A five minute strip of heartbeat data looks like nothing much: a row of small intervals, each one a fraction of a second longer or shorter than the last. That jitter is heart rate variability, and for two decades it has been pitched as a cheap window onto the aging brain, on the logic that the vagus nerve and the cortex are wired into the same regulatory loop. Last December the pitch took a hard hit. Sofia Marcolini and colleagues at the University Medical Center Groningen published a study of 269 cognitively unimpaired adults, selected by their coronary artery calcium score, whose baseline heart rate variability was tested against cognitive and brain outcomes measured roughly eight and thirteen years later. The finding, in their own title, was limited evidence. HRV at baseline, and the change in it over time, were not associated with the outcomes; the one signal that survived, in people with heavy coronary calcification, ran in a direction that surprised the authors. This section reported that verdict as it stood: a marker that looked promising and mostly did not deliver.
What arrived on 13 March 2026 is not a rebuttal but a reframe. Writing in the Journal of Alzheimer's Disease, a group of thirteen researchers led by Benson Opare Asamoah Botchway and Nicoletta Nicolaou, working across the University of Nicosia Medical School and the Cyprus Institute of Neurology and Genetics, published a narrative review of evidence from 2000 to 2025 on using electroencephalography and heart rate variability together, as one integrated biomarker rather than two competing ones. Their case rests on the central autonomic network, the distributed set of structures including the insula, anterior cingulate cortex and amygdala that coordinates thinking, feeling and the automatic housekeeping of the body. EEG catches one face of a failing network, in the slowing they describe as reduced alpha and increased theta activity. HRV catches the other, in a heart that has lost some of its flexibility. The review's central claim is that studies coupling the two distinguish mild cognitive impairment and early Alzheimer's more accurately than either measure achieves alone.
What that looks like in practice is a machine learning problem more than a clinical one. Among the work the review draws on is a 2024 paper in Digital Signal Processing that detected mild cognitive impairment from EEG and HRV recorded together during a standard cognitive battery, reporting average accuracy of 93.86 percent, sensitivity of 93.87 percent and specificity of 93.53 percent with a hybrid model. Numbers in the nineties are the kind that make headlines, and they deserve a cold read. Telling apart people who already carry a diagnosis from people who do not is a fundamentally different task from forecasting, in a room full of healthy 55 year olds, which of them will be different in thirteen years. The Groningen study attempted the second task. The classifier studies attempt the first. A review that stacks them beside each other is making an argument about mechanism, not delivering a test.
The review is candid about why that test does not exist yet. It identifies methodological gaps that currently limit reproducibility: recording conditions that differ from lab to lab, and no standardized way of analyzing the coupled signal. This is a genuine problem rather than a formality. Heart rate variability shifts with posture, breathing, caffeine, sleep, medication and the time of day; EEG shifts with electrode montage, eyes open or closed, and how a lab defines its frequency bands. Two studies can both report that brain and heart coupling breaks down in Alzheimer's and mean measurably different things by it. In response the authors propose a translational roadmap with short, mid and long term goals covering clinical validation, wearable integration and digital health applications, which is an honest way of saying that the clinic is at the far end of the road, not the near one.
For a family, the practical content of this is thin and worth stating plainly. Nobody is being offered a brain and heart reading that says yes or no. A narrative review does not pool data, does not weight studies by quality, and cannot generate a number you could apply to one person in one consulting room. What it does offer is a different way of thinking about a disappointment. The Groningen result did not show that the autonomic nervous system is irrelevant to dementia; it showed that one summary statistic, taken once and read in isolation, did not forecast much. Botchway and colleagues are betting that the information was never in the heartbeat alone, but in the relationship between the heartbeat and the cortex, and that measuring one half of a conversation was always going to sound like noise.
That bet is testable, which is the most useful thing about it. The specific study it demands has not been run: a large group of unimpaired people, EEG and ECG recorded at the same time under a protocol everyone agrees on, followed for a decade against the blood and imaging markers that the Groningen team already collected. Until someone does that, coupled EEG and HRV is a promising hypothesis with a roadmap attached, and the honest headline is that we still do not know.