Construction and external validation of a nomogram that predicts lymph node metastasis in early gastric cancer patients using preoperative parameters
Abstract
Objective: To create a nomogram to predict the incidence of lymph node metastasis (LNM) in early gastric cancer (EGC) patients and to externally validate the nomogram.
Methods: To construct the nomogram, we retrospectively analyzed a primary cohort of 272 EGC patients. Univariate analysis and a binary logistic regression were performed. A nomogram predicting the incidence of LNM in EGC patients was created. The discrimination ability of the nomogram was measured using the concordance index (c-index), and the nomogram was also calibrated. Then, another prospective cohort of 81 patients was analyzed to validate the nomogram.
Results: In the primary cohort, LNM was pathologically confirmed in 37 (13.6%) patients. In multivariate analysis, the presence of an ulcer, the maximum lesion diameter observed via gastroscopy, the thickness of the lesion observed via endoscopic ultrasonography, and the presence of enlarged lymph nodes on computed tomography (CT) were independent risk factors for LNM. A nomogram was then created based on the regression model with the c-index of 0.905, and the calibration curve of the nomogram fell approximately on the ideal 45- degree line. The cut-off score of the nomogram was 110, and the sensitivity, specificity, positive predictive and negative predictive values of the nomogram in the primary cohort were 81.1%, 86.0%, 47.6% and 96.7%, respectively, and in the prospective validation cohort were 75.0%, 91.0%, 60.0% and 95.5%, respectively. The calibration curve of the external validation cohort was almost on the 45-degree line.
Conclusions: We developed an effective nomogram predicting the incidence of LNM for EGC patients.
Keywords: Early gastric cancer; lymph node metastasis; nomogram; validation
Methods: To construct the nomogram, we retrospectively analyzed a primary cohort of 272 EGC patients. Univariate analysis and a binary logistic regression were performed. A nomogram predicting the incidence of LNM in EGC patients was created. The discrimination ability of the nomogram was measured using the concordance index (c-index), and the nomogram was also calibrated. Then, another prospective cohort of 81 patients was analyzed to validate the nomogram.
Results: In the primary cohort, LNM was pathologically confirmed in 37 (13.6%) patients. In multivariate analysis, the presence of an ulcer, the maximum lesion diameter observed via gastroscopy, the thickness of the lesion observed via endoscopic ultrasonography, and the presence of enlarged lymph nodes on computed tomography (CT) were independent risk factors for LNM. A nomogram was then created based on the regression model with the c-index of 0.905, and the calibration curve of the nomogram fell approximately on the ideal 45- degree line. The cut-off score of the nomogram was 110, and the sensitivity, specificity, positive predictive and negative predictive values of the nomogram in the primary cohort were 81.1%, 86.0%, 47.6% and 96.7%, respectively, and in the prospective validation cohort were 75.0%, 91.0%, 60.0% and 95.5%, respectively. The calibration curve of the external validation cohort was almost on the 45-degree line.
Conclusions: We developed an effective nomogram predicting the incidence of LNM for EGC patients.
Keywords: Early gastric cancer; lymph node metastasis; nomogram; validation