A method and system for constructing a risk stratification model for patients with EGFR mutant non-small cell lung cancer meningeal metastasis

Abstract

The application discloses a kind of risk stratification model construction method and system for EGFR mutant non-small cell lung cancer meningeal metastasis patient, belong to tumor prognosis evaluation technical field.The multicenter retrospective clinical database is first constructed and data preprocessing is completed, then the predictor is screened by LASSO Cox regression, and the five core prognostic factors of ECOG PS score, brain metastasis, MRI meningeal enhancement, CSFC result and intracranial pressure are determined by multi-factor Cox proportional risk regression and linear predictor is constructed, then it is converted into integer type weighted risk score, finally the best cutoff value is determined based on survival function, and the patient is divided into low-risk group and high-risk group.The application can clearly distinguish the patient group with significant difference in prognosis, the required indicators are all clinical routine items, the operation is simple, the accessibility is high, can accurately guide individualized treatment decision and clinical trial stratification, and has important clinical application value.

Description

Technical Field

[0001] This invention relates to the field of tumor prognostic assessment technology, and more specifically to a method and system for constructing a risk stratification model for patients with EGFR-mutant non-small cell lung cancer and leptomeningeal metastases. Background Technology

[0002] Leptomeningeal metastasis (LM) is a catastrophic and increasingly common complication in patients with epidermal growth factor receptor (EGFR)-mutated non-small cell lung cancer (NSCLC). Although the use of third-generation EGFR tyrosine kinase inhibitors (TKIs) has improved systemic control and overall survival, the central nervous system, acting as a refuge for these drugs, has led to an increase in the incidence of LM. Following a diagnosis of LM, patients experience a sharp decline in neurological function and quality of life; even with aggressive treatment, the historical median overall survival is only 3–11 months.

[0003] Currently, there is a high degree of uncertainty in the prognostic assessment of EGFR-mutant NSCLC patients with leukemia (LM) in clinical practice. There is a lack of precise risk stratification tools specifically designed for this molecular subtype and applicable to the context of contemporary third-generation TKI therapy. Some existing prognostic models are either derived from highly heterogeneous pan-cancer cohorts, fail to integrate key clinical, imaging, and cerebrospinal fluid indicators, or lack rigorous external validation, resulting in limited clinical applicability and predictive accuracy.

[0004] Therefore, how to provide a risk assessment model that can accurately predict patient survival using routine clinical data and thus guide individualized treatment decisions is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the present invention provides a method and system for constructing a risk stratification model for patients with EGFR-mutant non-small cell lung cancer (NSCLC) and leptomeningeal metastases. This model integrates five conventional clinical parameters—Patient Performance Status Score (ECOG PS), presence of brain metastases, leptomeningeal enhancement on MRI, cerebrospinal fluid cytology (CSFC) results, and intracranial pressure (ICP) level—to predict the overall survival of patients with EGFR-mutant NSCLC and leptomeningeal metastases (LM). The core of this model lies in providing a simple, easy-to-use, and quantifiable scoring system that clearly divides patients into low-risk and high-risk groups to guide the selection of clinical treatment intensity.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: A method for constructing a risk stratification model for patients with EGFR-mutant non-small cell lung cancer and leptomeningeal metastases includes: S1. Construct a multicenter retrospective clinical database and preprocess the clinical data of patients with EGFR-mutant non-small cell lung cancer with leptomeningeal metastasis in the database. S2. The LASSOCox regression model was used to screen the preprocessed clinical data for penalized variables to obtain predictors with non-zero coefficients. S3. Incorporate the screened predictors into a multivariate Cox proportional hazards regression model, with total survival as the dependent variable, determine the prognostic factors and their corresponding coefficients, and construct a linear predictor based on the prognostic factors and their corresponding coefficients. S4. Construct a weighted risk scoring system and convert the linear predictor into an integer risk score; S5. Determine the optimal cutoff value based on the survival function, and divide the patient's integer risk score into low-risk and high-risk groups according to the optimal cutoff value.

[0007] Optionally, S1 specifically includes: We collected clinical data from multiple medical centers on patients with EGFR-mutant NSCLC and LM. The collected clinical data should include at least: age, sex, smoking history, EGFR mutation subtype, LM type, brain metastasis status, pathological type, ECOG PS score, contrast-enhanced brain MRI results, intracranial pressure value, CSFC results, treatment history, overall survival and survival status; The collected clinical data were standardized, cleaned, and preprocessed, and the continuous variables were transformed to the optimal cutoff value.

[0008] Optionally, S2 specifically includes: The LASSO regression model achieves this by introducing an L1 regularization term into the loss function. Through 10-fold cross-validation, the λ value that minimizes the bias of the LASSO regression model is selected as the predictor with non-zero coefficients.

[0009] Optionally, the prognostic factors and their corresponding coefficients are specifically as follows: Factor 1: ECOG PS coefficient Factor 2: Brain metastasis, coefficient Factor 3: MRI meningeal enhancement, coefficient Factor 4: CSFC, coefficient Factor 5: ICP, coefficient .

[0010] Optionally, the linear predictor is specifically used to calculate the relative risk of an individual, using the following formula:

[0011] In the formula, As factor 1, As a factor of 2, The factor is 3. The factor is 4. It is a factor of 5.

[0012] Optionally, the weighted risk scoring system specifically refers to: Select a benchmark factor and set its risk score as follows: S base =100 points; For other factors, calculate their risk scores. S i Then, the values ​​are rounded to the nearest integer to obtain the final risk score system; The scores of each factor are summed to obtain the total risk score for each patient.

[0013] Optionally, the risk score S i The calculation formula is as follows: Optionally, S5 specifically includes: First, calculate the linear predictor LP value for each individual in the derivation queue; Then, the predicted survival function S(t) for each individual is calculated based on the multivariate Cox proportional hazards regression model, and the predicted survival probability at a specific time point is calculated. By iterating through all possible LP values ​​and employing an algorithm that maximizes the difference in survival curves between risk groups, the optimal LP cutoff value for dividing the population into low- and high-risk groups is determined. LP cut ; When the total risk score is less than the optimal LP cutoff value, it is classified as a low-risk group; when the total risk score is greater than or equal to the optimal LP cutoff value, it is classified as a high-risk group.

[0014] A risk stratification model construction system for patients with EGFR-mutant non-small cell lung cancer and leptomeningeal metastases includes: Data acquisition and preprocessing unit: used to receive and standardize the five core clinical parameters input and convert them into binary values; Weighted risk calculation engine: Connected to the data acquisition and preprocessing unit, it incorporates a weighted risk scoring system; this engine calculates the patient's total risk score based on the following preset, non-linear weight values: Dynamic risk stratification unit: The comparator in this unit performs the following judgments: If the total risk score is greater than or equal to the preset threshold, the group is classified as high-risk. If the total risk score is less than the preset threshold, the group is classified as low-risk. Prognostic prediction output unit: used to visualize or report risk assessment results; this unit directly outputs the qualitative results for high-risk or low-risk groups.

[0015] As can be seen from the above technical solution, compared with the prior art, the present invention discloses a method and system for constructing a risk stratification model for patients with EGFR-mutant non-small cell lung cancer leptomeningeal metastases, which has the following beneficial effects: 1. The model of this invention has a good ability to distinguish the risks of patients.

[0016] 2. The model of this invention has good universality and stability, and can be applied to different populations and medical centers.

[0017] 3. The model of this invention can clearly divide patients into two groups with significantly different prognoses: low-risk and high-risk. This precise stratification provides clear guidance for clinical decision-making: High-risk patients: intensive treatment plans can be prioritized to strive for longer survival benefits. Low-risk patients: moderate-intensity treatment plans can be considered to avoid the side effects and economic burden of overtreatment.

[0018] 4. The model of this invention can be used as an entry stratification factor in clinical trials to ensure baseline prognosis balance between the experimental group and the control group, thereby improving the reliability of the trial results.

[0019] 5. All five parameters required by the model of this invention are routine clinical examinations and assessments, requiring no additional, expensive, or hard-to-obtain molecular markers or special tests. It has extremely high clinical operability and promotional value, and is especially suitable for primary healthcare institutions. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0021] Figure 1 This is a schematic diagram of the method flow provided by the present invention. Detailed Implementation

[0022] 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.

[0023] This invention discloses a method for constructing a risk stratification model for patients with EGFR-mutant non-small cell lung cancer and leptomeningeal metastases, such as... Figure 1 As shown, it includes: S1. Construct a multicenter retrospective clinical database and preprocess the clinical data of patients with EGFR-mutant non-small cell lung cancer with leptomeningeal metastasis in the database. S2. The LASSO Cox regression model was used to screen the preprocessed clinical data for penalized variables to obtain predictors with non-zero coefficients. S3. Incorporate the screened predictors into a multivariate Cox proportional hazards regression model, with total survival as the dependent variable, determine the prognostic factors and their corresponding coefficients, and construct a linear predictor based on the prognostic factors and their corresponding coefficients. S4. Construct a weighted risk scoring system and convert the linear predictor into an integer risk score; S5. Determine the optimal cutoff value based on the survival function, and divide the patient's integer risk score into low-risk and high-risk groups according to the optimal cutoff value.

[0024] Specifically, S1 includes: a. Collect clinical data from multiple medical centers on NSCLC patients with EGFR mutations and LM.

[0025] b. The original set of variables collected should include at least: age, sex, smoking history, EGFR mutation subtype, LM type, brain metastasis status, pathological type, ECOG PS score, contrast-enhanced brain MRI results (meningeal enhancement status), intracranial pressure (ICP) value, CSFC results, treatment history, overall survival (OS), and survival status.

[0026] c. Standardize, clean, and preprocess the data, and perform optimal cutoff value transformation on continuous variables (such as ICP and age).

[0027] S2 specifically includes: a. In order to overcome the limitations of traditional Cox regression in handling high-dimensional collinear data and to achieve automated variable selection, the LASSO (Least Absolute Shrinkage and Selection Operator) Cox regression model is first used to perform punitive variable screening on all preprocessed variables.

[0028] b. LASSO regression achieves this by introducing an L1 regularization term (i.e., a penalty coefficient λ multiplied by the sum of the absolute values ​​of the coefficients) into the model's loss function. Ten-fold cross-validation is used to select the λ value that minimizes the model bias (i.e., minimizes the cross-validation error).

[0029] c. In this step, the model compresses the regression coefficients of some irrelevant or redundant variables to zero, thereby filtering out predictive variables with non-zero coefficients. This algorithmic step ensures that the variables ultimately entering the model are the most informative and least collinear robust predictors from the clinical data.

[0030] S3 specifically includes: a. Incorporate the non-zero coefficient variables selected by LASSO regression in step 2 into the multivariate Cox proportional hazards regression model, with overall survival (OS) as the dependent variable, for the final modeling.

[0031] b. This step obtains each independent predictor. x i regression coefficient β i And the corresponding risk ratio (HazardRatio, HR).

[0032] c. Based on the model output, five factors with independent prognostic significance and their coefficients are finally determined: Factor 1: ECOG PS coefficient ; Factor 2: Brain metastasis, coefficient ; Factor 3: MRI meningeal enhancement, coefficient ; Factor 4: CSFC, coefficient ; Factor 5: ICP, coefficient .

[0033] d. Thus, the formula for the linear predictor (LP) used to calculate individual relative risk is constructed:

[0034] In the formula, As factor 1, As a factor of 2, The factor is 3. The factor is 4. It is a factor of 5.

[0035] S4 specifically includes: a. To facilitate clinical application, the aforementioned continuous linear predictor LP is converted into an integer risk score. The specific conversion algorithm is as follows: Choose a benchmark factor (e.g., regression coefficient) The factor with the highest value (such as MRI meningeal enhancement) is assigned a risk score of 1. S base =100 points.

[0036] For other factors, their risk scores Si are calculated using the following formula and rounded to the nearest integer: .

[0037] in, These are the objective weights that the model learns from the data, reflecting the true biological impact of each risk factor; The selected reference weights are used to scale the objective weights proportionally to integer scores that are easy to remember and calculate, while maintaining the relative importance of each factor.

[0038] b. Using this weighted algorithm, the final risk score system is obtained (e.g., MRI meningeal enhancement = 100 points, ECOG PS ≥ 3 = 91 points, brain metastasis = 80 points, CSFC positive = 76 points, ICP > 220 = 73 points).

[0039] c. Each patient's total risk score is the sum of the scores for each factor: Total Score = S ECOG ×X ECOG +S BM ×X BM +S MRI ×X MRI +S CSFC ×X CSFC +S ICP ×X ICP .

[0040] Specifically, calculate the patient's total risk score: Risk score = 91 × (ECOGPS≥3? 1:0) + 80 × (Brain metastasis? 1:0) + 100 × (MRI enhancement? 1:0) + 76 × (CSFC positive? 1:0) + 73 × (ICP>220? 1:0); The weighted values ​​(91, 80, 100, 76, 73) are based on the regression coefficients of the Cox regression model in S3 using a specific formula. The calculation shows that the value was not arbitrarily assigned.

[0041] The question mark indicates yes or no, with yes = 1 and no = 0.

[0042] S5 specifically includes: First, calculate the linear predictor LP value for each individual in the derivation queue; Then, the predicted survival function S(t) for each individual is calculated based on the multivariate Cox proportional hazards regression model, and the predicted survival probability at a specific time point (e.g., 12 months) is calculated. Among them, through the formula Calculate individualized predicted survival probabilities for specific time periods (e.g., 12 months, 24 months); By iterating through all possible LP values ​​and employing an algorithm that maximizes the difference in survival curves between risk groups, the optimal LP cutoff value for dividing the population into low- and high-risk groups is determined. LP cut ; When the total risk score is less than the optimal LP cutoff value, it is classified as a low-risk group; when the total risk score is greater than or equal to the optimal LP cutoff value, it is classified as a high-risk group.

[0043] This step ensures that the risk grouping boundary is most relevant to the patient's actual survival probability, rather than an arbitrarily set integer. In practical clinical applications, this LP can be used... cut The value is then used to deduce a corresponding “total risk score” threshold (e.g., 320 points) using the formula in S4, in order to maintain ease of clinical use.

[0044] A risk stratification model construction system for patients with EGFR-mutant non-small cell lung cancer and leptomeningeal metastases includes: Data acquisition and preprocessing unit: used to receive and standardize the five core clinical parameters input and convert them into binary values; Weighted risk calculation engine: Connected to the data acquisition and preprocessing unit, it incorporates a weighted risk scoring system; this engine calculates the patient's total risk score based on the following preset, non-linear weight values: Dynamic risk stratification unit: The comparator in this unit performs the following judgments: If the total risk score is greater than or equal to the preset threshold, the group is classified as high-risk. If the total risk score is less than the preset threshold, the group is classified as low-risk. Prognostic prediction output unit: used to visualize or report risk assessment results; this unit directly outputs the qualitative results for high-risk or low-risk groups.

[0045] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0046] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for constructing a risk stratification model for patients with EGFR-mutated non-small cell lung cancer meningeal metastasis, characterized in that, include: S1. Construct a multicenter retrospective clinical database and preprocess the clinical data of patients with EGFR-mutant non-small cell lung cancer with leptomeningeal metastasis in the database. S2. The LASSO Cox regression model was used to screen the preprocessed clinical data for penalized variables to obtain predictors with non-zero coefficients. S3. Incorporate the screened predictors into a multivariate Cox proportional hazards regression model, with total survival as the dependent variable, determine the prognostic factors and their corresponding coefficients, and construct a linear predictor based on the prognostic factors and their corresponding coefficients. S4. Construct a weighted risk scoring system and convert the linear predictor into an integer risk score; S5. Determine the optimal cutoff value based on the survival function, and divide the patient's integer risk score into low-risk and high-risk groups according to the optimal cutoff value.

2. The method for constructing a risk stratification model for patients with meningeal metastasis of EGFR-mutated non-small cell lung cancer according to claim 1, characterized in that, S1 specifically includes: We collected clinical data from multiple medical centers on patients with EGFR-mutant NSCLC and LM. The collected clinical data should include at least: age, sex, smoking history, EGFR mutation subtype, LM type, brain metastasis status, pathological type, ECOG PS score, contrast-enhanced brain MRI results, intracranial pressure value, CSFC results, treatment history, overall survival and survival status; The collected clinical data were standardized, cleaned, and preprocessed, and the continuous variables were transformed to the optimal cutoff value.

3. The method for constructing a risk stratification model for patients with EGFR-mutant non-small cell lung cancer and leptomeningeal metastases according to claim 1, characterized in that, S2 specifically includes: The LASSO regression model achieves this by introducing an L1 regularization term into the loss function. Through 10-fold cross-validation, the λ value that minimizes the bias of the LASSO regression model is selected as the predictor with non-zero coefficients.

4. The method for constructing a risk stratification model for patients with EGFR-mutant non-small cell lung cancer and leptomeningeal metastases according to claim 1, characterized in that, The prognostic factors and their corresponding coefficients are as follows: Factor 1: ECOG PS coefficient Factor 2: Brain metastasis, coefficient Factor 3: MRI meningeal enhancement, coefficient Factor 4: CSFC, coefficient Factor 5: ICP, coefficient .

5. The method for constructing a risk stratification model for patients with EGFR-mutant non-small cell lung cancer leptospirosis according to claim 4, characterized in that, The linear predictor is specifically used to calculate the relative risk of an individual, using the following formula: In the formula, As factor 1, As a factor of 2, The factor is 3. The factor is 4. It is a factor of 5.

6. The method for constructing a risk stratification model for patients with EGFR-mutant non-small cell lung cancer leptospirosis according to claim 5, characterized in that, The weighted risk scoring system is specifically as follows: Select a benchmark factor and set its risk score as follows: S base =100 points; For other factors, calculate their risk scores. S i Then, the values ​​are rounded to the nearest integer to obtain the final risk score system; The scores of each factor are summed to obtain the total risk score for each patient.

7. The method for constructing a risk stratification model for patients with EGFR-mutant non-small cell lung cancer leptospirosis according to claim 6, characterized in that, The risk score S i The calculation formula is as follows: ； in, These are the objective weights learned by the model from the data, reflecting the true biological impact of each risk factor; The selected reference weights are used to scale the objective weights proportionally to integer scores that are easy to remember and calculate, while maintaining the relative importance of each factor.

8. The method for constructing a risk stratification model for patients with EGFR-mutant non-small cell lung cancer and leptomeningeal metastases according to claim 1, characterized in that, S5 specifically includes: First, calculate the linear predictor LP value for each individual in the derivation queue; Then, the predicted survival function S(t) for each individual is calculated based on the multivariate Cox proportional hazards regression model, and the predicted survival probability at a specific time point is calculated. By iterating through all possible LP values ​​and employing an algorithm that maximizes the difference in survival curves between risk groups, the optimal LP cutoff value for dividing the population into low- and high-risk groups is determined. LP cut ; When the total risk score is less than the optimal LP cutoff value, it is classified as a low-risk group; when the total risk score is greater than or equal to the optimal LP cutoff value, it is classified as a high-risk group.

9. A risk stratification model construction system for patients with EGFR-mutant non-small cell lung cancer and leptomeningeal metastases, characterized in that, A method for constructing a risk stratification model for patients with EGFR-mutant non-small cell lung cancer leptomeningeal metastases, according to any one of claims 1-8, includes: Data acquisition and preprocessing unit: used to receive and standardize the five core clinical parameters input and convert them into binary values; Weighted risk calculation engine: Connected to the data acquisition and preprocessing unit, it incorporates a weighted risk scoring system; this engine calculates the patient's total risk score based on the following preset, non-linear weight values: Dynamic risk stratification unit: The comparator in this unit performs the following judgments: If the total risk score is greater than or equal to the preset threshold, the group is classified as high-risk. If the total risk score is less than the preset threshold, the group is classified as low-risk. Prognostic prediction output unit: used to visualize or report risk assessment results; this unit directly outputs the qualitative results for high-risk or low-risk groups.

Images