Plasma Exchange: Young Blood Without the Young Blood

The full story

The cleanest experiment in this whole field is the one nobody can repeat in a person. In mice, researchers surgically joined the circulatory systems of a young animal and an old one - heterochronic parabiosis - and watched aged muscle stem cells wake up: Notch signalling restored, proliferation restored, regenerative capacity restored, with aged liver cells dividing again as well. That 2005 paper in Nature is the founding result behind every young-blood pitch you have ever scrolled past. It is also an animal experiment resting on a surgery no human will ever undergo.

Then the field turned its own question inside out. A 2016 mouse study in Nat Commun exchanged blood between young and old animals without sharing organs, and found effects in muscle, liver and hippocampus within a few days - with the inhibitory effect of old blood, in many cases, more pronounced than the benefit of young blood. If the trouble is what is in old blood rather than what is missing from it, you do not need a donor. You need a drain. That single pivot is why the clinics now sell plasma exchange rather than transfusions, and it is exactly where the human evidence stops being tidy.

Subtraction, not addition

The consequential mouse result came in 2020 in Aging (Albany NY): replacing half an old animal's plasma with saline plus 5% albumin - no young blood at all, a procedure the authors call neutral blood exchange - met or exceeded the effects of heterochronic exchange on muscle repair, liver adiposity and fibrosis, and hippocampal neurogenesis. The authors also note that ectopically added albumin does not appear to be the determinant, which points at the removal itself rather than at anything put back in.

A 2021 mouse study in Geroscience pushed it into behaviour: a single neutral blood exchange improved novel-object and whisker-discrimination performance in two-year-old animals and reduced CD68-positive activated microglia. In the same work the senolytic ABT-263 had limited effect on neuroinflammation and did not enhance hippocampal neurogenesis, which argues the mechanism is dilution rather than clearance of senescent cells. Elegant - and so far, a mouse hypothesis.

The one large human trial, and the endpoint it missed

Plasma exchange has been tested at scale against a brain disease in one major programme. A phase 2b/3 randomized, sham-controlled trial of plasma exchange with albumin replacement in Alzheimer's disease, reported in 2020 in Alzheimers Dement, randomized 347 patients out of 496 screened. Its registry record, filed by sponsor Instituto Grifols, prespecified two co-primary outcomes at 14 months. It met one: the functional scale ADCS-ADL, at P=.03, with 52% less decline. It missed the other: the cognitive scale ADAS-Cog landed at P=.06, a trend and nothing more. In patients with mild disease and a baseline MMSE of 22 to 26, there were no changes at all.

Everything downstream inherits that miss. Secondary analyses published in 2022 in Alzheimers Dement reported improved language fluency and processing speed in the mild cohort and improved short-term verbal memory in the moderate cohort - while progression of neuropsychiatric symptoms was no different from placebo. Those are subgroup-split secondary endpoints inside a trial whose cognitive co-primary was not met. Imaging substudies reported in 2022 in Eur J Nucl Med Mol Imaging, covering MRI in 198 participants and FDG-PET in 213, found fewer subcortical volume changes and less metabolic decline in the high-albumin-plus-immunoglobulin arm than in placebo. Exploratory, and labelled as such.

Two human trials in 2025, pointing opposite ways

In 2025, two human trials measured epigenetic clocks after plasma exchange and disagreed about the direction of the effect.

The first, in Aging Cell, was a small single-blinded placebo-controlled trial in 42 healthy adults over 50, with randomization based on entry date rather than a concealed sequence. It reported 15 epigenetic clocks showing rejuvenation versus placebo at a false-discovery rate under 0.05, with biweekly plasma exchange plus immunoglobulin the most effective schedule, and two adverse events requiring discontinuation.

The second, in Sci Rep, was a randomized crossover trial of eight versus four plasmaphereses over 18 weeks in healthy blood donors. It found no epigenetic rejuvenation whatsoever. What it found instead were increases - in DNAmGrimAge, the Hannum clock and the Dunedin Pace of Aging - and its authors concluded that the protocol "may accelerate epigenetic aging."

Both are human trials. Both are randomized on paper. One is single-blinded with date-based allocation; the other is a crossover on a different schedule in a different population. When two small human studies of the same procedure cannot agree on whether it moves a biomarker up or down, the honest reading is not "one of them is right." It is that the biomarker is not yet doing the job anyone needs it to do. A settled field does not publish both of these in the same year.

What the evidence does not show

Nothing in this literature shows that plasma exchange makes a healthy person live longer, function better, or fall ill less often. The Alzheimer's programme measured cognition and daily function in people with a diagnosis and split its verdict. The biological-age studies measured clocks, not outcomes - and the registry for the biological-age plasma-exchange study now under way says so out loud: 40 participants, with adverse events and the rate of change in a blood epigenetic clock as its prespecified primary outcomes. That is a biomarker study by design. It was also registered after the study had already started in September 2022, which is worth knowing before anyone reads its results as pre-planned.

There is also no validated protocol. A 2026 pilot in Geroscience - uncontrolled, open-label, 12 patients newly diagnosed with mild cognitive impairment plus biomarker evidence of Alzheimer's disease - exchanged 16 to 26 litres of plasma for young-donor plasma from 18-to-24-year-olds, across three protocol permutations of only three to five patients each. It reported feasibility and adverse events consistent with known plasma-exchange risks, and it explicitly calls for a larger controlled trial to test efficacy. That is a study behaving honourably. It is not a result.

The regulator has been blunter than any of the papers. In a statement dated 19 February 2019, the FDA said there is no proven clinical benefit of the infusion of plasma from young donors in the prevention of conditions such as aging or memory loss, that the dosing of these infusions is not guided by evidence from adequate and well-controlled trials, and that large-volume infusion carries risks of anaphylaxis, transfusion-related acute lung injury and circulatory overload. Plasmapheresis is a real procedure with a real complication profile, used for real indications. Selling it as an anti-aging service is a separate activity from studying it.

How to read the next headline about this

• Find the primary endpoint before you find the number. Registry records list it, and they are public. If the impressive figure is secondary, exploratory or confined to a subgroup, the study did not answer the question the headline implies.
• Check when the trial was registered. Registration after enrolment began means the prespecification you are being promised happened after the fact.
• Separate randomized from concealed. Allocation by entry date has a cost, and single-blinded is not double-blinded.
• Ask what got measured. A clock is a prediction; a cognitive scale, a walking speed, a hospitalisation is an outcome. Only the second kind tells you a person did better.
• Count arms and patients. Three to five people per protocol variant is a feasibility exercise. Twelve participants with no control group can neither detect a benefit nor rule out a harm.

The mouse work here is good science, and plasma dilution is a genuinely interesting idea that earned a genuine test. What it has received so far is one mid-sized trial that split its co-primary endpoints, two small human studies that cannot agree on the sign of the effect, and a dozen open-label patients whose own investigators are asking for the controlled trial nobody has run. None of that is a scandal; it is what a live scientific question looks like from the inside, well before anyone is entitled to a settled opinion. Anyone offering you the answer today is working from the same evidence you have just read, minus the parts that complicate it.