Method for adjusting dosing based on joint model according to adverse reaction risk

Abstract

The application discloses a method for adjusting a dosing amount based on a joint model according to an adverse reaction risk, and applies to the field of intelligent medical treatment.In the prior art, the relationship between a sirolimus steady-state trough concentration and a specific adverse reaction cannot be quantified, resulting in the defect that clinical medication lacks precise guidance.Through systematic research, the application first establishes a BKMR model to determine an individual target trough concentration range, obtains individual pharmacokinetic parameters of the patient through a population pharmacokinetic model, finally determines a medication time and a medication amount of the patient through a Bayesian negative feedback simulation, and uses the same to adjust a medication scheme.The application realizes prospective risk prediction of sirolimus adverse reactions, significantly improves the treatment safety of children with vascular diseases, and improves the long-term prognosis of patients.

Description

Technical Field

[0001] This invention relates to the field of intelligent healthcare, and more specifically, to a method for adjusting drug dosage based on adverse reaction risk using a combined model. Background Technology

[0002] Vascular diseases are a general term encompassing malformations of blood vessels and lymphatic vessels. Although their overall incidence in children is low, they exhibit high clinical and pathological heterogeneity. These diseases commonly occur in infancy and early childhood, affecting multiple systems throughout the body. They not only affect appearance but can also lead to serious complications such as pain, bleeding, functional impairment, and even high-output heart failure, posing a significant threat to the life, health, and quality of life of affected children. Existing research has widely confirmed the necessity of blood drug concentration monitoring in managing the safety of sirolimus treatment.

[0003] Nevertheless, a critical unresolved deficiency remains in current technology: the quantitative relationship between the steady-state trough concentration of sirolimus in children with vascular diseases and specific adverse reactions (especially myelosuppression, abnormal liver function, and dyslipidemia) is still lacking in a systematic and precise manner. Existing studies mostly remain at the level of qualitative or weak correlation descriptions. This technological gap directly results in: the lack of reliable, quantitative concentration target windows for clinical medication as guidance; and dosage adjustments relying heavily on physicians' personal experience and lagging clinical manifestations, leading to significant uncertainty and ambiguity.

[0004] Therefore, there is an urgent need in this field for a method that can systematically explore and quantify the complex relationship between sirolimus steady-state trough concentration and major adverse reactions, in order to overcome the above-mentioned deficiencies of existing technologies, provide clear medication guidance for clinical use, and ultimately improve the safety and efficacy of treatment for children with vascular diseases. Summary of the Invention

[0005] This invention aims to address at least one of the technical problems existing in the prior art. To this end, this invention provides a method for adjusting drug dosage based on adverse reaction risk using a combined model. The concentration range of sirolimus administered to a patient is determined based on a BKMR model, and individual pharmacokinetic parameters of the patient are determined using a PopPk model. Finally, the final dosing regimen for the patient is determined through Bayesian posterior feedback and Monte Carlo simulation.

[0006] The first aspect of this invention discloses a method for adjusting drug dosage based on adverse reaction risk using a combined model, comprising: S1: Obtain the lower limit of the target trough concentration of the target drug and clinical variables related to patient information, input the clinical variables related to patient information into the BKMR model to calculate the Bayesian confidence interval to determine the upper limit of the target trough concentration, and obtain the target trough concentration range; Furthermore, the method also includes data preprocessing, the data preprocessing process including: Obtain data on patients who have been prescribed sirolimus in their medical records; Select patient data diagnosed with vascular disease; Patient data lacking basic information, medication information, and laboratory test information were removed to obtain test samples; Optionally, the patient data may include: drug dosage, drug exposure, frequency of medication, and disease classification.

[0007] Furthermore, the clinical variables related to patient information are input into the BKMR model to calculate the Bayesian confidence interval to determine the upper limit of the target trough concentration, specifically as follows: The clinical variables were iterated through the BKMR model, and inter-group and intra-group importance comparisons were performed to obtain variables related to adverse reactions; The blood drug concentration of the variables that are strongly correlated with adverse reactions was determined by univariate exposure-response analysis to obtain the upper limit of the target trough concentration.

[0008] Furthermore, the types of adverse reactions include bone marrow suppression, abnormal liver function, and dyslipidemia.

[0009] Furthermore, the step of obtaining variables related to adverse reactions by iterating the clinical variables through the BKMR model and comparing their importance between and within groups specifically involves: iterating the clinical variables through the BKMR model and comparing their importance between groups using groupPIP, and comparing their importance within groups using CondPIP to obtain variables that are strongly correlated with predicting the occurrence of bone marrow suppression, abnormal liver function, and dyslipidemia.

[0010] Optionally, the clinical variables related to patient information include: drug exposure level, disease classification, basic patient information, and dosing strategy.

[0011] Optionally, the variables that are strongly correlated with adverse reactions are: drug exposure level and disease classification.

[0012] S2: Individual pharmacokinetic parameters of subjects calculated based on real-world clinical data of the target drug combined with the PopPK model; Furthermore, the individual pharmacokinetic parameters include: apparent volume of distribution and apparent clearance.

[0013] Furthermore, the method for calculating the individual pharmacokinetic parameters is as follows: Obtain the subject's current dosing regimen and blood drug concentration; The dosing regimen and blood drug concentration are input into the population pharmacokinetic model to obtain individual pharmacokinetic parameters.

[0014] Optionally, the population pharmacokinetic model is a PopPK model.

[0015] Furthermore, the specific steps of inputting the dosing regimen and blood drug concentration into the population pharmacokinetic model to obtain individual pharmacokinetic parameters are as follows: The dosing regimen and blood drug concentration are input into the PopPK model in the pharmacokinetic parameter calculation software to calculate the individual pharmacokinetic parameters of the patient.

[0016] Optionally, the pharmacokinetic parameter calculation software includes one or more of the following: NONMEM, Phoenix, Monolix.

[0017] S3: Obtain N simulated dosing regimens, where N is a natural integer; calculate the simulated blood drug concentration of the simulated dosing regimen based on the individual pharmacokinetic parameters of the subject; output the simulated dosing regimen in which the simulated blood drug concentration is within the target trough concentration range as the subject's dosing regimen.

[0018] Furthermore, S3 can be replaced by S3': obtaining a simulated dosing regimen, the simulated dosing regimen including drug dosage and dosing time; calculating the simulated blood drug concentration of the simulated dosing regimen based on the individual pharmacokinetic parameters of the subject; determining whether the simulated dosing regimen is the subject's dosing regimen based on the relative position of the simulated blood drug concentration and the target trough concentration range, the specific judgment criteria being: If the simulated blood drug concentration is within the target trough concentration range, then the simulated dosing regimen is determined as the subject's dosing regimen; If the simulated blood drug concentration is outside the target trough concentration range, the dosage or administration time is adjusted to obtain an updated dosing regimen. The simulated blood drug concentration is then recalculated using the updated dosing regimen until the simulated blood drug concentration is within the target trough concentration range.

[0019] Furthermore, the process of adjusting the drug dosage or administration time to obtain an updated dosing regimen, and recalculating the simulated blood drug concentration using the updated dosing regimen until the simulated blood drug concentration falls within the target trough concentration range, specifically involves: If the simulated blood drug concentration is below the target trough concentration range, the dosage is increased or the dosing time is decreased to obtain an updated dosing regimen. The simulated blood drug concentration is then recalculated using the updated dosing regimen until the simulated blood drug concentration is within the target trough concentration range. If the simulated blood drug concentration is above the target trough concentration range, the dosage is reduced or the dosing time is increased to obtain an updated dosing regimen. The simulated blood drug concentration is then recalculated using the updated dosing regimen until the simulated blood drug concentration is within the target trough concentration range.

[0020] A second aspect of this invention discloses a method for recommending the dosage of sirolimus, comprising: S61: Obtain the target valley concentration range of sirolimus, wherein the target valley concentration range is 5-7.5 ng / mL or 8-10.2 ng / mL; S62: Individual pharmacokinetic parameters of subjects calculated based on real-world clinical data of sirolimus combined with the PopPK population pharmacokinetic model; Furthermore, the method for calculating the individual pharmacokinetic parameters is as follows: Obtain the subject's current dosing regimen and blood drug concentration; The dosing regimen and blood drug concentration are input into the population pharmacokinetic model to obtain individual pharmacokinetic parameters.

[0021] Furthermore, the dosing regimen includes administering a first dose of sirolimus to the subject by a medical professional to obtain the patient's first blood dose.

[0022] Optionally, the first dose of sirolimus may be one of the following doses: 0.05 mg / q12h for ages 0-1.5 months, 0.1 mg / q12h for ages 1.5-4 months, 0.15 mg / q12h for ages 4-7 months, 0.2 mg / q12h for ages 7-10.5 months, 0.25 mg / q12h for ages 10.5-24 months, 0.5 mg / q12h for ages 2-4.5 years, 0.75 mg / q12h for ages 4.5-7 years, 1 mg / q12h for ages 7-11 years, 1.25 mg / q12h for ages 11-15 years, and 1.5 mg / q12h for ages 15-18 years.

[0023] S63: Obtain N simulated dosing regimens, where N is a natural integer; calculate the simulated blood drug concentration for each simulated dosing regimen based on the individual pharmacokinetic parameters of the subject, and output the simulated dosing regimen with the simulated blood drug concentration within the target trough concentration range as the subject's dosing regimen.

[0024] Furthermore, the calculation of the simulated blood drug concentration for each simulated dosing regimen based on the individual pharmacokinetic parameters of the subject specifically involves: calculating the simulated blood drug concentration for a single simulated dosing regimen based on the individual pharmacokinetic parameters in conjunction with a decision model that performs adaptive decision-making based on feedback signals.

[0025] Furthermore, the decision-making model based on feedback signals for adaptive decision-making includes: a population pharmacokinetic model and a Bayesian negative feedback model.

[0026] Furthermore, the calculation of the simulated blood drug concentration for a single simulated dosing regimen based on the individual pharmacokinetic parameters and the decision model that performs adaptive decision-making based on feedback signals specifically involves: calculating the ratio based on the apparent volume of distribution and apparent clearance to obtain the elimination rate constant; inputting the drug dose, dosing time, apparent volume of distribution, and elimination rate constant into the steady-state blood drug concentration formula to calculate the patient's blood drug concentration after the first dose and before the next dose; performing N simulations using a population pharmacokinetic model to generate a simulated blood drug concentration curve; calculating the steady-state trough concentration of the simulated blood drug concentration curve to obtain the simulated blood drug concentration of the simulated dosing regimen.

[0027] S63: Obtain N simulated dosing regimens, where N is a natural integer; calculate the simulated blood drug concentration for each simulated dosing regimen based on the individual pharmacokinetic parameters of the subject, and output the simulated dosing regimen in which the simulated blood drug concentration (steady-state trough concentration) is within the range of the target trough concentration as the subject's dosing regimen.

[0028] Furthermore, S63 can be replaced by S63': Obtain the simulated dosing regimen; Based on the individual pharmacokinetic parameters of the subjects, simulated blood drug concentrations for each simulated dosing regimen are calculated; the relative position of the simulated blood drug concentrations to the target trough concentration range of sirolimus is determined to determine whether the simulated dosing regimen is the subject's dosing regimen. The specific determination method is as follows: If the simulated blood drug concentration is within the target trough concentration range, then the simulated dosing regimen is determined as the subject's dosing regimen; If the simulated blood drug concentration is below the target trough concentration range, the dosage is increased or the dosing time is decreased to obtain an updated dosing regimen. The simulated blood drug concentration is then recalculated using the updated dosing regimen until the simulated blood drug concentration is within the target trough concentration range. If the simulated blood drug concentration is above the target trough concentration range, the dosage is reduced or the dosing time is increased to obtain an updated dosing regimen. The simulated blood drug concentration is then recalculated using the updated dosing regimen until the simulated blood drug concentration is within the target trough concentration range.

[0029] A third aspect of the present invention discloses a computer device, the device comprising: a memory and a processor; The memory is used to store program instructions; The processor is used to call program instructions, and when the program instructions are executed, the above method steps are implemented.

[0030] The fourth aspect of the present invention discloses a computer program product, including a computer program that is implemented by a processor using the above-described method steps.

[0031] The fifth aspect of the present invention discloses a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method steps.

[0032] Advantages of this invention: To address the core problem of existing technologies' inability to accurately quantify the relationship between sirolimus steady-state trough concentration and major adverse reactions, leading to a lack of precise guidance in clinical dosing regimen development and reliance on reactive adjustments based on lagging clinical manifestations, this invention provides an innovative solution, resulting in a series of significant beneficial effects. Addressing the fundamental deficiency of "unclear quantitative relationship," this invention, for the first time, constructs a precise exposure-response model between sirolimus steady-state trough concentration and bone marrow suppression, abnormal liver function, and dyslipidemia. Specifically, this application inputs patient information into a BKMR-based model to measure blood drug concentration, obtaining the upper limit of the target trough concentration; sets a lower limit of the target trough concentration based on clinical experience, obtaining an individual target trough concentration range; inputs the patient information into a PopPK model to obtain the patient's individual pharmacokinetic parameters; combines the patient's individual pharmacokinetic parameters with the individual target trough concentration range, and determines the patient's dosage through Bayesian negative feedback simulation. This invention significantly reduces the blindness and uncertainty of dosing, indirectly improving the long-term control rate of vascular diseases and the overall prognosis of children, providing crucial technical support for solving this clinical challenge. Attached Figure Description

[0033] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0034] Figure 1 This is a schematic flowchart of the method provided in the first aspect of the present invention; Figure 2 This is a graph showing the relationship between adverse reactions and quantitative blood drug concentration provided in an embodiment of the present invention; Figure 3 This is a specific reference index diagram of adverse reaction events provided in the embodiments of the present invention; Figure 4This is a diagram of adverse event evaluation criteria provided in an embodiment of the present invention; Figure 5 This is a schematic diagram of the method flow provided in the second aspect of the present invention; Figure 6 This is a schematic diagram illustrating the estimation of adverse reaction risk values ​​(compared to the median) under different blood drug exposure conditions based on the BKMR model provided in this embodiment of the invention. Figure 7 This is a schematic diagram of estimating the Bayesian confidence interval range (with 5.12 ng / mL as the lower limit of the target treatment window) based on the BKMR model according to an embodiment of the present invention. Figure 8 This is a schematic diagram of estimating the Bayesian confidence interval range (with 8.09 ng / mL as the lower limit of the target treatment window) based on the BKMR model according to an embodiment of the present invention. Figure 9 This is a diagram illustrating the complex interaction between adverse reactions and blood drug exposure based on the BKMR model, provided in this embodiment of the invention. Figure 10 This is a schematic diagram illustrating the BKMR-PopPK integrated framework-based clinical precision drug delivery provided in an embodiment of the present invention. Figure 11 This is a comparison chart of the methodological advantages and disadvantages provided in the embodiments of the present invention; Figure 12 This is a schematic diagram of a device for adjusting drug delivery dosage based on adverse reaction risk according to a joint model provided in an embodiment of the present invention. Detailed Implementation

[0035] To enable those skilled in the art to better understand the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0036] In some of the processes described in the specification, claims, and accompanying drawings of this invention, multiple operations appear in a specific order. However, it should be clearly understood that these operations may not be performed in the order they appear herein, or they may be performed in parallel. The operation numbers, such as S1, S2, etc., are merely used to distinguish different operations and do not themselves represent any execution order. Furthermore, these processes may include more or fewer operations, and these operations may be performed sequentially or in parallel.

[0037] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0038] Figure 1 This is a schematic flowchart of a method for adjusting drug dosage based on adverse reaction risk using a combined model, provided in an embodiment of the present invention. Specifically, it includes the following steps: S1: Obtain the lower limit of the target trough concentration of the target drug and clinical variables related to patient information, input the clinical variables related to patient information into the BKMR model to calculate the Bayesian confidence interval to determine the upper limit of the target trough concentration, and obtain the target trough concentration range; Furthermore, the method also includes data preprocessing, the data preprocessing process including: Obtain data on patients who have been prescribed sirolimus in their medical records; Select patient data diagnosed with vascular disease; Patient data lacking basic information, medication information, and laboratory test information were removed to obtain test samples; Optionally, the patient data may include: drug dosage, drug exposure, frequency of medication, and disease classification.

[0039] Furthermore, the patient information includes: age, weight, gender, dosage, drug exposure, frequency of medication, and disease classification.

[0040] Furthermore, the disease classification includes: simple vascular malformations and complex vascular malformations.

[0041] Furthermore, the types of adverse reactions include: predicted bone marrow suppression, abnormal liver function, and dyslipidemia.

[0042] Figure 2 This is a graph showing the relationship between adverse reactions and quantitative blood drug concentration provided in this embodiment of the invention. This invention retrieved all laboratory test indicators, blood drug concentration monitoring data, and demographic information of patients who received sirolimus prescriptions at Beijing Children's Hospital, affiliated with Capital Medical University, from January 2017 to May 2023. A total of 257 children were included in the analysis. Since no patients with renal abnormalities were found, modeling was only performed for three adverse reactions: bone marrow suppression, abnormal liver function, and dyslipidemia, with 766, 825, and 892 blood samples included for these three reactions, respectively. The entire inclusion and exclusion process is as follows: Figure 2-1As shown. Due to the retrospective nature of this study, it was approved by the Ethics Committee of Beijing Children's Hospital, Capital Medical University (

[2024] -Y-069-D), and informed consent from patients was waived.

[0043] In one specific embodiment, the inclusion and exclusion criteria for patients are as follows: ①Inclusion criteria: Diagnosed with vascular disease (according to ISSVA criteria); age ≤18 years; received oral sirolimus treatment for at least 7 days; at least one steady-state trough concentration (SCR) Test records and data from at least one laboratory safety assessment conducted on the same day (such as bone marrow suppression, liver function, and lipid-related laboratory indicators); complete patient basic information and medication information; ② Exclusion criteria: Concomitant use of drugs known to significantly affect the pharmacokinetics or target adverse reactions of sirolimus (such as potent CYP3A4, 3A45, P-gp inhibitors / inducers, hepatotoxic drugs, etc.); concomitant severe underlying diseases (such as severe hepatic or renal insufficiency, active infection, malignancy) that may significantly interfere with safety indicator assessment; missing key data during sirolimus treatment (patient basic information, medication information, laboratory test information). Based on the ISSVA classification criteria, the patient population was clearly divided into two categories: simple vascular malformations and complex vascular malformations.

[0044] Complex vascular malformations are further subdivided into several subcategories, including ① mixed vascular malformations, such as capillary-venous malformations; ② well-known vascular malformations, also known as "channel type" or "vascular trunk" vascular malformations; and ③ vascular malformations combined with other lesions, such as Klippel-Trenaunay syndrome and Sturge-Weber syndrome.

[0045] In one specific embodiment, the adverse event evaluation criteria adopted the National Cancer Institute's Common Adverse Event Evaluation Criteria, version 5.0 (NCI CTCAE v5.0). For the determination of lipid abnormalities, the relevant standards in the National Cholesterol Education Program (NCEP) established in 2012 were referenced. Reference was also made to the Chinese industry standards WS / T 780-2021 (Reference intervals of clinical biochemistry tests commonly used for children) and WS / T 779-2021 (Reference intervals of blood cell analysis for children). Specific reference indicators for adverse events are detailed below. Figure 4 The above reference indicators are used to determine whether adverse reactions have occurred. The Chinese industry standards WS / T 780-2021 and WS / T 779-2021 refer to the normal range of values ​​for patients. CTCAE v5.0 and NCEP are used to determine whether there are abnormalities and the degree of abnormality.

[0046] In one specific embodiment, since the same patient may not have undergone all blood routine and liver and kidney function tests during a single visit, missing values ​​are deleted in pairs to maximize data utilization.

[0047] Furthermore, the clinical variables related to patient information are input into the BKMR model to calculate the Bayesian confidence interval to determine the upper limit of the target trough concentration, specifically as follows: The clinical variables were iterated through the BKMR model, and inter-group and intra-group importance comparisons were performed to obtain variables related to adverse reactions.

[0048] The blood drug concentration of the variables that are strongly correlated with adverse reactions was determined by univariate exposure-response analysis to obtain the upper limit of the target trough concentration.

[0049] Furthermore, the step of obtaining variables related to adverse reactions by iterating the clinical variables through the BKMR model and comparing their importance between and within groups specifically involves: iterating the clinical variables through the BKMR model and comparing their importance between groups using groupPIP, and comparing their importance within groups using CondPIP to obtain variables that are strongly correlated with predicting the occurrence of bone marrow suppression, abnormal liver function, and dyslipidemia.

[0050] In one embodiment, the BKMR model is as follows: ,in, Represents the response variable. This can be understood as the hidden function value corresponding to the kernel function. The variables can reflect the importance weight of each exposed variable. Represents the coefficient vector. Represents a covariate vector. This represents the error term, in relation to C. min When modeling the quantitative relationship with adverse reactions, Only C min (This refers to the exposure variable, i.e., the important independent variable we want to explore), and the remaining variables are set as confounding factors (covariates); when discussing the relative contributions and joint effects among different variables, Set as all variables (see) Figure 9-2 and Figure 9-5 ).

[0051] In one embodiment, the step of inputting the test sample into a BKMR-based model to determine the blood drug concentration and obtain the upper limit of the target trough concentration specifically involves: grouping the patient information according to clinical significance into four groups: drug exposure level, disease classification, patient basic information, and dosing strategy; iterating the above four groups of variables through the BKMR model and comparing the importance between groups using groupPIP, and comparing the importance within groups using CondPIP to determine that the variables with strong correlations to predicting the occurrence of bone marrow suppression, abnormal liver function, and dyslipidemia are drug exposure level and disease classification; and measuring the blood drug concentration through univariate exposure-response analysis to obtain the upper limit of the target trough concentration.

[0052] In one embodiment, setting the target trough concentration lower limit based on clinical experience specifically involves setting 8 ng / mL as the baseline value based on experience, taking the 67.5th percentile of the BKMR model, and determining the target trough concentration lower limit as 8.09 ng / mL.

[0053] Furthermore, the clinical variables related to patient information include: drug exposure level, disease classification, basic patient information, and dosing strategy.

[0054] Furthermore, the variables that are strongly correlated with adverse reactions are: drug exposure level and disease classification.

[0055] In one specific embodiment, the drug used was sirolimus capsules, 0.5 mg in strength. The steady-state time for blood drug concentration is approximately 5-7 days; therefore, blood drug concentration monitoring was conducted at weeks 1, 2, 4, 8, 12, and 24 after the start of treatment. Blood samples of 2-4 mL were collected half an hour before administration. Sirolimus blood concentration was determined using fluorescence polarization immunoassay (FPIA), technically supported by Abbott Laboratories (Abbott Park, IL, USA), using a TDX / FLX device. The limit of quantitation was 2.5 ng / mL.

[0056] In one specific embodiment, there is currently no unified standard for the effective concentration of sirolimus for vascular diseases. Most research and medical institutions set 5 ng / mL as the effective concentration (reference: Saibene AM, Rosso C, Felisati G, et al. Sirolimus treatment for paediatric head and necklymphatic malformations: a systematic review. Eur Arch Otorhinolaryngol. 2023;280(8):3529-3540.), which is based on the drug's instructions. Based on our institution's accumulated clinical experience, we set 8 ng / mL as the effective concentration.

[0057] Furthermore, the quantile values ​​range from 67% to the 67% range.

[0058] Furthermore, the upper limit of the target valley concentration is taken as the value above the 67.5th percentile. The 95% Bayes confidence interval of the difference between each percentile and the baseline is calculated, and the upper limit of the target valley concentration is obtained as 10.19 ng / mL.

[0059] In a specific embodiment, based on the two different initial concentrations mentioned above, the corresponding BKMR models are selected to calculate the minimum drug exposure concentrations corresponding to statistically significant differences in the occurrence of the two adverse reactions in patients. This is subsequently referred to as the upper limit of the target trough concentration. Unlike probabilistic approaches (where a p-value less than or equal to a specific value, usually 0.05, is used to reject the null hypothesis), Bayesian statistics can accept or reject the null hypothesis. Determining the upper limit requires two steps: first, a statistical difference test, i.e., calculating the 95% Bayesian confidence interval (95% CrI) between the upper and lower limits. If it does not contain 0, then the risk of adverse reactions at the maximum tolerated concentration and the initial concentration is considered to be statistically significant; if it contains 0, it indicates that the direction of the effect is uncertain, and there is no statistically significant difference in the risk of adverse reactions between the two. Second, the lower limit of 95% CrI should be greater than 0.05 to ensure that the adjustment regimen has clinical significance.

[0060] In one specific embodiment, a quantitative analysis of the nonlinear relationship between blood drug concentration and the risk of abnormal liver function is performed based on a Bayesian nuclear mechanical regression (BKMR) model. 8 ng / mL is set as the baseline value (i.e., the lower limit of the target therapeutic window, corresponding to the 67.5th percentile, approximated as 8.09 ng / mL for simplified calculation), and gradually adjusted upwards to higher percentiles (e.g., 75%, 80%, etc.) as candidate upper limits. The 95% Bayesian confidence interval (CrI) between each percentile and the baseline is calculated; a difference is considered statistically and clinically significant when the CrI does not contain 0 and its lower limit is greater than 0.05. This process is used to determine the smallest upper limit of the therapeutic window that is significantly different from the lower limit and clinically significant. Within this range, adjusting the blood drug concentration is not considered to reduce the risk of adverse reactions.

[0061] In one specific embodiment, for continuous data, those conforming to a normal distribution are expressed as mean ± standard deviation (Mean ± SD), while those not conforming to a normal distribution are expressed as the median and interquartile range (IQR). Categorical data are expressed as frequency, rate, or proportion. For comparisons between groups of continuous data, those conforming to a normal distribution are expressed using... For non-normally distributed data, the rank-sum test is used; for categorical data, the chi-square test is used. (inspection) or Precise testing. After adjusting for confounding factors, multivariate logistic regression models and the RCS function were used to preliminarily explore the linear and nonlinear relationships between drug exposure and adverse events. Basic information on patients included in this study can be found in [link to patient data]. Figure 3 The standard of inspection is .

[0062] S2: Individual pharmacokinetic parameters of subjects calculated based on real-world clinical data of the target drug combined with the PopPK model; Furthermore, the individual pharmacokinetic parameters include: apparent volume of distribution and apparent clearance.

[0063] Furthermore, the method for calculating the individual pharmacokinetic parameters is as follows: Obtain the subject's current dosing regimen and blood drug concentration; The dosing regimen and blood drug concentration are input into the population pharmacokinetic model to obtain individual pharmacokinetic parameters.

[0064] Optionally, the population pharmacokinetic model is a PopPK model.

[0065] The current dosing regimen is the dosing regimen prior to the simulated dosing.

[0066] Furthermore, the specific steps of inputting the dosing regimen and blood drug concentration into the population pharmacokinetic model to obtain individual pharmacokinetic parameters are as follows: The dosing regimen and blood drug concentration are input into the PopPK model in the pharmacokinetic parameter calculation software to calculate the individual pharmacokinetic parameters of the patient.

[0067] Optionally, the pharmacokinetic parameter calculation software includes one or more of the following: NONMEM, Phoenix, Monolix.

[0068] Optionally, the population pharmacokinetic model is a PopPK model.

[0069] In one specific embodiment, patient drug exposure information is collected, and individualized pharmacokinetic parameters are calculated based on the PopPK model previously constructed and validated by the inventors (Reference: Liu B, Zhang X, Zhao Y, Xu X, Wang S, Wang X, Cheng X. Model-Informed individualized dosage regimen of sirolimus in pediatric patients with intractable lymphatic malformations. EurJ Pharm Sci. 2024 Sep 1;200:106837.). The formula is as follows: , , in, It represents the apparent volume of distribution in liters (L), indicating the extent to which the drug is distributed in the body, taking into account the bioavailability (F) after oral administration. A typical value is 292.57 L when the patient is 4.64 years old. Apparent clearance rate is expressed in liters per hour (L / h), representing the rate at which the body clears the drug. Bioavailability (F) is also taken into account, with a typical value of 5.56 L / h at an age of 4.64 years. Indicates age; 0.82 and 0.65 refer to exponential parameters, representing the relationship between age and... and The extent of the impact, specifically, for The influence index of age is 0.82, indicating that... It increases with age, but not non-linearly (power function relationship), for The influence index of age is 0.65, indicating that... It increases with age, but at a slower rate; and This refers to exponential random effects used to capture inter-individual variation, where... and It is a random variable with a mean of 0 and a variance of . The normal distribution and Ensure that the parameter values ​​are always positive and quantify the differences between individual patients and typical values.

[0070] S3: Obtain N simulated dosing regimens, where N is a natural integer; calculate the simulated blood drug concentration of the simulated dosing regimen based on the individual pharmacokinetic parameters of the subject; output the simulated dosing regimen in which the simulated blood drug concentration is within the target trough concentration range as the subject's dosing regimen.

[0071] Furthermore, S3 can be replaced by S3': obtaining a simulated dosing regimen, the simulated dosing regimen including drug dosage and dosing time; calculating the simulated blood drug concentration of the simulated dosing regimen based on the individual pharmacokinetic parameters of the subject; determining whether the simulated dosing regimen is the subject's dosing regimen based on the relative position of the simulated blood drug concentration and the target trough concentration range, the specific judgment criteria being: If the simulated blood drug concentration is within the target trough concentration range, then the simulated dosing regimen is determined as the subject's dosing regimen; If the simulated blood drug concentration is outside the target trough concentration range, the dosage or administration time is adjusted to obtain an updated dosing regimen. The simulated blood drug concentration is then recalculated using the updated dosing regimen until the simulated blood drug concentration is within the target trough concentration range.

[0072] Furthermore, the process of adjusting the drug dosage or administration time to obtain an updated dosing regimen, and recalculating the simulated blood drug concentration using the updated dosing regimen until the simulated blood drug concentration falls within the target trough concentration range, specifically involves: If the simulated blood drug concentration is below the target trough concentration range, the dosage is increased or the dosing time is decreased to obtain an updated dosing regimen. The simulated blood drug concentration is then recalculated using the updated dosing regimen until the simulated blood drug concentration is within the target trough concentration range. If the simulated blood drug concentration is above the target trough concentration range, the dosage is reduced or the dosing time is increased to obtain an updated dosing regimen. The simulated blood drug concentration is then recalculated using the updated dosing regimen until the simulated blood drug concentration is within the target trough concentration range.

[0073] In some specific embodiments, after S3, the method further includes: when the simulated blood drug concentrations calculated from the N simulated dosing regimens do not fall within the target trough concentration range, calculating the simulated blood drug concentration of the simulated dosing regimen closest to the target trough concentration range based on the individual pharmacokinetic parameters of the subject; determining whether the simulated dosing regimen is the subject's dosing regimen based on the relative position of the simulated blood drug concentration to the target trough concentration range, specifically based on the following criteria: If the simulated blood drug concentration is within the target trough concentration range, then the simulated dosing regimen is determined as the subject's dosing regimen; If the simulated blood drug concentration is outside the target trough concentration range, the dosage or administration time is adjusted to obtain an updated dosing regimen. The simulated blood drug concentration is then recalculated using the updated dosing regimen until the simulated blood drug concentration is within the target trough concentration range.

[0074] Figure 5 The present invention also discloses a schematic flowchart of a recommended method for sirolimus administration, comprising: S61: Obtain the target valley concentration range of sirolimus, wherein the target valley concentration range is 5-7.5 ng / mL or 8-10.2 ng / mL; S62: Individual pharmacokinetic parameters of subjects calculated based on real-world clinical data of sirolimus combined with the PopPK population pharmacokinetic model; Furthermore, the method for calculating the individual pharmacokinetic parameters is as follows: Obtain the subject's current dosing regimen and blood drug concentration; The dosing regimen and blood drug concentration are input into the population pharmacokinetic model to obtain individual pharmacokinetic parameters.

[0075] Furthermore, the dosing regimen includes administering a first dose of sirolimus to the subject by a medical professional to obtain the patient's first blood dose.

[0076] Optionally, the first dose of sirolimus may be one of the following doses: 0.05 mg / q12h for ages 0-1.5 months, 0.1 mg / q12h for ages 1.5-4 months, 0.15 mg / q12h for ages 4-7 months, 0.2 mg / q12h for ages 7-10.5 months, 0.25 mg / q12h for ages 10.5-24 months, 0.5 mg / q12h for ages 2-4.5 years, 0.75 mg / q12h for ages 4.5-7 years, 1 mg / q12h for ages 7-11 years, 1.25 mg / q12h for ages 11-15 years, and 1.5 mg / q12h for ages 15-18 years.

[0077] S63: Obtain N simulated dosing regimens, where N is a natural integer; calculate the simulated blood drug concentration for each simulated dosing regimen based on the individual pharmacokinetic parameters of the subject, and output the simulated dosing regimen with the simulated blood drug concentration within the target trough concentration range as the subject's dosing regimen.

[0078] Furthermore, the calculation of the simulated blood drug concentration for each simulated dosing regimen based on the individual pharmacokinetic parameters of the subject specifically involves: calculating the simulated blood drug concentration for a single simulated dosing regimen based on the individual pharmacokinetic parameters in conjunction with a decision model that performs adaptive decision-making based on feedback signals.

[0079] Furthermore, the decision-making model based on feedback signals for adaptive decision-making includes: a population pharmacokinetic model and a Bayesian negative feedback model.

[0080] Furthermore, the calculation of the simulated blood drug concentration of a single simulated dosing regimen based on the individual pharmacokinetic parameters and the decision model that performs adaptive decision-making based on feedback signals specifically involves: calculating the ratio based on the apparent volume of distribution and apparent clearance to obtain the elimination rate constant; inputting the drug dose, dosing time, apparent volume of distribution, and elimination rate constant into the population pharmacokinetic formula to calculate the patient's blood drug concentration curve and steady-state trough concentration after the first dose and before the next dose; performing N simulations using a posterior Bayesian negative feedback model to generate a simulated blood drug concentration curve; calculating the steady-state trough concentration of the simulated blood drug concentration curve to obtain the simulated blood drug concentration of the simulated dosing regimen.

[0081] S63: Obtain N simulated dosing regimens, where N is a natural integer; calculate the simulated blood drug concentration for each simulated dosing regimen based on the individual pharmacokinetic parameters of the subject, and output the simulated dosing regimen with the simulated blood drug concentration within the target trough concentration range as the subject's dosing regimen.

[0082] Furthermore, S63 can be replaced by S63': Obtain the simulated dosing regimen; Based on the individual pharmacokinetic parameters of the subjects, simulated blood drug concentrations for each simulated dosing regimen are calculated; the relative position of the simulated blood drug concentrations to the target trough concentration range of sirolimus is determined to determine whether the simulated dosing regimen is the subject's dosing regimen. The specific determination method is as follows: If the simulated blood drug concentration is within the target trough concentration range, then the simulated dosing regimen is determined as the subject's dosing regimen; If the simulated blood drug concentration is below the target trough concentration range, the dosage is increased or the dosing time is decreased to obtain an updated dosing regimen. The simulated blood drug concentration is then recalculated using the updated dosing regimen until the simulated blood drug concentration is within the target trough concentration range. If the simulated blood drug concentration is above the target trough concentration range, the dosage is reduced or the dosing time is increased to obtain an updated dosing regimen. The simulated blood drug concentration is then recalculated using the updated dosing regimen until the simulated blood drug concentration is within the target trough concentration range.

[0083] In some specific embodiments, after S63, the method further includes: when the simulated blood drug concentrations calculated from N simulated dosing regimens do not fall within the target trough concentration range, calculating the simulated blood drug concentration of the simulated dosing regimen closest to the target trough concentration range based on the individual pharmacokinetic parameters of the subject; determining whether the simulated dosing regimen is the subject's dosing regimen based on the relative position of the simulated blood drug concentration and the target trough concentration range, specifically based on the following criteria: If the simulated blood drug concentration is within the target trough concentration range, then the simulated dosing regimen is determined as the subject's dosing regimen; If the simulated blood drug concentration is outside the target trough concentration range, the dosage or administration time is adjusted to obtain an updated dosing regimen. The simulated blood drug concentration is then recalculated using the updated dosing regimen until the simulated blood drug concentration is within the target trough concentration range.

[0084] In one embodiment, the steady-state blood drug concentration formula is specifically as follows: , in, Indicates steady-state valley concentration. This refers to the elimination rate constant, where V represents the apparent volume of distribution, CL represents the clearance rate, and F represents bioavailability. The value represents the absorption rate constant, Dose represents the dosage, and t represents the administration time.

[0085] In one specific embodiment, the data was used to construct a Bayesian kernel regression model between different adverse events and drug exposures, and 20,000 iterations were performed using the Markov chain Monte Carlo algorithm, as shown in the following formula: , in, This refers to the result of adverse reactions. This refers to the exposed variables. It refers to a variable that contains a series of confounding factors. This refers to a feature with a high exposure-response function, using a Gaussian kernel. This represents the error term of an independent and identically distributed normal distribution.

[0086] In response to When modeling the quantitative relationship with adverse reactions, only... All variables were set as exposure factors, and the remaining variables were set as confounding factors. When discussing the relative contributions and combined effects among different variables, all variables were set as exposure factors. This study used the Posterior Predictive Check (PPC) to test the model's fit and generative ability, and Calibration Plot19 to test the accuracy of the model's predicted probabilities. This study used posterior inclusion probabilities (PIPs) to represent the degree of importance of different variables, ranging from 0 to 1. To assess the impact of key clinical variables at different levels on the risk of adverse reactions, this study performed risk estimation based on the posterior distribution obtained from MCMC sampling. Variables with higher PIPs were first selected as primary exposures, and these variables were fixed at their 25th, 50th, and 75th percentiles, respectively. The remaining variables maintained the observed data distribution or were set as reference values. For each quantile combination, Monte Carlo simulations were performed based on the posterior distribution of the BKMR model. Repeated sampling was used to generate predicted risk values, and the corresponding median and 95% Bayesian confidence interval (95% CrI) for each group were calculated to construct a quantile-risk relationship diagram.

[0087] In one specific embodiment, the therapeutic window was reached by simulating and comparing different dosages using the target trough concentration range and pharmacokinetic parameters to determine the precise dosage and regimen. Statistical analysis was performed using R studio (Version 2025.05.0+496) and IBM SPSS Statistics 25.0, and plotting was done using R studio (Version 2025.05.0+496), OriginPro (Version 2021), and Adobe Illustrator (2023). Key analytical code may be provided upon reasonable request.

[0088] In one specific embodiment, a total of 257 pediatric patients with vascular diseases were included, with a mean age of 4.18 ± 4.16 years, a male-to-female ratio of 141 / 116, and a mean weight of 19.98 ± 15.68 kg. The total daily dose of sirolimus was 1.06 ± 0.56 mg / day, and the total treatment duration was 210.92 ± 291.42 days. The percentages of patients who took the medication 1, 2, and 3 times / day were 82.29% (734 / 892), 17.49 (156 / 892), and 0.22% (2 / 892), respectively. Steady-state trough concentrations were observed during the follow-up period. The concentration was 7.68 ± 4.08 ng / mL, and the proportion of complex vascular malformations was 63.04% (162 / 257). Basic information of the children can be found here. Figure 3 According to established criteria, the incidence of adverse reactions was statistically analyzed. In the evaluable sample, the incidence rates of bone marrow suppression, abnormal liver function, and dyslipidemia were 30.14% (62 / 209), 43.44% (106 / 244), and 64.34% (157 / 244), respectively. The proportions of different severity levels and statistical results are shown below. Figure 2-2 . Figure 2-3 Correlation analysis between potential influencing factors and adverse reactions revealed that the occurrence of myelosuppression showed a negative correlation with treatment duration but a positive correlation with dosing frequency. For abnormal liver function, its occurrence was negatively correlated with age, weight, and daily dosage, but positively correlated with steady-state trough concentration. For dyslipidemia, its occurrence was also negatively correlated with age, weight, and daily dosage, but positively correlated with disease type, treatment duration, and steady-state trough concentration. This invention incorporates all potential influencing factors to construct a multifactorial model. Regression models were used to explore potential influencing factors of adverse reactions. See details below. Figure 2-4① Age, weight, drug exposure, and frequency of medication were statistically significant in increasing the risk of myelosuppression (P < 0.05); ② Age, weight, drug dosage, drug exposure, and frequency of medication were statistically significant in increasing the risk of abnormal liver function (P < 0.05); ③ Treatment duration, age, weight, gender, and drug exposure were statistically significant in increasing the risk of lipid abnormalities (P < 0.05). This invention incorporates factors with significant influence in a multivariate logistic regression model, constructing RCS models for steady-state trough concentration and three adverse reactions. The results show a significant nonlinear relationship between the two. (See attached data). Figure 2-5 (AC).

[0089] Figure 9 This is a diagram illustrating the complex interaction between adverse reactions and blood drug exposure based on the BKMR model, provided in this embodiment of the invention. Figure 9-1 This demonstrates the process of exploring covariates and binary dependent variables based on the BKMR model. Different variables are grouped according to their clinical significance into four groups: drug exposure level (…). Disease classification, patient basic information (gender, age, weight), and dosing strategy (daily dose, dosing frequency, treatment duration) were included. Group PIP and CondPIP were used to compare the importance of variables between and within groups. The BKMR model consistently confirmed that the steady-state trough concentration of sirolimus (…) Drug exposure and disease classification (especially complex vascular malformations) are the strongest and most stable drivers for predicting the occurrence of myelosuppression, abnormal liver function, and dyslipidemia (PIP close to 1). Drug exposure and disease classification play a dominant role in the occurrence of all three adverse reactions, while age, weight, daily dosage, and duration of treatment have different effects on the occurrence of each adverse reaction (see [link to relevant documentation]). Figure 9-2 The BKMR model for liver function abnormalities and dyslipidemia exhibited excellent overall goodness of fit (Posterior Predictive Check, PPC) and predictive accuracy, with robust and reliable results. The myelosuppression model accurately predicted the average risk trend, but the Calibration Plot test indicated high predictive uncertainty, especially in the intermediate-to-high risk range. This suggests that the occurrence of myelosuppression may be influenced by more individual factors that are not fully captured. The model results are more suitable for identifying risk trends and key drivers than for precise individual predictions of absolute risk. (See [link to relevant documentation]). Figure 9-3 Univariate exposure-response analysis revealed There is a significant non-linear association between [the risk of] adverse reactions and [the risk of] dyslipidemia. For example, for dyslipidemia, when [the risk of] adverse reactions increases... At that time, the probability of risk shows a steep upward trend, see Figure 9-4 For specific risk values ​​and 95% confidence intervals, please refer to [link / reference]. Figure 6To explore complex interactions among variables, single covariates with high PIP values ​​were fixed at specific quantiles (ranging from 0.25 to 0.75), while the remaining variables were fixed at the median. (See...) Figure 9-5 Taking myelosuppression as an example, with a fixed drug exposure level, the incidence of myelosuppression in patients with simple vascular malformations did not change significantly with age, while the incidence of myelosuppression decreased in patients with complex vascular malformations after reaching the 50th percentile. Furthermore, the incidence of myelosuppression decreased with increasing treatment duration in both patients with simple and complex vascular malformations. In summary, the multivariate exposure-response plot further demonstrates the effectiveness of drug exposure (…). The drug showed a high correlation with the occurrence of all three adverse reactions, but its effect on myelosuppression was lower compared to the other two, which may be related to the disease mechanism. Prolonged treatment significantly increased the risk of myelosuppression and dyslipidemia, especially in patients with complex vascular malformations, but its effect on liver function abnormalities was relatively weak, requiring a reassessment of whether prolonged treatment provides any benefit to patients. Age and weight showed a slight effect on certain adverse reactions (moderate PIP), but the direction / intensity of the effect needs to be explained in conjunction with specific curves. The effect of daily dosage on adverse reactions may be partially influenced by... The explanation is that its independent contribution to the model is relatively small; the increased risk of medication use due to complex vascular malformations, after adjustment... After considering other covariates, children with complex vascular malformations had a significantly higher risk of experiencing all three adverse events than children with simple lesions.

[0090] Figure 10 This is a schematic diagram illustrating the BKMR-PopPK integrated framework for assisting precise clinical drug delivery, as provided in an embodiment of the present invention. Figure 10-1 B and C show that the patient's liver function-related indicators showed an abnormally high trend; Figure 10-2 It demonstrates that, given a lower limit of the target valley concentration (8.09 ng / mL), the upper limit of the target valley concentration is determined to be 10.19 ng / mL; Figure 10-3 It demonstrates the use of real-world clinical data combined with the PopPK model to calculate individual patient parameters; Figure 10-4 Different dosing regimens were visualized, and subsequent treatment plans for patients were determined based on whether the target trough concentration range was reached. This framework quantifies the risk of adverse reactions in individual patients under different dosing regimens, providing clinicians with an actionable quantitative tool to select individualized doses that minimize adverse reaction risk under specific efficacy goals, achieving a precise risk-benefit balance. For a patient with lymphangiomas (admission age 3 months 27 days, weight 6.5 kg), a medication assessment was performed on day 25 of sirolimus administration. The assessment revealed a potential risk of abnormal liver function. (See...) Figure 10-1B, C. Currently, after comprehensive consideration with clinical practice, it is necessary to reduce the target trough concentration to approximately 8 ng / mL. How can the dosage be adjusted while minimizing the risk of abnormal liver function? The specific implementation steps are as follows: ① Based on the BKMR model, the target trough concentration range for abnormal liver function is determined to be 8.09-10.19 ng / mL (within this range, 95% CrI does not include 0, and the lower limit of 95% CrI is greater than 0.05); ② Based on real-world data and the PopPK model established by our research group, simulations are performed to determine the individual pharmacokinetic parameters for each patient: V / F approximately 43.725 L, CL / F approximately 0.847 L / h; ③ Using the individual pharmacokinetic parameters and the individual target trough concentration range, Bayesian negative feedback simulation is applied to determine the individualized treatment plan for each patient: adjusted to 0.1 mg q12h. The complete procedure is as follows: Figure 10 The lower limits of the target trough concentration were set at 5 ng / mL and 8 ng / mL, respectively. The upper limits of the corresponding target trough concentrations were calculated using the BKMR model. Within these ranges, the incidence of adverse reactions did not show statistical or clinical differences. The target trough concentration ranges for the three adverse reactions—bone marrow suppression, abnormal liver function, and dyslipidemia—were 5.12-15.96 ng / mL, 5.12-7.54 ng / mL, 5.12-7.84 ng / mL, 8.09-15.96 ng / mL, 8.09-10.19 ng / mL, and 8.09-10.19 ng / mL, respectively. The 95% CrI range is shown below. Figure 7 , Figure 8 In summary, the recommended target trough concentration ranges are 5-7.5 ng / mL or 8-10.2 ng / mL.

[0091] In one embodiment, the present invention successfully quantified the steady-state trough concentration of sirolimus for the first time in a cohort of pediatric vascular diseases. The nonlinear exposure-response relationship between [the disease] and key adverse reactions (myelosuppression, abnormal liver function, and dyslipidemia) was demonstrated in this invention using an innovative BKMR model. These findings reveal the strongest and most stable drivers of the three adverse reactions, highlighting the crucial role of disease complexity (especially complex vascular malformations) and the important discovery that prolonged treatment duration increases the risk of myelosuppression and dyslipidemia. These results fill critical knowledge gaps regarding the safety of sirolimus in the treatment of pediatric vascular diseases, providing a solid scientific basis for optimizing the risk-benefit ratio of this important therapy. More importantly, this invention pioneeringly integrates the BKMR safety model with the PopPK model, constructing a personalized treatment decision-making framework that can be translated into clinical practice, marking a substantial step towards precision medicine for pediatric vascular diseases.

[0092] In one embodiment, the data of the present invention irrefutably establishes... As the primary predictor of sirolimus-related adverse events (PIP close to 1), this aligns with sirolimus's safety profile in the transplant field, but this study provides the first quantitative evidence in the unique population of children with vascular diseases. Particularly noteworthy is the nonlinear relationship observed in this invention (e.g., Figure 2-5 10-4) has important clinical value: when Beyond a certain threshold (e.g., dyslipidemia >10.19 ng / mL), the risk of adverse reactions increases sharply. This strongly suggests a relatively narrow safe treatment window. In another long-term observational study of sirolimus in children with tuberous sclerosis, the research team pooled and analyzed two 10-year prospective cohort studies involving 1738 patients of all ages (aged 5 days to 69 years), controlling steady-state trough plasma concentrations at 5-10 ng / mL, and the results showed good patient safety. Combining the known efficacy concentration data (5-10 ng / mL) mentioned above, the results of this study provide a basis for establishing a TDM-based target concentration range (maintenance) for this population. The invention provides a direct and complete theoretical basis for balancing efficacy and specific safety at concentrations of 5-7.5 ng / mL or 8-10.2 ng / mL, significantly reducing the blindness of empirical drug use. Furthermore, beyond drug exposure itself, a key and novel finding is that complex vascular malformations are an independent and potent risk factor for adverse reactions. This highlights the inadequacy of a 'one-size-fits-all' safety strategy. The underlying mechanisms may involve broader lesion involvement, higher inflammatory burden, or more complex vascular endothelial dysfunction. This has significant clinical implications: for children with complex malformations, even... Even when maintaining the same control level, more intensive safety monitoring should be initiated (e.g., shortening the intervals between complete blood count, liver function tests, and blood lipid tests). Disease subtyping should be a core consideration in developing individualized monitoring plans. This invention is the first to quantitatively demonstrate that prolonged treatment time is an independent risk factor for bone marrow suppression and dyslipidemia, and is related to... Level-independent. This finding explains why some children still experience delayed adverse reactions at stable doses. Mechanistically, this may reflect the cumulative effect of sirolimus or metabolic / hematopoietic adaptive changes induced by long-term mTOR suppression. This is crucial for clinical practice: it alerts physicians to the need for continued vigilance with children on long-term medication (e.g., >6-12 months), even if their... Stable. Regular assessment of bone marrow reserves and lipid profiles should be a routine part of long-term management, and the need for preventative interventions (such as lipid management) should be considered.

[0093] In one embodiment, model validation ( Figure 9-3The study affirmed the powerful ability of the BKMR model to resolve such complex relationships. The excellent fit and predictive performance of the liver function abnormalities and dyslipidemia models make them directly applicable for risk prediction and clinical decision support. However, the high predictive uncertainty of the myelosuppression model (especially at the individual level) also reveals the inherent complexity of this endpoint—potentially influenced by more under-captured genetic, nutritional, or comorbid factors. Therefore, it is recommended that BKMR prediction of myelosuppression be more suitable for identifying high-risk trends and population risks (such as patients with high drug exposure intensity, long treatment duration, or complex conditions) rather than precise individualized prediction. Clinical decision-making should be dynamically adjusted based on model predictions and actual monitoring indicators.

[0094] In one specific embodiment, this invention represents a pioneering application of the BKMR model in the field of pediatric pharmacological epidemiology / precision medicine. Unlike conventional approaches, the BKMR model's unique advantages in handling multiple exposures (between drug exposure and multiple covariates), nonlinearity, and complex interactions perfectly address the common challenges of small sample sizes and confounding conditions in pediatric rare disease / specialty drug research. (See...) Figure 11 Compared to traditional logistic regression, BKMR offers a more comprehensive assessment of variable importance (PIP) and a more realistic characterization of exposure-response (linear and nonlinear relationships, additive and non-additive relationships), avoiding bias from pre-defined functional forms. This provides a powerful methodological blueprint for future exploration of the complex safety issues of other pediatric drugs (especially anticancer drugs and immunosuppressants).

[0095] In a specific embodiment, there are several clinical prediction models or exploratory analysis methods for binary outcomes. Compared with traditional research approaches, this study included a more specific population and disease, resulting in a smaller data volume (257 patients included, with 766-892 blood samples), making it difficult to construct machine learning models. (According to the study by Scott Silvey et al. (reference: SilveyS, Liu J. Sample Size Requirements for Popular Classification Algorithms in Tabular Clinical Data: Empirical Study. J Med Internet Res. 2024 Dec 17;26:e60231.), the median number of samples required to construct a well-fitting XGBoost, Random Forest, and Neural Network machine learning model is approximately 9960, 3404, and 12298, respectively). Conventional exploratory approaches (Logistic regression models and RCS functions for exploring binary outcomes) are greatly influenced by subjectivity. The former requires the assumption of linear correlation between variables, while the latter's point selection is subject to human intervention and will affect the final results. Most importantly, it is difficult to obtain quantitative numerical comparisons of different risk values ​​for both. Compared to the models mentioned above, the BKMR model can handle nonlinear and multivariate interactions, and simultaneously estimate exposure importance and synergistic effects without requiring the large sample size of traditional machine learning. However, this model is computationally intensive and has low adoption.

[0096] Furthermore, the Bayesian negative feedback algorithm is an iterative correction mechanism that combines group priors with patient information to continuously correct individual parameter estimates and gradually approximate the patient's true pharmacokinetic characteristics.

[0097] In one embodiment, the "negative feedback" does not refer to a negative value in a mathematical sense, but rather to automatically lowering parameters when the observed concentration is higher than the model prediction and vice versa, thereby achieving feedback correction similar to a control system. The rationality of the final solution is judged by the target treatment window: if the patient is in a steady state... If the dosing regimen falls within this range, it is considered appropriate; if it falls outside this range, further adjustments are required.

[0098] In one specific embodiment, the most transformative value of this invention lies in seamlessly linking the security relationships revealed by BKMR to the PopPK model ( Figure 10This integration goes beyond traditional PK / PD analysis, enabling the prediction of safety outcomes based on individualized PK simulation. It allows clinicians to quantify the potential adverse reaction risks of different regimens before treatment or during dose adjustments, thus achieving truly evidence-based 'dosing individualization'. This model-informed precision dosing (MIPD) strategy holds great potential in pediatric drug use, especially for drugs with narrow therapeutic windows and high safety concerns.

[0099] The present invention also discloses a computer device, the device comprising: a memory and a processor, such as... Figure 12 As shown: The memory is used to store program instructions; The processor is used to invoke program instructions, and when the program instructions are executed, the method steps are implemented.

[0100] The present invention provides a computer program product, including a computer program, which is implemented by a processor to implement the method steps.

[0101] The present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the method steps.

[0102] This embodiment also discloses a system for adjusting drug dosage based on adverse reaction risk using a combined model, the system comprising: The acquisition module 201 acquires the lower limit of the target trough concentration of the target drug and clinical variables related to patient information, inputs the clinical variables related to patient information into the BKMR model to calculate the Bayesian confidence interval to determine the upper limit of the target trough concentration, and obtains the target trough concentration range. The calculation module 202 is used or configured to calculate individual pharmacokinetic parameters of subjects based on real-world clinical data of the target drug combined with a population pharmacokinetic model. The prediction module 203 is used or configured to acquire N simulated dosing regimens, where N is a natural integer; calculate the simulated blood drug concentration of the simulated dosing regimen based on the individual pharmacokinetic parameters of the subject; and output the simulated dosing regimen in which the simulated blood drug concentration is within the target trough concentration range as the subject's dosing regimen.

[0103] This embodiment also discloses a sirolimus dosage recommendation system, the system comprising: The acquisition module 301 acquires the target valley concentration range of sirolimus, wherein the target valley concentration range is 5-7.5 ng / mL or 8-10.2 ng / mL. The calculation module 302 calculates the individual pharmacokinetic parameters of the subjects based on real-world clinical data of sirolimus combined with a population pharmacokinetic model. The prediction module 303 acquires N simulated dosing regimens, where N is a natural integer; calculates the simulated blood drug concentration for each simulated dosing regimen based on the individual pharmacokinetic parameters of the subject, and outputs the simulated dosing regimen with the simulated blood drug concentration within the target trough concentration range as the subject's dosing regimen.

[0104] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0105] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0106] The modules described as separate components may or may not be physically separate. Similarly, the components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0107] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The aforementioned integrated modules can be implemented in hardware or as software functional modules.

[0108] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, which may include: read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0109] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware, and the program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk.

[0110] The computer device provided by the present invention has been described in detail above. For those skilled in the art, there will be changes in the specific implementation and application scope based on the ideas of the embodiments of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for adjusting the dosage based on a combined model according to the target trough concentration range, characterized in that, include: S1: Obtain the lower limit of the target trough concentration of the target drug and clinical variables related to patient information, input the clinical variables related to patient information into the BKMR model to calculate the Bayesian confidence interval to determine the upper limit of the target trough concentration, and obtain the target trough concentration range; The clinical variables related to patient information are input into the BKMR model to calculate the Bayesian confidence interval to determine the upper limit of the target trough concentration. Specifically: The clinical variables were iterated through the BKMR model, and inter-group and intra-group importance comparisons were performed to obtain variables related to adverse reactions. Specifically, the clinical variables were iterated through the BKMR model and inter-group importance comparisons were performed using groupPIP, and intra-group importance comparisons were performed using CondPIP to obtain variables related to predicting the occurrence of myelosuppression, abnormal liver function, and dyslipidemia. The blood drug concentration of the variables related to adverse reactions was measured by univariate exposure-response analysis. The 95% Bayes confidence interval of the difference between each quantile and the baseline was calculated. When the 95% Bayes confidence interval does not contain 0 and its lower limit is greater than 0.05, the minimum drug exposure concentration corresponding to the above 95% Bayes confidence interval is determined as the upper limit of the target trough concentration. S2: Individual pharmacokinetic parameters of subjects are calculated based on real-world clinical data of the target drug combined with population pharmacokinetic models; S3: Obtain N simulated dosing regimens, where N is a natural integer; calculate the simulated blood drug concentration of the simulated dosing regimen based on the individual pharmacokinetic parameters of the subject; output the simulated dosing regimen in which the simulated blood drug concentration is within the target trough concentration range as the subject's dosing regimen.

2. The method for adjusting the dosage based on the target trough concentration range using a combined model according to claim 1, characterized in that, S3 can be replaced by S3': Obtain a simulated dosing regimen, which includes the drug dosage and dosing time; calculate the simulated blood drug concentration of the simulated dosing regimen based on the individual pharmacokinetic parameters of the subject; determine whether the simulated dosing regimen is the subject's dosing regimen based on the relative position of the simulated blood drug concentration and the target trough concentration range, with the specific judgment criteria being: If the simulated blood drug concentration is within the target trough concentration range, then the simulated dosing regimen is determined as the subject's dosing regimen; If the simulated blood drug concentration is outside the target trough concentration range, the dosage or administration time is adjusted to obtain an updated dosing regimen. The simulated blood drug concentration is then recalculated using the updated dosing regimen until the simulated blood drug concentration is within the target trough concentration range.

3. The method for adjusting the dosage based on a joint model according to the target trough concentration range according to claim 2, wherein adjusting the dosage or administration time to obtain an updated dosing regimen, and recalculating the simulated blood drug concentration using the updated dosing regimen until the simulated blood drug concentration falls within the target trough concentration range, specifically comprises: If the simulated blood drug concentration is below the target trough concentration range, the dosage is increased or the dosing time is decreased to obtain an updated dosing regimen. The simulated blood drug concentration is then recalculated using the updated dosing regimen until the simulated blood drug concentration is within the target trough concentration range. If the simulated blood drug concentration is above the target trough concentration range, the dosage is reduced or the dosing time is increased to obtain an updated dosing regimen. The simulated blood drug concentration is then recalculated using the updated dosing regimen until the simulated blood drug concentration is within the target trough concentration range.

4. The method for adjusting the dosage based on the target trough concentration range according to a combined model as described in claim 1, characterized in that, The clinical variables related to patient information include: drug exposure level, disease classification, patient basic information, and dosing strategy.

5. The method for adjusting the dosage based on the target trough concentration range using a combined model according to claim 1, characterized in that, The adverse reactions include: bone marrow suppression, abnormal liver function, and abnormal blood lipids.

6. The method for adjusting the dosage based on the target trough concentration range according to a combined model as described in claim 1, characterized in that, The variables related to adverse reactions are: drug exposure level and disease classification.

7. The method for adjusting the dosage based on the target trough concentration range according to a combined model as described in claim 1, characterized in that, The individual pharmacokinetic parameters include: apparent volume of distribution and apparent clearance.

8. The method for adjusting the dosage based on the target trough concentration range using a combined model according to claim 1, characterized in that, The method for calculating the individual pharmacokinetic parameters is as follows: Obtain the subject's current dosing regimen and blood drug concentration; The dosing regimen and blood drug concentration are input into the population pharmacokinetic model to obtain individual pharmacokinetic parameters; The specific steps for inputting the dosing regimen and blood drug concentration into the population pharmacokinetic model to obtain individual pharmacokinetic parameters are as follows: The dosing regimen and blood drug concentration are input into the population pharmacokinetic model in the pharmacokinetic parameter calculation software to calculate the individual pharmacokinetic parameters of the patient.

9. The method for adjusting the dosage based on the target trough concentration range according to a joint model as described in claim 8, characterized in that, The pharmacokinetic parameter calculation software includes one or more of the following: NONMEM, Phoenix, Monolix.

10. A recommended method for administering sirolimus dosage, characterized in that, include: S101: According to the method of claim 1, the target trough concentration range of sirolimus is obtained, and the target trough concentration range for the population is 5-7.5 ng / mL or 8-10.2 ng / mL, respectively; S102: Individual pharmacokinetic parameters of subjects calculated based on real-world clinical data of sirolimus combined with population pharmacokinetic models; S103: Obtain N simulated dosing regimens, where N is a natural integer; calculate the simulated blood drug concentration for each simulated dosing regimen based on the individual pharmacokinetic parameters of the subject, and output the simulated dosing regimen with the simulated blood drug concentration within the target trough concentration range as the subject's dosing regimen.

11. The method for recommending a sirolimus dosage according to claim 10, characterized in that, S103 can be replaced by S103': Obtain the simulated dosing regimen; Based on the individual pharmacokinetic parameters of the subjects, simulated blood drug concentrations for each simulated dosing regimen are calculated; the relative position of the simulated blood drug concentrations to the target trough concentration range of sirolimus is determined to determine whether the simulated dosing regimen is the subject's dosing regimen. The specific determination method is as follows: If the simulated blood drug concentration is within the target trough concentration range, then the simulated dosing regimen is determined as the subject's dosing regimen; If the simulated blood drug concentration is below the target trough concentration range, the dosage is increased or the dosing time is reduced to obtain an updated dosing regimen. The simulated blood drug concentration is then recalculated using the updated dosing regimen until the simulated blood drug concentration is within the target trough concentration range. If the simulated blood drug concentration is above the target trough concentration range, the dosage is reduced or the dosing time is increased to obtain an updated dosing regimen. The simulated blood drug concentration is then recalculated using the updated dosing regimen until the simulated blood drug concentration is within the target trough concentration range.

12. A computer device, comprising: Memory and processor; The memory is used to store program instructions; The processor is used to invoke program instructions, which, when executed, implement the method steps of any one of claims 1-11.

13. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the method steps of any one of claims 1-11.

14. A computer-readable storage medium storing a computer program thereon, characterized in that, When the computer program is executed by a processor, it implements the method steps of any one of claims 1-11.

Images