Adjusting for Selection Bias in Neuropathology Studies of Dementia: The Challenge of Trying to Extrapolate from the Dead to the Living

• Activity Number: 595
• Type: Contributed
• Date/Time: Wednesday, August 4, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Steven P. Millard*+ and Mary Lou Thompson
• Companies: Veterans Affairs Puget Sound Health Care System and University of Washington
• Keywords: Autopsy ; Alzheimer's Disease ; Cohort Study ; Dementia ; Neuropathology ; Selection Bias

Abstract

Autopsy studies are usually the only means of establishing association between potential risk factors and neuropathological outcomes. Alzheimer's disease, for example, can only be definitively established by brain autopsy. We will consider the analysis of neuropathological outcomes at a particular time point in a cohort study. Because of potential differential selection into autopsy, estimation based on autopsied subjects may be biased relative to the true parameter values in the original population. We will review standard approaches to adjusting for selection bias in epidemiological studies and discuss how these methods have been proposed for use in neuropathology studies of dementia. We will emphasize the assumptions that underlie any correction for selection bias, including whether estimates can reflect members of the cohort who are still alive.