Nomograms Involving HER2 for Predicting Lymph Node Metastasis in Early Gastric Cancer

Abstract

Objective: We aimed to establish a nomogram for predicting lymph node metastasis in early gastric cancer (EGC) involving human epidermal growth factor receptor 2 (HER2). Methods: We collected clinicopathological data of patients with EGC who underwent radical gastrectomy and D2 lymphadenectomy at Ruijin Hospital, Shanghai Jiao Tong University School of Medicine between January 2012 and August 2018. Univariate and multivariate logistic regression analysis were used to examine the relationship between lymph node metastasis and clinicopathological features. A nomogram was constructed based on a multivariate prediction model. Internal validation from the training set was performed using receiver operating characteristic (ROC) and calibration plots to evaluate discrimination and calibration, respectively. External validation from the validation set was utilized to examine the external validity of the prediction model using the ROC plot. A decision curve analysis was used to evaluate the benefit of the treatment. Results: Among 1,212 patients with EGC, 210 (17.32%) presented with lymph node metastasis. Multivariable analysis showed that age, tumor size, submucosal invasion, histological subtype, and HER2 positivity were independent risk factors for lymph node metastasis in EGC. The area under the ROC curve of the model was 0.760 (95% CI: 0.719-0.800) in the training set (n = 794) and 0.771 (95% CI: 0.714-0.828) in the validation set (n = 418). A predictive nomogram was constructed based on a multivariable prediction model. The decision curve showed that using the prediction model to guide treatment had a higher net benefit than using endoscopic submucosal dissection (ESD) absolute criteria over a range of threshold probabilities. Conclusion: A clinical prediction model and an effective nomogram with an integrated HER2 status were used to predict EGC lymph node metastasis with better accuracy and clinical performance.

Keywords: HER2; early gastric cancer; lymph node metastasis; nomogram; prediction model.

FIGURE 1

Flow diagram of patient enrollment and characteristics of patients in the training set and validation set.

FIGURE 2

Nomogram for predicting lymph node metastasis in EGC patients. EGC, early gastric cancer; pap, papillary adenocarcinoma; tub1, well-differentiated tubular adenocarcinoma; tub2, moderately differentiated tubular adenocarcinoma; sig, signet-ring cell carcinoma; muc, mucinous adenocarcinoma; por, poorly differentiated adenocarcinoma.

FIGURE 3

Assessment of the nomogram for predicting lymph node metastasis in the training set and validation set. (A) Calibration plot in the training set. After 2000 repetitions, the bootstrap-corrected calibration curve (solid line) lay close to the ideal reference line (dashed line), which demonstrated a perfect agreement between the predicted and actual outcomes (mean absolute error = 0.012); (B) ROC plot in the training set. The AUC of the ROC was 0.760 (95% CI, 0.719–0.800); (C) ROC plot in the validation set. The AUC of the ROC was 0.771 (95% CI, 0.714–0.828). ROC: receiver-operating characteristic; AUC: area under the ROC curve.

FIGURE 4

Clinical performance of the clinical model (nomogram) and ESD indications. Decision curve analysis on the clinical model (nomogram) (red line) and ESD absolute indications recommended by JGCA (solid line). The y-axis represents net benefits, calculated by subtracting the relative harms (false positives) from the benefits (true positives). The x-axis measures the threshold probability.