Bayesian Methodology Applied to A Phase II Relapsed Ovarian Cancer Study
by
Rudy Gunawan
University of Waterloo
Coauthors: Dr. Amit Oza (Princess Margaret Hospital) and Dr Siv Sivaloganathan (University of Waterloo)

Cancer Antigen - 125 (CA-125) has been demonstrated to be a surrogate marker in ovarian cancer. Additionally, CA-125 is crucial for screening efficacy in phase II clinical trials with small sample sizes. The standard measure for solid tumor is an objective radiographic response using RECIST (Response Evaluation Criteria for Solid Tumor); however, it can be difficult in ovarian cancer due to the intraperitoneal distribution of the disease. From a series of phase II studies (conducted at the Princess Margaret Hospital in Toronto, ON) investigating various molecularly targeted agents (MTAs) such as topotecan, sorafenib and gemcitabine in relapsed ovarian cancer patients, we attempt to tackle two main challenges : the concordance between the objective radiographic (RECIST) responses and the objective biochemical (CA-125) responses based on the clinical trials' data and the scientific validation whether the CA-125 response is a better surrogate for overall survival than the RECIST response in these clinical trials. We formulate the analysis as a problem in scientific inference to be solved using Bayes' theorem. No smoothing, interpolation, extrapolation, or other preprocessing of the data is necessary. Despite the limited sample size, the mean of the proportion of the patients responding to the MTA in each clinical trial is obtained, together with its uncertainty. Based on the relevant prior information and the current clinical trials' data, we do not observe any concordance between the RECIST response and the CA-125; furthermore, the CA-125 responders have a greater survival probability only for up to six months.

Date received: May 5, 2008