Statistical methods and designs in clinical oncology

Symposium of the ONCOSTAT team INSERM U1018 CESP, Univ. Paris-Sud, Univ. Paris-Saclay

 & Service de Biostatistique et Epidémiologie de Gustave Roussy

 Thursday November 9, 2017

The OncoStat symposium is dedicated to statistical methods for clinical trials. It focuses on recent or controversial approaches. This year, we propose two themes:

  • Adaptive designs Yes? Adaptive randomization No?
  • Beyond standard inference in survival analyses to account for design deviations

Adaptive design has come up as a very promising way of modifying several features of a design during the course of the trial based on interim results, without ruining statistical inference. In particular, multi-stage multi arms trials have been implemented to compare several treatments with the possibility to drop ineffective ones during the course of the trial. Statistical tools using Likelihood or Bayesian inference have been proposed to estimate treatment effect while controlling for the risk of over-estimation. However, in all those cases, the cornerstone of all comparisons relied on fair randomization. In parallel, several authors have proposed to adapt the randomization based on interim efficacy results, in order to give greater probability for a patient to receive the treatment arm where the observed response rate appears to be the highest. Such approaches are strongly debated. This will be our first topic.

More complex trials generally come together with more difficult analyses as it increases the risk that the conditions to apply the planned analysis are no longer valid. For instance, missing data, possibly due to cross-over (or treatment switch) are increasingly common, or interactions between the treatment effect and baseline covariates may create spurious association. Analysis tools need be adapted to tackle these issues. This will be our second topic.

Conditions d'inscription

This symposium is open to statisicians interested in advanced methods for clinical research in oncology.

The maximum number of attendees is 120. Registration is mandatory and the registration fee is

  • 60€ for academia
  • 160€ for members of pharma companies.
  • Free for students (but registration is mandatory)

This fee includes the access to the symposium as well as the lunch.

Click on the “registration” box on the left-hand side for online registration and payment (by credit card, bank transfer or purchase order). The registration fee for participants is nonrefundable, though substitutions may be made.

We wish this symposium to stimulate exchanges and discussions on recent statistical methods and designs in the field of clinical trials in oncology. Each speaker will be given 35 minutes including 5 to 10 minutes for interacting with the room.


For administrative questions, please contact  Françoise Terrier 

For questions on the scientific content, please contact Xavier Paoletti  or Stefan Michiels







Up sud







Personnes connectées : 1