Methods, devices, media, and procedures for constructing a recommended model for personalized addition of IS during low-dose immune tolerance induction therapy.

Abstract

This invention provides a method, device, medium, and program product for constructing a recommendation model for personalized addition of immunosuppression (IS) during low-dose immune tolerance induction therapy (ITI), relating to the field of intelligent healthcare. The method includes: acquiring clinical data X, treatment regimen T, and treatment outcome Y of a training set sample; training a filled model capable of predicting treatment outcomes for different samples using the clinical data and treatment outcome; training a weighted model capable of predicting the likelihood of different samples receiving treatment using the clinical data and treatment regimen; calculating a DR estimate based on the filled model and the weighted model; calculating a causal effect value based on the DR estimate; and training a regression model based on the clinical data and the causal effect value to obtain the recommendation model. This application establishes a recommendation model that can help clinicians decide whether to personally add IS during ITI treatment by utilizing causal inference and machine learning, and the model's judgment has a relatively satisfactory consistency with the clinician's judgment.

Description

Technical Field

[0001] This invention relates to the field of intelligent healthcare, and more specifically, to a method, apparatus, medium, and procedure for constructing a recommendation model for personalized addition of IS during low-dose immune tolerance induction therapy. Background Technology

[0002] Hemophilia A (HA) is an X-linked coagulation disorder, typically treated with factor (F)VIII replacement therapy. As a major complication, 25-30% of patients with severe HA (SHA) develop inhibitory FVIII-specific antibodies (FVIII inhibitors). These inhibitors render factor replacement therapy ineffective.3 Currently, induction of immune tolerance (ITI) is the only method to eradicate the inhibitors. Most HA patients with inhibitors will develop “immune tolerance” to FVIII after ITI, with a success rate between 60% and 90%. However, some patients struggle to tolerate and / or do not respond to ITI.

[0003] As an alternative second-line treatment, additional interventions, such as immunosuppressants (IS), are recommended during ITI. Rituximab, an anti-CD20 antibody, is presumably intended to promote immune tolerance in cases of resistant inhibitors by rapidly depleting B lymphocytes. However, the appropriate criteria for combining IS with ITI in patients with a poor response remain unclear.

[0004] Given economic constraints, high-dose ITI regimens are not widely adopted. Therefore, our previous experience adding an intradermal solution (IS) to LD-ITI (associated with poor ITI response) in patients with poor ITI response (LD-ITI+IS) achieved a success rate of 62.5% with no complications. However, the LDITI+IS regimen does not guarantee success for all patients with poor ITI response; conversely, some patients may succeed even with only LDITI without IS. Therefore, experience in managing the addition of IS during ITI is challenging and limited. Summary of the Invention

[0005] This invention aims to at least address one of the technical problems existing in the prior art. To this end, this invention provides a method, apparatus, medium, and procedure for constructing a recommendation model for personalized addition of IS during low-dose immune tolerance induction therapy; the method of this invention utilizes causal reasoning and machine learning to establish a recommendation model that can help clinicians decide whether to personally add IS during ITI treatment, and the judgment of this model has a relatively satisfactory consistency with the judgment of clinicians.

[0006] The first aspect of this application discloses a method for constructing a recommendation model for personalized addition of immune tolerance (IS) during low-dose immune tolerance induction therapy, the method comprising:

[0007] S101, Obtain clinical data X, treatment plan T, and treatment outcome Y from the training set samples;

[0008] S102, a filler model that can predict treatment outcomes for different samples is trained using clinical data and treatment results;

[0009] S103, a weighted model trained using clinical data and treatment plans to predict the likelihood of different samples receiving treatment;

[0010] S104, Calculate the DR estimator based on the filled model and the weighted model; calculate the causal effect value based on the DR estimator;

[0011] S105, Based on the clinical data X and the causal effect value, train the regression model to obtain the recommendation model.

[0012] In some embodiments, the treatment plan includes whether IS is added to the LD-ITI treatment process, where T=1 indicates that treatment is received and T=0 indicates that no treatment is received;

[0013] Optionally, the treatment outcome includes whether the LD-ITI treatment was successful, where Y=1 indicates a successful outcome and Y=0 indicates an unsuccessful outcome.

[0014] Optionally, the clinical data may include: immediate inhibitor titer before ITI, historical peak inhibitor titer, age at ITI initiation, and interval.

[0015] In some embodiments, the expression for the filling model is f(X,T);

[0016] Optionally, the weighted model is expressed as g(X) = P(T=1|X); where T=1 represents receiving treatment;

[0017] Optionally, the formula for calculating the DR estimate is: ; where 1 (T=t) is an indicator function, which is equal to 1 when T equals t, and 0 otherwise; T is a random variable representing "intervention", and t is the specific value of the random variable.

[0018] In some embodiments, the formula for calculating the causal effect value is: .

[0019] In some embodiments, the regression model includes any of the following: linear regression, vector regression, random forest regression, and gradient boosting regression tree.

[0020] A second aspect of this application discloses a recommended method for personalized addition of immunosuppressive disorder (IS) during low-dose immune tolerance induction therapy, the method comprising:

[0021] S201, Obtain clinical data from the subjects;

[0022] S202, The clinical data is input into the recommendation model constructed by the method described in the first aspect of this application to calculate the causal effect value;

[0023] S203, recommend treatment options for the subject based on the causal effect value; when the causal effect value is greater than 0, output the auxiliary prediction result of the subject accepting treatment; when the causal effect value is less than 0, output the auxiliary prediction result of the subject not accepting treatment.

[0024] In some embodiments, the clinical data includes any one or more of the following: immediate inhibitor titer before ITI, historical peak inhibitor titer, age at ITI initiation, and interval.

[0025] A third aspect of this application discloses a computer device, the device comprising: a memory and a processor; the memory being used to store a computer program; and the processor executing the computer program to implement the steps of the above-described method.

[0026] The fourth aspect of this application discloses a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described above.

[0027] The fifth aspect of this application discloses a computer program product, including a computer program that, when executed by a processor, implements the steps of the above-described method.

[0028] This application has the following beneficial effects:

[0029] This application innovatively discloses a method for constructing a recommendation model for personalized addition of IS during low-dose immune tolerance induction therapy. This method employs causal inference to estimate the target causal relationship (to resolve confounding biases and achieve unbiased estimation of the causal relationship), such as... Figure 6 As shown, this approach employs machine learning methods to simulate the correlations between variables. Specifically, the machine learning model is used to capture the correlations between treatment and outcomes, including causal relationships and spurious correlations caused by clinical characteristics. Subsequently, causal inference methods are used to accurately identify causal relationships from the obtained overall statistical associations. Finally, a regression model is used to learn the mapping from clinical characteristics to causal relationships, thereby enabling the prediction of causal relationships and corresponding optimal treatment strategies for any newly admitted patient. Unlike traditional machine learning algorithms that focus on accurately predicting outcomes using numerous input features, this approach aims to recommend the most beneficial treatment plan for each patient. Attached Figure Description

[0030] 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.

[0031] Figure 1 This is a schematic diagram of the method flow provided in the first aspect of the present invention;

[0032] Figure 2 This is a schematic diagram of the method flow provided in the second aspect of the present invention;

[0033] Figure 3 This is a schematic diagram of a computer device provided in an embodiment of the present invention;

[0034] Figure 4 This is a schematic diagram of the architecture of an exemplary computing device provided in an embodiment of the present invention;

[0035] Figure 5 This is a schematic diagram of the storage medium provided in an embodiment of the present invention;

[0036] Figure 6 This is a schematic diagram illustrating the process of training the recommendation model provided in the embodiments of the present invention, and its use to predict personalized results under different IS treatment plans.

[0037] Figure 7 This is a schematic diagram illustrating the relationship between clinical data X, treatment plan T, and treatment outcome Y provided in this embodiment of the invention.

[0038] Figure 8 This is a research flowchart provided in an embodiment of the present invention. Detailed Implementation

[0039] 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.

[0040] In some of the processes described in the specification, claims, and accompanying drawings of this invention, multiple operations appearing in a specific order are included. However, it should be clearly understood that these operations may not be executed in the order they appear herein, or may be executed in parallel. The operation numbers, such as 101, 102, etc., are merely used to distinguish different operations and do not represent any execution order. Furthermore, these processes may include more or fewer operations, and these operations may be executed sequentially or in parallel. It should be noted that the descriptions such as "first," "second," etc., in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to different types.

[0041] 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.

[0042] Ideally, we aim to use observational (historical) data to compare the potential outcomes of all possible treatments (e.g., adding or not adding IS). If we obtain both potential outcomes for all patients, we can calculate the difference between the different potential outcomes, i.e., the causal effect. By checking if the causal effect for a particular patient is greater than 0, we can derive a personalized optimal treatment plan. However, we can observe the outcomes of adding or not adding IS, but not both simultaneously. This is because once a treatment choice is made, we cannot turn back time to retract the treatment. In other words, the goal becomes estimating the causal effect (or missing potential outcome) based on the observed potential outcome. Unfortunately, patients' treatment choices and outcomes are often dependent on their clinical characteristics, leading to confounding effects on treatment outcomes and posing a challenge to accurately estimating the causal effect. Traditional observational research methods (e.g., correlation analysis) struggle to mitigate this confounding and cannot achieve an unbiased estimate of the causal effect.

[0043] Figure 1 This is a schematic flowchart of a method for constructing a personalized recommendation model for adding IS during low-dose immune tolerance induction therapy, provided by an embodiment of the present invention. Specifically, the method includes the following steps:

[0044] S101: Obtain clinical data X, treatment plan T, and treatment outcome Y from the training set samples; the training set samples include the treatment group and the control group;

[0045] In some embodiments, the treatment plan includes whether IS is added to the LD-ITI treatment process, where T=1 indicates that treatment is received and T=0 indicates that no treatment is received;

[0046] In some embodiments, the treatment outcome includes whether the LD-ITI treatment was successful, where Y=1 indicates a successful outcome and Y=0 indicates an unsuccessful outcome.

[0047] In some embodiments, the clinical data include: inhibitor titer immediately before ITI (inhibitor level measured immediately before the start of immune tolerance induction therapy (ITI)), historical peak inhibitor titer (or highest previous inhibitor titer, the highest inhibitor concentration ever reached in the patient's medical history), age at ITI initiation, and interval (from inhibitor diagnosis to ITI initiation).

[0048] S102, a filler model that can predict treatment outcomes for different samples is trained using clinical data and treatment results;

[0049] In some embodiments, the expression for the filling model is f(X,T).

[0050] S103, a weighted model trained using clinical data and treatment plans to predict the likelihood of different samples receiving treatment;

[0051] In some embodiments, the weighted model is expressed as g(X) = P(T=1|X); where T=1 indicates receiving treatment.

[0052] S104, Calculate the DR estimator based on the filled model and the weighted model; calculate the causal effect value based on the DR estimator;

[0053] In some embodiments, the formula for calculating the DR estimate is: ; where 1 (T=t) is an indicator function, which is equal to 1 when T equals t, and 0 otherwise; T is a random variable representing "intervention", and t is the specific value of the random variable.

[0054] In some embodiments, the formula for calculating the causal effect value is: .

[0055] S105, Based on the clinical data X and the causal effect value, train the regression model to obtain the recommendation model.

[0056] In some embodiments, the regression model includes any of the following: linear regression, vector regression (SVR), random forest regression, and gradient boosting regression tree (GBRT).

[0057] A second aspect of this application discloses a recommended method for personalized addition of immunosuppressive disorder (IS) during low-dose immune tolerance induction therapy, the method comprising:

[0058] S201, Obtain clinical data from the subjects;

[0059] In some embodiments, the clinical data includes any one or more of the following: immediate inhibitor titer before ITI, historical peak inhibitor titer, age at ITI initiation, and interval.

[0060] In some embodiments, the terms “subject” or “test subject” or “sample” as used herein refer to any animal (e.g., a mammal), including but not limited to humans, non-human primates, rodents, etc., which will become the recipient of a particular treatment. Generally, the terms “subject” and “patient” are used interchangeably herein when referring to human subjects. Preferably, the subject is a human.

[0061] In some embodiments, the sample to be tested is a patient undergoing prognostic assessment in a clinical setting.

[0062] S202, The clinical data is input into the recommendation model constructed by the method described in the first aspect of this application to calculate the causal effect value;

[0063] S203, recommend treatment options for the subject based on the causal effect value; when the causal effect value is greater than 0, output the auxiliary prediction result of the subject accepting treatment; when the causal effect value is less than 0, output the auxiliary prediction result of the subject not accepting treatment.

[0064] In some embodiments, the auxiliary prediction results include, but are not limited to, paper or electronic reports. These results are obtained by intelligent machines based on the relevant data of the subjects and are intended only as a reference for medical personnel, not as the final diagnosis results of the subjects.

[0065] In some embodiments, the threshold 0 is obtained through training on training set samples. It can be a specific threshold or an interval range. The specific form is not specifically limited in this embodiment.

[0066] Figure 3 This is a schematic diagram of a computer device provided in an embodiment of the present invention, such as... Figure 3 As shown, the device 2000 may include: one or more processors 2010 and one or more memories 2020; wherein the memories store computer-readable code that, when run by the one or more processors, can perform the methods described above.

[0067] The processor in this embodiment can be an integrated circuit chip with signal processing capabilities. The processor can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, operations, and logic block diagrams disclosed in this embodiment. The general-purpose processor can be a microprocessor or any conventional processor, and can be based on an x86 or ARM architecture.

[0068] In general, the various exemplary embodiments of this disclosure can be implemented in hardware or dedicated circuitry, software, firmware, logic, or any combination thereof. Some aspects can be implemented in hardware, while others can be implemented in firmware or software that can be executed by a controller, microprocessor, or other computing device. When aspects of embodiments of this disclosure are illustrated or described as block diagrams, flowcharts, or using some other graphical representation, it will be understood that the blocks, apparatuses, systems, techniques, or methods described herein can be implemented as non-limiting examples in hardware, software, firmware, dedicated circuitry or logic, general-purpose hardware or controllers or other computing devices, or some combination thereof.

[0069] For example, the method or apparatus according to embodiments of this disclosure can also be used by means of Figure 4 The architecture of the computing device 3000 shown is used for implementation. For example... Figure 4 As shown, the computing device 3000 may include a bus 3010, one or more CPUs 3020, a read-only memory (ROM) 3030, a random access memory (RAM) 3040, a communication port 3050 connected to a network, an input / output component 3060, a hard disk 3070, etc. The storage devices in the computing device 3000, such as the ROM 3030 or the hard disk 3070, may store various data or files used for processing and / or communication of the methods provided in this disclosure, as well as program instructions executed by the CPU. The computing device 3000 may also include a user interface 3080. Of course, Figure 4 The architecture shown is merely exemplary and can be omitted as needed when implementing different devices. Figure 4 One or more components in the computing device shown.

[0070] This invention also includes a computer-readable storage medium, such as... Figure 5The diagram illustrates a storage medium 4000 provided in an embodiment of the present invention. The computer storage medium 4020 stores computer-readable instructions 4010. When the computer-readable instructions 4010 are executed by a processor, the method described above according to embodiments of the present disclosure can be performed. The computer-readable storage medium in the embodiments of the present disclosure may be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. Non-volatile memory may be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory may be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), Synchronous Dynamic Random Access Memory (SDRAM), Double Data Rate Synchronous Dynamic Random Access Memory (DDRSDRAM), Enhanced Synchronous Dynamic Random Access Memory (ESDRAM), Synchronous Link Dynamic Random Access Memory (SLDRAM), and Direct Memory Bus Random Access Memory (DR RAM). It should be noted that the memory used in the methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0071] This disclosure also provides a computer program product or system, including a computer program that, when executed by a processor, implements the steps of the above-described method.

[0072] In some embodiments, this embodiment also discloses a system for constructing a recommendation model for personalized addition of IS during low-dose immune tolerance induction therapy, the system comprising:

[0073] The first data acquisition module is used or configured to acquire clinical data X, treatment plan T, and treatment result Y of the training set samples;

[0074] A filler model training module, used or configured to train a filler model capable of predicting treatment outcomes for different samples using clinical data and treatment results;

[0075] The weighted model training module is used or configured to train a weighted model that can predict the likelihood of different samples receiving treatment using clinical data and treatment plans;

[0076] The causal effect value training module is used or configured to calculate the DR estimate based on the imputed model and the weighted model; and to calculate the causal effect value based on the DR estimate.

[0077] The recommendation model training module is used or configured to train a regression model based on the clinical data X and the causal effect value to obtain the recommendation model.

[0078] In some embodiments, this embodiment also discloses a recommendation system for personalized addition of IS during low-dose immune tolerance induction therapy, the system comprising:

[0079] The second data acquisition module is used or configured to acquire clinical data from the subjects;

[0080] A causal effect value calculation module is used or configured to input the clinical data into the recommendation model constructed by the method described in the first aspect of this application to calculate the causal effect value;

[0081] The outcome prediction module is used or configured to recommend treatment plans to subjects based on causal effect values; when the causal effect value is greater than 0, it outputs an auxiliary prediction result that the subject will receive treatment; when the causal effect value is less than 0, it outputs an auxiliary prediction result that the subject will not receive treatment. Specific Implementation

[0083] 1. Methods and Materials

[0084] 1.1 Research Design and Population

[0085] This single-center retrospective cohort study was conducted at the BCH Comprehensive Hemophilia Care Center in China. We recruited children with and without ischemic attacks (IS) who received LD-ITI treatment between September 2016 and December 2023, with a follow-up period of ≥24 months. Using four key pre-ITI clinical features extracted from patients' electronic medical records, we trained and developed a recommendation model using causal inference to help clinicians decide whether to add IS individually during LD-ITI treatment.

[0086] Inclusion criteria included: (i) SHA (FVIII coagulation activity [FVIII:C] <1% before inhibitor production); (ii) age ≤14 years at ITI initiation; (iii) high-titer inhibitor (≥5 BU / mL); (iv) receiving LD-ITI with or without IS, with a follow-up period ≥24 months. Exclusion criteria included: (i) congenital or acquired bleeding defects other than hemophilia A; (ii) concomitant immune diseases.

[0087] 1.2 Data Collection

[0088] Clinical, laboratory, and treatment data were collected from medical records and interviews. Baseline clinical and laboratory information was collected prior to the ITI. During the ITI, patients visited every 1 to 2 weeks until the inhibitor titer trended downward after the initial peak caused by early repeated FVIII exposure, and then visited monthly until the end of the study. Following successful ITI and during the prophylactic phase, patients continued to visit every 3 months.

[0089] 1.3 ITI Treatment

[0090] All patients received LD-ITI (FVIII 50 IU / kg, every other day) and plasma-derived factor VIII / von Willebrand factor concentrate (pd-FVIII / VWF). Once a patient achieved success (as defined below), the FVIII dose was slowly reduced to 25–30 IU / kg three times a week for continued prophylaxis. During ITI, IS (rituximab and prednisone) was added (as LDITI+IS) based on the following criteria: (1) if the inhibitor titer was ≥40 BU / mL before or during ITI alone, and (2) if there was no downward trend (if the inhibitor decline during ITI was <20% within the first 3 months of ITI).

[0091] Infection prevention (IS) treatment consists of rituximab 375 mg / m² / week (maximum dose 600 mg) for 4 weeks, combined with prednisone 2 mg / kg / day (maximum dose 60 mg) for one month, followed by gradual tapering over 6 weeks. Following rituximab treatment for infection prophylaxis, intravenous immunoglobulin (IVIG, 200 mg / kg / month for 6 months) is administered for 6 months to compensate for acquired IgG deficiency.

[0092] 1.4 Result Definition

[0093] Success: Negative inhibitor titer and FVIII recovery rate ≥ 66% of the expected rate at the time of analysis. We were unable to conduct half-life studies for every patient achieving an FVIII recovery rate ≥ 66%, therefore we did not include a half-life > 6 hours in our definition of success. Partial success: Negative inhibitor titer, but FVIII recovery rate < 66% of the expected rate; Unsuccessful: Success criteria not met. Relapse: After success, inhibitor titers > 0.6 BU / mL twice consecutively, at least one month apart, regardless of whether the patient received subsequent FVIII prophylactic treatment.

[0094] 1.5 Laboratory Testing

[0095] During the ITI, inhibitor titers were measured at BCH every two weeks using the Bethesda assay (modified Nijmegen assay) until a stable decrease in inhibitor titers was observed, and then monthly thereafter.

[0096] Once the inhibitor titer is negative twice consecutively (one month apart), in vivo FVIII recovery is performed. Once the in vivo FVIII recovery rate is ≥66%, the monitoring frequency is reduced to once every 3 months.

[0097] 1.6 The Development Process of Causal Reasoning

[0098] To support clinical decision-making, we strive to establish causal relationships rather than correlational models between treatment and outcomes; therefore, we employ a causal inference approach. Causal inference is used to assess the causal impact of treatment regimens on outcomes, where causal impact refers to the difference in outcomes under different interventions. In this paper, intervention is defined as whether or not an inhibitor (IS) is added during LD-ITI treatment, and outcome is defined as the success of the patient's ITI treatment. Baseline clinical characteristics of enrolled patients were collected, including immediate inhibitor titer before ITI, historical peak inhibitor titer, age at ITI initiation, and interval (from inhibitor diagnosis to ITI initiation). Based on these characteristics, this paper analyzes the impact of IS on outcomes and develops a model to personally recommend more favorable treatment regimens for patients based on causal effect estimation.

[0099] Estimating causal effects using observational data always faces two challenges. One challenge is the missing data problem mentioned earlier, meaning we can only observe one patient's response to a particular treatment, and not both responses simultaneously; this is also known as the fundamental problem of causal inference. The other challenge is that differences in characteristics between the treatment and control groups in the observational data mean that the outcome is not entirely attributable to the difference in treatment. For example, patients in the control group often have a more optimistic condition than those in the treatment group, and they are more likely to achieve successful outcomes without treatment. In such cases, simply comparing the experimental results between the two groups may underestimate the causal effect of the treatment, leading to bias.

[0100] To address the two issues mentioned above, we employ a dual robustness (DR) approach to estimate causal effects, one of the most effective methods for causal inference, which has achieved significant success in practical applications. Specifically, the DR estimator comprises two parts: an imputed model and a weighted model. The imputed model aims to fill in missing patient outcomes, while the weighted model balances the characteristic distributions of the treatment and control groups through sample weighting. Statistical theory suggests that if the attribution values ​​or estimated weights are accurate, the DR estimator can serve as an unbiased estimator of the target causal effect. Based on the learned DR estimator, we are able to calculate the causal effect of adding an intervention (IS) on outcomes for all patients, corresponding to the potential difference in outcomes with or without IS. For patients in the training set, we can directly compare whether their corresponding causal effect estimates exceed zero to determine which treatment is more effective. To adapt to any new patient, we employ a regression model that estimates causal effects using clinical characteristics, enabling us to predict the difference in outcomes for any patient under different treatment regimens and thus recommend better treatment strategies.

[0101] Categorical variables were expressed as frequencies and percentages, compared using the chi-square or Fisher exact chi-square test. Continuous variables were expressed as means and 95% confidence intervals (CIs), compared using the Student's test (for normally distributed variables) or the Mann-Whitney U test (for non-normally distributed variables). Reported p-values ​​are two-tailed, and p-values ​​<0.05 were considered statistically significant. All statistical analyses were performed using SPSS version 26.0 (IBM Corp., Armonk, NY, USA).

[0102] 1.7 Dual Rugged Design

[0103] First of all, we Figure 7 The text explains the relationship between feature X, treatment T, and outcome Y, which explains the data generation process. For example... Figure 7As shown, the outcome Y is jointly determined by the feature X and the treatment plan T. Furthermore, the treatment T received by a patient is often influenced by its respective features. Without loss of generality, we assume that the intervention variable T and the outcome variable Y are both binary variables taking values ​​of 0 and 1. Here, T=1 indicates receiving treatment, otherwise 0; Y=1 indicates a successful outcome, otherwise 0. As described in the previous section, the DR estimator is used to estimate causality. Without loss of generality, we denote the imputed model as f, whose input includes the feature X and the treatment T. We aim to use the imputed model to predict the outcomes for different patients with different X and T, mathematically expressed as f(X,T). The imputed model f(X,T) can be trained in a data-driven manner by minimizing the difference between the output of the imputed model and the true outcome. For the weighted model denoted as g(X), the input is the feature X, and the output is a probability value between 0 and 1, representing the likelihood of receiving treatment given the patient's known features, denoted as g(X) = P(T=1|X);

[0104] Similarly, a weighted model g(x) is trained by fitting the treatment assignments in the training data. Based on the two trained models above, the estimated potential outcome of the DR estimator with features X and treatment t is derived as follows: ; where 1 (T = t) is an indicator function, which equals 1 when T equals t, and 0 otherwise. Clearly, when a patient with characteristic Xi ​​is observed to receive treatment t, the estimated potential outcome corresponding to treatment t is adjusted from the observed outcome by an imputation and weighted model. The potential outcome corresponding to unobserved treatment (1-t) is directly generated by the imputation model.

[0105] Finally, the target causal effect CE(Xi) caused by the DR estimator can be expressed as the average difference between the estimated potential outcomes in a population with characteristic Xi. The calculation method is as follows: ;

[0106] According to the definition of CE(Xi), a CE(Xi) greater than 0 indicates that a patient with the corresponding characteristic Xi ​​has a higher probability of success when receiving treatment at t=1 than when receiving treatment at t=0. In this case, the optimal treatment for the patient is at t=1. In other words, a positive or negative value of CE(Xi) represents the superiority or inferiority of different treatment methods. Furthermore, to predict the optimal treatment plan for newly admitted patients, we developed a regression model to establish a mapping relationship from patient characteristics to CE(Xi), thereby contributing to optimal treatment recommendations. Specifically, given patient characteristic X as input and the aforementioned CE(X) as output, we are able to learn a regression model, such as linear regression, using training data. Therefore, for any given characteristic X, we can predict its corresponding causal effect (CE), thereby determining whether treatment should be administered.

[0107] 1.8 Experimental Details and Model Specifications

[0108] The entire experimental dataset included 195 patients, with detailed records of each patient's four key clinical characteristics, specified IS treatment, and final outcome. Specifically, the four features—immediate inhibitor titer before ITI, historical peak inhibitor titer, age at ITI initiation, and interval—were considered as model input X, while LD-ITI or LD-ITI+IS treatment and outcome were considered as T and Y, respectively. Five-fold cross-validation was used to verify the effectiveness of the proposed model. We constructed a decision tree as the base model and utilized a gradient boosting-based ensemble learning method to implement the propensity model, imputation model, and prediction model.

[0109] 2 Results

[0110] 2.1 Study population and baseline clinical characteristics

[0111] Of the 291 patients screened, 96 (33.0%) did not meet the inclusion criteria, such as... Figure 8 As shown. A total of 195 patients with SHA and high-titer inhibitors were included and treated with LD-ITI. The median follow-up period after ITI initiation was 4.4 years. Of the 195 patients, 145 (74.4%) achieved inhibitor eradication, and 50 (25.6%) failed to eradicate the inhibitor. Of the 195 patients, 105 (53.8%) received LD-ITI, and 90 (46.2%) received LD-ITI+IS. A total of 13 patients (9.0%) relapsed at a median of 3.7 months after successful ITI. Of the 13 relapsed patients, 8 (61.5%) received LD-ITI+IS.

[0112] In the results, within the median (range), the successful group had a shorter interval between ITIs (2.7 months vs. 5.7 months, p = 0.030) and a younger age at the start of ITI (3.1 years vs. 6.0 years, p < 0.000) compared to the unsuccessful group. Furthermore, compared to the unsuccessful group, the successful group had lower immediate inhibitor titers before ITI (13.5 BU / mL vs. 28.7 BU / mL, p = 0.015), historical peak inhibitor titers (21.3 BU / mL vs. 66.9 BU / mL, p = 0.043), and peak inhibitor titers during ITI (55.7 BU / mL vs. 146.4 BU / mL, p < 0.000).

[0113] When focusing on the 145 successful patients, we compared the baseline characteristics of LD-ITI successful patients (n=82) and LD-ITI+IS successful patients (n=63). Compared with LD-ITI successful patients, LD-ITI+IS successful patients had higher pre-ITI inhibitor titers (42.2 BU / mL vs. 5.0 BU / mL, p<0.000), higher historical peak inhibitor titers (64.0 BU / mL vs. 10.4 BU / mL, p<0.000), and higher peak inhibitor titers during ITI (89.5 BU / mL vs. 39.2 BU / mL, p<0.000). No significant differences were found in age characteristics (age at initial inhibitor, age at ITI initiation, and age at ITI interval) between the LD-ITI and LD-ITI+IS groups.

[0114] Of the 50 patients whose inhibitors failed to clear, 23 (46%) received LD-ITI and 27 (54%) received LD-ITI+IS. Similarly, compared with patients who failed LD-ITI, patients who failed LD-ITI+IS had significantly higher inhibitor characteristics (immediate inhibitor titer before ITI, historical peak inhibitor titer, and peak inhibitor titer during ITI) and similar age characteristics.

[0115] 2.2 Recommendation Model Based on Causal Reasoning

[0116] Because the experimental data is retrospective, all patients in the dataset received treatment recommended by their clinicians. When the model-recommended treatment differed from the clinician-recommended treatment, the corresponding potential outcomes could not be observed, thus making it impossible to assess the accuracy of these recommendations. Therefore, to evaluate the accuracy of the model recommendations, we considered a subset of patients whose model recommendations were consistent with their clinician recommendations. Of the 195 participants, 119 (61.0%) received treatment recommendations consistent with their clinicians' recommendations. Of these 119 patients, 92 (77.3%) adhered to the model-recommended treatment plan and achieved success. For comparison, clinicians recommended treatment plans for 195 patients, of whom 145 were successful, resulting in an overall success rate of 74.5%. This indicates that our proposed recommendation model is comparable to clinicians in recommending treatment plans to patients.

[0117] 2.3 The rationality and potential mechanisms of the model recommendations

[0118] We conducted an in-depth analysis by comparing patient characteristics recommended for IS treatment by the DR model with those recommended by clinicians. This comparison aimed to validate the rationale behind the model's recommendations and explore its underlying mechanisms. Specifically, subgroup analyses were performed, focusing on four distinct subgroups. Of the 195 patients, 90 (46.2%) received IS treatment in LD-ITI based on clinician recommendations (Group A). ​​55.4% (108 / 195) of the patients were recommended by the model to add IS treatment to LD-ITI (Group B). Of the 108 patients, 61 (56.5%) received recommendations from both the model and clinicians (Group B1), and 47 (43.5%) received only model recommendations without clinician recommendations (Group B2).

[0119] No significant differences were found between groups A and B1 in inhibitor characteristics (median immediate titer before ITI: 45.0 BU / mL vs. 60.3 BU / mL, p=0.869; historical peak titer: 69.7 BU / mL vs. 76.0 BU / mL, p=0.356) and age characteristics (age at ITI initiation: 3.5 years vs. 3.1 years, p=0.492; interval: 3.8 months vs. 4.5 months, p=0.645). By combining group A, which received clinician-recommended treatment, with group B2, which received model-recommended treatment but not clinician-recommended treatment, it was observed that group B2 had significantly lower inhibitory characteristics (immediate titer before ITI: 15.7 BU / mL vs. 45.0 BU / mL, p=0.001; historical peak titer: 10.7 BU / mL vs. 69.7 BU / mL, p=0.001) and higher age characteristics (age at ITI initiation: 8.6 years vs. 3.5 years, p=0.012; interval: 12.2 months vs. 3.8 months, p=0.009).

[0120] Finally, ITI is currently the only clinically proven effective strategy for clearing inhibitors, especially for high-titer inhibitors. Although non-factor therapies (such as emicizumab) are increasingly used for patients who fail their first ITI treatment, emicizumab is not yet covered by health insurance in developing countries, and reports indicate that only 2.8% of HA patients with inhibitors can use this expensive product for prophylaxis. For patients with poor ITI response, the appropriate use of IS during LD-ITI is an important attempt in economically constrained developing countries, as these countries cannot afford higher-dose ITI regimens or non-factor therapies. However, current guidelines on IS use are lacking and inaccurate. The criteria for adding IS during LD-ITI treatment remain to be explored. To our knowledge, this is the first work to establish a recommendation model using causal reasoning to help clinicians make the right decisions regarding the combination of IS during LD-ITI treatment.

[0121] The model incorporates four key patient baseline clinical characteristics: immediate inhibitor titer before ITI, historical peak inhibitor titer, age at ITI initiation, and the interval between ITI initiation. Based on extensive prior research, these four clinical characteristics are important indicators of ITI outcomes. Multiple studies have shown that patients with higher inhibitor profiles (immediate ITI inhibitor titer, historical peak inhibitor titer) are more likely to fail. When the model detects a patient with a high inhibitor profile, a comprehensive algorithm will suggest more aggressive treatment approaches, such as combining with idiopathic inflammatory response (IS), in line with current international treatment practices.

[0122] Notably, 61.0% of patients received treatment recommendations consistent with their clinicians, indicating its potential reliability. While this percentage may seem low, the groups with inconsistent treatment recommendations allow us to identify previously overlooked predictors of additional IS treatment. The high similarity in characteristics between groups A and B1 suggests that the model's decision-making is not random, like a coin toss, but rather extracts useful clinical knowledge from the data to make its recommendations similar to those of clinicians. This demonstrates the rationality of the model's decision-making.

[0123] The comparison between group A and group B2 indicates that when inhibitor titer-related characteristics are relatively low, clinicians generally do not recommend IS treatment regardless of age-related characteristics such as ITI initiation age or interval. Conversely, when patients are relatively older but have relatively low inhibitor titers, the model still recommends adding IS. Based on this information, it can be inferred that older age at ITI initiation or longer intervals may be a new, previously unnoticed indicator for combining with IS treatment. While previous guidelines generally did not recommend older age as a criterion for increasing treatment intensity, some studies have shown that older age at ITI initiation and longer intervals lead to decreased ITI success rates. When older age is detected, our model recommends IS during LD-ITI. In the future, we plan to conduct a prospective large-scale cohort study to determine the cutoff value for age characteristics.

[0124] In this study, enrolled patients were considered "low-risk for ITI" because their median immediate pre-ITI titer was 16.8 BU / mL. A total of 74.4% of high-titer inhibitor SHA patients received LD-ITI+IS treatment on the advice of their clinicians and achieved success. Assuming the model's treatment recommendations are followed, the success rate could be as high as 77.3%. Rituximab has been used to achieve tolerability in inhibitor patients with a "low-risk status" since the early 2000s. Rituximab binds to B cells, leading to peripheral B cell depletion, but does not affect plasma cells or T cells. Rituximab achieved a negative inhibitor titer in 40% (6 / 15) of SHA patients. In the UK series, 80% (12 / 15) of patients were also receiving FVIII, and 27% (4 / 15) were using it in combination with prednisolone or other IS. It is perhaps not surprising that the success rate in the UK series appears lower than ours, as all patients included in the UK series had experienced ITI failure, with 33.3% experiencing more than one failure. However, in this study, rituximab was used when the IS criteria were met, and there was no evidence of waiting for ITI failure. Furthermore, our IS regimen included rituximab and prednisolone, not just rituximab. However, early identification of patients with “poorly positioned” conditions and co-administration of IS appeared to improve success rates. Notably, no immunosuppressant-related adverse events, such as infections, were observed with routine IVIG injections.

[0125] As E. Carlos noted, the application of artificial intelligence (AI) in hemophilia is still in its early stages. Machine learning models in hemophilia have been used with encouraging results in predicting disease severity, identifying factor V as an essential regulator of thrombin generation in mild to moderate HA, and determining predictors of long-term annualized bleeding rates. So far, we have developed this recommendation model based on baseline clinical information (immediate inhibitor titer before ITI, historical inhibitor titers, age at ITI initiation, and interval) to add IS during LD-ITI. The application of AI currently focuses primarily on predicting ITI outcomes. However, we hope that AI models can provide useful treatment recommendations. This requires developing optimal treatment recommendation models based on patient characteristics, not just focusing on outcome prediction. Traditional machine learning models for treatment recommendation aim to adapt to clinicians' decisions based on past data. That is, learning the mapping between patient characteristics and treatment plans under the clinician's judgment, without considering efficacy. However, the model's performance is highly dependent on the clinician's skill level, requiring the clinician to be very experienced with the disease for the model to perform perfectly. Limited by the characteristics of rare diseases, models inevitably make incorrect decisions. In summary, traditional machine learning models are ineffective at recommending optimal treatment plans for patients. Therefore, we employ a combination of causal inference and machine learning to address this issue. Causal inference has developed rapidly over the past decade and is widely used in fields such as statistics, computer science, and medicine. It can accurately estimate the impact of different treatment options on efficacy and recommend the optimal plan.

[0126] It should be noted that the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0127] In general, the various exemplary embodiments of this disclosure can be implemented in hardware or dedicated circuitry, software, firmware, logic, or any combination thereof. Some aspects can be implemented in hardware, while others can be implemented in firmware or software that can be executed by a controller, microprocessor, or other computing device. When aspects of embodiments of this disclosure are illustrated or described as block diagrams, flowcharts, or using some other graphical representation, it will be understood that the blocks, apparatuses, systems, techniques, or methods described herein can be implemented as non-limiting examples in hardware, software, firmware, dedicated circuitry or logic, general-purpose hardware or controllers or other computing devices, or some combination thereof.

[0128] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0129] 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 units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units 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 an indirect coupling or communication connection between apparatuses or units through some interfaces, and may be electrical, mechanical, or other forms.

[0130] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0131] The exemplary embodiments of this disclosure described in detail above are merely illustrative and not restrictive. Those skilled in the art will understand that various modifications and combinations can be made to these embodiments or their features without departing from the principles and spirit of this disclosure, and such modifications should fall within the scope of this disclosure.

Claims

1. A method for constructing a recommended model for personalized addition of IS during low-dose immune tolerance induction therapy, characterized in that, The method includes: S101, Obtain clinical data X, treatment regimen T, and treatment outcome Y from the training set samples; the clinical data includes: immediate inhibitor titer before ITI, historical peak inhibitor titer, age at ITI initiation, and interval; the treatment regimen includes whether IS is added during LD-ITI treatment, T=1 indicates that IS is added, and T=0 indicates that IS is not added; the treatment outcome includes whether LD-ITI treatment is successful, Y=1 indicates a successful outcome, and Y=0 indicates an unsuccessful outcome. S102, a filler model that can predict treatment outcomes for different samples is trained using clinical data and treatment results; S103, a weighted model trained using clinical data and treatment protocols to predict the likelihood of a sample receiving additional IS treatment; S104, Calculate the DR estimator based on the filled model and the weighted model; calculate the causal effect value based on the DR estimator; S105, Based on the clinical data X and the causal effect value, train the regression model to obtain a recommendation model that can output a personalized causal effect value when inputting the clinical data X of a new individual.

2. The method for constructing a recommended model for personalized addition of IS during low-dose immune tolerance induction therapy according to claim 1, characterized in that, The expression for the filling model is f(X,T).

3. The method for constructing a recommended model for personalized addition of IS during low-dose immune tolerance induction therapy according to claim 1, characterized in that, The weighted model is expressed as g(X) = P(T=1|X); where T=1 indicates receiving additional IS treatment.

4. The method for constructing a recommended model for personalized addition of IS during low-dose immune tolerance induction therapy according to claim 1, characterized in that, The formula for calculating the DR estimate is: ; where 1 (T=t) is an indicator function, which is equal to 1 when T equals t, and 0 otherwise; T is a random variable representing "intervention", and t is the specific value of the random variable.

5. The method for constructing a recommended model for personalized addition of IS during low-dose immune tolerance induction therapy according to claim 1, characterized in that, The formula for calculating the causal effect value is as follows: .

6. The method for constructing a recommended model for personalized addition of IS during low-dose immune tolerance induction therapy according to claim 1, characterized in that, The regression model includes any of the following: linear regression, vector regression, random forest regression, and gradient boosting regression tree.

7. A recommended method for personalized addition of IS during low-dose immune tolerance induction therapy, characterized in that, The method includes: S201, Obtain clinical data from the subjects; S202, input the clinical data into the recommendation model constructed by the method described in any one of claims 1-6, and calculate the causal effect value; S203, recommend treatment options for the subject based on the causal effect value; when the causal effect value is greater than 0, output the auxiliary prediction result of the subject receiving the additional IS treatment; when the causal effect value is less than 0, output the auxiliary prediction result of the subject not receiving the additional IS treatment.

8. A computer device, characterized in that, The device includes: a memory and a processor; the memory is used to store a computer program; the processor executes the computer program to implement the steps of the method according to any one of claims 1-7.

9. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the steps of the method as described in any one of claims 1-7.

10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the steps of the method described in any one of claims 1-7.

Images