Please join us for DFCI Ethics Grand Rounds

"Knights, Knaves or Pawns?  Physicians as Accountable Professionals"

Christine K. Cassel, MD MACP

President and CEO

American Board of Internal Medicine

Philadelphia, PA

Monday, January 23

12 noon - 1:00pm

Yawkey 306

Dana-Farber Cancer Institute

Also available via webcast at: 

http://video.dfcionline.org/accordent/EthicsGrandRounds012312

 Lunch will be served at 11:45; Rounds will begin at noon

Contact:  Jennifer Kesselheim, MD, MEd

Systemic Therapy for Brain Metastases from Breast Cancer

Friday, January 27th, 2012

Nancy U. Lin, M.D.
Medical Oncologist
Dana-Farber Cancer Institute
Assistant Professor
Harvard Medical School

Dana-Farber Cancer Institute Seminars in Oncology

"Target Discovery with Oncomine Discovering New Targets and Positioning Old Targets by Integrative Analysis of 10,000’s of Tumor Genomes and Transcriptomes”
Daniel Rhodes, Ph.D.
CEO

Compendia Bioscience

Ann Arbor, MI

Tuesday, January 24, 2012
4:00PM
Jimmy Fund Auditorium

Host: Myles Brown, M.D.

617-582-7646

Special Seminar in the Department of Biostatistics and Computational Biology

William Barry, Ph.D.
Assistant Professor, Department of Biostatistics and Bioinformatics
Duke University

 
Abstract:

In cancer clinical research, the role of biomarkers has increased dramatically in evaluating the efficacy of therapeutics.    Further, the availability and decreasing costs of high-throughput assays allow for interrogation of the entire genome when considering molecular biomarkers of disease.  In order to draw proper conclusions in both the retrospective and prospective setting, novel statistical methods are required to account for flexible trial designs and for the unique structures of molecular data.

To address this need in cancer clinical trials with integral biomarkers, we propose the Bayesian response-adaptive design of Zhou et al. (2008) for a randomized phase II study in metastatic breast cancer with multiple predictive biomarkers.  This design allows for a gradual and seamless approach to moving from randomized-block designs into biomarker marker-enrichment designs, and is more robust than traditional designs against model misspecification, such as marker prevalence.  Furthermore, by jointly modeling responses to treatments, a simulation study demonstrates how target populations are better identified for future definitive phase III studies.

For retrospective evaluations of multiplex molecular markers, we propose a novel extension to pathway-analysis as a first step for evaluating signatures in external datasets before translating predictive models into a new clinical context.  In this framework, inferences on the global association of signatures can properly account for the gene-gene co-regulation, and are not confounded by the particular challenges of integrating genomic data derived from different technologies, laboratories, and protocols. Illustrations are drawn from previously identified signatures in ovarian cancer with publicly available data, including The Cancer Genome Atlas prognostic signature (2011) that was externally validated, and the chemosensitivity signature of Dressman et al. (2007) that fails to validate in independent cohorts.