(System Info - 126673 TULL LORI 04/22/2010 11:31:03 TULL)
Data: October 22, 2009
Time: 11:00 AM
FDA Participants: Drs. Ball, Bross, George, Gupta, Hsu
External Consultant: Dr. Ralph D’Agostino, Boston University, Boston, MA
Subject: Discussion of Study D9902B Summary and Question
We asked Dr. D’Agostino if he needs any clarification of the summary or the question. He said he understood the issues in question.
Dr. D’Agostino said that he understood the concern about potential confounding of the overall survival endpoint by subsequent salvage therapy. As to additional analyses to be performed to address concerns of the use of the salvage treatment, he proposed stratifying time to salvage therapy and to provide the descriptive statistics. It may provide some light to the use of salvage therapy. It is equivalent to considering time to salvage therapy as a time dependent covariate in the Cox model. We also discussed the concern that this covariate is prognostic dependence. The Cox model would provide a test statistic and a p-value, and the descriptive statistics by stratifying time to salvage therapy will provide some insight into the survival trend.
Dr. D’Agostino also mentioned that delayed effect of the treatments by about 8-10 months raises a question of the violation of assumptions of parallelism for the Cox model. It may affect the estimate of the hazard rate and the power of the test. Consequently it would most likely cause a false negative conclusion (a type II error, not a false positive one). He did not have an analytical solution to the problem.He talked of using propensity score. However, he realized that it would not provide any substantive information due to the small sample size of only about 500 patients.
He stated that he would like to look at the complete clinical report and we agreed to send it to him immediately. We agreed to teleconference again at 1:00 PM on Friday, October 30, 2009.