Using tumor growth rate to inform treatment efficacy in pancreatic adenocarcinoma: From the metastatic to the neoadjuvant setting.

726 Background: The development of new treatments in oncology is a long, costly, and, too often, unsuccessful process. Methods for screening agents earlier in development and strategies for conducting smaller randomized controlled trials are needed. Methods: We used a mathematical model of tumor growth kinetics fit to serial radiographic tumor measurements or CA19-9 values to estimate rates of exponential tumor growth [g] and decay [d] during treatment for pancreatic ductal adenocarcinoma (PDAC). Results: We retrospectively collected and analyzed data from 2691 patients with stage III-IV PDAC who were enrolled in five clinical trials or were included in two large real-world data sets from Columbia University Irving Medical Center and Veterans Administration Medical Centers. Using log-rank comparison of Kaplan-Meier plots by quartile of g, we found that in patients with metastatic PDAC g correlates highly with overall (OS) and progression-free survival (PFS) (p<0.001), with slower g associated with improved survival in this population. Pairwise comparisons showed significantly slower median g in the experimental arm versus control arm in the pivotal trials analyzed (p<0.001). At the individual patient level, g was significantly faster for liver metastases as compared to primary pancreatic tumors and g consistently increased towards the end of therapy (often a threefold increase) suggesting development of chemoresistance. In addition to utility in the metastatic setting, a pilot analysis of data from a prospective study of patients treated with gemcitabine + docetaxel + capecitabine (GTX) in the neoadjuvant setting suggests that the emergence of a detectable g during neoadjuvant therapy may portend worse OS following surgery though sample size was small (n=45, median OS with detectable g 13.6 m v 33.1 without detectable g, p =0.35). Conclusions: We applied a tumor growth model to the data of over 2500 patients with PDAC and showed that g is inversely associated with survival in this population. Given the strong association between g and survival, g could be useful in clinical trials as an informative endpoint to expedite the assessment of novel therapies for the treatment of PDAC. Furthermore, g offers valuable patient-level data, including trends in resistance and variations in growth rate by metastatic disease site We plan to evaluate whether the emergence of g during neoadjuvant therapy should be considered a prompt to change treatment regimens.