Oral cancer patient postoperative dysphagia risk prediction method based on LASSO regression

Abstract

The invention discloses an LASSO regression-based oral cancer patient postoperative dysphagia risk prediction method, which comprises the following steps of collecting clinical data, performing dysphagia function evaluation on postoperative patients by adopting an oral cancer MAN-OC (MASA-OC), and dividing the patients into a dysphagia group and a dysphagia-free group according to an MASA-OC result; candidate variables are screened through LASSO regression analysis, and potential risk factor analysis of a single factor is carried out through LASSO regression analysis; analyzing independent danger prediction factors of the plurality of factors through logistic regression; independent risk prediction factors are obtained, a risk prediction model is constructed, a prediction result is displayed in a column graph form, and a risk probability is obtained; and evaluating the efficiency of the risk prediction model. And screening out key prediction factors through LASSO regression analysis. On the basis of the factors, a clinical risk prediction model is constructed and displayed in a column diagram form, and medical staff can conveniently recognize high-risk groups in an early stage in a clinical working environment and carry out early intervention through the visual model.

Description

Technical Field

[0001] The present invention relates to the technical field of dysphagia prediction after oral cancer surgery, and in particular to a method for predicting the risk of dysphagia after oral cancer surgery based on LASSO regression. Background Art

[0002] Oral cancer is one of the common malignant tumors of the head and neck. Its clinical symptoms include oral ulcers, pain, bleeding, dysphagia, etc., which seriously affect the patient's quality of life. Since the oral cavity is located in the oral and maxillofacial region of the human body, it is the beginning of the digestive tract and respiratory tract, and has the functions of chewing, swallowing, speaking, and maintaining maxillofacial morphology; at the same time, surgery may cause damage to specific anatomical parts or nerves, and patients are prone to swallowing disorders after surgery, which in turn affects the patient's long-term quality of life.

[0003] At present, the research on early postoperative dysphagia in patients with oral cancer is still in its infancy. In particular, the screening and evaluation tools for postoperative dysphagia in patients with oral cancer are mostly universal tools. There is still a lack of objective, convenient and targeted screening tools suitable for clinical medical staff. In addition, most research indicators are single and there is a lack of multi-indicator joint prediction models. Therefore, it is urgent to develop a risk prediction model for postoperative dysphagia in patients with oral cancer based on routine clinical testing items and multi-indicator joint construction, so as to improve the scientificity of risk prediction of postoperative dysphagia in patients with oral cancer and reduce the incidence of postoperative dysphagia in patients with oral cancer. Summary of the invention

[0004] The purpose of the present invention is to provide a risk prediction method for postoperative dysphagia in oral cancer patients based on LASSO regression in response to the above-mentioned problems. Through LASSO regression analysis, seven key predictive factors are screened out, including daily living activities, depression, tumor T stage, repair and reconstruction, tongue plasty, jaw resection, and nasopharyngeal ventilation tube retention. Based on these factors, a clinical risk prediction model is constructed and displayed in the form of a nomogram. Through the visual model, it is convenient for medical staff to identify high-risk groups early in the clinical working environment and conduct early intervention.

[0005] The technical solution of the present invention is as follows:

[0006] A method for predicting the risk of postoperative dysphagia in patients with oral cancer based on LASSO regression, comprising the following steps:

[0007] Clinical data were collected, and a cohort study design was adopted. The Mann Swallowing Scale for Oral Cancer (MASA-OC) was used to evaluate the swallowing function of patients at the bedside 48-72 hours after surgery. According to the results of MASA-OC, the patients were divided into the dysphagia group and the non-dysphagia group;

[0008] LASSO regression analysis was used to screen candidate variables, and LASSO regression analysis was used to analyze the potential risk factors for dysphagia after oral cancer surgery.

[0009] Logistic regression was used to analyze multiple factors as independent risk predictors of dysphagia after oral cancer surgery;

[0010] Through regression analysis, we obtained the independent risk predictors of dysphagia after oral cancer surgery, and constructed a risk prediction model for dysphagia after oral cancer surgery. The prediction results were presented in the form of a nomogram to obtain the risk probability of dysphagia after oral cancer surgery.

[0011] To evaluate the effectiveness of risk prediction models.

[0012] Furthermore, the screening of candidate variables by LASSO regression analysis specifically includes the following steps:

[0013] Install and run the R Studio software glmnet package to perform LASSO regression;

[0014] The predictive factors affecting postoperative dysphagia after oral cancer surgery were input, and the penalty coefficient λ gradually decreased with the increase in the number of variables. Through 10-fold cross-validation, the predictive factors that can be selected for subsequent regression analysis were obtained. The predictive factors included: patient gender, daily living activities, depression, frailty, hemoglobin count, blood potassium level, blood sodium level, history of respiratory diseases, history of nervous system diseases, history of head and neck diseases, tumor location, maximum tumor diameter, tumor T stage, neck lymph node dissection, repair and reconstruction, glossoplasty, maxillomandibular resection, salivary gland palpation, nasopharyngeal ventilation tube retention, gastric tube retention, operation time, and intraoperative blood loss.

[0015] Furthermore, the potential risk factor analysis of dysphagia after oral cancer surgery using LASSO regression analysis specifically includes:

[0016] A univariate analysis was performed with whether oral cancer patients had dysphagia after surgery as the dependent variable and the risk factors screened by the above LASSO regression as the independent variables to obtain the potential risk factors for dysphagia after oral cancer surgery.

[0017] Furthermore, the potential risk factors include gender, ability to perform daily activities, depression, tumor location, maximum tumor diameter, tumor T stage, neck lymph node dissection, repair and reconstruction, glossoplasty, maxillomandibular resection, retention of nasopharyngeal ventilation tube, retention of gastric tube, operation duration, and intraoperative blood loss.

[0018] Furthermore, the independent risk predictors of dysphagia after oral cancer surgery analyzed by logistic regression include:

[0019] The occurrence of postoperative dysphagia in patients with oral cancer was taken as the dependent variable, and the risk factors with P < 0.05 in univariate analysis were taken as independent variables. Multivariate logistic regression analysis was performed by including them into the equation according to the stepwise backward method of maximum likelihood ratio estimation.

[0020] Furthermore, the independent risk predictors are daily living activities, depression, tumor T stage, repair and reconstruction, glossoplasty, maxillomandibular resection, and retention of nasopharyngeal airway.

[0021] Furthermore, the risk prediction model is constructed by constructing a prediction model for postoperative dysphagia in patients with oral cancer through the rms function package of R Studio software. For each variable determined to enter the logistic regression model, the corresponding Nomo score can be obtained through different variable values ​​​​through the vertical line. Finally, the sum of all variable scores is the total score, thereby obtaining the risk probability of postoperative dysphagia in patients with oral cancer.

[0022] Furthermore, the evaluation of the effectiveness of the risk prediction model specifically includes the following steps:

[0023] Discrimination of risk prediction model: By running the R language function package, the area under the ROC curve of the prediction model was 0.933, the sensitivity was 90.9%, and the specificity was 81.7%;

[0024] Calibration of risk prediction model: The Hosmer-Lemeshow method was used to test the goodness of fit of the model.

[0025] x 2 =5.391, P=0.715 (>0.05).

[0026] Furthermore, the clinical data include general information, assessment of daily living activities, preoperative swallowing function assessment, consciousness status 24 hours after surgery, results of various test indicators, medical history, disease and treatment conditions all collected in the hospital medical record system, anxiety and depression status, and weakness conditions.

[0027] Furthermore, the general information includes age, gender, height, and weight, which are collected from nursing records; the anxiety and depression status is obtained by the Hospital Anxiety and Depression Scale (HADS); and the frailty status is obtained by the Fried frailty assessment method.

[0028] Compared with the prior art, the present invention has the following beneficial effects:

[0029] 1. A method for predicting the risk of postoperative dysphagia in patients with oral cancer based on LASSO regression provides a convenient and effective tool for assessing the risk of postoperative dysphagia in oral cancer and provides a reference for postoperative follow-up and diagnosis and treatment. Compared with traditional methods, the prediction model of the present invention can more accurately and quickly identify patients with high risk of postoperative dysphagia in oral cancer and provide personalized postoperative management plans;

[0030] 2. A method for predicting the risk of postoperative dysphagia in patients with oral cancer based on LASSO regression, which accurately predicts the risk of postoperative dysphagia and provides a prediction model based on LASSO regression, which can accurately assess the risk of dysphagia in patients with postoperative oral cancer and help doctors take preventive and intervention measures in a timely manner;

[0031] 3. A method for predicting the risk of postoperative dysphagia in patients with oral cancer based on LASSO regression simplifies the clinical operation process. Through simple data collection and the use of the visualization tool Nomogram of the prediction model in the present invention, there is no need to increase the additional economic burden on patients and the workload of medical staff. The prediction model involved in the present invention is easy to promote and apply in clinical practice;

[0032] 4. A method for predicting the risk of postoperative dysphagia in patients with oral cancer based on LASSO regression can improve patients' treatment experience and rehabilitation effect, and can help medical staff to identify the risk of dysphagia in advance in the clinical environment, improve the quality of life after surgery, and accelerate the rehabilitation process. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is a process diagram for increasing lambda screening variables in the Lasso regression model of the present invention.

[0034] Figure 2 It is the Lasso coefficient distribution diagram of 22 candidate variables of the present invention.

[0035] Figure 3 This is a nomogram model diagram for predicting the risk of postoperative dysphagia in oral cancer patients according to the present invention.

[0036] Figure 4 It is a ROC curve diagram of the risk prediction model for postoperative dysphagia in oral cancer patients of the present invention.

[0037] Figure 5 This is a calibration curve diagram of the risk prediction model for postoperative dysphagia in oral cancer patients of the present invention.

[0038] Figure 6 Receiver operating characteristic (ROC) curve diagram for internal validation of the present invention.

[0039] Figure 7 This is a calibration chart for internal validation of the present invention. DETAILED DESCRIPTION

[0040] It should be noted that relational terms such as "first" and "second" are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0041] The features and performance of the present invention are further described in detail below in conjunction with the embodiments.

[0042] See also Figure 1-7 , a method for predicting the risk of postoperative dysphagia in patients with oral cancer based on LASSO regression, comprising the following steps:

[0043] Clinical data were collected, and a cohort study design was used to evaluate the swallowing function of patients at the bedside 48-72 hours after surgery to obtain the Mann Swallowing Scale for Oral Cancer (MASA-OC). According to the results of MASA-OC, the patients were divided into a dysphagia group and a non-dysphagia group.

[0044] LASSO regression analysis was used to screen candidate variables, and LASSO regression analysis was used to analyze the potential risk factors for dysphagia after oral cancer surgery.

[0045] Logistic regression was used to analyze multiple factors as independent risk predictors of dysphagia after oral cancer surgery;

[0046] Through regression analysis, we obtained the independent risk predictors of dysphagia after oral cancer surgery, and constructed a risk prediction model for dysphagia after oral cancer surgery. The prediction results were presented in the form of a nomogram to obtain the risk probability of dysphagia after oral cancer surgery.

[0047] To evaluate the effectiveness of risk prediction models.

[0048] Screening candidate variables through LASSO regression analysis specifically includes the following steps:

[0049] Install and run the R Studio software glmnet package to perform LASSO regression;

[0050] The predictive factors affecting oral cancer postoperative dysphagia were input, and the penalty coefficient λ gradually decreased with the increase in the number of variables. Through 10-fold cross-validation, the predictive factors that can be selected for subsequent regression analysis were obtained. The predictive factors include: patient gender, daily living activities, depression, frailty, hemoglobin count, blood potassium level, blood sodium level, history of respiratory diseases, history of nervous system diseases, history of head and neck diseases, tumor location, maximum tumor diameter, tumor T stage, neck lymph node dissection, repair and reconstruction, glossoplasty, maxillomandibular resection, salivary gland palpation, nasopharyngeal ventilation tube retention, gastric tube retention, operation time, and intraoperative blood loss. Figure 1 and Figure 2 As shown, Figure 1 It is the process of increasing lambda to select variables in the Lasso regression model. The horizontal axis is the optimal tuning parameter logλ, and the vertical axis is the regression coefficient (coefficients); Figure 2 The Lasso coefficient distribution of 22 candidate variables was used with 10-fold cross validation. Vertical lines were drawn at the minimum criterion (lambda.min) and one standard error to the right of the minimum criterion (lambda.1se), which were 0.01366216 and 0.06643365, respectively. The horizontal axis is the optimal tuning parameter logλ, and the vertical axis is the mean squared error.

[0051] The potential risk factors for dysphagia after oral cancer surgery were analyzed by LASSO regression analysis, including:

[0052] A univariate analysis was performed with whether oral cancer patients had dysphagia after surgery as the dependent variable and the risk factors screened by the above LASSO regression as the independent variables to obtain the potential risk factors for dysphagia after oral cancer surgery.

[0053] Potential risk factors include gender, ability to perform daily living activities, depression, tumor location, maximum tumor diameter, tumor T stage, neck dissection, repair and reconstruction, glossoplasty, maxillomandibular resection, retention of nasopharyngeal ventilation tube, retention of gastric tube, operation time, and intraoperative blood loss.

[0054] The independent risk predictors of dysphagia after oral cancer surgery analyzed by logistic regression include:

[0055] The occurrence of postoperative dysphagia in patients with oral cancer was taken as the dependent variable, and the risk factors with P < 0.05 in univariate analysis were taken as independent variables. Multivariate logistic regression analysis was performed by including them into the equation according to the stepwise backward method of maximum likelihood ratio estimation.

[0056] Independent risk predictors were reduced ability to perform daily activities, depression, tumor T stage, revision surgery, glossoplasty, maxillomandibular resection, and retention of a nasopharyngeal airway.

[0057] The risk prediction model was constructed by using the rms function package of R Studio software to construct a prediction model for postoperative dysphagia in patients with oral cancer. For each variable determined to enter the logistic regression model, the corresponding Nomo score could be obtained through different variable values ​​of the vertical line. Finally, the sum of all variable scores was the total score, thereby obtaining the risk probability of postoperative dysphagia in patients with oral cancer.

[0058] like Figure 3 As shown in the figure, the prediction model is visualized. An application example is taken as follows: a patient who needs obvious help in daily living activities, has no depression, has a tumor T stage of T4, has undergone repair and reconstruction, glossoplasty, maxillomandibular resection, and did not retain a nasopharyngeal airway after surgery. The nomogram scores are as follows: daily living activities 66 points, depression 0 points, tumor T stage 53 points, repair and reconstruction 70 points, glossoplasty 74 points, maxillomandibular resection 55 points, no nasopharyngeal airway after surgery 0 points, with a total score of 318 points. The risk of dysphagia after surgery is 99.9%.

[0059] The evaluation of the effectiveness of the risk prediction model includes the following steps:

[0060] like Figure 4 As shown, the discrimination of the risk prediction model: by running the R language function package, the area under the ROC curve of the prediction model AUC was 0.933, the sensitivity was 90.9%, and the specificity was 81.7%;

[0061] like Figure 5 As shown in Figure 2, the calibration of the risk prediction model: the model was tested for goodness of fit using the Hosmer-Lemeshow method.

[0062] x 2 =5.391, P=0.715 (>0.05). This indicates that the model has good goodness of fit and good calibration. In addition, by drawing the calibration curve, it can be seen that the C-index is 0.934, and the calibration curve and the ideal curve are relatively close, which also shows that the model has good calibration.

[0063] Clinical data included general information, assessment of daily living activities, preoperative swallowing function assessment, consciousness status 24 h after surgery, results of various test indicators, medical history, disease and treatment conditions all collected from the hospital medical record system, anxiety and depression status, and frailty.

[0064] General information including age, gender, height, and weight were collected from nursing records; anxiety and depression status was obtained using the Hospital Anxiety and Depression Scale (HADS); and frailty was obtained using the Fried Frailty Assessment Method.

[0065] Validation of risk prediction models

[0066] The model was internally validated using the Bootstrap method, which involved sampling the original data 1,000 times with replacement, analyzing the area under the ROC curve and drawing a calibration curve. Internal validation refers to testing the stability and repeatability of the model by repeatedly extracting the original data based on the model. Figure 6 It can be seen that the AUC is 0.912, the sensitivity is 93.3%, and the specificity is 63.8%. At the same time, it can be seen from the calibration curve that the actual curve is very close to the calibration curve, indicating that the risk prediction model established by the present invention has good consistency in predicting postoperative dysphagia in oral cancer patients in internal validation and the actual occurrence of dysphagia. It can be seen that the prediction model for postoperative dysphagia in oral cancer patients established by the present invention has good internal discrimination and calibration, and the prediction ability is relatively stable after 1000 Bootstrap samplings.

[0067] Clinical use

[0068] Through the visualization prediction model, medical staff can combine various factors of patients to conduct personalized assessment and prediction of dysphagia in clinical work. The clinical application steps are as follows: Take a patient who needs obvious help in daily living activities, has no depression, tumor T stage is T4, undergoes repair and reconstruction, glossoplasty, jaw resection, and does not retain nasopharyngeal airway after surgery as an example. The Nomogram score is: daily living activities 66 points, depression 0 points, tumor T stage 53 points, repair and reconstruction 70 points, glossoplasty 74 points, jaw resection 55 points, no nasopharyngeal airway after surgery 0 points, the total score is 318 points, and the risk of dysphagia after surgery is 99.9%.

[0069] The above-mentioned embodiments only express the specific implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the protection scope of the present application. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the technical solution concept of the present application, and these all belong to the protection scope of the present application.

Claims

1. A method for predicting the risk of postoperative dysphagia in patients with oral cancer based on LASSO regression. It is characterized in that The following steps are involved: Clinical data were collected, and a cohort study design was adopted. The swallowing function of postoperative patients was evaluated using the Mann Swallowing Scale for Oral Cancer (MASA-OC). According to the results of MASA-OC, the patients were divided into a dysphagia group and a non-dysphagia group. LASSO regression analysis was used to screen candidate variables, and LASSO regression analysis was used to analyze the potential risk factors for dysphagia after oral cancer surgery. Logistic regression was used to analyze multiple factors for independent risk predictors of dysphagia after oral cancer surgery; Through regression analysis, we obtained the independent risk predictors of dysphagia after oral cancer surgery, and constructed a risk prediction model for dysphagia after oral cancer surgery. The prediction results were presented in the form of a nomogram to obtain the risk probability of dysphagia after oral cancer surgery. To evaluate the effectiveness of risk prediction models.

2. According to claim 1, a method for predicting the risk of postoperative dysphagia in oral cancer patients based on LASSO regression, It is characterized in that The screening of candidate variables by LASSO regression analysis specifically includes the following steps: Install and run the R Studio software glmnet package to perform LASSO regression; The predictive factors affecting postoperative dysphagia after oral cancer surgery were input, and the penalty coefficient λ gradually decreased with the increase in the number of variables. Through 10-fold cross-validation, the predictive factors that can be selected for subsequent regression analysis were obtained. The predictive factors included: patient gender, daily living activities, depression, frailty, hemoglobin count, blood potassium level, blood sodium level, history of respiratory diseases, history of nervous system diseases, history of head and neck diseases, tumor location, maximum tumor diameter, tumor T stage, neck lymph node dissection, repair and reconstruction, glossoplasty, maxillomandibular resection, salivary gland palpation, nasopharyngeal ventilation tube retention, gastric tube retention, operation time, and intraoperative blood loss.

3. According to claim 2, a method for predicting the risk of postoperative dysphagia in oral cancer patients based on LASSO regression, It is characterized in that The potential risk factor analysis of dysphagia after oral cancer surgery using LASSO regression analysis specifically includes: Univariate analysis was performed with whether dysphagia occurred after surgery for oral cancer as the dependent variable and risk factors screened by LASSO regression as the independent variables to obtain the potential risk factors for dysphagia after surgery for oral cancer.

4. According to claim 3, a method for predicting the risk of postoperative dysphagia in oral cancer patients based on LASSO regression, It is characterized in that The potential risk factors include gender, ability to perform daily living activities, depression, tumor location, maximum tumor diameter, tumor T stage, neck lymph node dissection, repair and reconstruction, glossoplasty, jaw resection, retention of nasopharyngeal ventilation tube, retention of gastric tube, operation duration, and intraoperative blood loss.

5. According to claim 4, a method for predicting the risk of postoperative dysphagia in oral cancer patients based on LASSO regression, It is characterized in that The independent risk predictors of dysphagia after oral cancer surgery analyzed by logistic regression include: The occurrence of postoperative dysphagia in patients with oral cancer was taken as the dependent variable, and the risk factors with P < 0.05 in univariate analysis were taken as independent variables. Multivariate logistic regression analysis was performed by including them into the equation according to the stepwise backward method of maximum likelihood ratio estimation.

6. A method for predicting the risk of postoperative dysphagia in oral cancer patients based on LASSO regression according to claim 1 or 5, It is characterized in that The independent risk predictors were daily living activities, depression, tumor T stage, repair and reconstruction, glossoplasty, maxillomandibular resection, and retention of nasopharyngeal airway.

7. The method for predicting the risk of postoperative dysphagia in oral cancer patients based on LASSO regression according to claim 1, It is characterized in that The risk prediction model is constructed by using the rms function package of R Studio software to construct a prediction model for postoperative dysphagia in patients with oral cancer. For each variable determined to enter the logistic regression model, the corresponding Nomo score can be obtained through different variable values ​​of the vertical line. Finally, the sum of all variable scores is the total score, thereby obtaining the risk probability of postoperative dysphagia in patients with oral cancer.

8. The method for predicting the risk of postoperative dysphagia in oral cancer patients based on LASSO regression according to claim 1, It is characterized in that The evaluation of the effectiveness of the risk prediction model specifically includes the following steps: Discrimination of risk prediction model: By running the R language function package, the area under the ROC curve of the prediction model was 0.933, the sensitivity was 90.9%, and the specificity was 81.7%; Calibration of risk prediction model: The Hosmer-Lemeshow method was used to test the goodness of fit of the model. x 2 =5.391,P=0.715(>0.05)。 9. The method for predicting the risk of postoperative dysphagia in oral cancer patients based on LASSO regression according to claim 1, It is characterized in that The clinical data include general information, assessment of daily living activities, preoperative swallowing function assessment, consciousness status 24 hours after surgery, results of various test indicators, medical history, disease and treatment conditions all collected from the hospital medical record system, anxiety and depression status, and weakness.

10. The method for predicting the risk of postoperative dysphagia in oral cancer patients based on LASSO regression according to claim 9, It is characterized in that The general information includes age, gender, height, and weight, which are collected from nursing records; the anxiety and depression status is obtained by the Hospital Anxiety and Depression Scale (HADS); and the frailty status is obtained by the Fried frailty assessment method.

Images