A USC Leonard Davis School of Gerontology study on Alzheimer’s screening methods indicates that swapping costly imaging for more accessible blood tests is a trade-off that calls for larger sample sizes and more caution against bias.
USC Leonard Davis Assistant Professor Eleanor Hayes-Larson was the senior author of the study, which was highlighted as an editor’s choice by the Journal of Gerontology, Series A: Medical Sciences.
In a LinkedIn post about the Editor’s Choice honor, the journal’s editor-in-chief, Professor Joyce Siette of Western Sydney University, called the article a “thoughtful, timely, and highly relevant contribution for researchers working in dementia, epidemiology, neuroimaging, and biomarker science, especially as our field moves toward more scalable and accessible approaches to Alzheimer’s disease research.”
USC Leonard Davis Assistant Professor Eleanor Hayes-Larson
Q&A with Assistant Professor Eleanor Hayes-Larson
This study compares two ways of screening for signs of Alzheimer’s disease: a brain scan and a blood test. What exactly are each of these tests measuring?
Hayes-Larson: “The brain scan looks directly for amyloid plaques in the brain, one of the hallmark signs of Alzheimer’s disease. Participants receive a small amount of a radioactive tracer that sticks to amyloid, making it visible on a [Positron Emission Tomography] PET scan. The blood test measures phosphorylated tau, or p-tau, a related Alzheimer’s-linked protein in blood. It is much easier to collect, but it is an indirect and noisier proxy for brain amyloid.”
How are these tests different for study subjects and for the researchers administering them?
Hayes-Larson: “PET scans are time- and resource-intensive and logistically complex. They require administration of the radioactive tracer, which carries a low dose exposure to ionizing radiation, followed by the scan itself, which typically occurs with a patient lying on a table that is slid into a scanner at a clinic. Study subjects also need an additional MRI or CT scan to map the brain more precisely and layer with the PET scan images. There are also complications from the administrative perspective about timing of the preparation of the radioactive tracers, which are only made by certain labs and decay over time, imposing geographic constraints.
In contrast, obtaining measures of blood-based biomarkers is much more straightforward. From a study subject perspective, it is identical to the routine blood draws most of us are used to at the doctor’s office. For research, these blood draws can be safely conducted almost anywhere, meaning they can reach participants who wouldn’t be able to get to a PET scan location due to geographic or mobility reasons. Being low-risk and easy to collect, combined with their much lower cost, make them extremely appealing for use in large population-based studies. However, they are ‘noisier,’ or more subject to error in measurements, which can lead to the problems we saw regarding power and bias.”
The study found that using blood-based measures instead of PET scans reduced the statistical “power” of the results. In lay terms, what does power mean and why is it important?
Hayes-Larson: “Power is the probability or chance that a study detects an effect when it really exists. As an example, if a particular exposure truly increases risk for Alzheimer’s pathology, a study with 90% power has a 90% chance of finding that the exposure increases risk. Conversely, 10% of the time, it will fail to detect the effect and wrongly conclude that the exposure has no significant effect. We want studies to be well powered (power of 80% or 90% are commonly used thresholds) so that we don’t miss important factors related to AD risk.”
Why does increasing the sample size help offset this reduction in power?
Hayes-Larson: “A study’s power depends on multiple things, and sample size is one of them. With more people in a sample, our estimates are more precise. When they are more precise, it is easier to conclude that an observed effect is statistically different from ‘no effect’.”
What are some other challenges/biases that arose when swapping PET scans for blood biomarkers?
Hayes-Larson: “When we studied how blood biomarkers relate to cognitive function, we found that the noisiness of blood biomarkers can make the connection between biomarkers and cognition look weaker than it really is. In other words, the results may underestimate the true relationship, not just make the estimates less precise.
Bigger picture: the blood biomarkers are not measuring the AD pathology burden directly in the brain. As a result, they are affected by other things such as liver and kidney function, and the literature is still evolving on the most appropriate ways to handle these factors.”
How might these results affect how Alzheimer’s researchers design future studies – does it give them additional factors to consider?
Hayes-Larson: “I think it shows there are reasons researchers might want to consider collecting both types of measures on at least a subset of the study sample. The blood biomarkers are great for the reasons we discussed in terms of costs and logistics, but having even a small group where there is also a PET scan might be useful for validating or correcting the blood biomarker findings later.”
What’s next for this avenue of your research? Does this study prompt new questions or goals?
Hayes-Larson: “I’m working on an analysis of how these challenges affect studies trying to understand whether exposures that reduce cognition do so by influencing AD pathology burden. This is called a mediation analysis, and similar problems to those we describe may result in us missing or underestimating the ways that exposures become biologically embedded.”
The study was supported by National Institute on Aging grants R00AG073454, R01AG082730, R00AG075317, R13AG064971, and R01AG072681. Hayes-Larson’s coauthors included first author Sarah F. Ackley of Brown University; Renaud La Joie and Michelle Caunca of the University of California, San Francisco; and Shubhabrata Mukherjee, Seo-Eun Choi, Emily H. Trittschuh, and Paul K. Crane of the University of Washington.
