ACTRIMS 2026 | The phenomenon of pseudo-atrophy when using brain volume measured by MRI in MS trials

The other thing I showed just using a data set that I had access to, and I was trying to encourage other people to do the same, but I think the pseudo-atrophy that is there, this post-baseline measure, if you look at the change from baseline, is more highly correlated to GAD lesions, which is perfectly understandable. A GAD lesion is telling you that there’s inflammation in the brain, is more highly correlated with relapses than it is with progression. And the association between that pseudo-atrophy and what are called T2 lesions, which are sort of the remnants of damage, not the current active damage that you see with the GAD lesions, that pseudo-atrophy is less correlated with that overall damage, and that overall damage is more correlated with progression. So it’s not really measuring the endpoint that has become more important. As we’ve found more and more treatments using anti-inflammatories in different pathways to control MS and MS relapses, the focus has become more on progressive MS, which is more of the confirmed disability worsening rather than the relapses that can be predicted. So in that sense, it provided usefulness maybe in the earlier days, but it’s losing its usefulness because we are now trying to really focus on progression. And I would add that studies are generally done over a short term in the life of MS, maybe two years, maybe three years. And that just may not be enough time for brain atrophy to show its face. It changes very, very slowly. So it may not be a bad outcome if you’re studying something for 10 or 20 years. But in terms of what we need from a clinical point of view, we need something that we can monitor the patient while they’re going through treatment, not in decades of their life. So the main problem that I see on the reproducibility of the MRI, not the technique itself, if you put somebody in a machine and don’t move them, do an MRI, and then 30 minutes later do another one, they’re going to agree, the problem is over time, like six months apart, they don’t agree because of variability in measurement. And if you look at the distribution of the change from baseline to say six months, this re-baselining and pseudo-atrophy, the distribution of those changes looks fairly normal. And when statisticians see normal distributions, we often think of measurement error. And when you see this and think about it as measurement error, then you should say, well, if the general decline in brain and brain atrophy occurs, even in normal, healthy individuals, we all lose brain volume over time, some of us faster than others, but it is a very small percentage of your brain that’s lost on an annual basis or a semi-annual basis. And when you look at this, these distributions in the pseudo-atrophy range, what you see is about, in the data set I looked at, it was about 35%, and I’ve seen other data sets with similar percents of people who have increases in brain volume. And that’s unusual, if not unexpected. And that tells me that it’s very difficult to then define using brain atrophy, who responds to therapy. So we don’t have a definition of what a responder is. So at the individual level, we really can’t say much about whether it’s working or not. And some of the reasons for that lack of precision is, again, not the MRI machine, although machines and the amount of brain atrophy vary from machine to machine. It might even vary within machine if they change some parameters of the machine or some of the hardware on the machine or software on the machine. But it’s also how much water the person has, if you measure somebody in the morning and they’re very well hydrated, you might actually get a different reading than if you measured them in the afternoon after they spent the day outside on a hot day. So the amount of atrophy is somehow related to the amount of water in the brain and other things that can affect that, how they were positioned when they went into the MRI machine, even though people are, quote, registered so that the brains are aligned, there can be minor differences that when you’re measuring something like the loss in brain volume, which is a very small fraction of your brain, even small deviations can make a lot of difference in that. So that’s the first major problem that I see with brain volume, is we just don’t have a good tool at the individual level. And usually, the FDA wants to have a clinically meaningful outcome whenever you use a continuous outcome, they want to know that, is that change clinically meaningful? And this makes it very difficult to define a clinically meaningful change. We can find changes that are so big that we know something bad is going on, but that’s pretty rare.

This transcript is AI-generated. While we strive for accuracy, please verify this copy with the video.