Digital Biomarkers and Alzheimer’s Research: Q&A With Liam Kaufman, VP of Clinical Sciences at Cambridge Cognition

Liam Kaufman
VP of clinical sciences
Cambridge Cognition

Researching and developing treatments for Alzheimer’s disease comes with unique challenges, due to the nature of the disease. Liam Kaufman, VP of clinical sciences at Cambridge Cognition, spoke with Pharmaceutical Executive about how digital biomarkers are being used to improve this process.

Pharmaceutical Executive: What are some of the difficulties measuring the impact of Alzheimer’s Disease treatments?
Liam Kaufman: The difficulty is that the traditional tools in Alzheimer’s research tend to be subjective and time consuming. These are pen and paper tests that have been in use for decades now. They often have very subjective components or they have components that are difficult to administer and people make mistakes. A typical mistake could be prompting someone too many times or giving subtle clues. It can also be difficult to score some of these assessments.

A test may ask to rate a patient’s word finding abilities on a scale of one to five. The person administering the assessment must think back over the past 15 to 30 minutes and decide whether the person is a three or four out of five. That’s just one example of some of the subjectivity.

There’s a lot of variability between people and the symptoms can be subtle early on.

PE: Can you discuss how biomarkers are having an impact on this?
Kaufman: There was a realization several years ago about these existing end points. There was a larger push to use newer technology to objectively measure some of these symptoms. There’s a number of different tools that people have provided.

Pharma’s a risk adverse industry, so it doesn’t change overnight. Over time, more evidence has been created to support these tools, so there’s been a shift from being concerned about putting digital biomarkers into clinical trials to greater widespread acceptance of these tools. Over time, you’ll see greater integration and eventually they’ll become part of a primary or secondary endpoint. Although now they are primarily used as experimental.

PE: Can you discuss how the technology behind these biomarkers works?
Kaufman: When people have something like Alzheimer’s disease, they have word finding difficulties. They’ll pause for a longer period of time, they’ll have trouble thinking of the correct words. They might use more pronouns instead of actual names or proper nouns. They’ll typically use simpler words. We measure those changes subjectively.

One of the things we do is to get people to describe a picture or illustration, which elicits speech. People describe a picture for about 45 seconds to a minute. Then we create a transcript, which use with a raw audio recording of them. We use natural language processing and large language models on the transcript to generate linguistic, semantic, and grammatical variables. We also run different kinds of analysis on the acoustics as well. That leaves us with about 800 different variables. We can then use each variable on its own to quantify changes, or we can use machine learning and AI to combine the variables to create simple measure. We can even create binary classifiers.

PE: What are some of the advantages that digital biomarkers bring to drug development?
Kaufman: One is objectivity. That removes some of the error you get from people administering these tests differently. It removes the variables in scoring these tests. There’s also things that computers can pick up that humans cannot, such as certain acoustic variables.

The other thing is the frequency of testing. Traditional tests can be a burden and are only done in a clinical setting. In the trial, a test might be administered every three months. With digital biomarkers, there’s an opportunity to self-administer them once a month, or even more frequently. Then you can get a better sense of how someone changes over time and potentially reduce the number of participants or the length of a clinical trial.