SC.06

Virginie Rondeau is researcher since 2001 and director of research since 2014 in the department of Biostatistics at the National Health and Medical Research Institute of the Bordeaux University. Dr. Rondeau‘s interests include statistical modeling, survival analysis, correlated survival data, prognostic models. She has been heavily involved in collaborations in the areas of cancer recurrences and environmental health effects.

1. Rondeau is well-suited for this course for different reasons:

• she already gave this kind of international course at the Karolinska Institutet (with Agnieszka Krol) in 2016 and also in 2014 at the ISCB conference (in Vienna).
• She as solid track record as principal investigator and co-investigator of many French government agency funded statistical methods and clinical research studies with >70 publications
• She has major contributions to the development of joint statistical models for recurrent event data in the presence of competing events, including statistical software development (e.g. R statistical package “Frailtypack”since 2005
• She has the ability to successfully collaborate with local and international clinical and statistical researchers (e.g. Sebastien Haneuse and Ina Jazic from Haward, Boston, USA; Bernard Rachet, from the London school of Hygiene and Tropical Medicine, London, UK; Takeshi Emura, from the National Central University, Taoyuan, Taiwan; Laurent Briollais, from the Dalla Lana School of Public Health, Ontario).
• She has ongoing participation in professional statistical societies, including (International Biometric Society, International Society for Clinical Biostatistics).

Joint Models for recurrent events and competing risks (e.g. death)

Development of joint models for multiple longitudinal markers and/or multiple times-to-event.This research axis focuses mainly on the development of models for time-to-event data taking account of incomplete and/or correlated observations. These models cover a wide field of event history analysis from survival models for truncated and/or censored data, with or without interval-censoring. Both parametric and semi-parametric penalized likelihood approaches were implemented. The developed methods were mainly applied to study recurrences of cancer. I supervised several PhD students on this topic and initiated new international collaborations for that (e.g., Laurent Briollais, from the Dalla Lana School of Public Health, Ontario).

Prognostic survival models and validation

Using estimated parameter from joint models it is quite easy to compute prediction probabilities to meet a health event in a window of prediction. However, a major step in the development of dynamic prognostic tools is the validation of model assumptions and the assessment of predictive performances. These steps are much more challenging with joint models and complex censoring schemes and extensions of existing approaches need to be considered. I supervised different PhD students on this topic and initiated new international collaborations for that (e.g., Bernard Rachet from the London School of Hygiene and Tropical Medicine).

Statistical software development for correlated survival data, including joint models for recurrent events with competing risks or terminating events (e.g. death).  R statistical software “Frailtypack”

I wanted to invest a great effort in the dissemination of these statistical methods for biostatisticians but also for epidemiologists and clinicians. This is why I worked on their implementation in the R-package “FRAILTYPACK”. Furthermore, I work on the organization of short courses in several conferences, publication of didactic papers in journals of epidemiology or medical specialties and participating in the writing of a book.

Denis Rustand