Analysis of the rate of missing data, the rate of discordance between readers, and the rate of site versus central discordance in clinical studies of recently approved breast cancer agents that have used blinded independent central review.

Abstract

Abstract Abstract #2081 Background: The United States Food and Drug Administration (USFDA) recommends blinded independent central review (BICR) for oncology registration studies when the primary study endpoint is based on tumor measurements, such as progression-free survival (PFS) or objective response rate (ORR). However, there is no published guidance regarding acceptable metrics during the BICR.
 Method: The USFDA Summary Basis of Approvals and Oncologic Drugs Advisory Committee (ODAC) transcripts for the following recently approved breast cancer therapies: Gemzar, Abraxane, Tykerb, Ixempra, and Avastin, were reviewed with attention to metrics referable to the BICR. The findings are summarized.
 Results: Up to 13% of subjects were reported as unevaluable at the BICR based on missing data, with missing data defined as missing radiographic evaluations. The rates of discordance between two readers during the BICR were reportedly between 34-51% depending on the number of adjudication variables. The rates of discordance between the site and BICR were reportedly between 34-49% for the date of progression and 24-29% for the progression status. Up to 9% of subjects enrolled with measurable disease at baseline did not have this confirmed by the BICR. The causes of discordance were reportedly related to differences in the methodologies used to perform the reviews, and clinical or image data that was missing at the time of the BICR. Additional reported causes included imaging requirements not detailed in the protocol, the BICR charter containing more stringent requirements for determining response and progression than was indicated in the protocol, protocol amendments issued midway through the study which changed the imaging eligibility criteria for the sites while the BICR used the same requirements throughout the study, differences in the choice and classification of lesions, radiographic exams only repeated at subsequent time points if deemed positive by the investigator resulting in missing data and unevaluable assessments at the BICR, variability in measurement of small lesions, and discordance in the determination of new lesions. Inter- and intra-reader variability was reported in a small population of subjects; 8% resulted in 1-step response discrepancies for inter-reader variability and 31% for intra-reader variability.
 Conclusion: Based on FDA approvals, metrics as described above may prove to be valuable benchmarks when using BICR data to determine endpoints in future breast cancer studies. Citation Information: Cancer Res 2009;69(2 Suppl):Abstract nr 2081.