BMC Gastroenterology (Jul 2022)

A dynamic predictive nomogram of long-term survival in primary gastric lymphoma: a retrospective study

  • Jinru Yang,
  • Tao Liu,
  • Ying Zhu,
  • Fangyuan Zhang,
  • Menglan Zhai,
  • Dejun Zhang,
  • Lei Zhao,
  • Min Jin,
  • Zhenyu Lin,
  • Tao Zhang,
  • Liling Zhang,
  • Dandan Yu

DOI
https://doi.org/10.1186/s12876-022-02419-2
Journal volume & issue
Vol. 22, no. 1
pp. 1 – 12

Abstract

Read online

Abstract Background Primary gastric lymphoma (PGL) is the most common extranodal non-Hodgkin lymphoma (NHL). Due to the rarity of the disease, it is important to create a predictive model that provides treatment and prognosis for patients with PGL and physicians. Methods A total of 8898 and 127 patients diagnosed with PGL were obtained from the SEER database and from our Cancer Center as training and validation cohorts, respectively. Univariate and multivariate Cox proportional hazards models were used to investigate independent risk factors for the construction of predictive survival nomograms, and a web nomogram was developed for the dynamic prediction of survival of patients with PGL. The concordance index (C-index), calibration plot, and receiver operating characteristics (ROC) curve were used to evaluate and validate the nomogram models. Results There were 8898 PGL patients in the SEER cohort, most of whom were married men over the age of 60, 16.1% of the primary tumors were localized in the antrum and pylori of the stomach, which was similar to the composition of 127 patients in the Chinese cohort, making both groups comparable. The Nomogram of overall survival (OS) was compiled based on eight variables, including age at diagnosis, sex, race, marital status, histology, stage, radiotherapy and chemotherapy. Cancer-specific survival (CSS) nomogram was developed with eight variables, including age at diagnosis, sex, marital status, primary tumor site, histology, stage, radiotherapy and chemotherapy. The C-index of OS prediction nomogram was 0.948 (95% CI: 0.901–0.995) in the validation cohort, the calibration plots showed an optimal match and a high area below the ROC curve (AUC) was observed in both training and validation sets. Also, we established the first web-based PGL survival rate calculator ( https://yangjinru.shinyapps.io/DynNomapp/ ). Conclusion The web dynamic nomogram provided an insightful and applicable tool for evaluating PGL prognosis in OS and CSS, and can effectively guide individual treatment and monitoring.

Keywords