Difficulties with Observational Medical Studies

WORKSHOP READING MATERIAL (PDF file)

Statistics Ready for a Revolution (PDF file): M. van der Laan and S. Rose

NISS will hold a workshop entitled "Difficulties with Observational Medical Studies" on June 16-17. The workshop will take place at NISS headquarters in Research Triangle Park, NC. 

Observational studies have gained prominence with funding of comparative effectiveness research (CER), as well as the growing efforts in personalized medicine. There is, however, a tension between finding the best drug and finding the best drug for each person. The difficulties with the analysis of observational studies are many: multiple biases (e.g., doctors' channeling patients to specific medicines), multiple testing (e.g., the many side effects of interest), and multiple modeling. There is need for researchers, policy makers and individuals to recalibrate their thinking away from randomized clinical trials (RCT), or perhaps to take lessons learned from RCT and apply them to observational studies.

The program begins with an introduction to the analysis of observational studies, in the form of a one-half day tutorial by Alan Brookhart of the Department of Epidemiology at the University of North Carolina at Chapel Hill. Technical sessions will cover bias, multiple testing and multiple modeling, and case studies will illustrate both the problems that happen in the real world and success stories. Finally, a panel of experts will lead a discussion of the ins and outs of observational studies. There will be multiple opportunities for formal and informal discussions among the participants.

FEE

There is no fee for workshop speakers, or for members of NISS Affiliates who attend.
Otherwise, the fee is $100.00.

SCHEDULE

Thursday, June 16, 2011

Alan Brookhart, UNC - Basics of Observational Studies   (2:00 – 5:00)

Friday, June 17, 2011

Welcome: Alan Karr, NISS  (8:30 -8:45)

Introduction: Stanley Young, NISS  (8:45-9:15) 

Technical Session (Multiple testing, multiple modeling, bias)

Moderator: Adam Polis, Merck

• Patrick Ryan, Johnson & Johnson – Problems with Multiple Modeling (9:15-10:00)

Break (10:00 – 10:30)

• Robert Obenchain, Risk Benefit Statistics – How to deal with bias  (10:30-11:15)
• Sujit Ghosh, N.C. State University – Meta analysis and multiple testing  (11:15-12:00)

Lunch (12:00-1:00)

Case Studies: Basic facts, literature claim, reaction, statistical issues, lessons learned

Moderator: Robert Obenchain, Risk Benefit Statistics

• Allen Heller, Bayer HealthCare – Aprotinin/Trasylol  (1:00 – 1:45)
• Adam Polis, Merck - Cancer Risk—Does ezetimibe or ezetimibe/simvastatin increase it?  (1:45-2:30)

Break (2:30-3:00)

• Timothy Vaughan, PatientsLikeMe – Drug Efficacy in the Wild (3:00-3:45)

Panel Discussion (3:45-4:45)

Where are we with data, technology, and analysis strategy?

Moderator: Alan Menius, GlaxoSmithKline

• Allen Heller, Bayer HealthCare
• Frank Rockhold, GlaxoSmithKline
• Martin Marciniak, GlaxoSmithKline
• Robert Obenchain, Risk Benefit Statistics

Patrick Ryan, Johnson & Johnson