Cancer Medicine (Oct 2023)

Development and validation of a network calculator model for safety and efficacy after pancreaticoduodenectomy in the elderly patients with pancreatic head cancer

  • Ming Cai,
  • Tong Guo,
  • Zixiang Chen,
  • Wuhan Li,
  • Tian Pu,
  • Zhiwei Zhang,
  • Xiaorui Huang,
  • Xinyi Guo,
  • Yahong Yu

DOI
https://doi.org/10.1002/cam4.6613
Journal volume & issue
Vol. 12, no. 19
pp. 19673 – 19689

Abstract

Read online

Abstract Background Benefiting from increased life expectancy and improved perioperative management, more elderly patients with pancreatic head cancer (PHC) underwent pancreaticoduodenectomy (PD). However, individualized predictive models for the safety and efficacy of PD is still lacking. this study aimed to developed three safety‐ and efficacy‐related risk calculators for elderly (> = 65 years) PHC patients. Methods This study was designed with two research cohorts, namely, the training cohort and the validation cohort, and comprises four general steps: (1) Risk factors were analyzed for the incidence of postoperative complications, cancer‐specific survival (CSS), and overall survival (OS) in the training cohort (N = 271) using logistic and Cox‐regression analysis. (2) Nomograms were then plotted based on the above results. (3) The accuracy of the developed nomogram models was then verified with the validation cohort (N = 134) data using consistency index (C‐index) and calibration curves. (4) We then evaluated the efficacy of these nomograms using decision curve analysis (DCA) in both the training and validation cohorts, and ultimately constructed three online calculators based on these nomograms. Results We identified ASA, diabetes, smoking, and lymph node invasion as predisposing risk factors for postoperative complications, and the predictive factors that affected both OS and CSS were ASA, diabetes, BMI, CA19‐9 level, and tumor diameter. By integrating the above risk factors, we constructed three nomograms on postoperative complication, CSS, and OS. The C‐index for complication, CSS, and OS were 0.824, 0.784, and 0.801 in the training cohort and 0.746, 0.718, and 0.708 in the validation cohort. Moreover, the validation curves and DCA demonstrated good calibration and robust compliance in both training and validation cohorts. We then developed three web calculators (https://caiming.shinyapps.io/CMCD/, https://caiming.shinyapps.io/CMCSS/, and https://caiming.shinyapps.io/CMOS/) to facilitate the use of the nomograms. Conclusions The calculators demonstrated promising performance as an tool for predicting the safety and efficacy of PD in elderly PHC patients.

Keywords