Prognostic nomogram for patients undergoing resection for adenocarcinoma of the pancreas

Abstract

Background: Predictive nomograms are becoming increasingly used to define and predict outcome. They can be developed at presentation or following treatment and include variables not conventionally used in standard staging systems.

Methods: We use a predictive nomogram based on prospectively collected data from 555 pancreatic resections for adenocarcinoma at a single institution. At last follow-up, 481 (87%) had died, defining a mature and comprehensive database. We used a 1-, 2-, and 3-year follow-up, as the number of patients alive beyond 3 years is sufficiently limited to provide insufficient events.

Results: Based on a Cox model, we then developed a nomogram that predicts the probability that a patient will survive pancreatic cancer for 1, 2, and 3 years from the time of the initial resection, assuming that there is not death from an alternate cause. Calibration between observed and corrected is good, and variables not conventionally associated with standard staging systems improved the predictivity of the model.

Conclusions: This nomogram can serve as a basis for investigating other potentially predictive variables that are proposed of prognostic importance for patients undergoing resection for adenocarcinoma of the pancreas.

FIGURE 1. Disease-specific (dotted line) and other cause (solid line) probability of death. Figures at top indicate number of patients at risk.

FIGURE 2. Nomogram for predicting 12-, 24-, and 36-month disease-specific survival probabilities.

FIGURE 3. Calibration of the nomogram. Bootstrapping was used to correct for optimistic bias. X-axis is nomogram-predicted probability of survival. Y-axis is observed disease-specific survival.

FIGURE 4. Comparisons of nomogram predictions with those of AJCC stage groupings. Note the heterogeneity of predicted probabilities of survival within each AJCC stage, especially in stages II and III.