PD-002 - Impact of the definition of time to event endpoint on randomized clinical trials results in oncology (DATECAN-2): an analysis of 9 pancreatic cancer...

Date 04 July 2015
Event WorldGI 2015
Session Posters
Topics Pancreatic Cancer
Presenter A. Pam
Citation Annals of Oncology (2015) 26 (suppl_4): 101-107. 10.1093/annonc/mdv234
Authors A. Pam1, S. Gourgou2, D. Vernerey3, B. Chauffert4, C. Louvet5, F. Levi6, H. Galjaard7, C. Bellera8, A. Anota1, F. Bonnetain1
  • 1University Hospital of Besançon, Besançon/FR
  • 2Comprehensive Cancer Centre and Data Center for Cancer Clinical Trials, Montpellier/FR
  • 3University Hospital of Besancon, Besancon/FR
  • 4CLCC Georges-François Leclerc, Dijon/FR
  • 5Institut Mutualiste Montsouris, Paris/FR
  • 6Paul Brousse Hospital & INSERM, Villejuif/FR
  • 7Regio Rotterdam, Rotterdam/NL
  • 8Comprehensive Cancer Centre, Bordeaux/FR



For randomized clinical trials (RCTs), some composite time-to event-endpoints such as progression-free survival (PFS) as potential surrogate of overall survival (OS) are frequently used. Such endpoints required to be precisely defined and surrogacy for OS should be assessed for internal validation. DATECAN-2 objective is to study the impact of the definitions proposed in DATECAN-1 initiative (F. Bonnetain et al., EJC 2014) on the results and conclusions from individual patients data enrolled in 9 pancreatic cancers RCT.


DATECAN-1 endpoints definition recommendations have been used to estimate effect of evaluated treatment using Kaplan Meier estimation and Cox analyses and have been compared with results from the definition used for the referenced publication. Agreement was measured by the Spearman correlation coefficient for the time-to-event duration and Cohen's Kappa coefficient for the events. Internal validation was check through PTE for surrogacy of OS.


Among the 9 databases, finally 7 RCTs were kept since 1 was not published and the other due to availability of updated database only. OS, PFS, Time to Progression and Diseases-Free Survival were considered 2, 1, 0 and 1 times as a primary endpoint and 5, 3, 2 and 0 times as a secondary endpoint, respectively. Only two time-to-event endpoints (Time to Local Progression (TLP) and Cancer-Specific Survival (CSS)) among the 14 from the recommendation of DATECAN-1 could have been formed. Many events considered in the criteria from the recommendation of DATECAN are missing and or not detailed in the trial database. A descending classification of the PTE for the three time-to-event endpoint has been done as follows PFS, CSS and TLP.


Our results show that RCT failed to collect precisely each event that constitute definition of composite time-to-event. Then we could not apply all DATECAN1 definitions. To be rigorous and to allow standardization we suggest in future RCT to collect each localization of progression and of recurrence. Moreover we will be able to classify all the criteria based on their substitution capabilities to OS.