The Longevity Podcast: Optimizing HealthSpan & MindSpan
Welcome to a new era of conversation—where artificial intelligence explores what it means to live longer and better. Created and guided by Dr. Trinh, The Longevity Podcast uses AI hosts to bring scientific discovery, health innovation, and human wisdom together. Through AI-driven discussions inspired by real research and medical insight, each episode reveals practical tools for optimizing your healthspan and mindspan—rooted in science, shaped by compassion.
Mind. Body. Spirit.
Powered by Science, Guided by Humanity.
The Longevity Podcast: Optimizing HealthSpan & MindSpan
Blood Based Biomarkers Predicting Memory Loss
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
Imagine learning that your brain is on a specific trajectory toward memory loss and you find out from a routine blood draw, not a spinal tap, not a hard-to-access PET scan. That future is arriving fast, and it’s creating a new challenge: we don’t have one “Alzheimer’s blood test” anymore. We have a growing menu of brain biomarkers, each reporting on a different kind of cellular crisis, and the only way to use them well is to match the marker to the underlying biology.
We walk through four of the biggest names in blood-based dementia diagnostics and brain health monitoring: p-tau217, p-tau181, NFL (neurofilament light chain), and GFAP. You’ll learn what each protein normally does, what it means when it shows up in plasma, and how ultrasensitive Simoa technology can measure these signals at almost unimaginable concentrations. Then we dig into a rigorous 2026 study that compares two very different cohorts, a South Korean memory clinic group and the North American ADNI research cohort, to test whether these biomarkers hold up across demographics, education levels, and disease stages.
The headline finding is a true plot twist: amyloid status flips the winners. In amyloid-positive people, p-tau217 emerges as the strongest predictor of cognitive decline, while GFAP can be a dominant signal for brain atrophy in symptomatic patients. In amyloid-negative people, NFL becomes the most reliable indicator of neurodegeneration and overall cognitive worsening, pointing to non-Alzheimer’s causes where general axonal injury is the key story. If you care about early Alzheimer’s detection, precision neurology, or the future of dementia screening, this is the framework to know. Subscribe, share this with someone navigating memory concerns, and leave a review with the biomarker question you want answered next.
This podcast is created by Ai for educational and entertainment purposes only and does not constitute professional medical or health advice. Please talk to your healthcare team for medical advice.
Never miss an episode—subscribe on your favorite podcast app!
A Blood Test For Brain Aging
SPEAKER_03Imagine if uh a single drop of blood drawn at your annual physical today could reveal the exact year your memory's gonna start failing.
SPEAKER_00Aaron Powell Yeah, a decade from now.
SPEAKER_03Right. Like it sounds like science fiction. Or I don't know, maybe the premise of some dystopian movie.
SPEAKER_00Oh, absolutely. But you know, as of 2026, this whole biological revolution of blood-based brain mapping, it's not just possible anymore.
SPEAKER_03Aaron Powell No, it's happening right now in lags all over the world. We are standing on the edge of this monumental shift in how we understand the human mind.
SPEAKER_00Aaron Powell We really are. Because for decades, if you or a loved one started showing signs of memory loss, getting a real definitive biological answer meant navigating a diagnostic landscape that was, well, honestly, it was brutal.
SPEAKER_03Oh, totally brutal. I mean, you were looking at an incredibly invasive spinal tap to draw a cerebrospinal fluid, right?
SPEAKER_00Right. A lumbar buncture.
SPEAKER_03Yeah. Or you had to somehow find a facility with a highly specialized PT skein machine, which is, you know, massively expensive and historically just very hard to access for most people.
SPEAKER_00Exactly. Those traditional methods created this massive bottleneck in neurological care.
SPEAKER_01A huge bottleneck.
SPEAKER_00Because a spinal tap involves literally inserting a needle between the vertebrae of your lower back. It carries risks of severe headaches, infection, and it requires a specialized clinical setting. You can't just do it at a corner clinic.
SPEAKER_03Right. And the PT scans aren't much easier.
SPEAKER_00No, they require injecting a patient with a radioactive tracer that binds to specific proteins in the brain. And then you have to lie completely still in this massive scanner.
SPEAKER_03And the infrastructure for that is insane, right? Yeah. Like, don't you need a cyclotron or something?
SPEAKER_00You do, yeah. You need cyclotron to generate those radioactive crecers in the first place. Which means that outside of major urban research hospitals, these scans are virtually unavailable to the global population.
SPEAKER_03So finding out what was actually happening inside your aging brain was reserved for like the lucky few who had the geography and the insurance to cover it.
SPEAKER_00Precisely. But this new era of blood testing, it completely shatters that bottleneck.
SPEAKER_03It's revolutionary. The idea that we can peer into the intricate microscopic health of the brain just by looking at a standard tube of blood drawn from your arm?
SPEAKER_00It is. But uh this massive breakthrough actually brings up a totally new, highly complex problem.
SPEAKER_03Aaron Powell Right, the new problem. Because we don't just have one blood test now.
SPEAKER_00Aaron Powell No, we have multiple. The market is suddenly flooded with all these different biological markers.
SPEAKER_03Aaron Powell So if you give a vial of blood, the question is which specific molecule is the truest mirror of your memory or you know your brain's physical structure?
SPEAKER_00Aaron Powell That is the multi-million dollar question driving the entire field of neurology today. Because not all biomarkers are created equal.
SPEAKER_03Aaron Powell Yeah, they reflect entirely different cellular crises happening within the brain.
SPEAKER_00Aaron Ross Powell Right. Some reflect structural collapse, some reflect immune panic, and some reflect the accumulation of toxic garbage.
SPEAKER_03Aaron Powell And that, for you listening, is the exact puzzle we are decoding on this deep dive today. We've got our hands on this landmark 2026 study published in the Journal of Prevention of Alzheimer's disease, or JPAD.
SPEAKER_00It's a fantastic paper.
SPEAKER_03It really is. It does something incredibly rigorous. It takes four specific blood-based biomarkers, these tiny biological breadcrumbs floating in our veins, and essentially pits them against each other in this massive battle royale.
SPEAKER_00Aaron Powell Well, I'd say it's less of a battle royale and more about, you know, finding the right tool for the right job.
SPEAKER_03Yeah.
SPEAKER_00But yes, they wanted to establish a definitive hierarchy.
SPEAKER_03Trevor Burrus, Jr.: Right, right. A hierarchy of predictive power. They wanted to see which single test is the absolute best at predicting two very scary, very real outcomes.
SPEAKER_00Aaron Powell The first being the actual physical shrinkage of the brain, which we call neurodegeneration. Trevor Burrus, Jr.
SPEAKER_03The hardware failure.
SPEAKER_00Trevor Burrus Exactly. And the second being the cognitive decline, the real-world memory loss, the software failure.
SPEAKER_03Aaron Ross Powell So whether you're a medical professional trying to keep up with the cutting edge or someone who tracks longevity science or just someone inherently curious about how our most complex organ ages, our mission today is to decode this new biological alphabet.
SPEAKER_00Aaron Powell By the end of this deep dive, you will understand exactly how doctors in the very near future will use a simple blood draw to map the precise trajectory of a patient's cognitive
Why Spinal Taps Became The Norm
SPEAKER_00health.
SPEAKER_03Because the goal is moving away from educated guesswork.
SPEAKER_00Right. Yeah, moving away from symptom-based diagnosis toward a highly personalized predictive science based on your unique molecular profile.
SPEAKER_03Okay, let's unpack the core science first. We need to meet the contenders in this. Well, I'm still gonna call it a battle royale.
SPEAKER_02Fair enough.
SPEAKER_03The researchers focused on four specific blood-based biomarkers. And I hear these names thrown around constantly in longevity podcasts and health blogs, usually as these ultimate villains of aging.
SPEAKER_00They do get a bad rap.
SPEAKER_03They do. But proteins don't exist in our bodies just to give us dementia. I want to know what they are actually supposed to be doing before things go wrong. We have PITO217, PITO 181, NFL, and GFAFA.
SPEAKER_00Let's start with the Tau proteins.
SPEAKER_03Okay, what is Tau actually doing in a healthy brain?
SPEAKER_00In a healthy functioning neuron, Tau is a vital structural component. To visualize this, you have to picture the inside of a brain cell. A neuron isn't just a blob of jelly.
SPEAKER_02Right.
SPEAKER_00It has a highly complex internal skeleton. And part of that skeleton includes these long, hollow tubes called microtubules.
SPEAKER_03Microtubules, okay.
SPEAKER_00You can think of these microtubules as a microscopic railway system. It transports essential nutrients, mitochondria, and chemical messengers from the center of the cell all the way down to the very distant synapses.
SPEAKER_03Where the neuron communicates with other cells. So it's basically a microscopic supply chain keeping the cell alive.
SPEAKER_00Precisely that. But those microtubule tracts are fragile. They need something to bind them together and keep them stable so they don't fall apart under pressure.
SPEAKER_03And that's where tau comes in.
SPEAKER_00That is exactly what tau protein does. It binds to the microtubules, acting like the wooden railroad ties, holding the steel tracks in perfect parallel alignment.
SPEAKER_03Okay, so as long as tau is doing its job, the supply chain runs smoothly, the trains run on time.
SPEAKER_00Yes, exactly.
SPEAKER_03So what triggers it to go rogue in Alzheimer's disease? Because the markers we are looking at in the blood test are P ta 217 and P tau 181. What does the P stand for?
SPEAKER_00The P stands for phosphorylated, and this is where we get into the cascade of cellular disaster. Exactly. Through a process that researchers are still intensely studying, certain enzymes in the aging or diseased brain become overactive. These enzymes are called kinases.
SPEAKER_01Kinases, right?
SPEAKER_00And these kinases start aggressively attaching phosphate molecules to the tau protein.
SPEAKER_03What does a phosphate molecule actually do to the railroad tie, though?
SPEAKER_00You can think of a phosphate group like a biological magnet with a really strong negative electrical charge. When too many of these negative magnets attach to the tau protein, a state we call hyperphosphorylation, the electrical repulsion forces the tau protein to radically change its three-dimensional physical shape.
SPEAKER_03Wait, so it twists itself into something else?
SPEAKER_00It contorts so violently that it physically snaps off the microtubule tract.
SPEAKER_01Oh wow.
SPEAKER_00Yeah. And once it detaches, these mangled tau proteins start magnetically clinging to each other. They form these insoluble, chaotic clumps inside the cell body.
SPEAKER_03And those are the tangles, right?
SPEAKER_00Yes. We call these neurofibrillary tangles. Meanwhile, without the tau to hold them together, the microtubule supply tracts just completely disintegrate.
SPEAKER_03So the neuron starves.
SPEAKER_00The neuron starves, its communication network collapses, and the cell eventually dies. The markers we measure in the blood, PA-217 and PA181, are the specific phosphorylated fragments of this destruction.
SPEAKER_03They leak out of the brain.
SPEAKER_00Exactly. They leak out of the dying brain cells, cross the blood-brain barrier, and make their way into the peripheral blood circulation. So they are a direct, highly specific echo of Alzheimer's pathology.
SPEAKER_03Okay, that makes complete sense. The PTA markers are giving us a window into the tangles and the collapse of the internal supply chain. Let's look at the next one on the list. NFL or neurofilament light chain.
SPEAKER_00Right, NFL.
SPEAKER_03How is this different from Tau? Because it's also structural, isn't it?
SPEAKER_00It is a structural protein, yes, but it serves an entirely different architectural purpose. If Tau is the railroad tie for the internal transport system, neurofilaments are the thick, load-bearing cables that give the neuron's axon its physical shape and caliber.
SPEAKER_03The axon being the long wire part of the cell.
SPEAKER_00Yes. The axon is the long wire-like projection that carries electrical impulses to other cells. And some axons in the human body can be incredibly long, up to a meter in some nerves.
SPEAKER_03A meter from a single microscopic cell. That's insane.
SPEAKER_00It is. So they require immense structural support to maintain their integrity. And neurofilaments provide that dense cable-like scaffolding.
SPEAKER_03So when an axon gets physically severed or damaged, these cables just snap.
SPEAKER_00Exactly. When an axon is injured, whether that's by a traumatic blow to the head, a lack of oxygen, or a neurodegenerative disease, the cell membrane ruptures.
SPEAKER_01And the cables spill out.
SPEAKER_00The internal contents of the axon, including high concentrations of these NFL proteins, spill out into the surrounding cerebrosminal fluid. And from there, they drain right into the bloodstream.
SPEAKER_03But the crucial distinction here, as I understand it from the paper, is that NFL is completely nonspecific.
SPEAKER_00That is the defining characteristic of NFL, and it's so important to remember. It is an exquisitely sensitive marker of axonal injury, but it tells us absolutely nothing about the underlying cause of that injury.
SPEAKER_02Right.
SPEAKER_00If a patient comes in with multiple sclerosis, where the immune system attacks the axon coating or Parkinson's disease, or if a professional football player suffers a severe concussion.
SPEAKER_03Or even ALS,
Four Biomarkers And What They Mean
SPEAKER_03right?
SPEAKER_00Yes, even ALS. In all those cases, their NFL levels will spike dramatically. It is a universal distress signal indicating that brain cells are actively dying, but it is entirely agnostic to the disease pathology driving the death.
SPEAKER_03It's just a broad alarm bell.
SPEAKER_00Okay.
SPEAKER_03Okay, let's look at the final contender, GFAP, glial fibrillary acidic protein.
SPEAKER_00GFAP, yes.
SPEAKER_03This one feels entirely different to me because we aren't even looking at the neurons anymore, are we?
SPEAKER_00We aren't. We are shifting our focus to the other half of the brain's cellular population, the glia. Specifically a type of glial cell called an astrocyte.
SPEAKER_03Astrocytes, the star-shaped ones.
SPEAKER_00Exactly. Astrocytes act as the caretakers of the brain environment. For a long time, neuroscience largely ignored them, thinking they were just passive structural glue holding the neurons in place.
SPEAKER_03But they do way more than that.
SPEAKER_00Oh, we now know they are incredibly dynamic. They regulate blood flow, they provide energy to neurons, they recycle neurotransmitters, and they act as the brain's resident immune responders.
SPEAKER_03Aaron Powell So they are basically the medics and the cleanup crew of the brain.
SPEAKER_00Aaron Powell That's a great way to put it.
SPEAKER_03But how does a helpful cleanup crew become a biomarker for cognitive decline? That sounds counterintuitive.
SPEAKER_00Aaron Powell It comes down to a process called astrogliosis.
SPEAKER_03Astrogliidosis, okay.
SPEAKER_00Trevor Burrus When astrocytes detect something abnormal in their environment, say a viral infection, a lack of oxygen, or the accumulation of toxic proteins like amyloid plaques, they enter a state of chronic alarm.
SPEAKER_01They panic.
SPEAKER_00They become what we call reactive.
SPEAKER_01Aaron Powell What does a reactive astrocyte actually look like? Like does it change physically?
SPEAKER_00Drastically. A resting astrocyte has fine, delicate branches that gently interact with thousands of nearby synapses. When it becomes reactive, it undergoes a massive morphological shift.
SPEAKER_01It hulks out.
SPEAKER_00It really does. The cell body swells, the branches become thick and rigid, and it starts heavily upregulating the production of structural proteins to support this new, larger shape.
SPEAKER_03And the primary protein it produces is GFAP. Ah, there it is. So GFAP is the scaffolding for the swollen, angry astrocyte.
SPEAKER_00Exactly. Furthermore, in this reactive state, the astrocyte stops doing its normal maintenance jobs. It stops being the helpful cleanup crew and starts secreting inflammatory chemicals called cytokines to attack the perceived threat.
SPEAKER_03But if they do that for too long.
SPEAKER_00Over time, this chronic neuroinflammation becomes deeply toxic. The inflammatory chemicals start causing collateral damage, harming the very neurons the astrocyte is supposed to protect.
SPEAKER_03So the medics start killing the patients.
SPEAKER_00Sadly, yes. And the excess GFAP being produced leaks into the bloodstream, serving as a direct, measurable proxy for this state of toxic neuroinflammation.
SPEAKER_03Okay, I want to try an analogy here to ground this for anyone listening who is trying to keep all these biological mechanisms straight. Let's think of the brain as a highly complex, high-performance car engine that is starting to show its age.
SPEAKER_00I like car analogies. Let's hear it.
SPEAKER_03If you plug a diagnostic computer into the dashboard and it gives you a highly specific proprietary error cone that says, like transmission control module failure, that is PTAW, it tells you exactly what type of proprietary Alzheimer's-specific pathology is happening.
SPEAKER_00I follow the logic.
SPEAKER_03Then you have NFL. That is the equivalent of walking out to your driveway in the morning and seeing a massive puddle of black oil under the car.
SPEAKER_00Uh yes.
SPEAKER_03You know absolutely for a fact that something structural is broken, the engine is leaking its internal fluids, but you have no idea if it's the transmission, a cracked engine block, or a blown head gasket. Trevor Burrus, Jr.
SPEAKER_00Right. It's just a general catastrophic loss of structural integrity.
SPEAKER_03Yeah.
SPEAKER_00A very apt description of nonspecific axonal damage.
SPEAKER_03And finally, GFAP. That is your engine temperature gauge on the dashboard flashing bright red. It's telling you the engine is overheating, the whole system is inflamed, and the cooling fans are working in overdrive to react to some kind of underlying stress.
SPEAKER_00Even if the structural parts haven't completely snapped yet.
SPEAKER_01Exactly.
SPEAKER_00That framework, the transmission code, the oil leak, the temperature gauge is going to be incredibly useful as we look at how these markers perform in the actual clinical data.
SPEAKER_03Aaron Powell But uh before we look at the patients, we need to address a mechanical reality here. How on earth are we finding these microscopic proteins in a standard blood draw? Because the blood brain barrier is notoriously strict about what it lets out of the central nervous system, right?
SPEAKER_00It is incredibly strict.
SPEAKER_03I was looking into this, and the concentration of these proteins in peripheral blood is vanishingly small compared to the cerebrospinal fluid. It's not like measuring cholesterol where the molecules are just everywhere. We are talking about femtograms per milliliter.
SPEAKER_00To give that some context, a femtogram is one quadrillionth of a gram.
SPEAKER_03That is hard to even conceptualize.
SPEAKER_00It's nearly impossible. Before the development of ultrasensitive assay technology, looking for these brain proteins in peripheral blood was biologically impossible. Standard laboratory tests, like lysis, simply weren't sensitive enough.
SPEAKER_03It was like trying to find a specific individual grain of sand scattered somewhere across an entire beach.
SPEAKER_00Precisely.
SPEAKER_03So how did the researchers in this JPAD study actually measure them? What's the tech?
SPEAKER_00They utilized a technology called single molecule array assays, or Samoa.
SPEAKER_01Samoa, okay.
SPEAKER_00Samoa is a masterclass in bioengineering. The process involves taking the patient's blood plasma and introducing millions of microscopic magnetic beads.
SPEAKER_03Magnetic beads into the blood sample.
SPEAKER_00Yes. Each of these beads is coated with thousands of highly
How Tau Turns Into Tangles
SPEAKER_00specific antibodies designed to act like molecular velcro, grabbing onto just one specific target, like PTA 217.
SPEAKER_03So the magnetic beads float around in the plasma, just grabbing the target proteins out of the blood.
SPEAKER_00Exactly. And once they capture their targets, the technicians use a magnet to physically pull the beads out of the plasma, washing away all the other billions of unrelated blood proteins.
SPEAKER_03That is so smart.
SPEAKER_00Then they introduce a second antibody that binds to the target protein, creating a sort of molecular sandwich.
SPEAKER_03A bead, the target protein, and then another antibody on top.
SPEAKER_00Right. And this second antibody is attached to an enzyme. Finally, they load these magnetic beads into a specialized disc containing hundreds of thousands of microscopic wells.
SPEAKER_03Like tiny little test tubes?
SPEAKER_00Yes, but they are designed so that only one single bead can fit into one single well.
SPEAKER_03Oh wow. So they separate them out individually?
SPEAKER_00Yes, and then they add a chemical substrate over the whole disc. If the target protein is present on a bead, the enzyme attached to it reacts with the substrate and emits a tiny microscopic flash of fluorescent light.
SPEAKER_01A flash of light.
SPEAKER_00A flash of light. The machine's camera then takes a picture and quite literally counts the individual flashes of light. They are counting individual protein molecules.
SPEAKER_03That is staggering what biological engineering can achieve today, counting individual molecules.
SPEAKER_00It is the sheer leap in technological sensitivity that has made the era of blood-based brain biomarkers a reality.
SPEAKER_03Okay, so we have our four diagnostic tools, our engine gauges, and we have the Samoa technology to measure them with like atomic precision. Now we need to look at the arenas where they tested these tools.
SPEAKER_00The cohorts.
SPEAKER_03Right. The researchers didn't just test these on one random group of volunteers. They set up two incredibly distinct cohorts. And honestly, this dual cohort design is the single biggest reason I trust the data in this paper.
SPEAKER_00It addresses a fundamental historical flaw in medical research.
SPEAKER_03Yes.
SPEAKER_01The weird problem.
SPEAKER_00Exactly. For decades, large-scale clinical trials and genetic studies, particularly in North America and Europe, have suffered from what sociologists call the weird problem.
SPEAKER_03Western, educated, industrialized, rich, and democratic.
SPEAKER_00Yes. If you develop a diagnostic test and you only ever validate it in populations that are predominantly white, upper middle class, highly educated, and living in industrialized nations, you are creating a massive blind spot.
SPEAKER_03Because biology isn't perfectly uniform.
SPEAKER_00Not at all. Biology has variations. Genetic backgrounds dictate how proteins fold and clear. Environmental exposures, lifetime diet, baseline metabolic health, and socioeconomic stress all alter the biological baseline.
SPEAKER_03So a biomarker might look incredibly predictive in a wealthy suburb of Boston, but completely fail to predict cognitive decline in a rural village in Asia due to, I don't know, differing rates of vascular comorbidities or genetic expressions.
SPEAKER_00Exactly. The background noise is different.
SPEAKER_03So the researchers here intentionally sought out two completely separate environments to see if the biological truths held up regardless of the demographic. Let's look at group number one, the SAMD Lab Covert.
SPEAKER_00The South Korean group.
SPEAKER_03Right. This group consisted of 150 participants recruited directly from a memory clinic in South Korea. They were 100% Asian descent. The average age was around 74, but the most vital detail to me is the clinical setting.
SPEAKER_00Because they were recruited from an active memory clinic, this group represents real-world symptomatic patients.
SPEAKER_03That clinical context shifts the baseline entirely. These are individuals who sought medical help because they or their families noticed a terrifying change in their daily function.
SPEAKER_00Yes. They forgot how to drive home, they left the stove on, they couldn't recognize a grandchild. Consequently, the Salmon D cohort had a much wider range of clinical severity. We're looking at a population with more advanced, deeply entrenched neurodegeneration.
SPEAKER_03And they measured that severity using something called the clinical dementia rating scale or CDR.
SPEAKER_00Right. And the SAMMD group had a widespread across that scale. Furthermore, their average years of formal education was about nine years.
SPEAKER_03Nine years, okay.
SPEAKER_00Education is a critical variable in Alzheimer's research because the concept of cognitive reserve, which we will definitely touch on when we discuss the memory tests.
SPEAKER_03We will. So contrast that SAMMD group with group number two, the ADI cohort.
SPEAKER_00ADNI. The Alzheimer's Disease Neuroimaging Initiative.
SPEAKER_03Right. This is a massive, incredibly famous, ongoing observational study in the U.S. and Canada. The researchers pulled data from 284 participants in this initiative. They were 94% white. The average edge was also around 74, so they were perfectly age-matched with the South Korean group.
SPEAKER_00But the clinical makeup is vastly different.
SPEAKER_03Vastly. Over 55% of the ADI group were classified as cognitively unimpaired, essentially healthy volunteers. Compare that to just 13% in the SAMD group.
SPEAKER_00Well, ADI is an observational research initiative, not a reactive clinical setting. People volunteer for ADI often out of just general interest in science or because they have a family history and want to be monitored.
SPEAKER_03They aren't necessarily there because they are actively experiencing severe deficits.
SPEAKER_00Precisely.
SPEAKER_03And the ADI group was highly educated, averaging over 16 years of schooling. This means most of them had college degrees or postgraduate education.
SPEAKER_00So the study design gives us two distinct testing grounds: a memory clinic in South Korea dealing heavily with symptomatic
NFL Damage Signal Versus GFAP
SPEAKER_00patients with lower average formal education, and a North American research initiative heavily weighted toward individuals early in the disease process or completely healthy with high levels of formal education.
SPEAKER_03I have to pause here though and pray devil's advocate because when I first read this methodology, my immediate thought was that this setup is a statistical nightmare.
SPEAKER_00It is certainly a challenge.
SPEAKER_03If the SAM D group is full of people whose brains are already heavily damaged and actively failing, and the ADI group is full of people who are mostly healthy, doesn't that completely skew the math? It's like trying to test the sensitivity of a smart alarm, but in one test house the living room is fully endulved in flames, and in the other test house, someone just struck a match in the kitchen. The alarm is obviously going to react differently. Isn't this an apples to oranges comparison?
SPEAKER_00It is the single most important critique of the initial study design, absolutely. And the authors of the paper dedicate significant space to addressing it. The disparity in disease stage distribution is a massive confounding variable.
SPEAKER_03So if PTAW seventeen performs differently in the South Korean cohort than in the North American cohort, you don't know why.
SPEAKER_00Exactly. The immediate question is is it because the biological mechanisms of Alzheimer's differ across ethnicities, or is it simply a mechanical result of the SAMD group having more advanced pathology?
SPEAKER_03The math gets messy if you don't control for the severity of the fire.
SPEAKER_00The researchers were acutely aware of this. We will delve deeply into the complex statistical matching techniques they deployed later in the analysis to artificially level the playing field, to solve this apples to oranges problem.
SPEAKER_02Right.
SPEAKER_00But the rationale for using these disparate groups initially is actually very sound. If a biomarker shows dominance in both the heavily symptomatic Korean clinic and the mostly healthy North American research trial, then we have discovered a fundamental, unshakable law of neurobiology.
SPEAKER_03A law that transcends demographics and clinical stages.
SPEAKER_00Exactly.
SPEAKER_03Okay. I will hold my skepticism on the cohort differences until we get to the statistical stress tests. Let's move to the finish line. We have our four blood tests and our two diverse populations. What exactly are we testing them against?
SPEAKER_01The outcomes.
SPEAKER_03Right. If I claim my blood test predicts brain disease, I have to prove it against a tangible reality. What are the specific outcomes they measured?
SPEAKER_00They categorize the outcomes into two broad domains: the physical reality of the brain tissue and the cognitive output of the mind. Hardware and software.
SPEAKER_03Hardware and software. Let's start with the hardware. They measured this by looking at what they term the AD signature ROI thickness. ROI means region of interest. This is the literal physical shrinkage of the brain, the cortical atrophy, measured using high-resolution MRI scans.
SPEAKER_00Correct.
SPEAKER_03But they didn't just calculate the total volume of the entire brain, did they?
SPEAKER_00No, they didn't. Because Alzheimer's disease is not a uniform attacker. It doesn't dissolve the brain evenly like, say, acid poured over a sponge.
SPEAKER_03It's more strategic than that.
SPEAKER_00Very. It is a highly localized progressive disease that originates in specific neural networks and spreads along very predictable pathways. Measuring the whole brain would actually dilute the data with healthy tissue.
SPEAKER_03So they zoom in?
SPEAKER_00Yes. Instead, they used MRI software to measure the cortical thickness in fractions of a millimeter in four highly vulnerable regions. The entohynal cortex, the inferior temporal cortex, the middle temporal cortex, and the fusiform gyrus.
SPEAKER_03Okay, why those four specific neighborhoods? What makes them the epicenter of the disease?
SPEAKER_00These regions form the core infrastructure for memory encoding and higher level sensory processing. The endohynal cortex is arguably the most critical.
SPEAKER_03Endocernal cortex.
SPEAKER_00You can think of it as the main sensory toll booth or gateway to the hippocampus, which is the brain's central memory filing system.
SPEAKER_03Okay, a toll booth.
SPEAKER_00Every single piece of sensory information, what you see, hear, and feel, has to pass through the endohinal cortex to be consolidated into a long-term memory.
SPEAKER_03So if the toll booth breaks down, the memories can't get into the filing cabinet. They just bounce off.
SPEAKER_00Exactly. And the internal cortex is almost universally the very first region of the brain to show physical accumulation of tau tangles and subsequent cellular death in Alzheimer's.
SPEAKER_03He's a ground zero.
SPEAKER_00It is ground zero. After the intornal cortex, the pathology spreads outward to the temporal lobes. The inferior and middle temporal cortices are deeply involved in semantic memory, our understanding of meaning, language, and facts about the world.
SPEAKER_03So losing that tissue means losing the vocabulary to even describe the world around you. Yes.
SPEAKER_00And finally, the fusiform gyrus. This area is highly specialized for complex visual processing, most notably facial recognition.
SPEAKER_01Oh wow.
SPEAKER_00When atrophy reaches the fusiform gyrus, patients often develop prosopegnosia, the terrifying inability to recognize the faces of their own spouses or children, even while their actual vision remains perfectly fine.
SPEAKER_03That is just the tragedy of the disease mapped out in tissue loss. So the researchers are using MRI to measure the literal thinning of the brain tissue in the exact spots Alzheimer's hits hardest. That's the hardware check.
SPEAKER_00Yes.
SPEAKER_03Now, how did they measure the software that total cognition and memory?
SPEAKER_00Aaron Powell Because the researchers were utilizing two historically distinct cohorts, they had to rely on the established psychometric testing protocols
Simoa Tech That Counts Molecules
SPEAKER_00utilized by those specific institutions. This requires careful harmonization of the data. Aaron Powell Right.
SPEAKER_03They didn't take the exact same test.
SPEAKER_00No. In the South Korean SAMD clinical group, they utilized the Cred K battery to measure total global cognition and a specifically designed composite memory score to isolate memory function.
SPEAKER_03What does a battery like Cred K actually ask a patient to do?
SPEAKER_00It's a very comprehensive neuropsychological evaluation. It involves tasks like verbal fluency, asking a patient to name as many animals as they can in 60 seconds.
SPEAKER_01Oh, that's harder than it sounds.
SPEAKER_00It is. It also involves naming line drawings of objects, copying complex geometric shapes on paper, and learning a list of 10 unrelated words, which they are then asked to recall several minutes later. It stresses different domains of the brain software simultaneously.
SPEAKER_03And what about the North American ADI group?
SPEAKER_00For ADI, the researchers looked at the ADS COG score for total cognition, which is a very similar, globally recognized metric involving word recall, following commands, and orientation to time and place.
SPEAKER_03And for just memory. Ah, this is the one where the examiner reads a short narrative story and the patient has to repeat back as many details as possible.
SPEAKER_00Exactly. They read a short paragraph containing specific details, names, locations, actions. They ask for an immediate recall, and then crucially, they wait 30 minutes.
SPEAKER_01What happens during those 30 minutes?
SPEAKER_00They actively distract the patient with other tasks and then ask them to recall the story again. This tests the brain's ability to not just hold information in the short-term working memory, but to successfully encode it, store it in the hippocampus, and retrieve it on demand.
SPEAKER_03It seems like a lot of work to measure both the MRI hardware and the cognitive software. I mean, if the MRI shows the enterhenal cortex has shrunk by two millimeters, doesn't that automatically mean the patient's memory test scores will be terrible? Why measure both?
SPEAKER_00Well, if human biology were a simple math equation, yes. But the relationship between structural brain damage and real-world cognitive output is one of the most confounding mysteries in neurology.
SPEAKER_03Really? It's not a one-to-one ratio.
SPEAKER_00Not at all. We frequently see a phenomenon where a patient's brain looks severely atrophied on an MRI. They are missing significant volume in the temporal lobes, yet they sit down for the ADS COG test and perform remarkably well, functioning normally in their daily lives.
SPEAKER_03Wait, how is that physically possible if the hardware is gone? Where is the software running?
SPEAKER_00It brings us back to the concept of cognitive reserve, which we mentioned earlier. It is heavily tied to education, lifelong learning, and complex occupational demands.
SPEAKER_03So using your brain builds up resistance.
SPEAKER_00Exactly. The theory is that a highly educated, constantly simulated brain builds millions of redundant synaptic pathways. It builds cognitive bridges. When the Alzheimer's pathology destroys the primary bridge, a brain with high cognitive reserve simply reroutes the electrical signals over a secondary bridge. They compensate for the hardware loss.
SPEAKER_03Until the disease destroys the secondary bridges, too, and then they crash rapidly.
SPEAKER_00Exactly. And conversely, you have patients with very mild structural atrophy on an MRI who present with devastating, profound memory loss. Therefore, a truly dominant, universally useful blood biomarker cannot just predict structural shrinkage, and it cannot just predict cognitive test scores.
SPEAKER_03It has to do both.
SPEAKER_00It must act as a reliable anchor for both the physical reality of the tissue damage and the lived reality of the memory loss. Okay.
SPEAKER_03The stage is completely set. We understand the four specific blood proteins we are tracking. We know the profound demographic differences between the Korean clinic and the North American research trial. And we understand the MRI and cognitive test defining our finish line. Let's get into the actual results.
SPEAKER_00Let's do it.
SPEAKER_03And reading through the paper, the initial data poll was incredibly messy.
SPEAKER_00Messy is putting it mildly. It looked like statistical noise.
SPEAKER_03Yeah. When the researchers first ran the numbers performing basic standard correlations, what statisticians call spearman partial correlations, they found that almost every single biomarker correlated significantly with both brain shrinkage and cognitive decline.
SPEAKER_00Right. It was as if every single warning light on the car dashboard started flashing simultaneously.
SPEAKER_03The initial data suggested everything was highly predictive. To give you specific numbers for you listening, in the SAMD cohort, when looking at the shrinkage of the cortical thickness, the correlation coefficient for PAW217 was negative 0.443.
SPEAKER_00Meaning that as the concentration of PTA 217 increased in the blood, the thickness of the brain tissue reliably decreased.
SPEAKER_03Right. But GFAP, the inflammation marker, was sitting right next to it with a correlation of negative .442, almost identical. NFL and PTA 181 also demonstrated statistically significant negative correlations.
SPEAKER_00They were all pointing in the same direction.
SPEAKER_03So if you are a doctor sitting in a clinic and you look at that raw correlation data, you might logically conclude, fantastic, these all work. I can just draw blood, test for whichever marker is cheapest or easiest to process, and get an accurate prediction. But that assumption is a massive trap, isn't it?
SPEAKER_00It is a massive trap, born of biological overlap. In statistics, this problem is known as interbiomarker collinearity.
SPEAKER_03Collinearity. Let's untangle that concept because it is the entire reason the researchers had to deploy such extreme mathematics in this paper.
SPEAKER_00Collinearity occurs when you are trying to predict an outcome using several different variables. But those variables are highly correlated with each other. Because these blood biomarkers reflect biological disasters that are happening simultaneously and exacerbating each other, their statistical signals merge together.
SPEAKER_03They overlap so much you can't tell them apart.
SPEAKER_00Right. Think about a massive house fire. The internal temperature is rising rapidly, thick black smoke is billowing out of the windows, and the structural wooden beams are snapping and collapsing.
SPEAKER_03All three things are happening at once.
SPEAKER_00If you bring in sensors to measure the heat, the volume of smoke, and the sound of breaking wood, all three of those measurements are going to correlate perfectly with the fact that the house is being destroyed.
SPEAKER_03Right. If the smoke increases, the destruction increases. If the heat increases, the destruction increases.
SPEAKER_00But that doesn't mean the smoke is causing the wood to snap.
SPEAKER_03Exactly.
SPEAKER_00If you want to know which specific sensor gives you the most uniquely valuable information about the core structural collapse of the house, you have to find a way to statistically untangle them.
SPEAKER_03And in the brain, toxic neuroinflammation represented by GFAP almost always happens alongside the structural destruction of the axons represented by NFL and the formation of the tangles represented by PTAL, they move as a pack.
SPEAKER_00Yes, they move as a pack. Standard correlation math simply tells you the pack is moving. It doesn't tell you who is leading the pack.
SPEAKER_03So the standard math couldn't solve this. The researchers couldn't just look at the raw correlations and declare a winner. They needed a mechanism to figure out which biomarker was actually driving the predictive power and which ones were just along for the ride.
SPEAKER_00Right. They needed to isolate the unique, independent contribution of each individual biomarker while aggressively, mathematically controlling for the noise generated by the other three.
SPEAKER_03And this leads us to the analytical hero of this paper, a statistical method I had to read up on called dominance analysis.
SPEAKER_00Dominance analysis is a phenomenal, rigorous approach to solving collinearity. It is rarely seen outside of advanced econometrics or highly specialized biological modeling.
SPEAKER_03I read through the methodology section describing
Two Cohorts And The WEIRD Trap
SPEAKER_03this, and I want to try to visualize this for the listener because it's fascinating. It's not just a simple comparison. Dominance analysis is essentially a brutal statistical tournament.
SPEAKER_00It really is.
SPEAKER_03I originally thought of it like a battle royale, but it's more precise than that. It's like listening to a choir where everyone is singing at maximum volume. Standard correlation math just tells you the choir as a whole is incredibly loud.
SPEAKER_00That's a strong analogy. The noise is overwhelming.
SPEAKER_03Dominance Analysis is the audio engineer sitting at the mixing board, methodically muting different singers one by one in hundreds of different combinations to figure out whose specific voice is actually carrying the melody and who is just harmonizing in the background to make it sound richer.
SPEAKER_00The researchers ran 256 separate subset linear regression models.
SPEAKER_03256.
SPEAKER_00The sheer volume of computation is vital to understanding the rigor here. They systematically added and removed every possible variable to observe the effect on the model's predictive accuracy.
SPEAKER_03So they build a mathematical model to predict memory loss, and they put PTOG 217 in the model all by itself. They measure how accurate it is, then they add the patient's age to the model, then they add their biological sex, then they add the inflammation marker, GFAP. They measure the accuracy again, then they mute GFAP, take it out of the equation, and put the structural marker NFL in over and over, cycling through 256 different combinations of variables. And in every single combination, they measure a specific metric called the R squared value.
SPEAKER_00Right. The R squared value is the statistical measure of how much of the variance in the outcome in this case, the brain shrinkage or the cognitive decline, can be directly explained by the variables in that specific model.
SPEAKER_03So if a model has an R squared of 0.30, it means it explains 30% of the reason the patient's brain is failing.
SPEAKER_00Correct. By running this tournament, they aren't just seeing if PTA 217 is a good predictor on its own. They are testing if its predictive power survives no matter who else is singing in the choir.
SPEAKER_03If PTA 217 still significantly predicts memory loss even when you unmute GFAP and NFL, then PA217 is dominant. And the researchers categorize this into a strict hierarchy. Complete dominance, conditional dominance, and general dominance.
SPEAKER_00Complete dominance is the holy grail of this statistical tournament. If biomarker A demonstrates complete dominance over biomarker B, it means that biomarker A provided a greater unique contribution to the R squared value in every single one of the 256 possible model combinations, without exception.
SPEAKER_03It beat it every single time.
SPEAKER_00Every single time. Conditional and general dominance indicate a strong hierarchy, but with minor statistical fluctuations in certain subsets.
SPEAKER_03Okay, so the researchers hit enter on the computer, ran this massive 256 subset tournament to find out which blood test is the undisputed champion of predicting brain decay and the results, the ultimate winner.
SPEAKER_00Well, this is where the paper delivers a massive biological plot twist.
SPEAKER_03A huge twist, because the entire dominance hierarchy fractured down the middle, didn't it?
SPEAKER_00It did. The ultimate winner depended entirely on one underlying hidden factor, the amyloid divide.
SPEAKER_01The amyloid divide.
SPEAKER_00The hierarchy of which blood test was best flipped completely upside down, depending on whether the patient had amyloid beta plaques present in their brain.
SPEAKER_03This is wild. This is the crux of the entire study, isn't it?
SPEAKER_00It is. And it represents a foundational shift in how we approach Alzheimer's diagnostics. You cannot blindly apply these advanced blood tests to any patient who walks into a clinic complaining of memory loss.
SPEAKER_03Aaron Ross Powell Because the biology of the brain is fundamentally chemically altered by the presence of amyloid?
SPEAKER_00Yes. You must first stratify the patients by their amyloid status before interpreting the blood markers.
SPEAKER_03Aaron Powell We need to clearly define this for the listener because amyloid is like the biggest buzzword in neuroscience. What exactly is amyloid and how do the researchers definitively prove it was inside these patients' brains?
SPEAKER_00Amyloid beta is a sticky, naturally occurring protein fragment that is produced as a byproduct of normal normal function.
SPEAKER_01So we all have it.
SPEAKER_00We all have it. In a young, healthy brain, the cellular waste disposal systems break down these amyloid fragments and flush them out into the cerebrospinal fluid while you sleep.
SPEAKER_03The blane takes out the trash at night.
SPEAKER_00Exactly. But in Alzheimer's disease, the disposal system fails. The amyloid fragments accumulate, they clump together, and they form hard, insoluble, microscopic plaques in the spaces between the neurons.
SPEAKER_03They essentially gum up the extracellular space.
SPEAKER_00Aaron Ross Powell Yes. And according to the leading theory in the field, the amyloid cascade hypothesis, the formation of these plaques is the primary initiating event of Alzheimer's disease. The plaques form first, and their toxic presence is what eventually triggers the astrocytes to become inflamed, the tau proteins detangle, and the axons to degenerate.
SPEAKER_03It is the biological match that starts the forest fire.
SPEAKER_00Precisely. Now to determine if that match had been lit in these specific patients, the researchers couldn't use a blood test. They had to look directly at the brain.
SPEAKER_03Back to the PDA scans.
SPEAKER_00Back to the PT scans. They subjected the participants to amyloid PT scans. This involves injecting a radioactive tracer that binds exclusively to amyloid plaques and then scanning the brain. They quantify the density of the plaques using a standardized measurement scale called centoloids.
SPEAKER_03Centeloids, what's the scale?
SPEAKER_00It ranges from zero to roughly a hundred. It translates a visual scan into a hard number.
SPEAKER_03And what number means you have Alzheimer's pathology?
SPEAKER_00The clinical consensus is that a centoloid score of 30 represents the threshold of pathology. If a patient's scan showed a centeloid score over 30, they were classified as amyloid positive. The Alzheimer's match had been lit, the plaques were definitively there.
SPEAKER_03And if they scored under 30, they were classified as amyloid negative. And when they split the entire study population into these two distinct camps, the amyloid positives and the amyloid negatives, the results of the 256 subset dominance tournament went in completely opposite directions.
SPEAKER_00Totally opposite.
SPEAKER_03Let's dive into the amyloid positive group first. These are the patients who definitively have the toxic Alzheimer's plaques forming in their cortices. When you run the dominance math on them, who wins?
SPEAKER_00We saw two major winners in this group, heavily dependent on whether we were predicting the hardware or the software. Let's start with predicting the physical shrinkage of the brain, the AD signature ROI thickness.
SPEAKER_01The structural loss.
SPEAKER_00Right. In the South Korean SAMD clinical cohort, the results delivered a major surprise. GFAP demonstrated complete dominance over all the other biomarkers.
SPEAKER_03Wait, GFAP, the temperature gauge. The marker for astrocyte inflammation beat out the tautangles and the structural damage.
SPEAKER_00Yes. GFAP was the undisputed champion of explaining why the cortex was shrinking in the Korean memory clinic patients, and the biological implications of this are profound.
SPEAKER_03Because GFAP is just the cleanup true getting angry.
SPEAKER_00Exactly. Remember what GFA represents. It is the primary structural protein of reactive, angry astrocytes. Astrocytes are highly sensitive to the presence of amyloid plaques. This data strongly suggests that in patients with active symptomatic Alzheimer's pathology, the early physical destruction of their brain tissue might be most directly driven by the toxic neuroinflammation caused by those reactive astrocytes rather than straight structural axonal collapse.
SPEAKER_03That is fascinating. The brain's own cleanup crew is so agitated by the amyloid plaques that their inflammatory response becomes the loudest, most dominant signal predicting the actual melting away of the cortical tissue.
SPEAKER_00It is a striking finding.
SPEAKER_03But that was in the SAMD Clinic cohort. What about the North American ADI observational cohort?
SPEAKER_00In the ADI cohort, when predicting brain shrinkage in the amyloid positive group, the winner shifted. Here, PTO 217 was dominant. Ah. Okay. And crucially, when we move away from predicting hardware and look at predicting the software, the total global cognition scores and the specific memory recall tests, PTA 217, was the undisputed champion across both the Salmon D and ADI cohorts in the amyloid positive groups.
SPEAKER_03So it won for memory in both groups.
SPEAKER_00It demonstrated complete dominance in almost every model for cognitive decline.
SPEAKER_03Let me synthesize this to make sure I'm translating the clinical utility correctly for anyone listening who might be dealing with this in their own family. If, you know, say through a preliminary PEET scan that your parent has amyloid plaques, they are amyloid positive. Right. What this massive statistical tournament tells us is that a high PTAR-217 level in their blood is the loudest, most accurate biological alarm bell indicating that their actual day-to-day memory is going to suffer. And a rapidly rising GFE level tells the doctor that their brain is actively inflamed and physically shrinking.
SPEAKER_00You have articulated the very foundation of precision molecular medicine. If the amyloid
MRI Atrophy Versus Memory Testing
SPEAKER_00pathology is present, PTAL 217 and GFAP are your premium dominant diagnostic tools.
SPEAKER_03They track the specific downstream cascades triggered by the amyloid.
SPEAKER_00Exactly. PTAL 217 tracks the intracellular tangles that directly disrupt the neural networks essential for cognition, and GFAP tracks the toxic extracellular inflammation driving the physical tissue loss.
SPEAKER_03That is half the plot twist, and it provides incredible clarity for Alzheimer's patients. But what about the other half of the data? What about the individuals who came into the clinic? They are experiencing terrifying memory issues. Their MRIs show their cortices are shrinking, but they get the PT scan and their centeloid score comes back at like 12.
SPEAKER_00They are amyloid negative.
SPEAKER_03They are amyloid negative.
SPEAKER_00This is an incredibly critical demographic in neurology, and historically one of the most misdiagnosed. These are individuals experiencing real, devastating cognitive decline and neurodegeneration.
SPEAKER_03Yeah.
SPEAKER_00But it is definitively not being driven by the Alzheimer's amyloid cascade.
SPEAKER_03It's something else entirely.
SPEAKER_00It could be vascular dementia where microstrokes are destroying the tissue. It could be Lewimati dementia, frontotemporal dementia, or degeneration driven by severe. Metabolic disease. The key is that the Alzheimer's match was never lit.
SPEAKER_03So no amyloid fire in the brain. When the researchers ran the brutal 256 subset dominance tournament on this amyloid negative group, who won the battle to predict the brain shrinkage and memory loss?
SPEAKER_00The hierarchy flipped completely. It wasn't the highly specific TAL markers, and it wasn't the astrocyte inflammation marker.
SPEAKER_03Who was it?
SPEAKER_00The undisputed winner was NFL neurofilament light chain.
SPEAKER_03The structural protein, the marker we compared to seeing black oil leaking on the driveway.
SPEAKER_00NFL demonstrated complete dominance over GFAP and both PTAL markers in explaining the variance in brain shrinkage and total cognition across both the Korean and North American cohorts in the amyloid negative groups.
SPEAKER_03The biological logic behind this makes perfect sense when we return to our analogies. If there is no amyloid fire in the house, looking at the smoke detector, GFAP isn't going to give you useful predictive data.
SPEAKER_00No, it won't.
SPEAKER_03And if there is no specific Alzheimer's transmission failure in the engine, the proprietary error code PTAW is totally useless. You need to measure the raw structural rotting of the wood or the volume of oil leaking on the driveway to understand why the house is collapsing.
SPEAKER_00That is the exact underlying mechanism. Without amyloid driving the specific Alzheimer's cascade, the brain is not primarily suffering from hyperphospholated tau tangles or amyloid-induced astrocyte reactivity. It is suffering from a different etiology of cellular destruction.
SPEAKER_03And NFL, being a direct, nonspecific structural component of the axonal cables themselves, becomes a much more trustworthy, universally informative metric for that general catastrophic damage.
SPEAKER_00It is the ultimate diagnostic pivot.
SPEAKER_03If it's not Alzheimer's, the doctor just needs a marker that accurately quantifies how many brain cells are breaking down, regardless of the cause. And NFL provides that quantifiable metric.
SPEAKER_00It does. However, to truly understand the rigor of this paper, we must discuss a fascinating psychometric nuance the researchers uncovered in this amyloid negative group, specifically regarding the memory testing in the North American AD and I cohort.
SPEAKER_03Ah, I have this circled heavily in my notes. The floor effect.
SPEAKER_00The floor effect, yeah.
SPEAKER_03This is a brilliant example of how real-world clinical testing can mess up advanced statistical models. Walk us through the floor effect.
SPEAKER_00In the amyloid negative AD and I group, when the researchers ran the dominance analysis specifically to predict the memory outcome, not total cognition, just the isolated memory test, PTAL 181, actually appeared to dominate over NFL.
SPEAKER_03Wait, that initially contradicts the biological logic we literally just established.
SPEAKER_00It does. But upon deeper investigation, the researchers pointed out a major mechanical flaw in the cognitive test being utilized, the logical memory delayed recall test.
SPEAKER_03This is the test where they read the short story and ask the patient to repeat the details 30 minutes later.
SPEAKER_00Correct. The scoring for this test is based on raw numbers. You get a point for every detail you remember. The critical flaw the researchers noted was that 12 participants in the ADI cohort scored a literal absolute zero on that test.
SPEAKER_03They sat there for 30 minutes and couldn't recall a single word of the story.
SPEAKER_00Nothing. And statistically, this creates what psychometricians call a floor effect. Because the test uses a raw score, it is mathematically impossible to score lower than zero.
SPEAKER_03You can't get negative points.
SPEAKER_00Exactly. But biologically, a patient's neural networks can continue to degrade, and their memory function in the real world can continue to worsen long after they hit zero on that specific pencil and paper test.
SPEAKER_03So the test maxed out its ability to measure their decline, it hit the floor.
SPEAKER_00Exactly. So as the patient's disease progresses, the NFL biomarker in their blood might continue to rise steadily, accurately indicating worsening structural destruction. But the cognitive score stays artificially flat at zero, because the test simply isn't hard enough or sensitive enough at the bottom end to capture the continued degradation.
SPEAKER_03Oh, that is so interesting. It artificially flattens the data curve. The statistics algorithm thinks the patient has stopped getting worse cognitively, which completely scrambles the correlation with the rising blood test.
SPEAKER_00It fundamentally skews the linear regression models utilized in the dominance analysis. It is a vital reminder that biological data generated by ultrasensitive million-dollar Samoa machines is only ever as reliable as the physical or cognitive test we use to anchor it in reality.
SPEAKER_03Wow. How did the South Korean SAMD cohort avoid this floor effect then? Did they use a different test?
SPEAKER_00They used a composite memory score, but more importantly, they analyzed it using Z-scores rather than raw scores. Z-scores. A Z-score is a statistical measurement that describes a value's relationship to the mean of a group of values, measured in terms of standard deviations. Z-scores are statistically normalized and do not have a hard artificial floor of zero.
SPEAKER_03So they can continue to track relative decline much more accurately.
SPEAKER_00And sure enough, in a SAMD group where Z-scores were utilized, NFL demonstrated clear dominance for predicting memory in the amyloid negative group, perfectly aligning with the biological logic.
SPEAKER_03That behind-the-scenes look at data harmonization is fascinating. You can have the most advanced bioengineering in the world, counting single molecules of protein, but if your clinical memory test isn't statistically scaled correctly, your supercomputer math spits out the wrong hierarchy.
SPEAKER_00Science is rarely clean. It requires constant contextual evaluation of the tools being utilized.
SPEAKER_03So we have established the primary groundbreaking findings of this study. If the patient has amyloid plaques, PTO 217 and GF8 are the dominant predictors for determining cognitive decline and the physical shrinkage of the brain. If the patient does not have amyloid plaques, NFL is the most reliable guide.
SPEAKER_00That is the core finding, yes.
SPEAKER_03But earlier, I brought up a major concern. I argued that because the Korean SAMD group was pulled from a clinic and heavily symptomatic, and the North American ADNI group was mostly healthy volunteers, the entire comparison was an apples to oranges scenario that might invalidate
Collinearity And Dominance Analysis
SPEAKER_03the mathematical results.
SPEAKER_00The authors of the study recognized that exact vulnerability. Good science never stops at the first exciting statistical output. The researchers aggressively stress tested their own data.
SPEAKER_03Right. They had to prove it wasn't just a fluke.
SPEAKER_00They ran three highly specific, incredibly rigorous sensitivity analyses to ensure that the demographic and clinical disparities between the two cohorts weren't secretly generating false dominance patterns.
SPEAKER_03Let's walk through these stress tests because this is where the paper earns its credibility. Sensitivity analysis number one. Adjusting for CDRSB. Explain what that means in practice.
SPEAKER_00CDRSB stands for the Clinical Dementia Rating Sum of Boxes. It is a highly detailed, globally recognized clinical metric used to quantify the absolute severity of a patient's dementia. It evaluates memory, orientation, judgment, community affairs, home and hobbies, and personal care. By mathematically forcing the linear regression models to adjust for this specific score, the researchers were essentially demanding that the algorithm ignore the fact that the SAMD group was, on average, clinically sicker.
SPEAKER_03They forced the math to level the playing field clinically.
SPEAKER_00Yes. They asked the models, if we assume the disease severity is equal, do these biomarkers still behave the same way?
SPEAKER_01And did they?
SPEAKER_00The core findings held beautifully. Even after mathematically adjusting for disease severity, GFI still demonstrated dominance for predicting brain shrinkage in the amyloid positive SAMD participants, and PTA 217 remained dominant in ADNI.
SPEAKER_03And the amyloid negative group.
SPEAKER_00Importantly, NFL maintained its dominance in the amyloid negative groups across the board.
SPEAKER_03Okay. The data survives test one. What was sensitivity analysis number two?
SPEAKER_00They escalated the rigor. Instead of just mathematically adjusting for severity across the whole population, they physically isolated a very specific narrow subgroup of patients from both cohorts, those with a global CDR score of exactly 0.5.
SPEAKER_03A global CDR of 0.5. That indicates very mild cognitive impairment, right? It's the liminal space between healthy aging and active dementia.
SPEAKER_00Exactly. By isolating this group, they stripped out all the completely healthy volunteers from ADI, and they stripped out all the severely demented patients from SAMD. They were left looking only at the patients sitting at the exact same, very mild stage of clinical progression.
SPEAKER_03Truly comparing apples to apples this time.
SPEAKER_00This yielded a subgroup of 106 people from the Korean clinic and 116 from the North American trial.
SPEAKER_01And when they ran the math on just them?
SPEAKER_00Once again, the amyloid-dependent dominance patterns remained structurally intact. PTO 217 and GFAV remained the strongest predictive signals in the amyloid-positive individuals. NFL remained a very strong signal in the amyloid negative group. Although because the sample size was significantly reduced, some of the statistical dominance shifted from complete dominance to conditional dominance.
SPEAKER_03But the overarching hierarchy dictating clinical use remained entirely intact.
SPEAKER_00Completely intact.
SPEAKER_03That provides incredible reassurance. But they deployed a third stress test, and this one involves my absolute favorite statistical methodology, sensitivity analysis number three, propensity score matching, or PSM.
SPEAKER_00PSM is one of the most powerful analytical techniques available in observational research when dealing with disparate populations.
SPEAKER_03I want to break down how PSM actually works because it's brilliant. It's essentially an algorithm designed to create statistical twins across two totally different populations.
SPEAKER_00That is the exact goal. The algorithm calculates a specific probability score for every single participant based on a set of defined covariates in this case, their age, biological sex, years of formal education, and their precise disease severity.
SPEAKER_03Let's visualize this. Imagine you have a 75-year-old Asian woman in the South Korean Samadhi cohort who has 12 years of education and mild cognitive impairment. The PSM algorithm scans the entirety of the North American ADI cohort looking for her sadistical twin. It searches for a 75-year-old white woman in AB and I with exactly 12 years of education and the exact same clinical impairment score. It matches them up one-to-one, or in the case of this study, one to two, creating pairs or triplets that look virtually identical on a clinical spreadsheet despite living on different continents.
SPEAKER_00It is a way of artificially constructing a perfectly balanced randomized controlled trial out of retrospective observational data.
SPEAKER_01Which is amazing.
SPEAKER_00They utilized PSM to successfully match 61 SAMD participants with 103 AD and I participants, completely erasing the demographic and clinical disparities that defined the raw cohorts.
SPEAKER_03They forced the apples and oranges to become identical pieces of fruit. And when they ran the massive 256 subset dominance analysis tournament on this new, perfectly balanced group of statistical twins, let me guess the plot twist survived again.
SPEAKER_00It did. Even when the demographic variables and clinical severity were almost entirely normalized via propensity score matching, the fundamental biology held firm. Pito 217 and GFE ruled the amyloid positive group for predicting both brain shrinkage and total cognition. And NFL ruled the amyloid negative group.
SPEAKER_03Reading a study this dense, there is always a part of me looking for a flaw in the methodology so I can say, aha, the math is fake. But this exhaustive sequence of stress testing, adjusting for the sum of boxes, isolating the 0.5 impairment group, and finally running propensity score matching it, gives us immense foundational confidence in the findings.
SPEAKER_00It is incredibly robust.
SPEAKER_03This hierarchy is not a statistical fluke generated by the fact that one group went to a local clinic and the other signed up for a massive research trial. This represents a fundamental, unshakable biological truth about how human brains react to neurodegenerative disease.
SPEAKER_00It proves that the underlying molecular pathology dictates the utility of the biomarker, regardless of the patient's ethnic background, their education level, or the clinical setting in which their blood is drawn.
SPEAKER_03So we have explored the biology, we have praised the study design, and we have stress tested the math. But we must discuss the horizon. What are the limitations of this data? What could this specific study not tell us? Because no research is perfect.
SPEAKER_00The most glaring limitation of this paper is that it relies entirely on cross-sectional data.
SPEAKER_03Aaron Powell Meaning it is a snapshot frozen in time.
SPEAKER_00Yes. The researchers drew the blood, measured the cortical thickness on the MRI, and administered the memory tests all within roughly the same temporal window. What they did not do is track these specific individuals longitudinally over a period of five, ten, or twenty years.
SPEAKER_03Aaron Powell Why is longitudinal data so crucial for blood biomarkers?
SPEAKER_00Because Alzheimer's is a disease of decades. To truly definitively validate these blood markers for long-term clinical prediction, we need to prove causality over time. We need to measure a perfectly healthy 50-year-old's PTA 217 levels today, and then prove that a specific elevation in that marker accurately predicts the physical shrinkage of their enterinal cortex 10 years from now.
SPEAKER_03Cross-sectional data proves strong correlation in the present moment. Longitudinal data proves predictive destiny.
Amyloid Status Flips The Winners
SPEAKER_00Exactly.
SPEAKER_03We need to see the movie, not just a photograph. What else is missing? I noticed the paper didn't spend much time discussing other bodily systems.
SPEAKER_00That is a highly relevant point. The researchers did not aggressively account for systemic metabolic comorbidities.
SPEAKER_02Like severe obesity, advanced heart disease, or type 2 diabetes.
SPEAKER_00Exactly. The human brain does not exist in a vacuum. It is deeply connected to the cardiovascular and renal systems. Chronic kidney disease, for instance, significantly impairs the body's ability to filter and clear proteins from the blood.
SPEAKER_03Wait, so a patient with poor renal function might show artificially inflated levels of NFL or PITO 217 in a blood draw simply because their kidneys aren't clearing the proteins efficiently. Yes. Not because their brain is rapidly degenerating.
SPEAKER_00Correct. So a failing kidney could cause a false positive for Alzheimer's on a blood test. It is a distinct possibility that future diagnostic models will need to mathematically account for.
SPEAKER_02That's a huge caveat.
SPEAKER_00Furthermore, severe cardiovascular disease impacts the microvascular health within the brain itself, causing chronic oxygen deprivation and subsequent tissue death that operates entirely independently of the amyloid cascade.
SPEAKER_03So future iterations of these dominance models will absolutely need to factor in a patient's overall metabolic panel to truly refine the accuracy of these neural blood tests.
SPEAKER_00They absolutely will.
SPEAKER_03And there was one final limitation mentioned at the very end of the paper regarding a completely new biomarker that wasn't included in this tournament, which sounds like it might be the next major disruptor in the field.
SPEAKER_00Ah, yes. They briefly noted the absence of a relatively new marker called TA243. Specifically, plasma endogenous microtubule binding region TA243.
SPEAKER_03That is a mouthful. What makes TA243 different from the PTA 217 we've been discussing?
SPEAKER_00Well, while PTA 217 is proving to be incredibly sensitive at detecting the early presence of amyloid plaques and the initial formation of tangles, emerging cellular research suggests that TA243 might be far more tightly correlated with the actual physical spread of the tail pathology throughout the cortex and the later stages of the disease.
SPEAKER_03It tracks the progression of the fire, not just the initial spark.
SPEAKER_00Exactly. Some early data implies it might be a superior marker for predicting real-time cognitive decline once the disease is firmly established.
SPEAKER_03So if researchers were to rerun this exact same 256 subset dominance tournament next year and introduce Tatu 43 into the models, it is entirely possible that it might dethrone some of the current statistical champions we've discussed today.
SPEAKER_00It is entirely possible. The pace of biological engineering is just moving incredibly fast.
SPEAKER_03So to summarize the limitations, we desperately need decades of longitudinal tracking. We need complex algorithms that adjust for kidney and heart function, and we need to continuously test the newest engineered markers like TA243. But even acknowledging those hurdles, the consensus established in this paper provides incredibly clear guidance on how these blood tests should be deployed right now in a clinical setting.
SPEAKER_00The overarching clinical takeaway is clear. These specific advanced blood tests, GFEP, PTAL 217, and NFL, are optimized for use after a patient has had their amyloid status confirmed.
SPEAKER_03They are not the very first screening tool a primary care doctor should run.
SPEAKER_00Not purely in isolation, no. A physician might utilize a highly accurate preliminary blood test designed solely to predict amyloid presence, or they might still rely on an amyloid PT scan if available.
SPEAKER_03But once that binary question is answered, once you know for certain that the patient is amyloid positive, you immediately deploy GFFP and PTA 217 as highly precise secondary staging tools.
SPEAKER_00You use them to map out the exact severity of the downstream damage.
SPEAKER_03To stage the disease, to quantify exactly how far the toxic inflammation has spread and how severe the tangled burden is.
SPEAKER_00And crucially, you use them to monitor the efficacy of pharmacological interventions. We are entering an era of novel anti-amyloid therapies. New drugs. If you prescribe a patient one of these new drugs, you can draw their blood every three months and watch their GFAP and Beta-217 levels drop in real time, biologically proving that the neuroinflammation is subsiding and the brain is beginning to stabilize.
SPEAKER_03Let's bring this all together for everyone listening. Let's wrap up this massive journey through neurobiology and statistics.
SPEAKER_00It's been quite a journey.
SPEAKER_03We started this deep dive with the hope of a simple single blood test arriving to replace the painful spinal taps and inaccessible PE scans. And what we've discovered is that properly measuring the decay of the human brain isn't about finding one single magic bullet blood test.
SPEAKER_00No, it is entirely about context.
SPEAKER_03It is about stratifying the underlying molecular pathology before you attempt to interpret the biomarkers.
SPEAKER_00If the amyloid plaques are present, if the specific Alzheimer's matches lit inside the brain, we look to GFAP to quantify the physical structural damage being driven by toxic astrocyte inflammation, and we look to PTA 217 to predict the terrifying cognitive decline driven by the collapsing neuronal supply chains.
SPEAKER_03But if amyloid is completely absent, if a patient is suffering from some other form of neurodegeneration, NFL stands alone as the most trustworthy, dominant guide to quantifying their structural collapse.
SPEAKER_00It is a sophisticated diagnostic triaging system built entirely on molecular biology and validated by rigorous statistical modeling.
SPEAKER_03And the reason we spend an hour deep diving into the cellular behavior of astrocytes and the intricacies of propensity score matching is because we are looking at a blueprint for the immediate future of neurology.
SPEAKER_00We really are.
SPEAKER_03When you or your parent or a loved one visits a memory clinic in the coming years, this exact stratified logical framework will dictate the medical care you receive. The era of precision medicine for
Limits Today And What Comes Next
SPEAKER_03the brain has definitively arrived.
SPEAKER_00We are no longer treating dementia as a vague, terrifying umbrella term.
SPEAKER_03We are utilizing atomic level diagnostic tools to target the exact misfolding protein, causing the exact localized damage in your unique brain.
SPEAKER_00This approach ultimately empowers both the patient and the physician with highly actionable, biologically specific data. But, you know, as we close this analysis, contemplating the sheer sensitivity of this technology does raise one final profound question for the future.
SPEAKER_03Oh, let's hear it.
SPEAKER_00We have established today that a few drops of peripheral blood drawn from your arm can accurately, statistically reflect the incredibly complex, localized cellular destruction happening inside unique microregions of your brain, like the entropinal cortex. Right. If our Samoa technology is already this sensitive, what other systemic secrets is our blood quietly hiding about the architecture of our minds?
SPEAKER_01What else is floating in the plasma?
SPEAKER_00Exactly. Could advanced blood biomarkers one day be used not just to stage neurodegeneration that has already begun, but to accurately predict our baseline cognitive potential or quantify our brain's resilience to future stress decades before the very first symptom of aging ever appears?
SPEAKER_03That is a staggering thought to end on. Are the biological clues to our mental destiny already circulating in our veins when we are 40 years old and when we are 30?
SPEAKER_00The biological data is almost certainly there, circulating right now. It is merely a matter of our engineering and statistical models catching up to decode it.
SPEAKER_03Well, until that day comes, I will gladly take a simple blood draw over a broken system of murky diagnoses and spinal taps any day of the week.