Blood pressure and dementia: what the pooled cohorts can and cannot separate
An individual participant data meta-analysis across the COSMIC cohorts links late-life blood pressure and antihypertensive use to dementia risk, and Fuchs and Whelton argue the association with cardiovascular disease is causal. Neither is a trial of lowering your blood pressure to prevent dementia, and the difference matters more than either result.
mixedStudies disagree, or the mechanism is contested.Revised
The measurement is easy and the question is not. Two documents, one consensus on what they do not establish.
Blood pressure is the risk factor with the least ambiguity attached to it. It is measurable, it moves, drugs lower it, and lowering it prevents strokes. The argument for treating it causally is in Hypertension, written by Fuchs and Whelton, and it is one of the cleaner cases in preventive medicine.
The question this article is actually about is the one that follows: if high blood pressure is causal for cardiovascular disease, what follows for dementia? Two documents, and the gap between them.
The cardiovascular case, stated as a claim rather than a hunch
Fuchs and Whelton do not hedge about the framing. They write that fragmented investigation has masked the overall picture for the causes of cardiovascular disease, and that among the risk factors, high blood pressure is associated with the strongest evidence for causation and has a high prevalence of exposure.
The biological argument is a distributional one, and it is worth following because it is unusual. They argue that cardiovascular disease is primarily caused by a right-sided shift in the population distribution of blood pressure — not by a subgroup with extreme values, but by the whole curve moving. That is why, they write, normal levels of blood pressure are considerably lower than what has typically been characterised as normal in research and clinical practice.
Then the mechanism is listed across outcomes: large cohort studies have shown high blood pressure is an important risk factor for heart failure, atrial fibrillation, chronic kidney disease, heart valve diseases, aortic syndromes and dementia, in addition to coronary heart disease and stroke.
And then the step that makes it a causal argument rather than an association. In multivariate modelling, the presumed attributable risk of high blood pressure for stroke and coronary heart disease has increased steadily as progressively lower values have been treated as normal. Meta-analysis of blood-pressure- lowering randomised controlled trials has demonstrated a benefit almost identical to that predicted from the cohort risk relationships.
That last sentence is the load-bearing one. It is the agreement between two independent kinds of evidence — observation predicts, intervention delivers. Where those two agree, causation is a reasonable inference.
The dementia evidence, pooled at the participant level
The 2024 Neurology paper is an individual participant data meta-analysis from COSMIC, the Cohort Studies of Memory in an International Consortium. Individual participant data is the strongest form of pooling available short of a trial: instead of taking published summary estimates from each cohort, the analysts work with the underlying records, which lets them adjust consistently and check for differences between cohorts that a summary-level meta-analysis would smooth away.
Its stated background is that previous randomised controlled trials and longitudinal studies have indicated that ongoing antihypertensive use in late life reduces all-cause dementia risk.
That framing is important, and it is also where the care is needed. “Previous randomised controlled trials have indicated” is a statement about a body of evidence that the paper takes as given and builds on. It is not a new trial. The paper adds pooled observational data at a scale and quality that makes the association difficult to dismiss — and the title reflects exactly that, which is risk, antihypertensive use, and late-life Alzheimer and non-Alzheimer dementia risk.
Where the two documents stop
Here is the honest division, and it is the reason this article exists.
The cardiovascular claim rests on intervention agreement. The trials that lower blood pressure show the predicted benefit. That is as close to causal proof as observational medicine gets.
The dementia claim rests on association. COSMIC is a consortium of cohorts. No blood-pressure-lowering trial has been powered, or designed, or run, with dementia as its primary endpoint. That trial does not exist. The individual participant data makes the association more credible; it does not convert it.
The direction of causation is not fully resolved for the brain. Blood pressure rises with age, vascular damage accumulates over decades, and the timing of exposure relative to the disease is exactly the thing a late-life cohort cannot nail down. This is the standard weakness of the design and it applies here.
Surrogate chains weaken twice. Blood pressure is a surrogate for cardiovascular events. If it is also a surrogate for dementia, that is a longer chain with more places to fail, and no trial has followed it to an outcome.
What this literature does not show
Not that lowering blood pressure prevents dementia. No trial has tested it. The cardiovascular trials did not measure dementia as their primary endpoint.
Not that the effect size in the brain matches the heart. A hazard ratio from a cohort is not interchangeable with the event reduction from a blood-pressure trial, and the two documents do not produce one common number — they produce two different kinds of evidence about two different outcomes.
Not that a normal reading in old age is protective. The Fuchs and Whelton argument is about population distributions and attributable risk, not about an individual threshold at which treatment becomes worthwhile.
Not that the mechanisms match. Antihypertensives lower pressure. Whether the pathway from pressure to neuron matters is a separate question from whether the pathway from pressure to coronary event does.
Not that the telomere literature settles it either. When this topic is discussed next to aging, the compound that arrives is Epithalon, and its record has the same shape as the two documents above: telomerase reactivated and telomeres elongated in a dish, longer-lived flies and mice, and no modern human trial. Worth knowing when the telomere papers get cited in a dementia discussion, because that is where the extrapolation happens.
The sentence worth keeping
High blood pressure is probably causal for cardiovascular disease, and the evidence for that is unusually strong because prediction and intervention agree. For dementia, the same risk factor shows the same association at unusually high data quality. Those two sentences are true together, and neither one licenses the other. Anyone who merges them is claiming a trial result that has not been run.