validation results, the initial PPK model is externally validated based on the model validation group data to obtain the external validation results. The external validation can evaluate the precision and accuracy of the initial PPK model, so the external validation results can indicate the prediction error of the initial PPK model.

[0140] In one embodiment, the initial PPK model is externally validated based on the model validation group data to obtain an external validation result, including:

[0141] Multiple relative prediction error values ​​were calculated based on the model validation group data and the initial PPK model;

[0142] Calculating a first probability value that the relative prediction error value is within a first error range, and calculating a second probability value that the relative prediction error value is within a second error range;

[0143] An external verification result is determined based on the first probability value and the second probability value.

[0144] By inputting the statistical characteristics corresponding to the children in the model validation group data into the initial PPK model, the predicted blood drug concentration value predicted by the initial PPK model can be obtained. Based on the predicted blood drug concentration value and the actual blood drug concentration value in the model validation group data, multiple relative prediction error values ​​corresponding to the model validation group data can be obtained. The calculation process of the relative prediction error value can be expressed as Formula 1:

[0145]

[0146] In formula 1, i represents the i-th data in the model validation group data, PE% i It can represent the relative prediction error value corresponding to the i-th data, pred i represents the i-th predicted value, obs i Represents the corresponding i-th actual value.

[0147] After calculating and obtaining a plurality of relative prediction error values, a first probability value that the relative prediction error value is within a first error range and a second probability value that the relative prediction error value is within a second error range may be calculated. For example, the first error range is 20% (±20%) and the second error range is 30% (±30%). The first probability value may represent the probability that the error range of the relative prediction error value is within the first error range, and the second probability value may represent the probability that the error range of the relative prediction error value is within the second error range.

[0148] After obtaining the first probability value and the second probability value, the external verification result can be determined based on the first probability value and the second probability value. In one embodiment, when the first probability value and the second probability value meet a predetermined condition, it indicates that the external verification result is better and the model is accurate and effective. For example, the predetermined condition is that the first probability value is greater than 35% and the second probability value is greater than 50%.

[0149] By calculating the first probability value and the second probability value corresponding to the relative prediction error value for external verification, the accuracy and precision of the model prediction can be fully characterized, thereby improving the accuracy of external verification of the initial PPK model.

[0150] In another embodiment, before determining the external verification result based on the first probability value and the second probability value, the method further includes:

[0151] The average prediction error and the average absolute prediction error were calculated based on the model validation group data and the initial PPK model;

[0152] Determining an external verification result based on the first probability value and the second probability value includes:

[0153] An external validation result is determined based on the first probability value, the second probability value, the average prediction error, and the average absolute prediction error.

[0154] The calculation process of the average prediction error can be expressed as Formula 2:

[0155]

[0156] In Formula 2, MPE can represent the mean prediction error, N can represent the amount of data in the model validation group, and pred i Can represent the i-th predicted value, obs i It can represent the corresponding i-th actual value.

[0157] The calculation process of the mean absolute prediction error can be expressed as formula 3:

[0158]

[0159] In Formula 3, MAPE can represent the mean absolute prediction error. Similarly, N can represent the amount of data in the model validation group. i Can represent the i-th predicted value, obs i It can represent the corresponding i-th actual value.

[0160] After obtaining the average prediction error and the average absolute prediction error, an external verification result may be determined based on the first probability value, the second probability value, the average prediction error, and the average absolute prediction error.

[0161] In one embodiment, when the first probability value, the second probability value, the average prediction error, and the average absolute prediction error all meet the preset conditions, the external verification result may indicate that the model prediction effect is better. For example, the preset conditions are: the first probability value and the second probability value are both greater than the corresponding preset values, and the average prediction error and the average absolute prediction error are both less than the corresponding preset values.

[0162] In another embodiment, determining the external validation result based on the first probability value, the second probability value, the average prediction error, and the average absolute prediction error includes:

[0163] Determine, based on the initial PPK model, a first weight corresponding to the first probability value, a second weight corresponding to the second probability value, a third weight corresponding to the average prediction error, and a fourth weight corresponding to the average absolute error;

[0164] Perform weighted calculation on the first probability value, the second probability value, the average prediction error and the average absolute prediction error based on the first weight, the second weight, the third weight and the fourth weight to obtain a verification value;

[0165] The external verification result is determined based on the verification value and the preset reference value.

[0166] The first weight, the second weight, the third weight and the fourth weight can be used to indicate the influence of the first probability value, the second probability value, the average prediction error and the average absolute prediction error on the external verification result respectively. The first weight, the second weight, the third weight and the fourth weight can be preset based on the actual verification needs.

[0167] After determining the first weight, the second weight, the third weight, and the fourth weight, the first probability value, the second probability value, the average prediction error, and the average absolute prediction error can be weighted to obtain a verification value, and an external verification result can be determined based on the verification value and a preset reference value. Specifically, the verification value can be compared with the reference value, and when the comparison result meets a predetermined condition, the external verification result can indicate that the initial PPK model has a good prediction effect.

[0168] Based on the first weight, the second weight, the third weight and the fourth weight, the influence of the first probability value, the second probability value, the average prediction error and the average absolute prediction error on the external verification result can be adjusted according to the needs of the actual scenario, which is conducive to improving the flexibility of external verification.

[0169] Verifying the initial PPK model based on the first probability, the second probability, the average prediction error and the average absolute prediction error can further improve the reliability and effectiveness of the verification results.

[0170] In step 150, the initial PPK model may be modified according to the internal verification result and the external verification result to obtain a target PPK model.

[0171] Since the internal validation results can indicate the degree of fit of the initial PPK model to the data, and the external validation results can indicate the prediction error of the initial PPK model, when the initial PPK model is modified according to the internal validation results and the external validation results, the parameters in the initial PPK model can be adjusted in the direction of improving the degree of fit and reducing the prediction error, thereby obtaining the target PPK model.

[0172] According to another embodiment of the present disclosure, a method for determining a dosing regimen is provided. Figure 3 As shown, the method includes:

[0173] Step 310: Obtain statistical characteristic data of the target child;

[0174] Step 320, determining the metabolic data of r-ATG in the target child based on the statistical characteristic data and the target PPK model;

[0175] Step 330: Determine a target dosing regimen for the target child based on the metabolic data.

[0176] The method for determining the dosing regimen in the embodiments of the present disclosure can be applied to a variety of diseases that require r-ATG administration, such as α-thalassemia, β-thalassemia, severe aplastic anemia, etc.

[0177] The target child may be a child who needs to be administered r-ATG. The statistical characteristic data have been described in detail in the above examples and will not be repeated here.

[0178] The target PPK model may be a PPK model constructed by any of the aforementioned methods for constructing a PPK model for children using r-ATG.

[0179] Inputting the statistical characteristic data into the target PPK model can obtain the r-ATG metabolism data in the target patient. By evaluating the r-ATG metabolism data, the r-ATG metabolism level of the target patient can be determined, thereby determining an accurate target dosing regimen for the target patient.

[0180] Therefore, the method for determining the dosing regimen provided in the embodiments of the present disclosure can accurately evaluate the r-ATG metabolic level of the target child, thereby improving the safety and accuracy of r-ATG administration.

[0181] The embodiments of the present disclosure are described in detail in combination with specific application scenarios.

[0182] The following describes in detail the method for constructing the PPK model for children using r-ATG provided by the embodiments of the present disclosure in conjunction with specific scenarios.

[0183] First, the statistical characteristics of the children were collected from the hospital medical record system and the nursing system, including demographic data, the number of various lymphocytes, and laboratory test data. Demographic data included: hospital number, sex, age, height, weight, body surface area, main diagnosis, transplant type, risk level, donor sex, and donor-recipient HLA locus matching. The number of various lymphocytes: recipient lymphocytes, recipient total T lymphocytes, recipient CD4 T cells, recipient CD8 T lymphocytes, recipient B lymphocytes, recipient NK cells, donor lymphocytes, donor CD34 T lymphocytes, donor CD3 T lymphocytes, donor CD4 T lymphocytes, donor CD8 T lymphocytes, donor B lymphocytes, and donor NK cells. Laboratory test data included ferritin, cardiac ejection fraction, high-sensitivity C-reactive protein, hemoglobin, total protein, albumin, globulin, immunoglobulin G, alanine aminotransferase, aspartate aminotransferase, total bilirubin, and serum creatinine.

[0184] Each child was given r-ATG according to the preset dosing regimen: Haplo / MUD: 1 mg / kg / day*3 days (-3 / -2 / -1 day); MSD: 6 mg / kg*3 days (-3 / -2 / -1 day); intravenous drip.

[0185] Blood samples were collected according to the preset sampling plan. The preset sampling plan was: sampling 1-4 hours before administration and 15 minutes after administration; within 3 months after implantation, samples were collected once a week - routine blood tests and drug concentration samples were intercepted; venous blood was collected before and after each administration, and venous blood was collected every other week for one month after administration.

[0186] After blood sample collection, the blood concentration of active r-ATG can be measured by quantitative flow cytometry. By examining the blood concentration-time graph, the blood concentration data that do not conform to the pharmacokinetic curve trend are eliminated, and the retained blood concentration data are randomly divided into a model construction group and a model verification group at a ratio of 4:1. After data elimination, a total of 265 children's data were retained, of which 198 were included in the model construction group and 67 were included in the model verification group.

[0187] The data of multiple children were sorted to obtain the pharmacokinetic data file of r-ATG children. For example, Table 1 shows some examples of the pharmacokinetic data file of r-ATG children. Among them, the identifier indicates the corresponding child, and each child corresponds to a unique identifier; the date and time indicate the date and time of the event; the dosage indicates the dosage of r-ATG, and if the dosage is empty, it means that no drug was given in this event; the drip rate indicates the intravenous drip rate of r-ATG during this drug administration time; the event type is a description of the current event (1 indicates a drug administration event, 0 indicates an observed event); the observed value indicates the blood concentration of r-ATG (μg / mL); the missing observation value is 1, indicating that there is no observation value for this event, and 0 indicates that there is an observation value for this event; gender, age, and weight indicate the corresponding indicators of the child.

[0188]

[0189] Table 1

[0190] Multiple candidate PPK models may include: one-compartment model, two-compartment model, three-compartment model and Michaelis-Menten model. Each candidate PPK model is fitted by model construction group data to obtain the objective function value, model drawing diagnostic graph, statistical test and precision of parameter estimation value corresponding to each candidate PPK model. The inter-individual variation of the model can be described by formula 4:

[0191]

[0192] In Formula 4, P i Can represent individual parameters, P TV It can be expressed as the population parameter, η i represents the random effect of the ith individual, η i The mean is 0 and the variance is ω 2The normal distribution of .

[0193] The residual variation can be calculated based on the proportional residual model. Specifically, it can be expressed as Formula 5:

[0194] Y=F×(1+ε1)+ε2 (Formula 5).

[0195] In Formula 5, Y can be expressed as the observed value, F is the individual predicted value based on the model, ε is the residual variation, which has a mean of 0 and a variance of σ. 2 The normal distribution of .

[0196] The basic model determined based on the objective function value, model-drawn diagnostic graphs, statistical tests, and the precision of parameter estimates is a two-compartment model. The parameter expressions of the basic model include the following formulas 6 to 9:

[0197]

[0198]

[0199]

[0200]

[0201] In Formula 6-9, CL i represents the clearance rate of the ith patient; V 1，i represents the apparent distribution volume of the central compartment of the ith patient; Q i represents the inter-compartment clearance rate of the i-th patient; V 2，i represents the apparent distribution volume of the peripheral compartment of the ith patient, and η represents the inter-individual variation of the corresponding parameter.

[0202] Since each child has a large number of statistical characteristics, covariates with a significant impact on the pharmacokinetics of r-ATG can be selected and included in the basic model. During the screening process, the effect of each characteristic on pharmacokinetic parameters (clearance, central compartment apparent distribution volume, inter-compartmental clearance, and peripheral compartment apparent distribution volume) can be examined.

[0203] Statistical characteristics include categorical variables and numerical variables. The effect of categorical variables on pharmacokinetic parameters can be expressed as formula 10:

[0204]

[0205] In Formula 10, θ ij It can represent the jth pharmacokinetic parameter corresponding to the i-th patient; θ tv，jIt can represent the typical value of the group corresponding to the jth pharmacokinetic parameter, which can reflect the average level of the pharmacokinetic parameter in the child population; θ j Can represent the impact factor; η i It can represent the random effect of the jth pharmacokinetic parameter. COV can represent the classification of statistical features. When the statistical feature is a binary variable, COV can be equal to 0 or 1.

[0206] The effect of numerical variables on pharmacokinetic parameters can be expressed as formula 11:

[0207]

[0208] In Formula 11, θ ij It can represent the jth pharmacokinetic parameter corresponding to the i-th patient; θ tv，j It can represent the typical value of the population corresponding to the jth pharmacokinetic parameter; θ j Can represent the impact factor; η i It can represent the random effect of the jth pharmacokinetic parameter; COV can represent the value of this statistical feature; COV Median It can represent the median value of statistical characteristics.

[0209] When screening covariates, the forward inclusion process is first required. Each statistical feature is included in the basic model one by one, the pharmacokinetic parameters in the basic model are adjusted, and the corresponding reduction in the objective function value of the basic model after the statistical feature is included is calculated. As shown in Table 2, some examples of the changes in the objective function value corresponding to each statistical feature after incorporating it into the basic model are indicated.

[0210] Model Number Parameters - Covariates Objective function value Lower value p-value 0 Basic Model 2051.176 - - 1 Clearance-age 2036.879 -14.297 <0.001 2 Apparent volume of distribution of the central compartment-age 2042.629 -8.547 <0.05 3 Clearance-body weight 2007.249 -43.927 <0.001 4 Apparent volume of distribution in the central compartment-body weight 2045.383 -5.793 <0.05 5 Clearance-Height 2044.234 -6.942 <0.05 6 Apparent volume of distribution in the central compartment - sex 2042.594 -8.582 <0.05 7 Clearance-Body Surface Area 2034.038 -17.138 <0.001 8 Apparent volume of distribution of the central compartment-body surface area 2044.974 -6.202 <0.05 9 Intercompartmental clearance-body surface area 2012.805 -38.371 <0.001 10 Clearance rate-transplant type 2031.038 -20.138 <0.001 11 Apparent volume of distribution in the central compartment - transplant type 2037.022 -14.154 <0.001 12 Clearance rate - donor lymphocytes 2046.289 -4.887 <0.05 13 Clearance rate - donor CD34 T lymphocytes 2046.289 -4.887 <0.05 14 Clearance rate - donor CD3 T lymphocytes 2043.59 -7.586 <0.05 15 Clearance rate - donor CD8 T lymphocytes 2041.136 -10.04 <0.05 16 Apparent volume of distribution in the central compartment - risk level 2043.771 -7.405 <0.05

[0211] Table 2

[0212] As shown in Table 2, after the clearance pharmacokinetic parameters were adjusted using the "age" feature, the objective function value became 2036.8, a decrease of 14.297 compared to before the adjustment; after the central compartment apparent distribution volume was adjusted using the "weight" feature, the objective function value became 2045.383, a decrease of 5.793 compared to before the adjustment. Figure 4 As shown, the correlation between each statistical feature and pharmacokinetic parameters is indicated.

[0213] When screening covariates based on reduction values, statistical features with reduction values ​​greater than 3.84 can be used as candidate covariates, and p is less than 0.05, which is statistically significant.

[0214] Adding candidate covariates to the basic model can obtain the full model.

[0215] After the candidate covariates are screened out through the forward inclusion process, the reverse elimination process can be used to screen out covariates that have a significant effect on the pharmacokinetic parameters from the candidate covariates. The reverse elimination process specifically includes: removing each candidate covariate from the full model and calculating the increase in the objective function value before and after the elimination. If the increase exceeds 10.83, that is, p is less than 0.001, the candidate covariate can be considered to be significant and retained in the model, otherwise it will be eliminated. The candidate covariates that are finally retained in the model can be used as covariates that have a significant effect on the pharmacokinetic parameters.

[0216] Through the process of forward inclusion and backward elimination, it can be determined that age, weight, and body surface area have a significant effect on the clearance rate and the apparent distribution volume of the central compartment of the model, and the transplant type also has a significant effect on the clearance rate between compartments and the apparent distribution volume of the peripheral compartment. Due to the collinear relationship between age, weight, and body surface area, in actual clinical applications, only the effect of age on the clearance rate and the apparent distribution volume of the central compartment can be retained.

[0217] After the forward inclusion and backward elimination process, the covariates with significant effects on the pharmacokinetic parameters have been included in the model to obtain the initial PPK model.

[0218] The initial PPK model is verified through internal and external verification processes.

[0219] The internal validation process can include: bootstrap validation, model fit diagnostic plot validation, and predictive value calibration visual prediction test.

[0220] The bootstrap validation process is: fit 1000 bootstrap data to obtain the reference PPK model, calculate the median value and 95% confidence interval of the reference PPK model. If the parameters of the initial PPK model are close to the median value and within the 95% confidence interval, it can be proved that the model has good stability and predictive efficiency. The comparison of the various parameters of the initial PPK model and the various parameters of the reference PPK model can be expressed as Table 3:

[0221]

[0222] Table 3

[0223] Model fitting diagnostic plots include: dependent variable-group predicted value scatter plot, dependent variable-individual predicted value scatter plot, conditional weighted residual-group predicted value diagnostic plot, and conditional weighted residual-time diagnostic plot.

[0224] The process of verification using the model fitting diagnostic plot includes: evaluating the model's description of the central tendency of the data by comparing the degree of overlap between the trend line corresponding to the dependent variable-group predicted value scatter plot and the reference line (y=x); evaluating the model's central tendency and dispersion of individual data by comparing the degree of overlap between the trend line corresponding to the dependent variable-individual predicted value scatter plot and the reference line (y=x); evaluating the prediction performance of the initial PPK model by comparing the distribution of each scatter point on both sides of the reference line (y=0) in the conditional weighted residual-group predicted value diagnostic plot and the distribution of each scatter point on both sides of the reference line (y=0) in the conditional weighted residual-time diagnostic plot.

[0225] like Figure 5A and Figure 5B As shown, the four graphs are the dependent variable-group predicted value scatter plot, the dependent variable-individual predicted value scatter plot, the conditional weighted residual-group predicted value diagnostic plot, and the conditional weighted residual-time diagnostic plot. Figure 5A is the model fit diagnostic plot for the base model without covariate fitting, Figure 5B This is the model fit diagnostic graph corresponding to the initial PPK model after covariate fitting. Figure 5A and Figure 5B It can be found that: through the dependent variable-group prediction value scatter plot and the dependent variable-individual prediction value scatter plot, it can be seen that the overlap between the trend line and the reference line has been greatly improved, that is, the model fitting effect has been improved. It can be seen from the conditional weighted residual-group prediction value diagnostic plot and the conditional weighted residual-time diagnostic plot that compared with Figure 5A , Figure 5B In the figure, most of the data are evenly distributed on both sides of the reference line, which shows that the prediction performance of the model is greatly improved.

[0226] The prediction value calibration visualization prediction test includes: generating 1000 simulated data sets based on the basic model, parameter estimates and observed data. Calculating the distribution characteristics of the median, mean, 5% and 95% quantiles of the observed data set and the simulated data set at each time point. Figure 6 As shown in the figure, the solid line represents the 50th percentile of the observed data; the dotted line represents the 5th and 95th percentiles of the observed data; the shaded area represents the 90% prediction interval; the hollow origin represents the observed value corresponding to r-ATG. It can be seen that most of the observed values ​​are within the 90% prediction interval of the simulated data, and the 5th, 50th, and 95th percentile curves of the observed data all fall within the corresponding prediction intervals, which indicates that the model has good prediction performance.

[0227] The external validation process can be verified by calculating the first probability of the relative prediction error value with an error range of 20% and the second probability of the error range of 30%, the average prediction error and the average absolute prediction error. The external validation results are shown in Table 4:

[0228] drug Average prediction error Mean absolute prediction error First Probability Second probability r-ATG 15.88％ 26.29％ 37.30％ 55.00％

[0229] Table 4

[0230] When the external validation results meet the following conditions: the average prediction error is less than or equal to 20%; the average absolute prediction error is less than or equal to 30%; the first probability is greater than 35%; the second probability is greater than 50%, it indicates that the model is effective and has good prediction results.

[0231] Based on the internal and external verification of the initial PPK model, the final target PPK model can be obtained. The target PPK model can be expressed as the following formula 12-formula 15:

[0232]

[0233]

[0234]

[0235]

[0236] In the target PPK model, age (0.62-17.91) had an important effect on clearance (CL i ) and central distribution volume (V 1,i ), so age (AGE) was added to the model in the form of a power function (Formula 12 and Formula 13), and its influence factor was 0.669. Therefore, compared with younger children, older children have a greater clearance rate, and the dosing regimen for children can be adjusted according to age. For children with a larger body surface area (BSA), the clearance rate of r-ATG in the peripheral compartment is faster.

[0237] In addition, three different transplant types (TTPs) were screened in the target PPK model: HLA-matched sibling fully homozygous donor (MSD), haploidentical donor (HAPLO), and unrelated (MUD) hematopoietic stem cell transplantation. When the transplant type was MSD, the clearance rate of the child was 0.817 times that of the children who underwent other types of transplantation. Therefore, the dosage for children who underwent MSD was twice that of HAPLO and MUD patients.

[0238] It is to be understood that, although the steps in the above-mentioned flowcharts are sequentially displayed according to the characterization of arrows, these steps are not necessarily executed in sequence according to the order of arrow characterization. Unless there is a clear description in the present embodiment, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the above-mentioned flowcharts can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of the steps or stages in other steps.

[0239] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the terms used in this way can be interchangeable where appropriate, so that the embodiments of the present disclosure described herein can, for example, be implemented in an order other than those illustrated or described herein. In addition, the terms "comprises" and "comprising" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process or method comprising a series of steps is not necessarily limited to those steps clearly listed, but may include other steps that are not clearly listed or inherent to these processes or methods.

[0240] It should be understood that in the present disclosure, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0241] It should be understood that in the description of the embodiments of the present disclosure, the meaning of multiple (or multiple items) is more than two, greater than, less than, exceed, etc. are understood to not include the number, and above, below, within, etc. are understood to include the number.

[0242] In the several embodiments provided in the present disclosure, it should be understood that the disclosed method can be implemented in other ways.

[0243] It should also be understood that the various implementations provided in the embodiments of the present disclosure can be combined arbitrarily to achieve different technical effects.

[0244] The above is a specific description of the implementation methods of the present disclosure, but the present disclosure is not limited to the above implementation methods. Technical personnel familiar with the art can also make various equivalent modifications or substitutions without violating the spirit of the present disclosure. These equivalent modifications or substitutions are all included in the scope defined by the claims of the present disclosure.

Claims

1. A method for constructing a PPK model for children using r-ATG, characterized in that: The method comprises: Obtaining statistical characteristics of multiple children, wherein the statistical characteristics include age information and allo-HSCT transplant type information; Performing r-ATG drip infusion on the multiple children according to a preset dosing plan, and collecting blood samples from the multiple children according to a preset sampling plan to obtain blood drug concentration data of the multiple children; Dividing the blood drug concentration data of the plurality of children into model building group data and model verification group data, and constructing an initial PPK model based on the statistical characteristics and the model building group data; Performing internal verification on the initial PPK model based on the model verification group data to obtain an internal verification result, and performing external verification on the initial PPK model based on the model verification group data to obtain an external verification result; The initial PPK model is modified according to the internal verification result and the external verification result to obtain a target PPK model.

2. The method according to claim 1, characterized in that: The external verification of the initial PPK model based on the model verification group data to obtain an external verification result includes: Calculate a plurality of relative prediction error values ​​according to the model validation group data and the initial PPK model; Calculating a first probability value that the relative prediction error value is within a first error range, and calculating a second probability value that the relative prediction error value is within a second error range; An external verification result is determined based on the first probability value and the second probability value.

3. The method according to claim 2, characterized in that Before determining the external verification result based on the first probability value and the second probability value, the method further includes: Calculate the average prediction error and the average absolute prediction error according to the model validation group data and the initial PPK model; The determining of the external verification result based on the first probability value and the second probability value includes: An external validation result is determined based on the first probability value, the second probability value, the average prediction error, and the average absolute prediction error.

4. The method according to claim 3, characterized in that The determining of the external verification result based on the first probability value, the second probability value, the average prediction error and the average absolute prediction error comprises: Determine, based on the initial PPK model, a first weight corresponding to the first probability value, a second weight corresponding to the second probability value, a third weight corresponding to the mean prediction error, and a fourth weight corresponding to the mean absolute error; Performing weighted calculation on the first probability value, the second probability value, the average prediction error, and the average absolute prediction error based on the first weight, the second weight, the third weight, and the fourth weight to obtain a verification value; An external verification result is determined according to the verification value and a preset reference value.

5. The method according to any one of claims 1 to 4, characterized in that The internal verification of the initial PPK model based on the model verification group data to obtain an internal verification result includes: Obtain a preset number of bootstrap data, and obtain a reference PPK model based on the bootstrap data by fitting, and obtain reference model parameters corresponding to a preset quantile interval of the reference PPK model; The initial PPK model is internally verified based on the reference model parameters and the parameters of the initial PPK model to obtain an internal verification result.

6. The method according to claim 5, characterized in that Before performing internal verification on the initial PPK model based on the reference model parameters and the parameters of the initial PPK model to obtain an internal verification result, the method further includes: Draw a model fit diagnostic graph based on the model validation group data and the initial PPK model; The internal verification of the initial PPK model based on the reference model parameters and the parameters of the initial PPK model to obtain an internal verification result includes: The initial PPK model is internally verified based on the model fitting diagnostic graph and the comparison result between the reference model parameters and the parameters of the initial PPK model to obtain an internal verification result.

7. The method according to claim 1, characterized in that The constructing of the initial PPK model based on the statistical features and the model building group data comprises: Obtain multiple candidate PPK models; Fitting each of the candidate PPK models based on the model building group data to obtain multiple fitting models; Calculating an objective function value corresponding to each fitting model, and determining a basic model among the multiple candidate PPK models according to the objective function value; A plurality of model covariates are determined from the statistical features, and an initial PPK model is determined according to the plurality of model covariates and the basic model.

8. The method according to claim 7, characterized in that Determining a plurality of model covariates in the statistical features includes: Incorporating each of the statistical features into the basic model one by one, and evaluating the reduction value of the objective function value of the basic model when each feature is incorporated into the basic model; determining a plurality of candidate covariates among the plurality of features based on the reduction values; Eliminating the candidate covariates one by one from the basic model incorporating the statistical features, and evaluating the increase in the objective function when each candidate covariate is eliminated; A plurality of model covariates are determined among the plurality of candidate covariates based on the elevated values.

9. The method according to claim 1, characterized in that: The dividing the blood drug concentration data of the plurality of children into model building group data and model verification group data comprises: Determining a first amount of model building group data and a second amount of model validation group data based on the number of the plurality of children and a preset division ratio; A first number of blood drug concentration data are randomly determined from the blood drug concentration data of the multiple children as model building group data, and a second number of blood drug concentration data are determined from the blood drug concentration data of the multiple children excluding the model building group data as model verification group data.

10. A method for determining a dosing regimen, characterized in that: The method comprises: Obtain statistical characteristic data of target children; Determine the metabolic data of r-ATG in the target child based on the statistical characteristic data and the target PPK model, wherein the target PPK model is a PPK model constructed by the method for constructing a PPK model of r-ATG for children according to any one of claims 1 to 9; A target dosing regimen for the target child is determined based on the metabolic data.

Images