A prediction model for response to first-line glucocorticoid therapy based on aGVHD biomarker

Abstract

The present invention relates to the field of hematopoietic stem cell transplantation technology, and discloses a first-line glucocorticoid treatment response prediction model based on aGVHD biomarker, characterized by comprising the following steps: monitoring aGVHD biomarker indicators (sST2, REG3α) of aGVHD patients one week after first-line glucocorticoid treatment using a Luminex platform as basic data; correlating the above-mentioned basic data of the patients and whether aGVHD is severe at the time of occurrence with the clinical aGVHD first-line glucocorticoid treatment response, and using logistic regression technology to establish a first-line glucocorticoid resistance / dependence prediction model for aGVHD. The first-line glucocorticoid treatment response prediction model based on aGVHD biomarker predicts and stratifies whether aGVHD patients develop hormone resistance / dependence after first-line glucocorticoid treatment by establishing a machine learning model, providing a reference for clinical decision-making.

Description

Technical Field

[0001] The present invention relates to the technical field of hematopoietic stem cell transplantation, and in particular to a first-line glucocorticoid treatment response prediction model based on aGVHD biomarker. Background Art

[0002] Acute graft-versus-host disease (aGVHD) is a major cause of fatal complications and mortality after allogeneic hematopoietic stem cell transplantation (HSCT). Most patients require initial treatment with glucocorticoids after developing aGVHD. However, many patients remain unresponsive to steroid therapy (steroid resistance), or their steroid doses cannot be reduced or reactivate during dose reduction (steroid dependence). These patients typically have a poor prognosis and low survival rates. Furthermore, some patients require continuous observation until the fourth week after treatment to determine the response to initial treatment. This long time span hinders clinicians from modifying treatment strategies in a timely manner. Therefore, early prediction of glucocorticoid resistance / dependence can assist physicians in developing better aGVHD treatment strategies.

[0003] In terms of predicting steroid resistance, James Ferrara et al. found that biomarkers produced one week after aGVHD treatment could better predict the long-term prognosis of steroid-resistant patients than clinical standards. Alam N et al. combined donor IL6 and IFNG genotypes with whether a patient developed gastrointestinal aGVHD to establish a risk model for predicting steroid-refractory aGVHD. The model achieved an area under the curve (AUC) of 0.734 in a population of 268 patients. Dander E et al. used PTX3 plasma levels at the onset of clinical symptoms of aGVHD to predict whether a patient would respond to steroid treatment four weeks later. This factor achieved an AUC of 0.7 in a population of 61 pediatric patients. Minculescu L et al. found that in patients with grade II-IV aGVHD after HSCT, the C-reactive protein (CRP) level at diagnosis was an effective predictor of steroid-refractory disease (p = 0.001). Currently, there are not many studies on the prediction of steroid resistance / dependence, and the predictive effect still needs to be improved. In addition, existing studies rarely utilize clinical diagnostic information of aGVHD in patients before steroid treatment. Summary of the Invention

[0004] (1) Technical problems solved

[0005] In response to the shortcomings of the existing technology, the present invention provides a first-line glucocorticoid treatment response prediction model based on aGVHD biomarker, which has the advantages of higher accuracy and solves the problem that existing studies rarely utilize the clinical diagnostic information of aGVHD patients before hormone treatment.

[0006] (2) Technical solution

[0007] To achieve the above objectives, the present invention provides the following technical solution: a first-line glucocorticoid treatment response prediction model based on aGVHD biomarker, characterized by comprising the following steps:

[0008] 1) 72 patients with aGVHD, of whom approximately 29% were steroid-resistant or -dependent, were selected as observation samples;

[0009] 2) Using the Luminex platform, monitor the aGVHD biomarker indicators (sST2, REG3α) of the observation sample mentioned in step 1) after one week of first-line glucocorticoid treatment as basic data strips;

[0010] 3) Correlate the basic data of the above patients and whether aGVHD is severe at the time of onset with the clinical response to first-line aGVHD glucocorticoid treatment, and use logistic regression technology to establish a prediction model for first-line aGVHD glucocorticoid resistance / dependence;

[0011] 4) By assigning different weights to each modeling factor, the risk value p (model score value) of the patient developing hormone resistance / dependence is calculated as shown in the following formula 1:

[0012]

[0013] 5) Group the patient groups mentioned in step 1) by searching by the threshold (the model score value mentioned in step 4)

[0014] Furthermore, one piece of aGVHD biomarker data is provided for one patient, and if multiple pieces of data exist, they are randomly selected.

[0015] Furthermore, in step 4), the larger the p value, the higher the probability that the patient will develop hormone resistance / dependence. The model represented by the above formula 1 has an AUC of about 0.75 for predicting the response to first-line glucocorticoid treatment, and a sensitivity and specificity of about 0.7. The model can effectively predict clinical first-line glucocorticoid resistance / dependence.

[0016] (3) Beneficial effects

[0017] Compared with the prior art, the present invention provides a first-line glucocorticoid treatment response prediction model based on aGVHD biomarker, which has the following beneficial effects:

[0018] This first-line glucocorticoid treatment response prediction model based on aGVHD biomarkers uses a machine learning model to predict and stratify whether patients with aGVHD will develop hormone resistance / dependence after first-line glucocorticoid treatment, providing a reference for clinical decision-making. In addition, the present invention combines aGVHD biomarkers with clinical aGVHD diagnosis to further improve the accuracy of predicting hormone resistance / dependence. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 The steps for executing the first-line glucocorticoid treatment response prediction model based on aGVHD biomarker of the present invention are as follows;

[0020] Figure 2 The distribution differences of sST2, REG3α, and model score between the two groups of patients with steroid resistance / dependence and steroid sensitivity one week after the initial treatment of aGVHD;

[0021] Figure 3 This is the ROC curve of the first-line glucocorticoid treatment response prediction model on the model training set;

[0022] Figure 4 The percentage of hormone-resistant / dependent people in each group at different thresholds. DETAILED DESCRIPTION

[0023] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0024] Example 1: Figure 1 As shown, the present invention includes three steps: aGVHD biomarker monitoring and data screening and grouping, first-line glucocorticoid treatment response prediction model training, and risk stratification:

[0025] Step 1: Monitoring aGVHD biomarkers and screening and grouping data. The Luminex platform was used to monitor aGVHD biomarker (sST2, REG3α) levels in a multicenter patient population one week after first-line glucocorticoid treatment. Patients were grouped based on whether they developed steroid resistance or dependence after first-line aGVHD glucocorticoid treatment to form a model dataset. It should be noted that during the screening process, one piece of aGVHD biomarker data must be obtained for each patient.

[0026] Step 2: First-line glucocorticoid treatment response prediction model training, the aGVHD biomarker indicators of the model data set mentioned in step 1 and whether the clinical aGVHD is severe at the time of occurrence are associated with whether the clinical aGVHD first-line glucocorticoid treatment causes hormone resistance / dependence. Logistic regression technology in the field of machine learning is used to establish a prediction model for hormone resistance / dependence. The distribution differences of sST2, REG3α, and model score of the two groups of model data sets are as follows: Figure 2 As shown in Figure 2, the AUC of the single first-line glucocorticoid treatment response prediction model on the model training set is around 0.75 (model score), and the sensitivity and specificity can reach around 0.7. Figure 3 shown.

[0027] Among them, when different basic data pieces are substituted into formula 1, the larger the p-value, the higher the probability that the patient will develop hormone resistance / dependence. The model can effectively predict clinical first-line glucocorticoid resistance / dependence.

[0028] Step 3: Risk stratification: By searching different thresholds (the model score values ​​mentioned in step 2), the patient groups mentioned in step 1 are divided into high-risk and low-risk groups, such as Figure 4 As shown in the data, when the threshold was 0.30631, 52% of the high-risk group developed hormone resistance / dependence, and 17% of the low-risk group developed hormone resistance / dependence, which was statistically significant. Risk stratification can further assist clinical decision-making.

[0029] The above results indicate that the method of the present invention combines aGVHD biomarker indicators with clinical aGVHD diagnosis to establish a model. The model can predict the severity of hormone resistance / dependence and risk stratify patients one week after first-line aGVHD glucocorticoid treatment. The results can provide a reference for clinical decision-making.

[0030] The beneficial effects of the present invention are as follows: the first-line glucocorticoid treatment response prediction model based on aGVHD biomarkers predicts and stratifies whether patients with aGVHD will develop hormone resistance / dependence after first-line glucocorticoid treatment by establishing a machine learning model, providing a reference for clinical decision-making. In addition, the present invention combines aGVHD biomarkers with clinical aGVHD diagnosis to further improve the accuracy of predicting hormone resistance / dependence.

[0031] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for constructing a first-line glucocorticoid treatment response prediction model based on aGVHD biomarker, characterized in that: The following steps are involved: 1) 72 patients with aGVHD, of whom 29% were steroid-resistant or -dependent, were selected as observation samples; 2) Using the Luminex platform, monitor the aGVHD biomarker indicators sST2 and REG3α of the observation sample mentioned in step 1) after one week of first-line glucocorticoid treatment as basic data strips; 3) Correlate the basic data of the above patients and whether aGVHD is severe at the time of onset with the clinical response to first-line aGVHD glucocorticoid treatment, and use logistic regression technology to establish a prediction model for first-line aGVHD glucocorticoid resistance / dependence; 4) By assigning different weights to each modeling factor, the risk value p of the patient developing hormone resistance / dependence is calculated as shown in the following formula 1: ； The larger the p value, the higher the probability that the patient will develop hormone resistance / dependence. The model represented by the above formula 1 has an AUC of 0.75 for predicting the response to first-line glucocorticoid treatment, and a sensitivity and specificity of 0.

7. 5) Group the patient groups mentioned in 1) by searching with different thresholds, i.e., the p-values ​​mentioned in 4).

2. The method for constructing a first-line glucocorticoid treatment response prediction model based on aGVHD biomarker according to claim 1, characterized in that: One piece of aGVHD biomarker data was used for each patient. If multiple pieces of data existed, they were randomly selected.

Images