Method and system for constructing postoperative insulin resistance risk prediction model

Abstract

The invention provides a construction method and system of a postoperative insulin resistance risk prediction model, and belongs to the technical field of disease risk prediction. The method comprises the steps that dependent variable information and independent variable information of a patient are collected and assigned according to a preset rule; analyzing the patient data through single-factor logistic regression to obtain an independent variable of a preset target as a candidate factor and a statistical value; inputting the obtained candidate factors and statistical values into multi-factor regression analysis to obtain independent risk factors and statistical values thereof; analyzing all the independent risk factors of insulin resistance after the stomach cancer operation according to a preset analysis algorithm, and calculating a target P value; and forming a Nomoh map according to a comprehensive analysis result, and further constructing a prediction model. The risk prediction model constructed by the invention can effectively predict the postoperative insulin resistance of the gastric cancer gastric resection surgery patient, has an auxiliary effect on medical decision of clinicians, and has objective clinical value and economic benefit.

Description

Technical Field

[0001] The present invention belongs to the technical field of disease risk prediction, and particularly relates to a method and system for constructing a postoperative insulin resistance risk prediction model. Background Art

[0002] Gastric cancer (GC) is a common malignant tumor, and surgical operations (such as radical total or partial gastrectomy) are the main treatment methods for early gastric cancer. However, patients face various metabolic complications after surgery, and insulin resistance (IR), as its core pathological manifestation, has become a key issue affecting the quality of life of patients. Insulin resistance is manifested as a decrease in the responsiveness of peripheral tissues (such as the liver, skeletal muscle, and adipose tissue) to insulin, resulting in an imbalance in blood glucose homeostasis.

[0003] Constructing a prediction and evaluation model for IR after gastrectomy has important clinical significance: by integrating IR-related indicators and surgical parameters, an individualized postoperative monitoring plan can be formulated. For example, high-risk patients need to initiate blood glucose management in advance; optimizing the intervention timing, early use of metformin or lifestyle intervention can reduce the risk of cardiovascular events in IR patients; systematic metabolic management improves the prognosis and increases the postoperative survival rate.

[0004] It can be seen from this that constructing a prediction and evaluation model for IR after gastrectomy has important clinical significance, but currently, there is a lack of a postoperative insulin resistance risk prediction model with accurate prediction. Summary of the Invention

[0005] In view of this, the purpose of the present invention is to provide a method and system for constructing a postoperative insulin resistance risk prediction model, aiming to solve at least one technical problem in the background art.

[0006] The present invention is implemented as follows: The first aspect of the present invention provides a method for constructing a postoperative insulin resistance risk prediction model, which includes the following steps: Collect the dependent variable information and independent variable information of patients and assign values according to preset rules. The dependent variable is the postoperative HOMA-IR index, and the independent variables include age, gender, past medical history, albumin, surgical method, surgical category, preoperative HOMA-IR index, and preoperative insulin resistance index; Analyze the patient data through univariate logistics regression to obtain the independent variables of the preset target as candidate factors with statistical significance and statistical values; Input the candidate factors and statistical values of the univariate logistics regression analysis into the multivariate regression analysis to obtain independent risk factors and their statistical values; Analyze all independent risk factors of postoperative insulin resistance in gastric cancer according to the preset analysis algorithm to calculate the target P value; A nomogram is formed based on the results of comprehensive analysis, and then a prediction model is constructed.

[0007] Further, the step of assigning values to the dependent variable information according to a preset rule specifically includes: Obtain the postoperative HOMA-IR index and compare it with a preset standard value; Assign a value to the postoperative HOMA-IR index according to the comparison result. When the postoperative HOMA-IR index is less than the preset standard value, it is assigned a first threshold; when the postoperative HOMA-IR index is greater than or equal to the preset standard value, it is assigned a second threshold.

[0008] Further, the step of assigning values to the independent variables according to a preset rule specifically includes: Obtain the data of age, albumin, preoperative HOMA-IR index, and preoperative insulin resistance index respectively, and compare them with the preset standard values; Assign values according to the comparison results. When the data of the independent variable is less than the corresponding preset standard value, it is assigned a first threshold; when the data of the independent variable is greater than or equal to the corresponding preset standard value, it is assigned a second threshold; Obtain the patient's past medical history of hypertension and diabetes respectively. If not ill, it is assigned a first threshold; if ill, it is assigned a second threshold; Obtain the patient's gender. If male, it is assigned a first threshold; if female, it is assigned a second threshold; Obtain the patient's surgical method. The distal gastrectomy group is assigned a first threshold, the proximal gastrectomy group is assigned a second threshold, and the total gastrectomy group is assigned a third threshold; Obtain the patient's surgical category. The laparoscopic surgery is assigned a first threshold, and the open surgery is assigned a second threshold.

[0009] Further, the calculation formula of the preoperative insulin resistance index CFRESH is as follows: CFRESH ; In the formula, FCP is the fasting C-peptide level, with the unit of ng / mL; FINS is the fasting insulin level, with the unit of μIU / mL.

[0010] Further, the calculation formula of the target P value is as follows: ; In the formula, P is the significant influencing factor; is the initial coefficient, represents the influence coefficient of the independent risk factor independent variable on the risk of postoperative insulin resistance; represents the specific assignment of the independent risk factor independent variable.

[0011] Further, the prediction model adopts a nomogram model; There are 9 scales in this nomogram: The first scale represents the scale corresponding to the scores on the second to seventh scales; the scale value of the first scale is 0 - 100, 0 is at the leftmost end, 100 is at the rightmost end, and the scale of the scale is equally divided; The second scale represents the age of the patient; The third scale represents the surgical name of the patient; The fourth scale represents the surgical method of the patient; The fifth scale represents the preoperative HMOA - IR of the patient; The sixth scale represents the preoperative CFRESH of the patient; The seventh scale represents the total risk score, and the total risk score = the risk score corresponding to age + the risk score corresponding to the surgical name + the risk score corresponding to the surgical method + the risk score corresponding to the preoperative HMOA - IR + the risk score corresponding to the preoperative CFRESH, that is, the sum of the scores corresponding to the scales on the second to sixth scales; The ninth scale represents the diagnostic probability of insulin resistance after the operation of the patient, and it is predicted that the postoperative HOMA - IR of the patient = the linear predicted value corresponding to the score on the seventh scale on the eighth scale.

[0012] Further, the construction method further includes: Judging the sensitivity and specificity of the prediction model through the ROC curve to distinguish patients with and without events; Evaluating the performance of the prediction model through the calibration curve; Evaluating the clinical decision - making ability of the prediction model through the DCA curve.

[0013] The second aspect of the present invention provides a construction system for a postoperative insulin resistance risk prediction model, which is used to implement the above - mentioned construction method of the postoperative insulin resistance risk prediction model. The system includes: An information acquisition module, which is used to collect the dependent variable information and independent variable information of the patient and assign values according to preset rules. The dependent variable is the postoperative HOMA - IR index, and the independent variables include age, gender, past medical history, albumin, surgical method, surgical category, preoperative HOMA - IR index, and preoperative insulin resistance index; A univariate analysis module, which is used to analyze the patient data through univariate logistics regression to obtain the independent variables of the preset target as candidate factors with statistical significance and statistical values; A multivariate analysis module, which is used to input the candidate factors and statistical values of the univariate logistics regression analysis into the multivariate regression analysis to obtain independent risk factors and their statistical values; A target calculation module, configured to analyze all independent risk factors of insulin resistance after gastric cancer surgery according to a preset analysis algorithm, and calculate a target P value; A target model construction module, configured to form a nomogram based on the results of comprehensive analysis, and further construct a prediction model.

[0014] Further, the system further includes a model evaluation module, configured to judge the sensitivity and specificity of the prediction model through an ROC curve to distinguish patients with and without events; evaluate the performance of the prediction model through a calibration curve; evaluate the clinical decision-making ability of the prediction model through a DCA curve.

[0015] Compared with the prior art, the present invention proposes a preoperative insulin resistance index CFRESH as a new candidate factor for the prediction model based on conventional independent variable factors (such as age, gender, past medical history, albumin, surgical method, surgical category, preoperative HOMA-IR index), and uses single-factor logistics regression analysis and multi-factor regression analysis to obtain characteristic variables for accurately evaluating the risk of postoperative insulin resistance, and constructs a risk prediction model for postoperative insulin resistance of patients undergoing gastrectomy and a construction method thereof, which can effectively predict the postoperative insulin resistance of patients undergoing gastric cancer gastrectomy, has a certain auxiliary effect on the medical decision-making of clinicians, and has objective clinical value and economic benefits. Description of the Drawings

[0016] Figure 1 It is a flowchart of the construction method of the postoperative insulin resistance risk prediction model in Embodiment 1; Figure 2 It is a structural block diagram of the construction system of the postoperative insulin resistance risk prediction model in Embodiment 2; Figure 3 It is a list diagram formed by the prediction model constructed by the present invention; Figure 4 It is a receiver operating characteristic curve of the training set in Embodiment 1; Figure 5 It is an ROC curve of the training set and the control example in Embodiment 1; Figure 6 It is a receiver operating characteristic curve of the validation set in Embodiment 1; Figure 7 It is an ROC curve of the validation set and the control example in Embodiment 1; Figure 8 It is a comparison diagram of the receiver operating characteristic curves of the training set and the validation set in Embodiment 1; Figure 9 It is a calibration curve of the training set in Embodiment 1; Figure 10 It is a calibration curve of the validation set in Embodiment 1; Figure 11Decision curve for the training set of Example 1; Figure 12 Decision curve for the validation set of Example 1. Detailed implementation manners

[0017] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0018] Example 1 A method for constructing a postoperative insulin resistance risk prediction model, the flowchart of which is as Figure 1 shown, and includes the following steps: S1. Collect the dependent variable information and independent variable information of patients and assign values according to preset rules. The dependent variable is the postoperative HOMA-IR index, and the independent variables include age, gender, past medical history, albumin, surgical method, surgical category, preoperative HOMA-IR index, and preoperative insulin resistance index; S11, data acquisition; Retrieve gastric cancer (GC) patients who were treated in the Second Affiliated Hospital of Nanchang University from May 2020 to April 2024 from the electronic medical record system. A total of 366 eligible patients were included as the research population; the research population was divided into a training set (n = 256) and a validation set (n = 110) according to a ratio of approximately 7:3.

[0019] The inclusion criteria are as follows: (1) Patients diagnosed with GC by preoperative imaging [computed tomography (CT or enhanced CT)], gastroscopy, and pathology; (2) Aged between 20 and 75 years old, without metabolic diseases such as intestinal obstruction; (3) Without severe heart, lung, liver, and kidney dysfunction; (4) No history of radiotherapy or chemotherapy in the past; (5) Patients with complete medical records; (6) GC patients underwent radical gastrectomy; (7) Obtained voluntary informed consent.

[0020] The exclusion criteria are: (1) Detection of distant metastasis or inability to perform gastrectomy; (2) History of cancer or other systemic tumors in the past; (3) Patients with severe heart, lung, liver, and kidney dysfunction; (4) Past radiotherapy or chemotherapy; (5) Use of antibiotics or probiotics within the past 2 weeks; (6) Incomplete medical records.

[0021] S12, assign values to the independent variable data; specifically include: (1) If the age < 60 years old, assign a value of 0; if the age ≥ 60 years old, assign a value of 1; (2) If the gender is male, assign a value of 0; if the gender is female, assign a value of 1; (3) If not suffering from hypertension, assign a value of 0; if suffering from hypertension, assign a value of 1; (4) If not suffering from diabetes, it is assigned a value of 0; if suffering from diabetes, it is assigned a value of 1; (5) If albumin < 40 g / L, it is assigned a value of 0; if albumin ≥ 40 g / L, it is assigned a value of 1; (6) If the surgical method is the distal gastrectomy group, it is assigned a value of 0; if the surgical method is the proximal gastrectomy group, it is assigned a value of 1; if the surgical method is the total gastrectomy group, it is assigned a value of 2; (7) If the surgical category is open surgery, it is assigned a value of 1; if the surgical category is laparoscopic surgery, it is assigned a value of 0; (8) If the preoperative HOMA-IR index < 1.085, it is assigned a value of 0; if preoperative HOMA-IR ≥ 1.085, it is assigned a value of 1; (9) If the preoperative insulin resistance index CFRESH < 0.095, it is assigned a value of 0; if CFRESH ≥ 0.095, it is assigned a value of 1.

[0022] The basic information of the training set (n = 256) and the validation set (n = 110) was compared, as shown in Table 1.

[0023] Table 1

[0024] In Table 1, χ² is the symbol for the chi-square test.

[0025] S13. Among the dependent variables, a postoperative HOMA-IR index higher than 2.69 is considered insulin resistance and reaches the final outcome; other cases are considered not to reach the final outcome; The outcome variable of patients who reach the final outcome is assigned a value of 1; the outcome variable of patients who do not reach the final outcome is assigned a value of 0.

[0026] HOMA-IR = fasting insulin level (μIU / mL) × fasting blood glucose level (mmol / L) / 22.5.

[0027] S2. By univariate logistic regression analysis of patient data, the independent variables with P < 0.001 are used as candidate factors and statistical values with statistical significance; First, using the characteristics of newly diagnosed postoperative insulin resistance, through univariate logistic regression analysis, as shown in Table 2 below, the independent variables with P < 0.001 in the training set are selected as candidate factors and statistical values with statistical significance, specifically including: Age (< 60 years old: OR = 0.23, 95%CI = 0.14~0.39, P < 0.001; < 60 years old is the reference object); Surgical name (open surgery: OR = 3.55, 95%CI = 2.11~5.95, P < 0.001; open surgery is the reference object); Surgical method (distal gastrectomy group: OR = 0.22, 95%CI = 0.12 - 0.39, P < 0.001; distal gastric cancer resection as the reference object); Preoperative HOMA-IR (<1.085: OR = 0.08, 95%CI = 0.04 - 0.14, P < 0.001; <1.085 as the reference object); Preoperative CFRESH (<0.095: OR = 0.08, 95%CI = 0.04 - 0.14, P < 0.001; <0.095 as the reference object).

[0028] S3. Input the candidate factors and statistical values of the univariate logistics regression analysis into the multivariate regression analysis to obtain the independent risk factors and their statistical values; Input the results of the univariate regression analysis in step S2 into the multivariate regression analysis for further analysis. Analyze the independent risk factors for high intensity of inflammatory response after gastric cancer surgery. As shown in Table 2 below, from the data in Table 2, the independent risk factors and related statistical values include: Age (<60 years old: OR = 0.33, 95%CI = 0.17 - 0.66, P = 0.002; <60 years old as the reference object); Surgical name (open surgery: OR = 3.85, 95%CI = 1.84 - 8.05, P < 0.001; open surgery as the reference object); Surgical method (distal gastrectomy group: OR = 0.30, 95%CI = 0.14 - 0.66, P = 0.003; distal gastric cancer resection as the reference object); HOMA-IR (<1.085: OR = 0.26, 95%CI = 0.11 - 0.60, P = 0.002; <1.085 as the reference object); Preoperative CFRESH (<0.095: OR = 0.20, 95%CI = 0.09 - 0.46, P < 0.001; <0.095 as the reference object).

[0029] In univariate regression, β reflects the impact of this variable alone on the result. In multivariate regression, β reflects the independent impact of this variable after controlling other variables. S.E represents the uncertainty (fluctuation range) of measuring the β estimate value, reflecting the sampling error. The smaller S.E is, the more accurate the estimate of β is, and the more stable the result is. P represents the significant influencing factor, OR represents the risk ratio, and 95%CI represents the 95% confidence interval.

[0030] Table 2

[0031] S4. Analyze all independent risk factors for insulin resistance after gastric cancer surgery according to the preset analysis algorithm and calculate the target P value; All independent risk factors for insulin resistance after gastric cancer surgery were comprehensively analyzed, and the calculation formula for multivariate regression was summarized as follows: ; In the formula, P is the significant influencing factor; is the initial coefficient, represents the coefficient of influence of independent risk factor variables on the risk of postoperative insulin resistance; Indicates the specific value of the independent risk factor independent variable. β indicates the intensity and direction of the influence of the corresponding independent variable (such as age, surgical method, etc.) on the risk of postoperative insulin resistance; if β is positive, the risk of insulin resistance increases when the variable value increases; if β is negative, the variable will reduce the risk.

[0032] In this embodiment, the independent risk factors screened out include age, surgery name, surgical method, preoperative HOMA-IR, preoperative CFRESH and other five independent variables. ; In the formula, is -1.76; is 1.11; is 1.35; is -1.21; is 1.35; is 1.62; Assign a value to age; Assign a value to the surgery name; Assign values ​​to surgical methods; Assign a value to the preoperative HOMA-IR; Assign a value to the preoperative CFRESH.

[0033] S5. Based on the results of comprehensive analysis, a nomogram is formed, such as Figure 3 As shown, a prediction model is constructed.

[0034] Figure 3 The nomogram is in the form of a nomogram, which contains 9 scales: The first scale represents the scale corresponding to the scales on the second to seventh scales; the scale values ​​of the first scale are 0 to 100, with 0 at the far left and 100 at the far right, and the scales of the scale are equally divided; The second scale represents the patient's age; The third scale indicates the patient's surgery name; The fourth scale indicates the surgical approach of the patient; The fifth scale represents the patient’s preoperative HMOA-IR; The 6th scale represents the patient's preoperative CFRESH; The 7th scale represents the total risk score, and the total risk score = the risk score corresponding to age + the risk score corresponding to the surgical name + the risk score corresponding to the surgical method + the risk score corresponding to preoperative HMOA-IR + the risk score corresponding to preoperative CFRESH, that is, the sum of the scores corresponding to the scales on the 2nd to 6th scales; The 9th scale represents the diagnostic probability of the patient's postoperative insulin resistance, and the patient's postoperative HOMA-IR = the linear predicted value corresponding to the score on the 7th scale on the 8th scale.

[0035] S6. Model evaluation S61. Determine the sensitivity and specificity of the model through the ROC curve to distinguish patients with and without events; Use, such as Figure 4 , Figure 6 , Figure 8 shown, the area under the receiver operating characteristic curve (AUC) in the training set and the validation set is quantified, and the 95% CI of each AUC is calculated. Generally, it is considered that when the AUC is above 0.9, there is high accuracy; when the AUC is between 0.7 and 0.9, there is certain accuracy; when the AUC is between 0.5 and 0.7, there is low accuracy; when the AUC = 0.5, it means that the diagnostic method is completely ineffective and has no diagnostic value.

[0036] Taking a single index (age, surgical name, surgical method, preoperative HOMA-IR, preoperative CFRESH index) as the independent variable for the prediction of the control example, calculate the area under the ROC curve, P value, and 95% CI of the corresponding training set and validation set, and compare them with this embodiment. The ROC curve of the comparison between the training set and the control example is as Figure 5 shown, and the comparison of various prediction data is shown in Table 3; the ROC curve of the comparison between the validation set and the control example is as Figure 7 shown, and the comparison of various prediction data is shown in Table 4.

[0037] Table 3

[0038] Table 4

[0039] From the data in Table 3 and Table 4 combined with Figure 5 and Figure 7 it can be seen that the accuracy of the prediction model in Example 1 of the present invention is much higher than that of the control example with a single index as the independent variable for prediction.

[0040] S62. Further evaluate the performance of the model through the calibration curve; First, use the constructed nomogram to predict the probability of each research subject, and arrange them in a queue from low to high. Group them according to the quintile method, dividing the queue into five groups. Then, calculate the corresponding predicted occurrence probability for each research subject in each group, and sum up the predicted occurrence probabilities of all individuals in each group and take the mean. The abscissa is the predicted occurrence probability, and the ordinate is the actual occurrence probability (the actual occurrence probability of all research subjects in this group). The ideal calibration curve represents the calibration curve under the ideal state, that is, the probability predicted by the constructed model and the actual occurrence probability. The performance (NONparametric) curve is the calibration curve drawn by performing loess fitting on all the existing data, and the bias-corrected (logistic-calibration) curve is obtained by using all the existing data, taking 1000 samples within a certain range to calculate the fitting mean error, and then subtracting the fitting error from the NONparametric data. The calibration curves of the training set and the validation set are respectively as Figure 9 and Figure 10 shown. The results show that the prediction ability of the model in both the training set ( Figure 9 ) and the validation set ( Figure 10 ) has good consistency with the actual observation results.

[0041] S63. Evaluate the clinical decision-making ability of this model by using the DCA curve; Specifically: Use the R language rmda software package to draw the DCA curve to evaluate the clinical decision-making ability.

[0042] The abscissa of the DCA curve is the threshold probability. When various evaluation methods reach a certain value, the risk probability of postoperative insulin resistance in patients is denoted as Pi; when Pi reaches a certain threshold (denoted as Pt), it is defined as positive, and a certain intervention measure (such as a treatment measure) is taken. Then, the treatment measure naturally changes the balance between the benefits of anti-insulin resistance effect and the side effects of anti-insulin resistance drugs. The ordinate is the net benefit rate (NetBenefit, NB) after subtracting the disadvantages from the benefits. NB = A × P - B × L. A is the true positive proportion; P in the formula is the benefit value of applying the intervention to true positive patients. B is the false positive proportion; L is the loss value of applying the intervention to false positive patients. The decision curves of the training set and the validation set are respectively as Figure 11 and Figure 12As shown, the results indicate that the DCA curve values of the nomogram model in the training set and the validation set are in the range of 0.1 - 1.0, above the lines of no predictor and all predictors. The results of this model are good. The line of all predictor values means that all samples are positive and all receive interventions, and the net benefit curve is a curve with a negative slope, indicating the net benefit of this batch of samples in this model at different thresholds. The line of no predictor values means that all samples are negative (Pi < Pt) and there is no net benefit from interventions.

[0043] Example 2 As Figure 2 shown, a system for constructing a postoperative insulin resistance risk prediction model, which is used to implement the method for constructing a postoperative insulin resistance risk prediction model in Example 1. The system includes: An information acquisition module 100, which is used to collect patient dependent variable information and independent variable information and assign values according to preset rules. The dependent variable is the postoperative HOMA-IR index, and the independent variables include age, gender, past medical history, albumin, surgical method, surgical category, preoperative HOMA-IR index, and preoperative insulin resistance index; A univariate analysis module 200, which is used to analyze patient data through univariate logistics regression to obtain independent variables of a preset target as candidate factors with statistical significance and statistical values; A multivariate analysis module 300, which is used to input the candidate factors and statistical values of the univariate logistics regression analysis into a multivariate regression analysis to obtain independent risk factors and their statistical values; A target calculation module 400, which is used to analyze all independent risk factors of postoperative insulin resistance in gastric cancer according to a preset analysis algorithm and calculate the target P value; A target model construction module 500, which is used to form a nomogram based on the results of comprehensive analysis and then construct a prediction model; A model evaluation module 600, which is used to judge the sensitivity and specificity of the prediction model through an ROC curve to distinguish patients with and without events; evaluate the performance of the prediction model through a calibration curve; and evaluate the clinical decision-making ability of the prediction model through a DCA curve.

[0044] The above embodiments only represent several implementation manners of the present invention. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the patent of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several deformations and improvements can still be made, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention should be subject to the appended claims.

Claims

1. A method for constructing a postoperative insulin resistance risk prediction model, characterized in that, The construction method includes the following steps: Collect the dependent variable information and independent variable information of the patient and assign values according to preset rules. The dependent variable is the postoperative HOMA-IR index, and the independent variables include age, gender, past medical history, albumin, surgical method, surgical category, preoperative HOMA-IR index, and preoperative insulin resistance index; Analyze the patient data through univariate logistic regression to obtain the independent variables of the preset target as candidate factors with statistical significance and statistical values; Input the candidate factors and statistical values of the univariate logistic regression analysis into the multivariate regression analysis to obtain the independent risk factors and their statistical values; Analyze all the independent risk factors of insulin resistance after gastric cancer surgery according to the preset analysis algorithm to calculate the target P value; Form a nomogram according to the results of the comprehensive analysis, and then construct a prediction model.

2. The method for constructing a postoperative insulin resistance risk prediction model according to claim 1, wherein The step of assigning values to the dependent variable information according to the preset rules specifically includes: Obtain the postoperative HOMA-IR index and compare it with the preset standard value; Assign a value to the postoperative HOMA-IR index according to the comparison result. When the postoperative HOMA-IR index is less than the preset standard value, it is assigned the first threshold; when the postoperative HOMA-IR index is greater than or equal to the preset standard value, it is assigned the second threshold.

3. The method for constructing a postoperative insulin resistance risk prediction model according to claim 1 or 2, characterized in that, The step of assigning values to the independent variables according to the preset rules specifically includes: Respectively obtain the data of age, albumin, preoperative HOMA-IR index, and preoperative insulin resistance index, and compare them with the preset standard values; Assign values according to the comparison results. When the data of the independent variable is less than the corresponding preset standard value, it is assigned the first threshold; when the data of the independent variable is greater than or equal to the corresponding preset standard value, it is assigned the second threshold; Respectively obtain the patient's past medical history of hypertension and diabetes. If not ill, it is assigned the first threshold; if ill, it is assigned the second threshold; Obtain the patient's gender. If male, it is assigned the first threshold; if female, it is assigned the second threshold; Obtain the patient's surgical method. The distal gastrectomy group is assigned the first threshold, the proximal gastrectomy group is assigned the second threshold, and the total gastrectomy group is assigned the third threshold; Obtain the patient's surgical category. Laparoscopic surgery is assigned the first threshold, and open surgery is assigned the second threshold.

4. The method for constructing a postoperative insulin resistance risk prediction model according to claim 1, wherein, The calculation formula of the preoperative insulin resistance index CFRESH is as follows: CFRESH ; In the formula, FCP is the fasting C-peptide level, with the unit of ng / mL; FINS is the fasting insulin level, with the unit of μIU / mL.

5. The method for constructing a postoperative insulin resistance risk prediction model according to claim 1, characterized in that, The calculation formula of the target P value is as follows: ； Wherein, P is a significant influencing factor; is the initial coefficient, indicating the influence coefficient of the independent risk factor independent variable on the risk of postoperative insulin resistance; indicating the specific assignment of the independent risk factor independent variable.

6. The method for constructing a postoperative insulin resistance risk prediction model according to claim 1, characterized in that The prediction model adopts a nomogram model; There are 9 scales in this nomogram: The first scale represents the scale corresponding to the scores on the second to seventh scales; the scale value of the first scale is 0 to 100, 0 is at the leftmost end, 100 is at the rightmost end, and the scale of the scale is equally divided; The second scale represents the patient's age; The third scale represents the patient's surgical name; The fourth scale represents the patient's surgical method; The fifth scale represents the patient's preoperative HMOA-IR; The sixth scale represents the patient's preoperative CFRESH; The 7th scale represents the total risk score, and the total risk score = the risk score corresponding to age + the risk score corresponding to the surgical procedure name + the risk score corresponding to the surgical method + the risk score corresponding to preoperative HMOA-IR + the risk score corresponding to preoperative CFRESH, that is, the sum of the scores corresponding to the scales on the 2nd to 6th scales; The 9th scale represents the diagnostic probability of postoperative insulin resistance in patients, and it is predicted that postoperative HOMA-IR in patients = the linear predicted value corresponding to the score on the 7th scale on the 8th scale.

7. The method for constructing a postoperative insulin resistance risk prediction model according to claim 1, wherein The construction method further includes: Judging the sensitivity and specificity of the prediction model through the ROC curve to distinguish patients with and without events; Evaluating the performance of the prediction model through the calibration curve; Evaluating the clinical decision-making ability of the prediction model through the DCA curve.

8. A system for constructing a risk prediction model for postoperative insulin resistance, characterized in that, A system for implementing the construction method of the postoperative insulin resistance risk prediction model according to any one of claims 1 to 7, the system includes: An information acquisition module, configured to collect patient dependent variable information and independent variable information and assign values according to a preset rule, where the dependent variable is the postoperative HOMA-IR index, and the independent variables include age, gender, past medical history, albumin, surgical method, surgical category, preoperative HOMA-IR index, and preoperative insulin resistance index; A univariate analysis module, configured to analyze patient data through univariate logistics regression to obtain independent variables of a preset target as candidate factors and statistical values with statistical significance; A multivariate analysis module, configured to input the candidate factors and statistical values of the univariate logistics regression analysis into a multivariate regression analysis to obtain independent risk factors and their statistical values; A target calculation module, configured to analyze all independent risk factors of postoperative insulin resistance in gastric cancer according to a preset analysis algorithm and calculate a target P value; A target model construction module, configured to form a nomogram based on the results of comprehensive analysis, and further construct a prediction model.

9. The construction system of the postoperative insulin resistance risk prediction model according to claim 8, characterized in that, The system further includes a model evaluation module, configured to judge the sensitivity and specificity of the prediction model through the ROC curve to distinguish patients with and without events; evaluate the performance of the prediction model through the calibration curve; evaluate the clinical decision-making ability of the prediction model through the DCA curve.

Images