The signal increased steadily in years before symptoms appeared
In the long struggle to catch Alzheimer's disease before it steals the self, researchers have found a new kind of signal written in the blood — circular RNA molecules that rise quietly in the years before memory falters. A study published in Nature Medicine, drawing on samples from over 1,200 individuals, suggests that these molecular loops, combined with an existing tau biomarker, can identify early Alzheimer's pathology with near-clinical precision. The promise is not a cure, but something almost as consequential: the chance to intervene while the mind is still whole.
- A 34-circRNA blood signature detected early Alzheimer's pathology with 95% accuracy — surpassing the established pTau217 biomarker, which managed 88% on the same task.
- When circRNAs were combined with pTau217, accuracy climbed to 97–98%, and the model outperformed even amyloid-PET imaging in predicting who would progress from cognitively normal to symptomatic disease.
- The stakes are high: Alzheimer's pathology begins decades before symptoms appear, and current early-detection methods — lumbar punctures and costly PET scans — are invasive, expensive, or inaccessible at scale.
- The circRNA signal appears largely specific to Alzheimer's, showing low predictive value for Parkinson's, Lewy body dementia, and frontotemporal dementia, and held consistent across sex, ancestry, and genetic risk subgroups.
- The study remains largely retrospective, and researchers caution that prospective validation in larger, more diverse cohorts is essential before this approach could enter routine clinical practice.
A new study published in Nature Medicine has identified 34 circular RNA molecules in blood that can detect Alzheimer's disease years before cognitive symptoms emerge — a finding that could reshape how medicine approaches one of its most stubborn adversaries. Analyzing blood samples from more than 1,200 people, including hundreds with confirmed Alzheimer's and hundreds who were cognitively normal, researchers found that these circRNAs rose steadily in the years preceding symptom onset, functioning as a kind of biological early-warning system.
Tested alone, the 34-circRNA model achieved 95% accuracy in distinguishing early Alzheimer's pathology from healthy aging — outperforming phosphorylated tau-217, a biomarker already used in clinical settings, which reached 88% on the same task. The real breakthrough came in combination: pairing the circRNA signature with pTau217 pushed accuracy to 97–98%, and in a separate cohort of cognitively normal participants, the model predicted future progression to symptomatic disease better than amyloid-PET imaging, long considered the gold standard.
The significance lies partly in what it could replace. Early Alzheimer's detection currently depends on lumbar punctures or expensive PET scans — barriers that limit who gets screened and when. A blood test is simple, scalable, and cheap. The circRNA signal also appeared specific to Alzheimer's, showing little predictive power for Parkinson's disease, Lewy body dementia, or frontotemporal dementia, and results held broadly consistent across sex, ancestry, and genetic risk groups.
The researchers are careful about the limits of their work. The study was largely retrospective, drawing on stored samples from existing research cohorts. Prospective trials — following healthy individuals forward over time to see whether circRNA levels predict who develops symptoms — will be needed before this approach can move into clinical practice. Questions also remain about how common comorbidities like cardiovascular disease or diabetes might influence circRNA levels. But if the findings hold, the possibility of identifying Alzheimer's pathology early enough to intervene — before irreversible damage sets in — could prove genuinely transformative.
Researchers have identified a set of circular RNA molecules in blood that can detect Alzheimer's disease years before a person shows any cognitive decline, potentially opening a new avenue for early intervention in a disease that currently offers no cure. The finding, published in Nature Medicine, suggests that measuring these circRNAs alongside existing biomarkers could help doctors identify who is at highest risk of developing dementia and who might benefit most from emerging treatments.
The study analyzed blood samples from over 1,200 people, including 405 with confirmed Alzheimer's disease and 816 cognitively normal adults. Using RNA sequencing and other molecular techniques, the researchers isolated 34 distinct circRNAs—small loops of genetic material—that showed a strong association with Alzheimer's status. What made these findings particularly striking was that the signal from these circRNAs increased steadily in the years leading up to symptom onset, suggesting they could serve as an early warning system. When the team tested their 34-circRNA model alone, it achieved 95 percent accuracy in distinguishing between people with early Alzheimer's pathology and those without. That outperformed phosphorylated tau-217, a blood biomarker already in clinical use, which achieved 88 percent accuracy on the same task.
But the real power emerged when the researchers combined the circRNA signature with the tau-217 measurement. That hybrid approach reached 97 to 98 percent accuracy—a level of precision that could meaningfully change how doctors screen for the disease. Among participants in a separate cohort who were cognitively normal at the start of the study, the circRNA model proved better at predicting who would eventually progress to symptomatic disease than either tau-217 alone or amyloid-PET imaging, a costly scan that has long been considered the gold standard for detecting Alzheimer's pathology in the brain.
The implications are substantial. Alzheimer's disease is the leading cause of dementia, and the pathological changes that define it—accumulation of amyloid plaques and tau tangles—begin years or even decades before memory loss becomes noticeable. If doctors could identify those changes early, patients could potentially start treatment before irreversible cognitive damage occurs. Current approaches to early detection rely on either lumbar puncture to sample cerebrospinal fluid, an invasive procedure, or positron emission tomography scans, which are expensive and not widely available. A blood test, by contrast, is simple, scalable, and inexpensive.
The circRNAs identified in this study appear to be preferentially expressed in the brain, though the researchers could not definitively prove that the circRNAs detected in blood actually originate there. What matters clinically is that they correlate with Alzheimer's disease status and progression risk. The circRNA signal also seemed to capture aspects of disease biology that other biomarkers miss—the model's scores were associated with dementia severity and could track the dynamic progression of cognitive decline over time. When the team tested whether the circRNA signature was specific to Alzheimer's or would light up in other neurodegenerative diseases, they found it was largely selective for Alzheimer's, showing low predictive value for Parkinson's disease, dementia with Lewy bodies, or frontotemporal dementia.
The findings held up across different populations. Sensitivity analyses stratified by sex, ancestry, and genetic risk factors yielded consistent results, suggesting the circRNA signature might be robust across diverse groups—though the researchers noted that some ancestry subgroups in their analysis were small and would benefit from larger representation in future studies. This is an important caveat, as many biomarker studies have been conducted primarily in European ancestry populations, and it remains unclear whether findings always generalize.
Still, the work is preliminary. The researchers themselves emphasize that these findings need validation in larger, prospective clinical cohorts before circRNA testing could move into routine clinical practice. The current study was largely retrospective, analyzing stored blood samples from people already enrolled in research studies. A prospective trial—following cognitively normal people forward over time and seeing whether circRNA levels predict who develops symptoms—would provide stronger evidence. The researchers also note that future work should explore how other conditions that often accompany Alzheimer's disease, such as cardiovascular disease or diabetes, might affect circRNA levels in blood.
If the findings hold up in larger studies, the clinical landscape for Alzheimer's detection could shift significantly. Doctors could potentially offer a simple blood test to identify people with early Alzheimer's pathology, enabling enrollment in clinical trials of new treatments or initiation of approved therapies before cognitive symptoms appear. For a disease that has resisted treatment for decades, that kind of early detection capability could prove transformative.
Notable Quotes
CircRNAs are highly stable, tissue-specific, and can be measured in blood, offering a non-invasive alternative to cerebrospinal fluid sampling or expensive PET scans— Study findings
The Hearth Conversation Another angle on the story
Why does it matter that these circRNAs show up in blood years before symptoms? Couldn't doctors just wait until someone has memory problems and test them then?
Because by the time memory problems appear, the damage is often irreversible. The brain has already lost significant numbers of neurons. If we can identify the disease process while it's still in the preclinical stage, we have a window to intervene with treatments that might slow or halt progression. That window closes once cognitive decline becomes obvious.
But the study combined circRNAs with tau-217 to get the best results. Doesn't that mean circRNAs alone aren't enough?
They're complementary. CircRNAs alone outperformed tau-217 alone, which was surprising. But together they captured different aspects of the disease biology. CircRNAs seem to reflect broader changes in how the brain is functioning, while tau-217 is a more direct measure of one specific pathological protein. The combination gives you a fuller picture.
The researchers say they couldn't prove the circRNAs come from the brain. So how do we know they're actually measuring Alzheimer's in the brain and not something else?
That's a fair question, and it's why validation matters. What we know is that these circRNAs correlate strongly with Alzheimer's disease status and predict progression. They're also preferentially expressed in brain tissue. Whether they're literally brain-derived or whether they're a downstream consequence of brain pathology—either way, they're tracking something real about the disease.
The study looked at over 1,200 people, but only 405 had Alzheimer's. Isn't that a small sample for something this important?
It's a reasonable starting point, but you're right that larger studies are needed before this moves into clinical practice. The researchers were explicit about that. What's encouraging is that the findings replicated across multiple independent cohorts—the Knight ADRC and the A4 study—which suggests the signal is real and not just a statistical fluke in one dataset.
What happens to someone who gets tested and finds out they have early Alzheimer's pathology but no symptoms yet?
That's the practical question that matters most. Right now, there are a few FDA-approved monoclonal antibodies that target amyloid, and they show modest slowing of cognitive decline in early symptomatic disease. Whether they work better in truly asymptomatic people is still being studied. So a positive test would likely lead to enrollment in a clinical trial or close monitoring, not necessarily immediate treatment. But that could change as more effective therapies emerge.