A diagnostic model for predicting the prognosis of gastric cancer CAR-T cell immunotherapy and its application

Abstract

The present invention relates to the field of medical testing. Specifically, the present invention provides a diagnostic model for predicting the prognosis of CAR-T cell immunotherapy for gastric cancer and its application. For the first time, the present invention systematically analyzed the content and spatial distribution map of the microenvironment components of CLDN18.2-positive gastric cancer tissue samples by integrating multiple fluorescent immunohistochemistry detection and whole-slice digital imaging quantitative analysis methods. Through multidimensional evaluation of the tumor microenvironment under the treatment background of CT041, a diagnostic model for predicting the prognosis of CAR-T cell immunotherapy for gastric cancer was established for the first time, which enables accurate screening of beneficiary populations and risk stratification before treatment, guides treatment selection, and improves treatment efficacy.

Description

Technical Field

[0001] The present invention relates to the field of medical testing, and in particular, to a diagnostic model for predicting the prognosis of gastric cancer CAR-T cell immunotherapy and its application. Background Art

[0002] Chimeric antigen receptor T cell (CAR-T) therapy is a revolutionary, personalized treatment innovation. It genetically engineered a patient's own T cells to specifically recognize and target tumor cells. CAR-T therapy has become a research hotspot in recent years due to its remarkable efficacy and long-lasting anti-tumor activity, particularly in hematologic malignancies. However, the application of CAR-T therapy in solid tumors is still in its infancy, with limited efficacy, significantly lagging behind that in hematologic malignancies.

[0003] Claudin18.2 (CLDN18.2), a member of the tight junction protein family, has restricted expression under healthy conditions and is localized exclusively to the gastric mucosa. However, during tumorigenesis, disruption of intercellular tight junctions leads to abnormal exposure of the CLDN18.2 epitope on the cell surface, particularly in gastrointestinal tumors. This specific expression pattern has made it a promising therapeutic target and has rapidly fueled the development of therapies such as antibody-drug conjugates and monoclonal / bispecific antibodies. However, existing therapies targeting CLDN18.2 have limited clinical benefits, stringent enrollment criteria (e.g., a high CLDN18.2 expression threshold), and poor tolerability resulting from treatment-related adverse events (e.g., vomiting), hindering their clinical application. Innovative strategies are urgently needed to optimize efficacy.

[0004] Our team previously reported in a Phase I clinical trial (NCT03874897) the positive efficacy and manageable safety of CLDN18.2-specific CAR-T therapy (CT041) in the treatment of gastrointestinal tumors. For refractory / recurrent advanced gastric cancer, CT041 achieved an objective response rate and disease control rate of 54.9% and 96.1%, respectively, significantly outperforming existing therapies. The median progression-free survival and overall survival reached 5.8 months and 9.0 months, respectively. The success of CT041 represents a breakthrough in CAR-T therapy for the treatment of solid tumors and establishes its potential as a new strategy for solid tumor treatment.

[0005] Despite the proven efficacy of CT041, some patients still experience primary drug resistance, and approximately 50% of patients experience progression within 6 months of achieving initial disease control. The highly heterogeneous and complex tumor microenvironment (TME) of CLDN18.2-positive gastric cancer may be a key factor influencing treatment response and prognosis. However, no studies have yet conducted multidimensional analyses of the TME of patients receiving CT041 treatment, including cellular components and spatial relationships. Traditional immunobiological analysis remains limited to analyzing the content of single or a few markers. More importantly, there is currently a lack of methods or combinations of indicators that can be used to predict the efficacy and prognosis of CT041 treatment, hindering the improvement of treatment efficacy and patient prognosis. Summary of the Invention

[0006] In order to fill the gaps in the prior art, the present invention provides the following technical solutions:

[0007] A first aspect of the present invention provides a method for constructing a diagnostic model for predicting the prognosis of gastric cancer CAR-T cell immunotherapy, the method comprising the following steps:

[0008] 1) Recruit sufficient gastric cancer tissue specimens from patients undergoing CAR-T cell immunotherapy for gastric cancer to serve as a development cohort;

[0009] 2) Follow patients in the development cohort for more than 166 days and use imaging to assess whether they are responders or non-responders to treatment;

[0010] 3) Differences in the tumor microenvironment of patients in the development cohort were detected by mIHC. Univariate Cox proportional hazards regression analysis was used to evaluate the relationship between cellular components and progression-free survival. After screening, the percentage of TGF-β1+IQGAP3+ cells, the percentage of CD8+TCF1+GZMK+ cells, and the effective score between TGF-β1+IQGAP3+ cells and FAP+ cells were determined as calculation variables for CAR-T treatment efficacy and prognostic models;

[0011] 4) Calculate the weights for each indicator using the univariate Cox proportional hazards regression algorithm, and construct a model to predict the efficacy and prognosis of CAR-T therapy by weighted sum:

[0012] CTRS = [Ln(3.42) (normalized percentage of TGF-β1+IQGAP3+ cells) + Ln(0.21) (normalized percentage of CD8+TCF1+GZMK+ cells) + Ln(1.46) (normalized efficacy score between TGF-β1+IQGAP3+ cells and FAP+ cells)] (-1).

[0013] In one embodiment, patients with a progression-free survival of ≥166 days are defined as treatment responders, and the rest are treatment non-responders.

[0014] In one embodiment, the effective score is defined as the ratio of the number of pairs between the central cell and surrounding cells within a radius of 20 microns to the total number of central cells.

[0015] In one embodiment, the CTRS threshold / cutoff value is -0.68. When the CTRS is greater than the threshold, it is determined to be responsive to CAR-T therapy; conversely, when the CTRS is less than the threshold, it is determined to be unresponsive to CAR-T therapy.

[0016] In one embodiment, CTRS is divided into high and low score groups according to the cutoff value, and the progression-free survival and overall survival of the high CTRS score group are longer than those of the non-response group.

[0017] A second aspect of the present invention provides a model for predicting the prognosis of gastric cancer CAR-T cell immunotherapy, wherein the model is:

[0018] CTRS = [Ln(3.42) (Normalized percentage of TGF-β1+IQGAP3+ cells) + Ln(0.21) (normalized percentage of CD8+TCF1+GZMK+ cells) + Ln(1.46) (normalized efficacy score between TGF-β1+IQGAP3+ cells and FAP+ cells)] (-1);

[0019] Wherein, CAR-T is the treatment response score (CAR-T Response Score, CTRS); Ln(X) is the logarithm of X with the mathematical constant e as the base.

[0020] In one embodiment, the CTRS threshold / cutoff value is -0.68. When the CTRS is greater than the threshold, it is determined to be a response to CAR-T therapy; conversely, when the CTRS is less than the threshold, it is determined to be unresponsive to CAR-T therapy.

[0021] The third aspect of the present invention provides the use of a prognostic model constructed according to the aforementioned construction method for evaluating the prognosis of gastric cancer CAR-T cell immunotherapy.

[0022] A fourth aspect of the present invention provides use of the aforementioned model for evaluating the prognosis of gastric cancer CAR-T cell immunotherapy.

[0023] Compared with the prior art, the present invention has the following significant improvements:

[0024] 1. This study systematically analyzed the content and spatial distribution of microenvironmental components in CLDN18.2-positive gastric cancer tissue samples by integrating mIHC and whole-slice digital imaging quantitative analysis methods for the first time.

[0025] 2. This paper is the first to construct a model based on the spatial omics characteristics of CLDN18.2-positive gastric cancer TME to predict the prognosis of patients receiving CAR-T treatment.

[0026] 3. This study, through a multidimensional assessment of the TME in the context of CT041, a chimeric antigen receptor T cell immunotherapy targeting CLDN18.2, established for the first time a set of ideas and methods for screening gastric cancer patients who are responsive to CT041. This allows for precise pre-treatment screening of those who will benefit, guides treatment selection, and improves treatment efficacy. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0028] Figure 1 Schematic diagram of CTRS score calculation and multiplex fluorescent immunohistochemical staining multi-channel and single-channel staining results;

[0029] Figure 2 Multiple fluorescent immunohistochemical staining cell count results;

[0030] Figure 3 Multiplex fluorescence immunohistochemistry image analysis process;

[0031] Figure 4 . Principles of spatial analysis (A) and examples of effective scoring calculations (B);

[0032] Figure 5 Differences in CTRS scores between the responder and non-responder groups;

[0033] Figure 6 Waterfall chart visualizing each patient’s CTRS score and its association with CAR-T therapy response;

[0034] Figure 7 The predictive efficacy of the CTRS score for CAR-T therapy response was assessed based on the receiver operating characteristic curve and area under the curve.

[0035] Figure 8 The predictive efficacy of the CTRS score for progression-free survival in gastric cancer patients receiving CAR-T therapy.

[0036] Figure 9 The predictive efficacy of the CTRS score for overall survival in patients with gastric cancer receiving CAR-T therapy. DETAILED DESCRIPTION

[0037] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0038] Example 1 Systematic Analysis of the Content and Spatial Distribution of Microenvironmental Components in CLDN18.2-Positive Gastric Cancer Tissue Samples (Discovery Set)

[0039] 1. Specimen Collection

[0040] Inclusion criteria were: (1) histologically confirmed advanced gastric cancer; (2) CLDN18.2 positive expression (immunohistochemical staining intensity ≥2+ and the proportion of positive tumor cells ≥40%); (3) patients who had failed at least one line of standard treatment.

[0041] Exclusion criteria included: (1) the presence of target lesions ≥4 cm during lymphocyte apheresis; (2) active infection; and (3) extensive metastatic lesions in the lung, liver, or brain.

[0042] A total of 48 formalin-fixed paraffin-embedded (FFPE) gastric cancer tissue specimens from 48 patients with histologically confirmed gastric cancer, collected between March 2019 and April 2021, were ultimately included. Of the 48 enrolled patients, 42 (87.5%) had received at least two lines of therapy or at least triple therapy with a fluoropyrimidine, oxaliplatin, and paclitaxel, and 20 (41.7%) had previously received anti-PD-1 / PD-L1 antibody therapy. Of these, 24 tissue specimens were randomly selected as discovery set samples, and 24 tissue specimens were selected as validation set samples. This study was approved by the Ethics Committee of Peking University Cancer Hospital, and all participants or their legal guardians provided written informed consent.

[0043] Primary tumor samples were obtained by surgical resection or endoscopic biopsy before CAR-T therapy. Efficacy was assessed using the RECIST 1.1 criteria. Progression-free survival (PFS) was defined as the time from the first infusion to disease progression, regardless of cause, and overall survival (OS) was defined as the time from the first infusion to death. Patients with a PFS ≥ 166 days (the median PFS for patients included in this study) were defined as responders; the remainder were considered non-responders.

[0044] 2. Multiplex fluorescence immunohistochemistry (mIHC):

[0045] A. Tumor Immunoarray (Panel) Setup:

[0046] a. Tumor tissue and cell structure composition: used for tumor and stroma segmentation, including cytokeratin for cell membrane labeling and 4',6-diamidino-2-phenylindole for cell nucleus labeling;

[0047] b. CD8-positive T cell subset markers: cell identity marker CD8, immune checkpoint markers LAG-3 and PD-1, T cell killing marker GZMK, T cell activity marker CXCL13, T cell exhaustion markers TCF1 and CD39.

[0048] B. Technical Process:

[0049] To investigate the infiltration characteristics of various components within the TME, we quantitatively analyzed the content and spatial distribution of cellular components in 24 discovery set FFPE specimens using mIHC staining. The specific steps are as follows:

[0050] a. Baking: Bake the FFPE sample in an oven at 65°C for 20 minutes.

[0051] b. Dewaxing: Use xylene for 5 minutes each time, repeat 3 times;

[0052] c. Hydration: Soak the slides in graded ethanol: 100% ethanol for 5 minutes, 95% ethanol for 5 minutes, and 70% ethanol for 2 minutes.

[0053] d. Antigen retrieval: Immerse the slides in alkaline antigen retrieval solution (PANOVUE, Cat. 0019020500). Microwave on high power until boiling, then switch to low power for 15 minutes and allow to cool naturally to room temperature.

[0054] e. Block nonspecific binding sites: Add primary antibody blocking solution (PANOVUE; Cat. 0018001120) and shake at room temperature for 10 minutes.

[0055] f. Primary antibody incubation: For example, using GZMK, use GZMK monoclonal antibody (ab282703, 1:200 dilution) and incubate at room temperature for 30 minutes.

[0056] g. Secondary antibody incubation: Add horseradish peroxidase (HRP)-conjugated secondary antibody working solution (PANOVUE; Cat.0013001010) and incubate at room temperature for 10 minutes.

[0057] h. Tyramide signal amplification (TSA): Add PPD520 fluorescently labeled tyramide (1:100 dilution) and shake at 25°C for 10 minutes. Wash with TBST buffer (pH 7.6) for 5 minutes each, repeat three times.

[0058] i. Circulating antigen retrieval and antibody labeling: Microwave-mediated cyclic antigen epitope retrieval (pH 9.0, 95°C for 10 minutes) was used to regenerate antigen activity between rounds of antibody labeling. After each round of retrieval, the sample was washed three times with Tris-EDTA buffer (pH 8.0) to eliminate cross-reactivity.

[0059] j. Nuclear counterstaining: DAPI staining solution (0.1 μg / mL, containing anti-quencher) was added dropwise and incubated at room temperature in a dark humidified chamber for 5 minutes. The cells were then washed with TBST buffer (containing 0.1% Tween-20, pH 7.4) with shaking for 3 times, each time for 5 minutes.

[0060] k. Mounting: Wash the slides with sterile water for 2 minutes. Once the slides are slightly dry, use a pipette to add a drop of super-strong anti-fade mounting medium to the slides, submerging the sample area. Apply a coverslip and seal the perimeter with a neutral resin.

[0061] 1. Image interpretation: Multi-channel fluorescence images were acquired and interpreted using a laser confocal microscope;

[0062] m. Antibody Panel Optimization and Verification: Based on the antibody combination optimization system established in preliminary experiments (Table 1), an orthogonal experimental design was used to screen for the optimal labeling sequence and fluorescence spectrum matching scheme.

[0063] The pre-experimental steps are as follows:

[0064] (1) Optimize and clarify the dilution ratio of primary and secondary antibodies through conventional immunohistochemistry;

[0065] (2) Set up experimental controls to verify the antibody titer and the reliability of the detection system. These include positive controls, using homologous and heterologous tissue sections that have been confirmed to contain the target antigen; negative controls, using tissues that have been confirmed to not contain the target antigen; and negative reagent controls, i.e., blank controls;

[0066] (3) Determine the matching principle between TSA fluorescent dye and antigen: antigens with high abundance should be matched with TSA dyes with weak fluorescence intensity, while antigens with low abundance should be matched with TSA dyes with high fluorescence intensity;

[0067] (4) Determine the order of multi-label staining and set the primary antibody staining order to cell membrane--cytoplasm--nucleus.

[0068] Table 1. Antibody information used for marker detection

[0069]

[0070] 3. Multispectral microscopy and quantitative analysis:

[0071] A. Panoramic scanning of multiple rounds of stained samples was performed using the Vectra 3.0 multispectral imaging system (PerkinElmer). The spectral channel configuration is as follows:

[0072] B.CY5 channel: covers DAPI, PPD650, and PPD690 signal acquisition;

[0073] C. FITC channel: specifically captures PPD520 fluorescence signals (excitation 488 nm / emission 525 nm);

[0074] D.Texas Red channel: Analyze the PPD620 emission spectrum (excitation 587 nm / emission 615 nm);

[0075] Results: The results of multi-channel and single-channel staining are as follows: Figure 1 As shown in the figure, CLDN18.2 and IQGAP3 are mainly expressed in the gastric epithelium, COL3A1 and FAP are mainly expressed in the interstitial area, and the other markers are spatially co-localized, which is consistent with experimental expectations. In addition, the percentage of COL3A1+ cells and CD39+ cells in the detected cell types is significantly higher ( Figure 2 ).

[0076] 4. Multispectral imaging:

[0077] A. Images were captured using the PerkinElmer Mantra quantitative pathology imaging system.

[0078] B. Two experienced pathologists were responsible for selecting representative fields of view and acquiring multiple fields of view under a 20x objective for subsequent in-depth analysis;

[0079] C. Tumor Center Selection: Based on previously acquired full-slide scans, a fixed-size marker tool was used in the PerkinElmer phenotype map to precisely identify the tumor center. As many viable regions as possible were selected in each specimen, while minimizing overlap between regions. All processed data underwent rigorous quality control by a pathologist.

[0080] D. Use inForm Image Analysis Software 2.4 (analysis process as shown in Figure 3 As shown in the figure, the multispectral images were analyzed. First, a spectral library was constructed based on the single-channel staining images of each fluorescent dye, and the autofluorescence spectrum of the tissue was extracted using unstained sections. Subsequently, each DAPI-stained cell was identified based on the characteristics of the fluorophore and the cell morphological characteristics related to the DAPI signal. Using inForm software, the acquired images were segmented into tissue components to distinguish between tumor cells and interstitial areas. Figure 3 It can be seen that the results of tissue and cell segmentation of multispectral images by inForm software are highly consistent with H&E (hematoxylin and eosin) pathological sections, indicating the accuracy of the method.

[0081] 5. Digital analysis of tissue microenvironment

[0082] InForm software (Perkin Elmer) was used to perform basic segmentation of the multispectral images. Tumor and stroma tissue segmentation was performed based on cytokeratin expression, and cell segmentation was performed based on DAPI staining. Further multi-marker co-expression data analysis was performed based on the staining results as follows:

[0083] (1) Cell ratio = phenotype positive cell count / total cell count

[0084] (2) Cell density = phenotype-positive cell count / effective detection area, unit: cells / mm²

[0085] (3) Distance:

[0086] Cell coordinate data: The single-cell spatial coordinate matrix comes from the cell_seg_data file exported by the inform software, where the Cell X Position and Cell Y Position columns are the horizontal and vertical coordinates of the cell in the field of view image, respectively;

[0087] Intercellular distance data: Intercellular distance data was analyzed using a two-dimensional Euclidean coordinate system. The coordinate base generated raw pixel-level spatial positioning data based on a rectangular image acquisition area. Spatial scale calibration was based on the optical objective magnification and applied a pixel-to-micrometer conversion factor. The formula is physical length = number of pixels / system calibration parameter, in pixels / μm. Finally, the following Euclidean distance calculation formula was used:

[0088] (4) Effective score: It is defined as the ratio of the number of pairs of the central cell with the surrounding cells within a radius of 20 μm to the total number of the central cell. Examples of spatial analysis and effective score calculation are as follows: Figure 4 shown.

[0089] Example 2 Construction of CAR-T Therapeutic Efficacy and Prognosis Model

[0090] Univariate Cox proportional hazards regression analysis evaluated the relationship between cellular composition and progression-free survival (PFS) (Table 2). The results showed that the percentage of TGF-β1+IQGAP3+ cells in panel 1 and the difference between TGF-β1+IQGAP3+ cells and FAP+ cells were associated with shorter PFS after CAR-T therapy. In contrast, the percentage of CD8+TCF1+GZMK+ cells was associated with longer PFS after CAR-T therapy. The direction and magnitude of these effects were maintained after multivariate adjustment.

[0091] Table 2. Univariate Cox proportional hazards regression analysis of the association between cellular components and progression-free survival in patients with CLDN18.2-positive gastric cancer treated with CAR-T.

[0092]

[0093] a: Multivariate Cox proportional hazards regression model was analyzed by including peritoneal metastasis (presence vs. absence), Lauren classification (diffuse vs. mixed vs. intestinal), and the positive rate of Claudin18.2 protein detected by immunohistochemistry; bd: The cutoff values for high and low risk were as follows: 5.63%, 0.01%, and 0.03%; Abbreviations: HR, hazard ratio; CI, confidence interval.

[0094] Based on this, we selected the percentage of TGF-β1+IQGAP3+ cells, the percentage of CD8+TCF1+GZMK+ cells, and the effective score between TGF-β1+IQGAP3+ cells and FAP+ cells as the calculation variables for the CAR-T treatment efficacy and prognosis model. The weight of each indicator was calculated using the univariate Cox proportional hazards regression algorithm, and a weighted sum method was used to construct a model for predicting the efficacy and prognosis of CAR-T treatment:

[0095] CTRS = [Ln(3.42) (normalized percentage of TGF-β1+IQGAP3+ cells) + Ln(0.21) (normalized percentage of CD8+TCF1+GZMK+ cells) + Ln(1.46) (normalized efficacy score between TGF-β1+IQGAP3+ cells and FAP+ cells)] (-1).

[0096] Wherein, CAR-T is the treatment response score (CAR-T Response Score, CTRS); Ln(X) is the logarithm of X with the mathematical constant e as the base.

[0097] The CTRS threshold / cutoff value is -0.68. When the CTRS is greater than the threshold, it is judged to be responsive to CAR-T treatment; conversely, when the CTRS is less than the threshold, it is judged to be unresponsive to CAR-T treatment.

[0098] Example 3 Verification of the validity of the model

[0099] To further verify the effectiveness of the model, the efficacy of CAR-T treatment was evaluated in 24 validation set patients to test the effectiveness of the CTRS score:

[0100] The results showed that the CTRS score in the CAR-T treatment response group was significantly higher than that in the non-response group ( Figure 5 ); display the CTRS score of each patient separately, through the waterfall chart ( Figure 6 ) It was found that the treatment responders were concentrated in the group with higher CTRS scores on the left side, further proving that the CTRS score was associated with the efficacy of CAR-T treatment in gastric cancer patients; the receiver operating characteristic curve (ROC) and area under the curve (AUC) were used to evaluate the predictive efficacy of the CTRS score for the response to CAR-T treatment ( Figure 7The maximum AUC was found to be 0.889, with a 95% confidence interval of 0.752-1.000, a specificity of 0.833, a sensitivity of 1.000, an accuracy of 0.875, and a Youden index of 0.833.

[0101] CTRS was divided into high and low score groups according to the cutoff value. It was found that the PFS (HR 0.39, 95%CI0.17-0.92, P = 0.032) and overall survival (OS) (HR 0.36, 95%CI 0.13-0.96, P = 0.041) of the high CTRS score group were significantly higher than those of the non-response group. CTRS can effectively predict prognosis ( Figure 8-9 ).

[0102] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A method for constructing a diagnostic model for predicting the prognosis of gastric cancer CAR-T cell immunotherapy, characterized in that: The CAR-T cell immunotherapy is CT041 treatment, and the construction method includes the following steps: 1) recruiting sufficient gastric cancer tissue specimens from patients undergoing gastric cancer CAR-T cell immunotherapy to serve as a development cohort group; 2) Follow patients in the development cohort for more than 166 days and use imaging to assess whether they are responders or non-responders to treatment; 3) Differences in the tumor microenvironment of patients in the development cohort were detected by mIHC. Univariate Cox proportional hazards regression analysis was used to assess the relationship between cellular components and progression-free survival. After screening, the percentages of TGF-β1+IQGAP3+ cells, CD8+TCF1+GZMK+ cells, and the effective score between TGF-β1+IQGAP3+ cells and FAP+ cells were determined as variables for calculating the efficacy and prognostic model of CAR-T therapy. The effective score was defined as the ratio of the number of paired cells with surrounding cells within a 20-micron radius of the central cell to the total number of central cells. 4) Calculate the weights for each indicator using the univariate Cox proportional hazards regression algorithm, and construct a model to predict the efficacy and prognosis of CAR-T therapy by weighted sum: CTRS = [Ln(3.42) (normalized percentage of TGF-β1+IQGAP3+ cells) + Ln(0.21) (normalized percentage of CD8+TCF1+GZMK+ cells) + Ln(1.46) (normalized efficacy score between TGF-β1+IQGAP3+ cells and FAP+ cells)] (-1).

2. The construction method according to claim 1, wherein Patients with progression-free survival ≥166 days were defined as treatment responders, and the rest were defined as treatment non-responders.

3. The construction method according to claim 1, wherein The CTRS cutoff value is -0.

68. When the CTRS is greater than the cutoff value, it is judged to be responsive to CAR-T treatment; conversely, when the CTRS is less than the cutoff value, it is judged to be unresponsive to CAR-T treatment.

4. The construction method according to claim 3, wherein: CTRS was divided into high and low score groups according to the cutoff value. The progression-free survival and overall survival of the high CTRS score group were higher than those of the non-response group.

Images