Relative contribution of baseline variables in a nomogram to predict survival in patients treated with nab-paclitaxel plus gemcitabine or gemcitabine alone for metastatic pancreatic cancer

Abstract

Introduction: Nomograms for the prediction of survival outcomes have been developed in various cancers, including, but not limited to, breast, lung, and gastrointestinal. However, none have been created to predict survival in patients with metastatic pancreatic cancer (MPC) treated with nab-paclitaxel plus gemcitabine or gemcitabine alone. The objectives of this analysis were to develop a prognostic nomogram for the prediction of overall survival (OS) using baseline data from the phase III MPACT trial of nab-paclitaxel plus gemcitabine vs gemcitabine for the first-line treatment of MPC and to assess the relative contribution of each variable. Methods: A multivariable Cox model was developed using baseline variables that were significantly predictive of OS by univariable analysis (P<0.10) or considered clinically important and retained in the model after stepwise selection (P < 0.10). This final set of identified variables was used to create a nomogram (primary) that assigned points equal to the weighted sum of relative significance of each variable; the most predictive variable was assigned a maximum point value of 100, and other variables' points were determined based on comparison with this most influential variable. Patients were classified by prognosis into low-, medium-, and high-risk groups. Subsequent to creating the primary nomogram, the effect of individually adding 5 variables that were not statistically predictive, but were believed to be clinically important (carbohydrate antigen 19-9 [CA19-9], age, number of metastatic sites, number of lesions, lung metastasis), was examined to determine how much these variables would contribute to the predictive ability of the nomogram if forced into the model. All nomograms were internally validated using bootstrapping, a concordance index (c-index), and calibration plots. Results: Data from 861 patients were used for all analyses, except the CA19-9 analysis (n=634), which excluded CA19-9 nonsecretors. For the primary analysis, 7 variables were retained in the multivariable model. In order of magnitude, neutrophil-tolymphocyte ratio, albumin level, Karnofsky performance status, sum of longest tumor diameters, presence of liver metastases, treatment arm, and analgesic use were the key OS predictors. The resulting primary nomogram (Table) distinguished between low-(n=216), medium- (n=430), and high-risk (n=215) groups (c-index, 0.69; 95% CI, 0.67 - 0.71), with median OS values of 12.9, 8.2, and 3.7 months, respectively. CA19-9, number of metastatic sites, and lung metastasis individually contributed only 1 point when added to the primary nomogram. Age contributed up to 7 points; number of lesions, up to 10 points. Conclusion: This nomogram includes 7 baseline variables to predict OS in patients with MPC treated with first-line nab-paclitaxel plus gemcitabine or gemcitabine alone. There may be some incremental value in adding the number of lesions or age to the primary nomogram, but there is no perceived value in adding CA19-9, number of metastatic sites, or the presence of lung metastasis. The nomogram may help physicians and patients make informed decisions in MPC care. (Table Presented).