Cancer Medicine (Mar 2023)

Clinical models to predict lymph nodes metastasis and distant metastasis in newly diagnosed early esophageal cancer patients: A population‐based study

  • Hong Chen,
  • Junxian Wu,
  • Wanting Guo,
  • Lihang Yang,
  • Linbin Lu,
  • Yihong Lin,
  • Xuewen Wang,
  • Yan Zhang,
  • Xi Chen

DOI
https://doi.org/10.1002/cam4.5334
Journal volume & issue
Vol. 12, no. 5
pp. 5275 – 5292

Abstract

Read online

Abstract Background Patients with early esophageal cancer (EC) receive individualized therapy based on their lymph node metastasis (LNM) and distant metastasis (DM) status; however, deficiencies in current clinical staging techniques and the issue of cost‐effectiveness mean LNM and DM often go undetected preoperatively. We aimed to develop three clinical models to predict the likelihood of LNM, DM, and prognosis in patients with early EC. Method The Surveillance, Epidemiology, and End Results database was queried for T1 EC patients from 2004 to 2015. Multivariable logistic regression and Cox proportional hazards models were used to recognize the risk factors of LNM and DM, predict overall survival (OS), and develop relevant nomograms. Receiver operating characteristic (ROC)/concordance index and calibration curves were used to evaluate the discrimination and accuracy of the three nomograms. Decision curve analyses (DCAs), clinical impact curves, and subgroups based on model scores were used to determine clinical practicability. Results The area under the curve of the LNM and DM nomograms were 0.668 and 0.807, respectively. The corresponding C‐index of OS nomogram was 0.752. Calibration curves and DCA showed an effective predictive accuracy and clinical applicability. In patients with T1N0M0 EC, surgery alone (p < 0.01) proved a survival advantage. Chemotherapy and radiotherapy indicated a better prognosis in the subgroup analysis for T1 EC patients with LNM or DM. Conclusions We created three nomograms to predict the likelihood of LNM, DM, and OS probability in patients with early EC using a generalizable dataset. These useful visual tools could help clinical physicians deliver appropriate perioperative care.

Keywords