53 - Séance de prix
British Journal of Surgery Session BJS Lecture & BJS-Paper Session inkl. Award ceremony
3 juin 2021, 15:25 - 17:30, Stream 1 - 4
External validation of three nomograms predicting survival using an international cohort of patients with resected pancreatic head ductal adenocarcinoma
M. Schneider1, I. Labgaa1, D. Vrochides2, A. Zerbi3, G. Nappo3, J. Perinel1, 4, M. Adham4, S. van Roessel5, M. Besselink5, J. S. D. Mieog6, J. V. Groen6, N. Demartines1, M. Schäfer1, G.-R. Joliat1, 7, Presenter: M. Schneider1 (1Lausanne, 2Charlotte, North Carolina, 3Milan, 4Lyon, 5Amsterdam, 6Leiden, 7Bern)
Lymph node ratio (LNR, positive lymph nodes/collected lymph nodes during surgery) was identified as an important prognostic factor of survival in resected pancreatic cancer. Several nomograms based on LNR were recently proposed to predict survival after pancreatoduodenectomy (PD). The present study aimed to externally validate 3 published nomograms using an international cohort.
Consecutive patients with ductal adenocarcinoma of the pancreatic head who underwent PD without neoadjuvant treatment from 6 tertiary centers in Europe and the USA were retrospectively collected from 2000 to 2017. Patients with metastases at diagnosis, R2 resection, missing data regarding LNR, and who died within 90 postoperative days were excluded. The 3 selected nomograms were the updated Amsterdam nomogram (including LNR, adjuvant therapy, margin status, and tumor grade), the nomogram by Pu et al. (including LNR, age, tumor grade, and T stage) and the nomogram by Li et al. (including LNR, age, tumor location, grade, size, and TNM stage). Overall survivals (OS) were calculated using Kaplan-Meier method. For the validation, calibration (Hosmer-Lemeshow test), discrimination capacity (ROC curves for 3-year OS), and clinical utility (sensitivity and specificity at the value of Youden index) were assessed.
After exclusion of 95 patients with metastases, R2 resection, and who died within 90 postoperative days, 1167 patients were included. Median OS of the entire cohort was 23 months (95% confidence interval: 21-24). For the 3 nomograms, Kaplan-Meier curves showed significant diminution of OS with increasing scores (p<0.01 for the 3 nomograms). All nomograms showed good calibration (non significant Hosmer-Lemeshow goodness-of-fit tests). For the updated Amsterdam nomogram, the area under the ROC curve (AUROC) for 3-year OS was 0.66. Sensitivity and specificity were 73% and 50%. Regarding the nomogram by Pu et al., the AUROC was 0.67. Sensitivity and specificity were 65% and 60%. For the nomogram by Li et al., the AUROC was 0.67, while sensitivity and specificity were 56% and 71%.
The 3 selected nomograms were validated using an external international cohort and displayed interesting and comparable predictive values. Those nomograms may be used in clinical practice to estimate survival after PD for ductal adenocarcinoma.