YT YouTube2Text

YouTube2Text → Transcripts → 9/19/25 - Biomarker Clocks in Alzheimer’s Disease: Predicting Dementia Risk and Timing

9/19/25 - Biomarker Clocks in Alzheimer’s Disease: Predicting Dementia Risk and Timing — Transcript

by UW Department of Medicine · 10,015 words · 1,481 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00All right. Well, we are at 8:00 am, so
0:03why don't we get started um and
0:06delighted to have uh Dr. Steve Barzy,
0:10our uh interimm division chief for
0:12geriatrics and gerontology, introduce
0:15Dr. Toby Bet Bedhouser, another faculty
0:18member in geriatrics and gerontology.
0:21So, Steve,
0:23>> thanks Len and welcome everyone to
0:25department of medicine grand rounds for
0:27today. I'm excited really to be able to
0:30um to talk a little bit about our uh my
0:33wonderful colleague and um
0:36neuroscientist
0:37um who uh has been within our group now
0:40for some time. Um uh so Dr. Toby Bhauser
0:45is a medical physicist. He is a
0:48neuroscientist and as I learned earlier
0:50today previously an audio engineer
0:54um in a a different phase of his life
0:58and uh he today will be talking about I
1:01think some really pivotal work he's done
1:04um but to start he has a background uh
1:08in medical physics uh where he had he
1:10had done his undergraduate training in
1:12at Sunni Platsburg and then uh joined us
1:16um here at the university where he has
1:20um served in various capacities but has
1:22worked with Dr. Sterling Johnson who
1:24most all of you are well aware uh and
1:27really has uh done some some pivotal
1:30work pioneering work I'll say in the
1:32area of biomarker modeling um not only
1:36is he a wonderful scientist however
1:38though he is also a a really amazing
1:40colleague who uh collaborates with many
1:43of those both within the Alzheimer's
1:45disease research center but also really
1:48across campus and across the country. um
1:51and he's going to be sharing a little
1:53bit about, you know, we've all changed
1:55our frame of reference with regards to
1:58how we think about dementia now much
2:00more so as a chronic illness that has
2:02decades of leadin. And for us to
2:05understand what that trajectory is for
2:08the illness early or upstream and how we
2:11can then influence risk factors and and
2:14better understand the illness is going
2:16to be really pivotal as we move the
2:18needle on managing persons living with
2:22memory illness. And so without further
2:24ado, Toby, thank you for being here
2:26today and uh please share on biomarker
2:30clocks in Alzheimer's predicting
2:32dementia risk and timing.
2:35>> Thanks Steve and thanks uh for the
2:38invitation to present today. Um so I
2:41wanted to just show conflicts of
2:42interest. Um I will be discussing some
2:45uh drugs in particular pet tracers that
2:48are not yet um FDA approved. Um, and
2:51I'll be discussing some topics that are
2:53kind of at the razor's edge of uh,
2:55research and um, so these are some
2:57things that um, probably not quite into
2:59the clinical standard of care. Uh, for
3:03learning objectives today um, we kind of
3:04have our standard objective of
3:06understanding some current updates um,
3:08and how to apply these things in
3:10clinical practice and research. Um but
3:12more specifically I want to go over how
3:15uh we can interpret amaloid and tow PET
3:17imaging as well as plasma phosphat 217
3:20uh in a longitudinal and dementia
3:22progression context. Uh and then also
3:25give you some information about kind of
3:26the typical progression of these
3:28biomarkers um how these things can be a
3:30little heterogene heterogeneous between
3:33uh individuals as well as how these
3:35things can differ between some
3:36populations.
3:38Um, so before I dive into all of that, I
3:40just want to briefly acknowledge um the
3:43many people that contribute to these
3:45studies. Um, anybody that's been in this
3:47field for more than about a month knows
3:49that this takes a massive amount of
3:51resources. Um, not just funding, but um,
3:54staff and um, a number of investigators.
3:56And we're in we're very fortunate at
3:58Wisconsin to have a very collaborative
4:01and large investigator team uh, and a
4:03lot of staff and other leaders that uh,
4:05really drive this research. And of
4:07course importantly uh perhaps most
4:09importantly uh the research participants
4:12um and the work I'm presenting today uh
4:14is several thousand uh participants
4:16across multiple studies.
4:19And additionally I just also want to
4:21disclose um that I received research
4:22funding from NIH and I've received
4:24travel support from the Alzheimer's
4:26Association NIH and I'm a consultant for
4:28the Harvard Aging Brain Study.
4:31Um so uh I also wanted to give just a a
4:34brief background of my connection to
4:35dementia. So, this photo was taken in
4:372010. On the left is my late
4:40grandfather, Grandpa Lyle, and my
4:42grandma June. Um, both of whom developed
4:45um clinical impairment over the course
4:48of of their lifetime. Um, my grandma's
4:50was probably more vascular dementia. My
4:52grandfather was diagnosed with
4:53Alzheimer's disease in 2013 um and died
4:56about nine years later. So, um, you
4:59know, when I see this photograph as a
5:01dementia researcher now, uh, the thing
5:03that I wonder is when did this disease
5:05start for him? Um, and knowing what I
5:08know now, um, it's quite likely that
5:11unbeknownst to us at the time that this
5:13photo was taken, it's quite likely that
5:15he had, uh, proteins developing in his
5:17brain of Alzheimer's disease at that
5:19time. And this really drives a lot of my
5:21motivation to understand um the timing
5:24of pathological events in Alzheimer's
5:26disease and how these influence um
5:28symptoms that we see. Uh and then also
5:30just kind of a brief overview of of
5:32Alzheimer's disease and where we're at
5:34today. Um so these are facts and figures
5:36that the Alzheimer's Association pub
5:38publishes annually. And it'll be no
5:41surprise to anybody in this audience
5:42that um Alzheimer's disease and dementia
5:45is a growing uh public health uh
5:47concern. Um these are uh kind of where
5:50we're at today in 2020 with about 6 uh
5:53one million uh people living with um
5:56Alzheimer's disease. Uh and this is
5:58projected to increase over the coming
6:00decades um and become uh especially
6:03amongst those uh in the older
6:05demographic in our population.
6:08Uh along with uh this increased uh
6:11number of people with dementia, there's
6:12a substantial financial uh uh drain uh
6:17to uh care for people with Alzheimer's
6:19and dementia. Um estimated to be $384
6:22billion in 2025. Uh importantly, about a
6:26quarter of these costs um come out of
6:28pocket. Um and this also doesn't include
6:31the estimated almost 12 million unpaid
6:34caregivers. Again, anybody that's had um
6:36any relationship with people with
6:38dementia know that this doesn't just
6:40affect that person, but really their
6:41entire uh care uh support team um
6:45including family, friends, and others.
6:46So, um and specifically in Wisconsin, uh
6:49we uh it's estimated that there are
6:51about 205,000
6:53uh uh unpaid uh caregivers um caring for
6:56people with um Alzheimer's and dementia.
7:01um when compared to some other uh causes
7:04of death um one of the things we've seen
7:06over the past 20 years is that uh the uh
7:09death due to Alzheimer's disease and
7:11associated causes um has steadily uh
7:14increased over the this time span um
7:16whereas other uh leading causes of death
7:19um have remained relatively flat or have
7:21seen marginal increases or even
7:22decreases in some uh some cases.
7:26Uh when we focus on the disease
7:27pathology um the disease
7:30neuropathologically is defined by
7:32aggregated beta amalloid plaques which
7:34are in the extracellular space and
7:37neuropibrillary tangles composed of tow
7:39protein aggregated tow protein and these
7:42are within uh the neurons
7:45and um again just kind of going over
7:47some of the basics uh we know that
7:49increasing number of uh tangles and
7:52nuritic plaques and isocortex
7:55um are associated ated with u more uh
7:57advanced cognitive decline. This is a
8:00figure from Peter Nelson's paper from
8:012012. Um and then we also see um
8:04increased levels of atrophy, brain
8:06atrophy in those with um with AD
8:08pathology.
8:10I also just want to briefly go over to
8:12pathology because it's relevant to the
8:13biomarkers we're going to discuss. The
8:15the idea is that um the tow protein has
8:18a normal role in stabilizing
8:20microtubules
8:21um but in disease becomes
8:23hyperphosphorilated which leads to
8:25microtubial destabilization.
8:28This ultimately forms these um soluble
8:31protoibbrals and then later this leads
8:33to aggregation of paired helilical
8:35filaments which are insoluble um and
8:38then later goes on to form these
8:39neuropibrillary tangles. This is
8:41important as we get into understanding
8:43some of the current biomarkers.
8:45And when we think about um tow spatial
8:48progression um we see this kind of
8:50stereotypical pattern wherein um tow uh
8:54aggregates start really in the medial
8:56temporal lobe particularly in entrancal
8:58cortex and hippocampus and then spread
9:00kind of outwardly into the temporal
9:02cortex and larger neoortex um in
9:04advanced stages. Um
9:08so shifting to biomarkers um we're
9:10fortunate in the past 20 or so years
9:12that there's been great developments in
9:14biomarkers. So we can measure amaloid
9:16and tow pathology in vivo uh using PET
9:19imaging. So on the top showing examples
9:21of amaloid PET uh scans in an amaloid
9:24negative and an amaloid positive case.
9:26On bottom showing uh tow images for a
9:29tow negative and a tow positive case. Um
9:31and uh we've also been able to measure
9:33these in cerebral spinal fluid. And more
9:35recently there's been a lot of
9:36development in um the blood biomarker
9:39space. Um and I think the thing to keep
9:41important in your mind here is that
9:44these are really kind of macroscopic
9:46measures of really like microscopic
9:48detail, right? So um there's always
9:50going to be some detection sensitivities
9:52and other things involved in these. Um
9:54and it's uh there's also some
9:57pathological inference that we can make.
9:58Um but it sometimes we don't always know
10:00what's driving these biomarker signals.
10:04when we think of the typical biomarker
10:06progression in Alzheimer's disease um
10:08this was made famous by um Cliff Jack um
10:11so the idea is that we can uh detect
10:14levels of abnormal amaloid first uh in
10:16in typical disease followed by tow
10:19positivity uh neurodeenerative changes
10:22and then um later we see clinical and
10:24cognitive impairment
10:26and we largely break this disease into
10:28kind of these two phases uh the
10:30pre-clinical phase meaning uh before
10:32clinical symptoms Um and this is the
10:35phase where we can measure abnormal
10:36levels of these biomarkers uh reflecting
10:39amaloid and tow pathology. Um and uh but
10:43we don't yet have uh meet the bar for
10:45clinical uh impairment versus the
10:48clinical phase being um we can now
10:50detect clinical uh symptoms within
10:52patients.
10:55And uh these uh research over the past
10:5920 years has led to kind of these
11:00evolving frameworks of how we think
11:02about diagnosing Alzheimer's disease. So
11:05um this was just updated again um last
11:07year uh which is part of an Alzheimer's
11:09Association working group with um input
11:12from the um NIA and uh we largely now
11:16kind of focus on this rubric for
11:17defining AD in terms of a clinical stage
11:20and a biological stage. Um so just to
11:22kind of go over this rubric and how it's
11:24applied um across kind of the the
11:27horizontal axis here. Um we have
11:29clinical stages. Um stage zero I'll I'll
11:32describe in a moment. But uh clinical
11:34stages one and two are essentially in
11:35the pre-clinical phase where um stage
11:39two is kind of a transitional phase
11:41where people are maybe starting to
11:42notice some subjective cognitive
11:44complaints and mild cognitive um
11:47symptoms. But these don't yet let uh
11:49meet the level for clinical impairment.
11:52Whereas uh clinical stages three uh
11:54through six are kind of advancing into
11:57um clinically detectable levels of
11:59impairment and um and progressive um
12:02dementia. Um and then on the vertical
12:05axis uh is um defining biological stages
12:09uh based on amalloid and ta pathology.
12:11So um the idea is that stage A uh and
12:15and the diagnosis of Alzheimer's disease
12:17can be made by a positive amalloid
12:19biioarker. Um I don't have I could give
12:21a whole like hourong lecture on just
12:23this uh framework but um this is kind of
12:26a little bit of a simplification but um
12:28the idea is that people will be amaloid
12:30positive but tow negative in this um
12:33early preclinical phase and uh that
12:36towel will continue to spread uh again
12:38beginning in the medial temporal lobe
12:39and then out into the greater neoortex
12:42and these represent the more advanced
12:44biological stages.
12:46And the way that this rubric is applied
12:49is that anybody that goes along this
12:50diagonal is thought to follow kind of
12:52the stereotypical progression of
12:54Alzheimer's disease. So um we expect
12:56kind of these biological core stage
12:58biological stages on the left to
13:00correspond to the clinical stages on the
13:02top. Um but it also allows for people to
13:05have more advanced clinical stage for
13:07their given biomarker stage. Um and uh
13:11the opposite people can have a less
13:13advanced clinical stage given their
13:15biological stage. So it allows us to um
13:17kind of characterize uh people that have
13:20more advanced progression given their
13:22biology or uh their their biological
13:24stage um which is a key thing that we're
13:27trying to understand as a field as to
13:28why one person goes through an
13:30accelerated path whereas maybe other
13:32people are more resilient to pathology
13:36in terms of uh drug approvals. Um it's
13:40been a very exciting time to be in the
13:41field. Um the past couple of years we've
13:44seen um two different drugs approved for
13:47anti-amalid therapy. Um and uh kind of
13:51the notable difference in in these drugs
13:53versus what's been previously approved
13:55is that these um seem to actually slow
13:57the progression of disease rather than
13:59just treating symptoms. Um the other
14:01pattern kind of in clinical trials over
14:03the past um 10 to 15 years is kind of
14:05moving earlier and earlier in disease to
14:08um to treat and what we can treat. So um
14:10we still do not have any approved drugs
14:12for uh the pre-clinical phase of AAD
14:14although there are ongoing clinical
14:16trials um and just to kind of summarize
14:18the results from the um daanimab and
14:21lacanab trials. um silkanab on the left,
14:24danab on the right. Uh so each of these
14:27is showing amaloid burden measured in a
14:30unit called centaloid and centaloid
14:33essentially is a standardized unit
14:34across all amaloid tracers where um zero
14:38effectively represents the average of a
14:40young control group thought to be
14:42amaloid negative and 100 is kind of the
14:44typical level of amaloid that a dementia
14:48group would have. Um most research will
14:50say a positivity threshold around um 20
14:53to 25 centloids is is reasonable. And
14:56what you can see from these trials is
14:58that um both of these monoconal
15:00antibodies targeting beta amalloid
15:02effectively remove plaques over the uh
15:05over the scope of the trial. Um and uh
15:08and both of these drugs met their
15:11primary endpoints uh in terms of a
15:13slowing of um of clinical uh progression
15:17based on um the clinical dementia
15:19grading some of boxes. Um, and another
15:23factor that um came out of the Dynamanab
15:26trial was that it appeared that people
15:27that had low to medium tow um had a
15:30slightly slightly greater treatment
15:32effect compared to when they combined
15:34the entire population uh in the in the
15:36treatment arm. So, kind of some
15:38indication that um again kind of
15:40treating earlier um in disease is is
15:43likely beneficial.
15:46Uh and then this past summer this was
15:47updated. Um I couldn't find a nicer
15:50graphic unfortunately but um in a study
15:54from ESI they presented at the
15:56Alzheimer's Association International
15:58Meeting uh where they took uh
16:00participants from the Alzheimer's
16:02disease neuroiming initiative as well as
16:04the biofinder study uh and matched these
16:07two participants that were in the
16:09lacanab trial. Uh and the idea was they
16:12wanted to see if um people that were on
16:15the lacanab treatment arm um continued
16:18to show uh a clinical benefit um
16:21extending well beyond the the trial. Um
16:24the the caveat with these results is you
16:26know we don't have an actual control
16:27group anymore in this study. So um we're
16:30kind of extrapolating from observational
16:32cohort studies what um we think that
16:34group those groups would have done. um
16:35but still uh somewhat promising that um
16:37we think that these um drugs may have a
16:40lasting effect um even after the the the
16:42initial treatment um window.
16:45So this kind of brings us to to current
16:47day um where we're at which is um we
16:49really need to understand um when in
16:51disease uh to treat uh especially with
16:54these anti-amaloid therapies um to
16:56maximize the effectiveness of these
16:58treatments um and uh I would say the
17:01field in general is moving towards a
17:03more personalized approach to medicine
17:04which um really shouldn't be surprising
17:06to anybody um particularly if you look
17:08at oncology um where uh the more the
17:11better tests that we have the more
17:13predictive we can figure out those tests
17:15tell us about disease progression um the
17:17better we can personalize our uh uh
17:19treatment approaches and and prognosis
17:22um and in order to do this um I would
17:24argue that we really need to understand
17:26the things that accelerate and slow um
17:28these biological and clinical
17:30progressions.
17:32So um my lab um I started in the middle
17:35of COVID um about a week from when Lynn
17:37started um as chair of the department.
17:41Um so, uh it was an interesting time to
17:43start a lab, but I'm I'm very excited
17:45about the work that we're doing. Um the
17:47overall goal of our lab is really to
17:49develop accurate, generalizable and
17:51clinically relevant tools um that we can
17:53share uh in an open- source way um that
17:56will really inform um AD prognosis and
17:59allow us to take these biomarker results
18:02um see what replicates across multiple
18:04different cohorts and develop um risk
18:07models that are tailored to um different
18:09population groups. Uh in addition um
18:12we're really fortunate to have the uh
18:14resources and study cohorts that we can
18:16also um leverage some post-mortem
18:19validation studies where we can look at
18:21people uh that have had um either PET
18:23imaging or uh fluid biomarkers before
18:26they died and come to our brain donor
18:27program. And this really helps us anchor
18:30and understand why we see particular
18:32longitudinal patterns in our biomicro
18:34data.
18:35So, um, getting more specific into this
18:37idea of how we study time and and
18:39contextualizing this, um, it's kind of
18:42been established for for over a decade
18:44that, um, this pre-clinical phase of
18:46Alzheimer's disease is thought to last
18:4820 or more years. Um, and to put this
18:50into context with some of the
18:52developments here, I'm showing uh, based
18:55timeline of of FDA approvals and then on
18:58the top um, showing kind of when certain
19:00discoveries happen. So for example,
19:02amaloid pet imaging has only been around
19:05uh in humans since about 2004. Um so
19:08again if we think about the fact that
19:09this disease has this extended 20-year
19:12period of a preclinical phase um this
19:15really poses a challenge for studying
19:16Alzheimer's disease because even had we
19:18started massive cohorts the day that uh
19:21amalid pet became available with bib um
19:24we still would potentially just be
19:26catching people converting from uh the
19:28preclinical to clinical phase.
19:31Um, and to further contextualize this,
19:34if we look at the discovery of TAPED in
19:36humans, this has really only been around
19:38for a little over 10 years. So again, if
19:40we're trying to understand this full uh
19:4220 plus year disease timeline, um, we
19:45even haven't really even had the tools
19:46to do it um to be quite honest. Um and I
19:49will mention um you know if we look at
19:52from time from when these things are
19:54discovered to FDA approvals we also see
19:55kind of you know this offset um here
19:58with amaloid pet with taet and then more
20:00recently um just this year with some of
20:02the first blood tests being FDA approved
20:07now uh contextualizing this further if
20:10we look at our like observational cohort
20:12studies what we see is that we don't
20:14typically have even more than about 10
20:17to 15 years of longit longitudinal data,
20:19right? So, um this is kind of what
20:21facilitates the needs for these methods
20:23to understand this prolonged period of
20:25disease. And really the idea is that we
20:27want to take these longitudinal
20:29observations and turn this into some
20:30biomarkerbased timeline. And so that
20:33kind of brings up this idea of what is a
20:35biomarker clock? Um and kind of an
20:38oversimplified version is it allows us
20:39to take these agebased longitudinal
20:42measures and recast disease into a
20:46biomarker specific timeline. based on
20:48how long that biioarker has been
20:50abnormal. Um, and I would say the bulk
20:52of this work in our field has been done
20:54in um, amaloid pet imaging. Um, there's
20:58been several methods proposed to do this
20:59over the past decade or so. Um, and um,
21:04largely these models show very
21:06consistent things which is that you can
21:07use multiple different methods to get at
21:09u this idea that there's kind of a
21:12normal pattern of of accumulation once
21:14positivity occurs.
21:16Um and the method that I developed is
21:17called sampled iterative local
21:19approximation. And I just will go
21:21through kind of a toy example here of
21:23how this works. So um on the left we
21:26have our observed um amaloid versus age
21:28data. This is data from the Wisconsin
21:30registry for Alzheimer's prevention. Um
21:33we are able to apply the sila algorithm
21:35which outputs a curve which tells us how
21:37amaloid changes over time. And then we
21:40set the zero point of this to be at some
21:43um level in which is indicative of an
21:46abnormal level. So we you know call this
21:48amaloid positivity or positivity
21:50threshold. And then in uh once we have
21:52this model trained we can apply this to
21:55individual level data. So um in this
21:57case I'm showing uh this is an
21:59individual scan from a single
22:01participant. Um so here shown on the
22:03agebased timeline, here shown on the
22:05amaloid based timeline and we're able to
22:08take that value of amaloid at that time,
22:11project that down to figure out how long
22:14they've been amaloid positive and then
22:16uh by subtracting that from their age at
22:18that amaloid scan, we can estimate when
22:21amaloid positivity occurred in this
22:23individual. So the really um kind of
22:26neat thing about these methods is that
22:28um while the models themselves are
22:30trained on longitudinal data, we really
22:32only need one observation to be able to
22:34estimate when um when people became
22:36biomarker positive.
22:38Um and in the first study that we had on
22:41this um we had several um aims. One was
22:43just to compare multiple different
22:45methods of doing this um and largely saw
22:48that um this was consistent. We tested
22:50this in three different cohorts. Um so
22:52again ad uh the Alzheimer's disease
22:54neuroim I imaging initiative um the
22:56Baltimore uh longitudinal study on aging
22:58which is BLSA and then the Wisconsin
23:00registry for Alzheimer's prevention and
23:04uh amongst the things that we did to
23:05validate these methods we were able to
23:07show that um we can estimate um
23:10anticedant um amaloid levels with high
23:13accuracy uh in people that were observed
23:15to convert from negative to positive. we
23:18were able to show that we can estimate
23:20um that onset age usually within about a
23:22year or two and um importantly for these
23:25models um in all three of these cohorts
23:28we didn't see an effect of um APOE um or
23:31sex in terms of um how fast people
23:33accumulate amaloid um and in ADNE where
23:35we had enough uh sample to look at
23:37dementia this also was not associated
23:39with differences in amaloid accumulation
23:41trajectories
23:43uh applying these methods then we
23:44started to look at factors that
23:46associate with when amaloid positivity
23:48occurs. Um not surprisingly we found an
23:50effective APOE genotype. So um these are
23:53shown on the left here for individuals
23:55that became amaloid positive. These are
23:57the ages that they became amaloid
23:59positive on the left um organized by apo
24:02genotype um kind of going from least
24:04risk with um E2, E3 and E2 uh E2
24:09carriers um to E3 homozygot and then to
24:12E4 carriers and homozygot. Um so uh we
24:15see this expected pattern where people
24:17with um increasing number of AP4 alles
24:21um have an earlier amaloid onset uh age.
24:24But something that jumped out in the
24:25results as we were looking at this was
24:27the wide variability in when people
24:29become amaloid positive even within say
24:32uh apo genotypes. Um so here we're
24:35seeing roughly a 40year span in which um
24:38people that are E4 homozygot become
24:41amoid positive.
24:43Uh, another thing that these methods
24:45allow us to do is to study the time from
24:46amaloid positivity to dementia and
24:49things that that um speed up or slow
24:51down um that timeline. So really getting
24:53at um things that accelerate or or slow
24:56the preclinical um time frame. So uh in
24:59this study this is um ADNE data again
25:01where we had enough participants convert
25:02to dementia during the course of study.
25:04Uh we saw that um u apo4 carriage uh an
25:10older age at amaloid onset and um and
25:13females had a uh earlier uh transition
25:17from amaloid positivity to dementia. Um
25:20and notably when we kind of compare
25:21these effects uh across these different
25:23variables, we can see in particular
25:25people that became amaloid positive um
25:28kind of in in their 70s and early 80s um
25:31had a about a four-year shorter time
25:34frame from amaloid positivity to
25:35dementia um which contrast with um the
25:38APOE based and sex-based risks um was
25:41about twice as much uh in terms of how
25:43fast it uh how much it accelerated
25:45progression.
25:47Uh the other thing that jumped out to us
25:49was that um E4 carriers kind of have
25:51this double hit. So um not only do they
25:54have a faster preclinical time frame um
25:56but they also get amaloid earlier in
25:58life. So um if we're thinking about
26:00lifetime dementia risk right um the
26:03earlier that you develop pathology the
26:05the more likely it is that you'll
26:06develop dementia. But then also by being
26:08an E4 carrier um you you have both of
26:11these things combined to um really
26:13shorten that time window um and and
26:15increase your lifetime dementia risk. Um
26:18and these are uh results that have
26:19largely replicated across multiple
26:21cohorts. Um so I'm just throwing here um
26:24three examples from three different
26:25studies. Um and again another theme that
26:28pops out is the amount of variability in
26:30the time from amaloid positivity
26:32dementia. So for example in ADNE here we
26:34have some people that um are
26:36experiencing dementia essentially at the
26:39same time that they're getting amaloid
26:40positivity. Uh these are maybe not um
26:44due to um um Alzheimer's disease but
26:48then we also have people um on the other
26:50side of the spectrum who can make it 15
26:5220 or more years with amaloid um without
26:55developing dementia. And so these
26:57sources of variability are things that
26:59we're really trying to understand in our
27:01current research.
27:03Um so I shown you that applications with
27:05amaloid PET. Um but there's several
27:07other biomarkers that we could
27:08potentially develop clocks for that
27:10would allow us to study disease. Um and
27:13so um this is just kind of a brief
27:15overview of the things that we've
27:17attempted and and what has worked. Um
27:19and I'll focus kind of the rest of my
27:21talk on um applications with TAPET
27:23imaging as well as plasma phosphottow
27:25217.
27:27Um so if we look at the um tow petbased
27:29models um so similar to amaloid we're
27:32not seeing effects of ageoe or sex in
27:35terms of um influencing these overall
27:38trajectories but we do see some regional
27:40differences in performance. So for
27:41example um if we look at kind of how
27:44well people align to these two modeled
27:46curves um as well as in the statistics
27:48over here that um uh if we're using kind
27:51of a larger region of interest that
27:53includes um medial temporal and
27:55neoccortical brain regions um the model
27:58performance is a bit better compared to
28:00when we just use uh for example entrinal
28:02cortex and trying to get at kind of
28:04these earliest changes uh in disease. Um
28:07when we look at people that were
28:08observed to convert from town negative
28:10to positivity, we're able to see that um
28:12these estimates are usually within one
28:14to two years of when they uh when the
28:17window in which they they converted. Um
28:19the caveat being we don't know the
28:20actual age they convert just because of
28:23um variability in data and sampling
28:24frequencies and these types of things.
28:27Um and then something that was unique to
28:29tow imaging that um we didn't observe
28:31with amalloid data is this effect of um
28:34dementia particularly in the entrrenal
28:36cortex. So um here each of these lines
28:38is showing uh a modeled curve from
28:41t-fold cross validation and we see kind
28:44of this divergence in antennal cortex
28:46but not really metatmporal uh ROIs. when
28:49we get rid of um or when we exclude
28:51people with dementia from the analysis,
28:53we see much better concordance uh
28:55amongst those in the entrinal cortex. So
28:58um in looking into this a little bit
29:00further to try to understand what's
29:02going on here, we're also seeing
29:03basically that people with dementia are
29:05much more likely to have um negative
29:08changes in their um tow burden in the
29:10entrinal cortex. Um and so this is
29:13likely u making it challenging to
29:15estimate time estimates in and around
29:17cortex for people with dementia.
29:20And one of the things uh kind of getting
29:22back to this idea that um how do we
29:24understand these longitudinal changes in
29:26a pathological context uh we have a
29:28study ongoing where um we're taking
29:31individuals that came to brain donation
29:33um that had antimmal imaging to
29:36understand the um pathological
29:39correlates of these biomarkers. And our
29:41current hypothesis is that um uh
29:45essentially this this decrease that
29:46we're seeing in tapet signal might
29:48actually be due to the fact that the
29:49tracer doesn't bind as well to um ghost
29:52angles which um occur after the neurons
29:54die. But this is ongoing work that um
29:57we're continuing to to try to figure
29:58this out. Um in another study that um
30:02was led by uh Margot H who was a
30:04posttock in my lab um we'd apply these
30:06tow clocks to understand when tow
30:08positivity occurs. um what factors are
30:11associated with that and if there are
30:13any factors that um speed up the time
30:15from tow positivity to dementia um and
30:17also contrast some of these different
30:19timelines so age amaloid and and tow
30:21based timelines.
30:23So again this is using data from ADNE.
30:25Um so if we look at kind of the overall
30:27population um the the median town
30:30negative survival or I guess the age at
30:32which um half of the population became
30:34to positive was 88 years um which is
30:37probably a good thing just based on um
30:39overall dementia risk. Um and then we
30:41looked at factors that were associated
30:43with tow onset age. Um we saw again an
30:45E4 effect. So um individuals with higher
30:48numbers of E4 alals um had an earlier
30:51tow onset age. Um individuals that had
30:54higher levels of amaloid uh at tow
30:56positivity um also had a slightly
30:58earlier tow onset age. Um females
31:02compared to males um had a slightly
31:04earlier tow positive uh onset age. Um
31:07and then we also see kind of this um
31:10well hypothesized as a selective
31:12vulnerability um and also kind of a a
31:15survivor bias in which people that were
31:18in the uh 75 to 85 year age bracket um
31:22have this this higher level of risk of
31:24developing towel kind of within the next
31:26um 5 to 10 years after their baseline
31:28tow scan.
31:30Um in addition we were also interested
31:33in trying to understand if um select
31:35social determinance of health impacted
31:38uh uh when tow positivity occurs. Um and
31:42so we looked at two different measures.
31:44the um American National Adult Reading
31:46Test um shown on the left where um high
31:49higher uh total errors um are associated
31:52with lower exposure to um literacy
31:56enrichment. Um and then on the right um
31:58just based on educational attainment and
32:01um the results are rather consistent um
32:03but somewhat modest effect sizes which
32:05is um higher total um errors on inard as
32:08well as um uh lower uh education were
32:13associated with uh earlier tow onset
32:16age. So um this is an active area of
32:19research where we're trying to
32:20understand um some potential connections
32:22between these social determinance of
32:24health and biological mechanisms that
32:26might influence um biomarkers and when
32:29they become abnormal.
32:31Um and then if we again start to look at
32:33the variability so similar to amaloid
32:35we're seeing high variability in tow
32:37onset age. Again um APOE was kind of the
32:40the strongest predictor amongst all of
32:41these things. And we still see kind of
32:44that 30 to 40year window in which um
32:46people develop towel positivity even
32:48within for example um the E4 homozygous
32:51genotype.
32:53Now shifting to uh looking at the time
32:56from tow positivity to dementia. Um here
32:59we see a much more truncated timeline.
33:01Um so half of the uh the people that
33:04became tow positive in this study um
33:07experienced dementia within um just
33:10under seven years of of tow positivity.
33:13Uh and when we looked at factors that um
33:16accelerated this we saw uh an effect of
33:18amalloid um such that people with higher
33:21levels of amaloid burden um had a faster
33:24um decline from amalo or from tow
33:26positivity to dementia. Um and then also
33:29uh people that developed tow pathology
33:32at an older age um also had a much
33:34shorter time from amaloid pos or sorry
33:36from tow positivity to dementia.
33:40>> And then kind of in the last aim of the
33:42study we wanted to contrast these
33:44different um uh timebased metrics in in
33:47in thinking about how u how this can
33:50inform dementia risk. So uh this first
33:52plot again is showing um CDR sum of
33:54boxes. So higher levels u indicating
33:58worse dementia um and a level about 4.5
34:01kind of um being the threshold for mild
34:03dementia. So here just looking at the
34:05data you can see there's kind of a wide
34:07range in which people cross this four
34:09and a half line. Um it spans several
34:12decades. If we start to look at an
34:15amaloid based timeline and then a towb
34:17based timeline, we see kind of this
34:19progressive um shrinking of this window
34:21in which um people can develop dementia
34:24after amaloid positivity and after tow
34:26positivity. Um and if we restrict this
34:28to people that um that were observed to
34:30become positive on both amalloid and
34:32taet um we can kind of more directly
34:35compare the the ranges of these. So for
34:38example, if we're looking at age of
34:40dementia onset in this group, um we see
34:42this spans about 37 years. Um whereas um
34:46on an amaloid based timeline, this
34:47shrinks down to about 25 years. Um and
34:50in a tow based timeline, it's um shrinks
34:52again down to about 15 years.
34:55Um and now this is kind of really where
34:57we're trying to head with these models
34:59is to um get timebased predictions uh or
35:03or probabilities for when people will
35:05develop dementia. Um currently these are
35:08anchored to these these biomarker
35:09timelines. Um so on the left again this
35:12data from ADNE showing um time from
35:15amaloid to dementia on the left um time
35:17from towo positivity uh to dementia on
35:20the right um and then the y- axis here
35:23are showing probabilities of remaining
35:25dementia free. So one being everybody
35:27remains dementia free and zero being
35:29everybody um has dementia at that point.
35:32Um and if we contrast these two
35:33timelines again we see this expected
35:35pattern but it's really quite dramatic.
35:37So for example five years after amaloid
35:39positivity about 90% of the the study
35:42population remains dementia free whereas
35:445 years after tow positivity uh less
35:47than uh well about 30% of of people
35:50experience dementia. So about 75% remain
35:53dementia free. Um if we push that out to
35:5515 years after amaloid positivity we see
35:58about um a fourth of the population has
36:01has um experienced dementia whereas at
36:0315 years of tow positivity virtually
36:05nobody remains dementia free. Um so this
36:08is kind of where we're headed in terms
36:09of the prognostic value of these these
36:11methods is to um be able to put people
36:13on these timelines to give them a sense
36:15of what's the uh chance that you'll
36:17develop dementia in the next three five
36:1910 years.
36:21Um if we kind of stratify these curves a
36:24little bit more on the left now we've
36:25separated people out in terms of those
36:27that are amaloid positive but remain
36:29negative and those that are amaloid
36:31positive and become positive we can see
36:33a dramatic difference in the dementia
36:35free survival. Um so for example 20
36:37years after amalloid we're seeing um
36:40almost three-4s of the the participants
36:42remaining dementia free um and uh versus
36:46those that developed tow it's less than
36:48than half. So um really kind of
36:51developing TOW obviously is a is a big
36:53um additional risk factor in terms of
36:56accelerating dementia risk. The other
36:58thing we can see um in both the tow
37:00based and amaloid based timelines is
37:02that um these colors indicate uh the age
37:05at which either amaloid positivity or
37:07tow positivity became abnormal. And you
37:10see again this pattern where um people
37:12that are younger at uh amalator tow
37:15positivity um essentially can make it
37:18longer to uh dementia after these events
37:21um but those that are older um decline
37:24much faster um so again that this is
37:27something that we're continuing to see
37:28consistently across biomarkers um and I
37:31think something that will have high
37:32prognostic value as well um to
37:35contextualize these results again
37:36thinking about um these kind of
37:38timelines so this is now showing towa
37:40burden as a function of amaloid positive
37:42years. Um and again we can see kind of
37:44in this first 5 to 10 years that there
37:46are not that many people that are um uh
37:48tow positive and we can see quite a bit
37:50of variability um even before and after
37:53that in terms of when individual people
37:55start to become positive uh after
37:59amaloid positivity.
38:01Okay. And then um so getting to the um
38:04plasma uh kind of shifting gears um this
38:07paper was recent rec recently accepted
38:09um this week actually which um shows
38:12that we can also apply these models to
38:14plasma phosphottow
38:17um so uh in this work um led by Carly
38:19Cody who's a a former um PhD uh student
38:23in our neuroscience program in
38:24Sterling's lab um has uh led some really
38:28awesome analyses showing that um we can
38:31we can apply these models to two
38:33different um assays that measure plasma
38:35phosphate 217
38:38um and that um perhaps with a little bit
38:41more uh variability we can estimate the
38:43age of plasma phosphotype 217 positivity
38:46as well. Um and one of the neat things
38:49that this allows us to do is now to
38:51contextualize biomarkers in in uh with
38:55respect to each other. Right? So here on
38:57the left we're showing plasma
38:58phosphatile 217 levels as a function of
39:00how long people have been amaloid pet
39:02positive. Um and you can see for the
39:04most part uh people remain um phospatau
39:08217 negative um prior to amaloid
39:10positivity and for the most part people
39:12become positive on these um after
39:15amaloid positivity. Um and if we do the
39:17converse and we show uh amaloid PET uh
39:20burden as a function of plasma phosphat
39:22217 positivity, we again see that um
39:26people becoming uh amaloid positive
39:28before plasma phosphotype 217 um but
39:31also quite a bit of spread um in terms
39:33of of when these times happen.
39:36We can then uh use these timelines to
39:38look at other aspects such as um taipet
39:41positivity and entrinal cortex and
39:43getting out into the neo uh neo cortex
39:46in inferior temporal lobe. Um and um in
39:50in these are data from from RAP and our
39:52Wisconsin ADRC uh showing kind of this
39:55typical progression of amaloid PET first
39:57followed by plasma phosphotype 217
39:59positivity and terinal positivity and
40:02then spreading out into the neoortex. Um
40:05and uh if we recontextualize this based
40:08on how long people have been plasma
40:09phosphotachi 17 positivity, we can see
40:12kind of this progressive pattern of
40:14plasma first followed by um entrinal tow
40:17and inferior temporal tow. Um so again
40:19the the nice thing about these models is
40:21that um rather than kind of having
40:23things in like a biomarker concentration
40:26or these kind of um SUV centid units
40:28that require a bit of expertise to
40:30interpret um these timebased metrics
40:33really kind of give us a sense of how
40:34long it takes somebody to progress
40:36through disease on average.
40:38Um and similar to to past studies, we're
40:41seeing that um compared to an age-based
40:43timeline, um amaloid and phosphotype
40:4617based timelines um have a stronger
40:49association with pre-clinical cognitive
40:51decline. Um and in this case, um we saw
40:54that um depending on the assay um plasma
40:57phosphat 217 time either had a similar
40:59or higher association with preclinical
41:01cognitive decline um compared to amloid
41:04PAT. And then um the last study I want
41:07to highlight or last group of studies is
41:09um is related to Down syndrome. So I
41:12mentioned way at the beginning um this
41:14idea of stage zero. Um so stage zero is
41:17really meant clinical stage zero is
41:19meant to represent people that are
41:20thought to be genetically predisposed to
41:22having disease having Alzheimer's
41:24disease. Um and down syndrome uh uh is
41:28involve is characterized by a
41:30triplication of chromosome 21 which
41:33encodes the amaloid precursor protein uh
41:36gene and uh effectively this results in
41:39an overp production of a um and results
41:42in individuals with down syndrome
41:44getting amaloid uh earlier in their life
41:47um and nearly uh 100% penetrance of
41:50dementia during their lifetime um with
41:53most individuals with Down syndrome
41:54developing den dementia by age 55. Um,
41:58so this is some um ongoing work um that
42:02um I've recently started working with um
42:04Dr. Brad Christian's lab and and Dr.
42:06Matt Zamut um in in starting to contrast
42:10uh the the time course of biomarkers and
42:12Alzheimer's disease in sporadic late uh
42:14AD in a neurotypical population to our
42:17down syndrome uh cohorts. Um so in two
42:20recent studies um one published by um um
42:24Dr. McVy or PhD candidate McVey um uh
42:29we're seeing actually in observed data
42:31that uh we can measure this uh
42:33accelerated amaloid uh uh uh production
42:37rate um and and aggregation rate. So uh
42:41here showing a faster accumulation of
42:43amaloid in individuals with Down
42:44syndrome compared to our Wisconsin
42:46Registry for Alzheimer's prevention
42:48cohort. Um and in a more recent paper um
42:51that Dr. Zam and I are working on. Um we
42:53also see this when we apply sila to
42:56understand these overall um amaloid
42:58accumulation trajectories where we're
43:00seeing down syndrome on a on a shorter
43:02timeline.
43:04Uh when we apply um these these methods
43:07and contrast um down syndrome to our our
43:10local RAP cohort again we can clearly
43:12see this difference in amaloid positive
43:14age. So amongst those that became
43:16amaloid positive in these studies uh the
43:19down syndrome group uh became amaloid
43:21positive on average um just over age 40
43:24uh whereas in um this uh subop of rap um
43:28we're seeing people becoming amaloid
43:30positive um in kind of their mid60s um
43:33late late7 or early 70s um and this um
43:38stark difference is depicted on the
43:39right here in these Kaplan Meyer plots
43:41again showing kind of these median um
43:43amaloid free survival uh curves um with
43:46a with a really large um separation um
43:49of almost three decades.
43:51Um perhaps more interestingly now is uh
43:54contrasting plasma phosphot 17 and ta
43:57pet outcomes um across these groups. So
43:59the the um like pink or salmon color
44:02being the down syndrome group and the um
44:04teal color being our neurotypical rat
44:06population. Um so these are showing
44:09biomarker positivity along the age
44:12spectrum. So again we see this
44:13accelerated uh pattern in in plasma and
44:16in taped outcomes uh with respect to
44:18age. Um but more surprisingly is uh if
44:22we put these on an amaloid based
44:24timeline um we see that um a suggestion
44:28although this was not significant that
44:30um down individuals with down syndrome
44:33uh after amaloid positivity develop
44:35biomarker positivity faster. So here
44:37with um plasma phosphotype 217 um and
44:40then these were significant in the
44:42messial temporal and universal ROIs for
44:45TAPET positivity. Um so again we're uh
44:48in individuals with Down syndrome we're
44:50not just seeing kind of this earlier
44:52shift in in uh when people develop
44:55positive biomarkers but um even after
44:57amalloid positivity we're seeing a
44:59shorter time frame uh from amaloid
45:01positivity to um the the um when these
45:04other biomarkers become positive. Um and
45:07another thing that stood out to us was
45:09in down syndrome we don't really see
45:11that staggered effect of amaloid
45:13positivity phosphotal positivity and
45:15then ta pet um we're really seeing these
45:17events kind of happen almost
45:19coincidentally um I would say within the
45:22margin of error here so um in the other
45:25thing in in uh we were fortunate this
45:28week we finally got an NOA for this um
45:30so Matt and I will continue to study
45:32this but um in gearing up for Down
45:35syndrome uh clinical trial trials with
45:37anti-amalid therapies. Um it's really
45:40looking like we're going to have to
45:42treat very early on uh in order to to be
45:45able to affect this disease in Down
45:47syndrome. So this is one of the things
45:48we're trying to understand. Um and then
45:51on the right here um again just kind of
45:53showing these in a different way that um
45:55in down syndrome we really see the
45:57plasma um and messial temporal and um
46:00neoccortical tow um all kind of start to
46:03to um uh become positive really along
46:06the same timeline whereas in RAP we see
46:08this kind of staggered progression
46:10pattern. Uh and the other thing that's
46:12fascinating is a a paper also that was
46:15published from um using the same or
46:17similar data from this um Alzheimer's uh
46:19down syndrome consortium.
46:21Um we see that kind of this uptick in
46:24plasma phosphotau and misial temporal
46:26and and universal taip um also tends to
46:29correspond with the inflection point for
46:31when um we start to see um uh cognitive
46:35decline in individuals with Down
46:36syndrome. So again, um, thinking of
46:38dementia risk models, um, it really
46:40seems like in Down syndrome, we have a
46:42pretty limited window to intervene.
46:45Um, so to, um, summarize all of my
46:48findings here, um, several biomarkers,
46:50Alzheimer's biomarkers are highly
46:52predictable once we get above abnormal
46:54levels. Um, we've shown this with
46:56amaloid PET, TAPET, um, with some
46:58caveats about, uh, clinical progression
47:00and and region selection as well as
47:02plasma phosphotype 217.
47:05um we do need to kind of continue to
47:08evaluate uh the extent to which we can
47:10apply a model in other populations um as
47:13observed with the down syndrome study.
47:15Um the other thing that's very nice
47:17about being able to model these things
47:19with high accuracy is that we can now
47:21estimate when biomarkers became abnormal
47:23at an individual level. Um, and we're
47:25really hoping that this will drive um
47:27discovery of uh factors that um
47:31accelerate disease as well as allow us
47:33to generate uh dementia risk models that
47:36are easy to intuit it and um are based
47:38in years of uh exposure.
47:41Um and then the other kind of consistent
47:43thing that we're seeing across all of
47:44these studies is that people that are
47:46older when they become uh biomarker
47:48positive have a shorter uh pre-clinical
47:51time frame. And so when we're thinking
47:52of lifetime dementia risk, we really
47:54need to kind of weigh competing risk of
47:56death of of other causes with um kind of
47:59this shorter timeline that we observe um
48:01for for individuals that are older with
48:02possible biomarkers. Um, in terms of
48:05kind of some future and and ongoing
48:07work, um, the big thing we're working on
48:09right now, um, is a large multi-cohort
48:12study, um, with over 5,000 participants
48:14from multiple cohorts to develop a
48:17standard, um, amaloid and then, uh,
48:19hopefully tow based biomarks um, that
48:22will will have a dementia prognosis
48:24component. Um, continue to characterize
48:27consistent sources across multiple
48:29cohorts. uh and also understand the uh
48:32the extent to which these models uh will
48:34generalize across different population
48:36groups. Um so we have an ongoing
48:38collaboration for example with the um
48:40health uh and aging brain study uh
48:42health disparities at um University of
48:44North Texas looking at this across
48:46ethnoracial groups. Um and then lastly
48:50um really uh pathological correlates are
48:52going to be the key for us to really
48:54understanding these longitudinal
48:55biomarker uh measures and putting them
48:57into context. Um so with that I will
49:00thank you all uh for attending and for
49:03um for your kind attention. And again I
49:05want to thank everybody that's
49:06contributed to these studies. It's um
49:08it's a huge uh privilege to be able to
49:10present all of this work and we have
49:12such amazing teams here um at Wisconsin
49:15and and outside of Wisconsin that make
49:17it possible.
49:20>> Great. Thank you Toby. Maybe you can
49:22stop sharing for a few minutes. we take
49:27some questions. Um, and I do remember
49:30your start if I remember talking to you
49:33in your basement with the drum set in
49:36the background there. Um,
49:38>> dark days, you know, positive but dark.
49:41Yeah. Yeah.
49:43>> Um, yeah. From your your
49:46and your picture with your grandparents.
49:48Uh, I think it was more bandlike or or
49:52uh
49:52>> the long hair uh definitely was a was a
49:55remnant of being in metal bands. Yes.
49:57>> Yes. Yes. Great. Um, so there's a couple
50:00of questions. I'm going to see if I can
50:02interpret them. Uh, very creative use of
50:05data from going from longitudinal to
50:07cross-sectional in the past to
50:09subsequent clinical changes. So there I
50:13think what's not is there seems to be
50:14some baseline amaloid and tow
50:17measurements being able to measure. Is
50:20it ever normal? Is it is some
50:22accumulation part of normal aging or is
50:25it always pathological?
50:28>> Yeah, I mean I think this is something
50:30that's changed um like interpretations
50:32have evolved over time, right? Um I
50:34think uh earlier you know maybe five or
50:3810 years ago kind of the there was this
50:40idea that you know people can have
50:42amaloid but remain cognitively
50:43unimpaired and potentially this is some
50:47age related thing and not a disease. Um,
50:50you know, I think largely now that's not
50:52really um I think that that idea is kind
50:55of falling out of favor in terms of um
50:58um you know, these things really do um
51:02the biomarkers really do represent
51:03pathological entities and it really
51:06seems to be the case that if you have
51:08these things long enough, you're going
51:09to develop dementia. So, um it seems to
51:12be kind of like a beating the clock kind
51:14of thing, right? Are you are you going
51:16to make it uh long enough in life that
51:18you would develop these things or are
51:20you going to die from some other cause
51:21before they can kind of have their
51:22effect? Um
51:23>> yeah, so they're pathologic. They're not
51:26part of normal aging.
51:28>> That would be my opinion.
51:29>> The current especially the the tow
51:31measures um you know there's stuff with
51:34like primary age related toopathy. Um
51:36but largely um like especially to PET
51:39doesn't really detect that kind of small
51:41level of of tow.
51:43>> Great. And then um some of your your uh
51:46levels of amaloid had some plateauing or
51:50perhaps some decline in their levels.
51:52Does that correlate with any changes in
51:54their clinical uh phenotype?
51:58>> Um you know this part is kind of getting
52:00fascinating. You know it's kind of the
52:02dogma of the field is that amaloid
52:04plateaus right? Um when we look at
52:07observational data, we don't actually
52:09see a big um influence of this, but it's
52:12it's kind of a you know, it's kind of
52:16been hypothesized for a while that we
52:17can't necessarily catch that plateau
52:19phase all that well because people are
52:21progressing to dementia and we can't
52:23continue to follow them longitudinally,
52:24right? So it's,
52:26you know, the the error bars become
52:28pretty large out in that range and it's
52:30a little tough to say kind of what's
52:31going on there.
52:32>> Okay, great. Um and uh okay from Bob
52:37Pellski. Bob FDG PET hypom metabolism
52:41shows a positive correlation with
52:43necrosis tow deposition and loss of
52:46synaptic density. So have you looked at
52:48how FDG PET findings support your
52:52findings?
52:53>> Yeah. um you know we haven't done as
52:55much in terms of modeling with um with
52:57FDG um and and obviously yeah this is
52:59highly clinically relevant right it's
53:01one of the most used uh um diagnostic
53:04tools in in a in our tool belt um you
53:08know I would say kind of the the
53:10consensus in the field right is that um
53:13changes in FDG are probably um either
53:16co-occurring with Tao or a little bit
53:18downstream of that so um
53:22you know I I think um they they are
53:24pretty highly correlated in that sense.
53:26Um if we're thinking of kind of
53:28prognosis in the preclinical time frame
53:31um you know like David Jones has a lot
53:34of work on this to to kind of suggest um
53:36at what point those those patterns
53:38become meaningful. Um but I think the
53:40other thing for FDG is is not just with
53:42Alzheimer's, right? there's a lot of
53:44other dementias that we don't have good
53:45biomarkers for um that we can pick up on
53:48some hypometabolic patterns in in other
53:50neuro degenerative diseases and um so
53:52there's a really nice I think um
53:54differential diagnosis if you
53:56incorporate a few of these tools
53:58>> great um and then a couple of questions
54:01uh which is I think the burning question
54:03so once you you've detected amaloid or
54:06tow um what are the modifiers of
54:11deposition lifestyle
54:14um hospitalizations
54:16or ICU uh hospitalizations etc. Do any
54:20of these slow it down or accelerate?
54:23>> So we we haven't directly like we
54:26meaning like our local team haven't
54:28directly looked at this with these
54:29models. Um, you know, there was a a a
54:33trial that was presented this past
54:35summer, um, kind of the the US pointer
54:38uh, uh, study, um, where they did see
54:40kind of a modest effect of of lifestyle
54:43um, in terms of kind of um,
54:47say delaying progression uh of of um of
54:52cognitive decline. Um and I think the
54:55the other thing to keep in in mind there
54:57is um like past work in like Sterling's
55:01group we've seen for example like um
55:04things like Libra and heart health and
55:05other things um that are kind of these
55:08big risk models don't seem to really
55:10influence the like amaloid and ta
55:12biomarkers all that much. Um but that
55:15certainly doesn't mean that they can't
55:17influence like the time from amaloid to
55:19dementia or somebody's like cognitive
55:21health like obviously these things are
55:23very important for for people's
55:24cognitive health. So um whether or not
55:26you have Alzheimer's disease you should
55:28do these things um you know just to
55:30maintain a healthy brain
55:32>> maintain a healthy lifestyle don't don't
55:34get admitted to the hospital and don't
55:35come into the ICU. I think those are all
55:37generally good good guidelines to look.
55:40Yeah, I I mean I think it'll be nice. I
55:42we'll have to look actually um for the
55:45hospitalization and ICU data. I don't I
55:47don't know that there's going to be a
55:48lot of that out there, but um I think
55:50it'll be in in terms of people that have
55:53like big longitudinal biomarker cohorts,
55:55but um that's something fascinating to
55:57look at.
55:58>> Yeah. And there's a whole body of
55:59literature of um what the impact of
56:00delirium if you're in the ICU and poor
56:03outcomes. So um be interesting. Um, and
56:07then how early do you think we're going
56:10to need to intervene? Um, the big
56:13question. Um yeah,
56:16>> you know, I'm I'm
56:18uh I'll say optimistic that I think some
56:20of the secondary prevention studies that
56:23are going on now, um I I mean the more
56:26we understand about these longitudinal
56:28biomarker progressions, um I don't know,
56:31I think there's a lot of parallels with
56:32cancer where it's like, you know, the
56:35earlier you can detect an abnormality,
56:37the earlier you can treat it, um
56:39probably the better off you are in being
56:41able to slow or stop that disease. So we
56:43we haven't been able to stop the disease
56:46process yet, right? Um and I'm I would
56:49say I'm hopeful that um treating in that
56:51pre-clinical phase, especially before,
56:53you know, a lot of tow deposition
56:55occurs, um is really going to be the key
56:57to um to to really moving the needle in
57:00terms of when people progress. Well, and
57:02that brings us to the next question,
57:04which is in patients who have um people
57:06who have Down syndrome where you know
57:08they're going to be developing, should
57:11you be treating them, I don't know, at
57:14birth, at you know, adolescence like
57:17>> Yeah. I I mean, I think this is where
57:18the um the the blood biomarkers will
57:21really be influential. Um you know, if
57:25we're thinking about screening, for
57:26example, right? Um, you know, I could
57:29envision a world where we might, if
57:31people have Down syndrome, we might want
57:33to start doing blood tests people in
57:34their 30s even, right? Um, and we would
57:37typically never do this in, you know,
57:39uh, in the context of sporadic AD. Um,
57:42but I think that's the other piece where
57:43where, um, you know, those biomarkers
57:46are coming along. Now, that being said,
57:48um, I don't know. I the extent to which
57:51we'll be able to to move the needle and
57:53down syndrome after we're detecting
57:54things in plasma is you know we won't
57:57know until we're doing trials. So um um
58:00but I I yeah there's there's been a lot
58:02of conversation about this in the
58:03autotosomal dominant AD space too um and
58:05like the Diane study right um looking at
58:08people with autotosomal dominant
58:10mutations and um you know treating
58:12unfortunately the the drug that they
58:14used wasn't as effective at removing
58:16amaloid and so um it was a negative uh
58:19um they didn't meet their primary end
58:21points but um hopefully with some of
58:23these more effective amaloid clearing
58:26drugs that'll maybe shift when we can
58:28treat with those.
58:30>> Great. Okay. All right. I think we are
58:34um just at time and uh um done with the
58:40questions. So, uh thank you Toby. Thank
58:44you Steve um for the intro. Um great
58:47work Toby. Um an impressive body that
58:50you've you've been able to um to
58:54complete even in challenging times. So,
58:58um, many kudos to you.
59:00>> Yeah. No, thank you, Lynn. I appreciate
59:02it.
59:03>> All right. Take care, everybody.

About this transcript

This page contains the full transcript of 9/19/25 - Biomarker Clocks in Alzheimer’s Disease: Predicting Dementia Risk and Timing by UW Department of Medicine, generated from the public captions YouTube serves with the video. The transcript has 10,015 words across 1,481 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.