12. Statistical methods for phase I/II trials of molecularly targeted agents in oncology

Team: Dr Xavier Paoletti (Institute Gustave Roussy), Dr Andreas Kaiser (Bayer)
Clinical Advisor: Dr Christophe Le Tourneau
is an expert in Oncology at the Institut Curie and an internationally acclaimed leader in phase I oncology studies.
Location: INSERM (Paris, France)

Conventional dose-finding approaches mainly consider severe toxicities observed during the first cycle. This may not be relevant for molecularly targeted agents usually administered in chronic schedules, and for which more than 50% of severe toxicities occurred after the first cycle20 . This project focuses on the development of innovative adaptive dose finding methods using longitudinal data of both toxicity and activity measurements. As clinical responses are observed in less than 5% of patients, biomarker of activity such as prostate specific antigen, lymphatic T-cell in immunotherapy, or even continuous tumour measurements from CT or PET-scans are appealing. Continuous measurements of activity will be considered as they carry richer information than categorical measurements.

Adaptive mixed effect models based on longitudinal ordinal outcome21 will be extended to combined outcomes. A general model for the mixed correlated repeated ordinal and continuous responses will be proposed using a probit relation through a latent approach22 . Alternatively models of longitudinal outcomes with a shared random effect can be explored under the hypothesis that the constrained correlation structure is acceptable. The parameters of models will be estimated in a Bayesian framework. Dose finding designs based on estimates of outcomes trajectories will be proposed. After each group of observations, the dose that maximizes the activity under the constraint of maximum toxicity level is recommended for the next group of patients.

Due to the small sample sizes in phase I trials, the operating characteristics of the proposed designs will be evaluated in extensive computer simulations studies. A cornerstone to the evaluation is the realisation of a dedicated software or library that can be easily released to the community. Assessment of applicability of the methods on retrospective analyses may use databases from the pharmaceutical companies participating in this consortium that sponsors phase I trials in oncology.

Meet our Early Stage Researcher: Maria-Athina Altzerinakou, INSERM


I am from Athens, Greece. In 2012 I graduated from National and Kapodistrian University of Athens with a Bachelor in Mathematics. Two years later in 2014, I completed my Master in Biostatistics, also at the University of Athens. I am currently working on my PhD thesis at the Gustave Roussy Cancer Campus in Paris, France.  The main part of my thesis revolves around dose finding in Phase I/II trials of molecularly targeted agents in oncology. In my free time I enjoy meeting with my friends, reading books, watching films, traveling all around the world and going swimming.

20 Postel-Vinay S, Gomez C, Molife R, et al. (2011) J Clin Oncol, 29:1728-35.
21 Doussau A, Thiebaut R, Paoletti X. (2013) Stat Med, 32(30):5430-47
22 Molenberghs and Verbeke. (2005) Models for Discrete Longitudinal Data. Springer, Series in Statistics.


Project News 1 of 5 pages

Meet the IDEAS Network partners

The IDEAS network brings together leading institutions in the training of medical statisticians, large and small sized pharmaceutical companies and is supplemented by clinical experts.