Establishment of a LASSO-Logistic Regression-based Risk Prediction Model for Early Recurrence of Siewert Ⅱ/Ⅲ Adenocarcinoma of Esophagogastric Junction Post-Surgery

Objective To investigate the risk factors for early relapse after curative resection of Siewert type Ⅱ/Ⅲ adenocarcinoma of esophagogastric junction (AEG) and construct a visual predictive model.

Methods A retrospective analysis was conducted on the clinicopathological data of patients diagnosed with Siewert type Ⅱ/Ⅲ AEG who underwent curative resection at the Second Hospital of Lanzhou University from January 2016 to March 2021. The samples were randomly divided into a training group and a validation group in a 7∶3 ratio. The LASSO-Logistic regression method was used to select variables predictive of early recurrence of Siewert type Ⅱ/Ⅲ AEG and construct a predictive model for early recurrence. The model was validated through 1000 bootstrap resampling. Receiver operating characteristic (ROC) curves were drawn, and area under the curve (AUC), calibration curves, and decision curve analysis (DCA) were used to evaluate the model's stability.

Results According to the inclusion and exclusion criteria of this study, a total of 320 Siewert type Ⅱ/Ⅲ AEG patients were included, with 122 experiencing recurrence within two years. LASSO-Logistic regression analysis revealed AJCC staging, degree of differentiation, CA199, CEA, NLR, and tumor maximum diameter as independent predictive factors for early recurrence of Siewert type Ⅱ/Ⅲ AEG. A predictive model was constructed with these factors and depicted as a nomogram. For the training group, the AUC of the ROC curve was 0.836(95% CI: 0.785-0.887), with a sensitivity of 81.4% and a specificity of 85.6%;for the validation group, the AUC was 0.812(95% CI: 0.711-0.912), with a sensitivity of 80.6% and a specificity of 87.7%. Calibration curves for both the training and validation groups displayed curves close to the reference line, indicating high model stability. The DCA curve showed that the model provided a good net benefit with threshold probabilities between 0.05 and 0.75.

Conclusions A multivariate model developed using LASSO-Logistic regression could predict early relapse in patients with Siewert type Ⅱ/Ⅲ AEG, which may be instrumental in assessing patient prognoses and in guiding postoperative surveillance and management for patients with Siewert type Ⅱ/Ⅲ AEG.