A moving target: How reliable are dementia assessments?

This post is from ScIU’s archives. It was originally published by Emily Byers in February 2017 and has been lightly edited to reflect current events.

“We’re going to do a few tests to see whether your mother is showing typical signs of dementia.” The word conjures chilling images of loved ones’ lives reduced to confusion and fear as memories and independence slip away. While loss of physical independence is unfortunate, it can be more devastating to lose a loved one’s verbal companionship. Dementia reduces one’s ability to name objects, people, or recollect specific memories. As a result, dementia sufferers use language less and become removed from conversations happening around them.

As scientists, our goal is to ascertain what can be done to slow the progress of dementia and mitigate its symptoms. This takes the cooperation from many teams to break down the individual symptoms and see how these symptoms respond to treatment. In Speech and Hearing Sciences, our focus is on preserving language faculties in patients with dementia so that they can communicate effectively for as long as possible. Patients’ progress (or decline) is measured at various intervals to see how well selected treatments are working. For many years these assessments were primarily behavioral in nature. For example, therapists might count how many words a patient with dementia could name in one minute. Behavioral tests work well because they are inexpensive and easy to administer; however, patients’ performance may vary considerably from day to day based on fatigue, emotional state, or other factors.

More recently, there has been a movement to supplement behavioral studies with neuroimaging techniques to observe how brains process and select language in patients with dementia. Specifically, scientists want to see if certain language tasks activate particular brain areas. Functional magnetic imaging techniques (fMRI) produce images of the brain while a patient performs a language task (e.g. word naming). This allows scientists to look for relationships between the type of task and the brain region that is activated. Ideally, combining imaging and behavioral assessments will help us correlate the physical progress of the disease with decline in communicative skill–thereby enabling us to target better treatments as the disease progresses.

Because neuroimaging is expensive and requires access to medical equipment, neuroimaging studies typically test subjects only once. This deviates from normal scientific protocol–namely, the ability to replicate results gives us confidence in our measurements. In addition, studies that perform “pre-test” and “post-test” scans on patients operate on the assumption that a patient’s scans don’t vary widely from day to day. But how reliable is an fMRI scan? Like behavioral assessments, can language activation in the brain vary day to day? Assuming wide-scale variability doesn’t exist can be problematic when the conclusions have medical implications for treatment of progressive disease.

The notion that brain imaging scans may not have high standards of test-retest reliability is an unpleasant proposition. However, it is information that must be known. To answer this question, Eun Jin Paek, who graduated from the Department of Speech & Hearing Sciences at Indiana University and is now at the University of Tennessee, investigates test-retest reliability of fMRI images taken from patients with dementia across multiple testing sessions. Using a word-retrieval task (e.g. object naming, action labelling, etc.) she is working to establish the validity and reliability of fMRI measurements taken from the same person at various time intervals. Paek will be among the first generation of scientists to empirically determine whether imaging scans vary widely in the same patient across different testing sessions.

Lastly, Paek’s studies investigate whether people with dementia show more variability across testing sessions than we might observe in healthy adults. We do not yet know if progressive diseases such as dementia cause more variability between brain scans than one would observe in healthy participants. Obtaining physiological evidence to discover if dementia is a reliable, steady decline or a variable series of ups and downs in performance is particularly important for designing new treatments. If certain language skills appear to have more stable activation areas in the brain, then Paek’s research will establish a starting point for treating language skills in progressive decline. Understanding the limitations of the role that neuroimaging can play in designing language-based treatments will help us develop better understanding of how to combat dementia in our aging population.

Edited by Benjamin E. Draper and Noah Zarr