Assessment of intermediate-term mortality following pancreatectomy for cancer.

Abstract

Pancreatic cancer remains highly lethal, and resection represents the only chance for cure. Although patients are counseled regarding short-term (0-3 months) mortality, little is known about mortality 3-6 months (intermediate-term) following surgery. We assessed predictors of intermediate-term mortality, evaluated hospital-level variation, and developed a nomogram to predict intermediate-term mortality risk. Patients undergoing pancreatic cancer resection were identified from the National Cancer Database (2010-2020). Multivariable logistic regression identified predictors of intermediate-term mortality and assessed differences between short-term and intermediate-term mortality. Multinomial regression grouped by intermediate-term mortality quartiles evaluated hospital-level variation. A neural network model was constructed to predict intermediate-term mortality risk. All statistical tests were 2-sided. Of 45 297 patients, 3974 (8.9%) died within 6 months of surgery of which 2216 (5.1%) were intermediate-term. Intermediate-term mortality was associated with increasing T category, positive nodes, lack of systemic therapy, and positive margins (all P < .05) compared with survival beyond 6 months. Compared with short-term mortality, intermediate-term mortality was associated with treatment at high-volume hospitals, positive nodes, neoadjuvant systemic therapy, adjuvant radiotherapy, and positive margins (all P < .05). Median intermediate-term mortality rate per hospital was 4.5% (interquartile range [IQR] = 2.6-6.5). Highest quartile hospitals had decreased odds of treatment with neoadjuvant systemic therapy, neoadjuvant radiotherapy, and adjuvant radiotherapy (all P < .05). The neural network nomogram was highly accurate (accuracy = 0.9499; area under the receiver operating characteristics curve = 0.7531) in predicting individualized intermediate-term mortality risk. Nearly 10% of patients undergoing pancreatectomy for cancer died within 6 months, of which one-half occurred in the intermediate term. These data have real-world implications to improve shared decision making when discussing curative-intent pancreatectomy.