Trends and predictors of time to initiation of treatment for non-metastatic pancreatic cancer: a national cancer database analysis

Abstract

Background: Time to treatment initiation (TTI) is an important quality metric in oncology and a delayed TTI is believed to negatively influence survival. Trends and predictors of TTI are not well understood in patients with non-metastatic pancreatic cancer. We aim to evaluate trends in TTI and determine predictors of TTI in non-metastatic pancreatic cancer patients. Methods: The National Cancer Database (NCDB) was queried for the period 2004 through 2015 to identify all patients with a diagnosis of pancreatic cancer. Only those patients with a recorded TTI under 365 days were included as were those with non-metastatic pancreatic cancer. Patient, disease, and hospital characteristics were assessed for statistically significant differences in median TTI using Mann-Whitney U and Kruskal-Wallis tests and, where appropriate, assessed for correlation with TTI using Spearman’s rho test. Significant variables were then fitted into a zero-inflated negative binomial regression model to determine their significance in predicting TTI. This model was chosen as “0-days” comprised 19.9% of the TTI data element. Results: A total of 109,627 cases met the study criteria. The median age was 67 years (Interquartile range [IQR] = 58–74 years) with an equal gender ratio and a Caucasian predominance (84.5%). Most patients had Medicare health insurance (53%), attended an academic center for treatment (50.7%), and lived within 10 miles of the center of initial diagnosis (40.8%). The tumors were mainly located within the pancreatic head (61.3%) and of AJCC stage II (43%). Surgical intervention as a single treatment modality was the most common management (26.6%). The median overall TTI was 21 days (IQR = 6–38 days). A significant (<0.001) upward trend in TTI days was noted over the years between 2004 and 2015 from 18 days (IQR = 1–36 days) to 24 days (IQR = 11–40 days) respectively. Univariate analyses of the provided characteristics did show significant differences in relation of TTI, however, the Spearman rho test did not show a correlation with TTI. Furthermore, multivariate analysis using a zero-inflated negative binomial regression model demonstrated that ethnicity, Charlson comorbidity score, insurance status, zip code level of education, living distance from treatment facility, facility type, AJCC cancer stage, and treatment modality had significant differences in TTI (p < 0.05). Conclusion: There has been an increase in TTI for PDAC over the period 2004-2015 with a number of significant factors linked to TTI including patient, hospital, and treatment specific characteristics. The only modifiable factor identified is the type of facility with integrated network cancer programs having the lowest TTI and thus choice of facility might reduce TTI and influence outcomes.