We are interested in developing methods for
- conducting early phase clinical trial
- validating prognostic and predictive biomarkers.
- accounting for correlation induces by clustered observations (a cluster can be: a patient, a clinical trial)
- estimating treatment effect from observational data
- quantifying long term responders under immunotherapy
- inference from meta-analysis of individual patient data
To ensure reproducibility of our findings, our code can be found in our team’s
- Funded project for 2021 : “Using CA125 biomarker to predicting early progression and treatment failure in advanced ovarian cancer” (PI X. Paoletti) in collaboration with Pr. I. Ray-Coquard (Lyon)
Identification of early resistance to treatment in newly diagnosed ovarian cancer is crucial as it affects the overall treatment strategy. Currently, relapse within 6 months post treatment is used as the key endpoint, which may be too late, given the very poor prognosis of patients. It is usual practice to follow-up patients by serial measurements of the circulating CA-125 marker. Decrease in CA-125 levels after systemic treatment initiation is deemed to reflect response to treatment. Yet, the prognostic value of CA-125 and its predictive value to predict response to treatment resistance as well as the best modeling approach is strongly debated. Although the French NCI (INCa) recommended its use, this is not without highlighting the lack of clear scientific evidences. All the more as new classes of treatments have recently been demonstrated to be strongly active.
In this project, we aim at investigating
- (i) the prognostic and predictive value of CA125 trajectory to identify treatment resistance,
- (ii) its validity as a surrogate marker and
- (iii) to validate those results on new therapeutic classes of agents (among which PARP inhibitors) using a very large database of individual patients data from 13 randomized trials with 9106 women.
For each woman, we have collected the repeated measures of CA-125 in addition to the baseline characteristics and the clinical endpoints. This unique resource will be enriched with the most recent PARP inhibitors trials.
We will extend joint modeling approaches of longitudinal biomarkers and time to event endpoint to the field of meta-analysis. Those techniques are now well-established in the context of prostate cancer to evaluate prognostic markers and perform dynamic predictions in single trials/ cohorts, not for meta-analyses. We will apply those methodological tools on our meta-analysis database of women who received first line/ maintenance treatments.
The expected outcomes are to better clarify the role of CA-125, which may be over-interpreted or misinterpreted by patients and their physicians, and to extend those results to the most recent investigated classes of treatments