Construction and validation of a clinical risk model based on machine learning for screening characteristic factors of lymphovascular space invasion in endometrial cancer

Comparison of affected person clinical dataSignificant statistical variations had been noticed in myometrial infiltration depth, tumor diameter, pathological kind, histological grade, and cervical interstitial involvement between the LVSI and non-LVSI teams when analyzing affected person information (P < 0.05, Table 1).Table 1 Analysis of the distinction in clinical information between LVSI and non-LVSI teams in the coaching group.Comparison of affected person laboratory informationThe laboratory indicators of sufferers with LVSI had been in contrast with these with out the coaching group. The outcomes confirmed vital statistical variations in lymphocyte depend (LYM), monocyte depend (MONO), albumin (ALB), fibrinogen (FIB), monocyte-to-lymphocyte ratio (MLR), and platelet-to-lymphocyte ratio (PLR) between the 2 teams (P < 0.001, Table 2).Table 2 Comparison of laboratory indicators between sufferers with and with out LVSI in the coaching group.Multivariable logistic regression evaluationWe assigned values to the indices with variations in the univariate evaluation (Table 3), the place the metric information had been dichotomized utilizing Cut-off values as cut-off points16, and then discovered by multivariable logistic regression evaluation that myometrial infiltration depth (P < 0.001, OR = 17.876, 95% CI 5.546–57.619), cervical interstitial involvement (P = 0.003, OR = 8.028, 95% CI 2.037–31.639), LYM (P = 0.001, OR = 0.159, 95% CI 0.055–0.465), MONO (P = 0.001, OR = 0.149, 95% CI 0.048–0.460), ALB (P = 0.002, OR = 0.175, 95% CI 0.059–0.524), and FIB (P < 0.001, OR = 11.853, 95% CI 4.285–32.788) had been unbiased risk factors for LVSI in EC (Table 4).Table 3 Assignment desk.Table 4 Multivariable logistic regression.LASSO regression analysisIn the current examine, we used LASSO regression to display the characteristic factors resulting in LVSI in EC. LASSO regression recognized 22 characteristic factors when = λmin (0.0046564) and 19 characteristic factors when = λ0.1se (0.010757) (Fig. 1A). Considering the generalization efficiency of the model, 19 characteristic factors had been recognized when = λ0.1SE was chosen to assemble the model. A complete of 19 characteristic factors had been screened out, together with age, menarche, menopause, gravidity, parity, historical past of diabetes, myometrial infiltration depth, tumor diameter, pathological kind, histological grade, cervical interstitial involvement, adnexal metastasis, FIGO staging, NEUT, LYM, MONO, ALB, FIB, and PLR (Fig. 1B).Figure 1LASSO regression screening for characteristic factors of lymphovascular space invasion in endometrial cancer. (A) LASSO regression for the screening of characteristic factors resulting in lymphovascular space invasion in endometrial cancer; (B) 19 characteristic factors when = λ.1se.Risk model buildingWe constructed two risk fashions by logistic and LASSO regression. The logistic regression fashions had been constructed utilizing the β coefficient (Table 4), and the LASSO regression additionally used this coefficient building model (Table 5). By comparability, we discovered that in each fashions, the risk rating of sufferers in the LVSI group was considerably increased in contrast with that of sufferers in the non-LVSI group in the coaching group, with a statistical distinction (P < 0.001, Fig. 2A). Through Delong take a look at evaluation, the realm beneath the curve (AUC) of the risk model, which was constructed based on logistic regression was considerably decrease in contrast with that of the LASSO-constructed model (P < 0.001, Fig. 2B).Table 5 LASSO characteristic variables.Figure 2Comparison of risk scores and predictive efficiency between LVSI and non-LVSI patients in the training group. (A) Comparison of the patient scores in the training group calculated by logistic regression and LASSO regression. (B) ROC curve analysis of the AUCs of logistic regression and LASSO regression risk scores in predicting lymphovascular space invasion in endometrial cancer. Note LVSI: lymphovascular space invasion; ROC: receiver operating characteristic; AUCs: areas under the curves; C represents the non-LVSI group and P represents the LVSI group.Modeling validationA comparison of patient baseline data between the training and validation groups revealed no statistical difference between them (P > 0.05, Table 6). We calculated the logistic and LASSO risk scores of sufferers in the validation group. LVSI sufferers had increased logistic and LASSO risk scores in contrast with the non-LVSI sufferers in the validation group, with vital statistical variations (P < 0.001, Fig. 3A). Subsequently, we discovered by the Delong take a look at that the AUC of the logistic regression risk model was considerably decrease in contrast with that of the LASSO risk model (P < 0.001, Fig. 3B, Tables 7, 8).Table 6 Comparison of the clinical information between coaching and validation teams.Figure 3Comparison of risk scores and predictive effectivity between LVSI and non-LVSI sufferers in the validation group. (A) Comparison of the risk rating of sufferers in the validation group calculated by logistic regression and LASSO regression; (B) ROC curve evaluation of the AUCs of logistic regression and LASSO regression risk scores in predicting lymphovascular space invasion in endometrial cancer. Note LVSI: lymphovascular space invasion; ROC: receiver working characteristic; AUCs: areas beneath the curves; C represents the non-LVSI group and P represents the LVSI group.Table 7 ROC Curve parameters.Table 8 Delong take a look at parameters.
https://www.nature.com/articles/s41598-024-63436-7