Plasma Exchange: 2.6 Years, Gone by Session Six

Forty Clinics, Four Hundred Planned, One Trial

There is a sentence on the internet that costs about sixty thousand dollars to believe. It reads: FDA-cleared Therapeutic Plasma Exchange clinically shown to reduce biological age by 2.6 years. That is the homepage of Circulate Health, a longevity company that launched in January 2024, raised $14 million, built out roughly forty clinic locations, and was acquired by Viome Life Sciences in August 2026 in a deal reported between $15 and $50 million. Viome says the combined network could eventually reach four hundred sites.¹⁵

I want to take that sentence apart slowly, because both halves of it are doing quiet work. The device is FDA-cleared. Apheresis machines have been cleared for decades, and therapeutic plasma exchange is legitimate, unglamorous, fifty-year-old medicine. The American Society for Apheresis grades 166 indications across 91 conditions: thrombotic thrombocytopenic purpura, Guillain-Barré syndrome, myasthenia gravis, and a long tail of others.¹⁸ If you have TTP, plasma exchange will save your life. Nobody is disputing any of this.

Aging is not on that list. The longevity indication is off-label, self-pay, and uninsured, and it costs between $6,000 and $12,000 per session depending on the clinic. Global Apheresis, owned by the physician who is also the co-founder and chief scientific officer of Circulate, lists $8,000 a session, dropping to about $6,000 in multi-session packages.¹⁷ The protocol is six sessions. Do the arithmetic: $36,000 to $72,000.

And "clinically shown"? That phrase is pointing at exactly one study. Forty-four people enrolled, forty-two finished, four arms, about ten people in each. I have read it closely, and I am going to spend most of this issue on it, because this single paper is load-bearing for an entire industry.

The Mouse That Got Younger on Salt Water

Here is the part of this story I find genuinely good, and I want to give it full credit before I start taking things apart.

For two decades the dominant idea in blood-and-aging research was addition. Stitch a young mouse and an old mouse together so they share circulation, and the old mouse's tissues improve. The obvious reading was that young blood contains youth factors, and that if we could find them and bottle them, we would have something. That reading launched companies and, eventually, a clinic in California charging $8,000 a liter for plasma from teenagers.

In 2020, Irina Conboy's lab at Berkeley published a result that reframed the whole field. They took old mice and replaced roughly half the plasma volume with saline and five percent albumin. Nothing young was added. No youth factors, no donor, no teenager. Just dilution and a carrier protein. The result met or exceeded the effects of young-blood parabiosis on muscle repair, liver adiposity and fibrosis, and hippocampal neurogenesis.⁷

If that holds, the active ingredient was never the youth in young blood. It was the removal of whatever accumulates in old blood: pro-inflammatory signaling proteins, senescence-associated factors, the molecular exhaust of several decades. Dilute the exhaust and the tissue responds. This is a coherent, testable, mechanistically respectable hypothesis, and it is the single best argument for taking plasma exchange seriously as a longevity intervention.

It is also a mouse study of tissue repair markers. The human follow-up from the same collaboration, published in GeroScience in 2022, had eight plasma exchange patients, five young comparison donors, five old comparison donors, no randomization, no placebo, and no clinical outcome of any kind. It was a proteomics hypothesis generator, and it was honest about being one.⁸

Ten People Per Arm and a Curtain Over the Machine

The paper is Fuentealba and colleagues, published in Aging Cell on 28 May 2025.¹ Forty-four enrolled, forty-two completed. Four arms of roughly ten or eleven: biweekly plasma exchange plus intravenous immunoglobulin, biweekly plasma exchange alone, monthly plasma exchange, and a sham.

The sham deserves praise. Dark curtains over the apheresis machine, a water-filled container standing in for the collection bag, a theatrical rinseback, saline IVs in both arms. Blinding an apheresis procedure is genuinely hard, and they took it seriously. Blood was drawn before treatments one, four, and six, and run against thirty-five to thirty-six epigenetic clocks.

The headline: the plasma exchange plus immunoglobulin arm showed a 2.61-year reduction in average clock-based age acceleration versus sham. Monthly plasma exchange showed 1.32 years. Immune cell composition and senescence-associated proteins shifted in what the authors characterize as a younger direction.

Now the problems, all of which are verifiable in the paper itself or in the trial registry.

The effect was measured in the middle and was gone by the end. The 2.61 years is from timepoint two, drawn before the fourth of six treatments. At timepoint three, before the sixth, the paper states plainly that there were no significant differences in biological age compared to sham in any group. The protocol being sold in forty clinics is the six-session protocol. The benefit lives at session three and has left by session six, and the paper offers no explanation beyond a gesture at unmeasured compensatory mechanisms.

Nothing moved from baseline once the statistics were corrected. The authors say so themselves: no biologic clock showed a significant change from baseline to timepoint two or timepoint three with false discovery rate correction, so they used nominal uncorrected p-values instead, citing low sample size and high variance in human samples. Every significant result in this paper is a between-arm comparison against a sham group whose clock readings carry standard deviations as wide as twenty-six years.

It was registered after it finished. The registry record shows an actual start date of 12 September 2022 and actual primary completion on 30 November 2023. It was first submitted to ClinicalTrials.gov on 17 July 2024 and posted 2 August 2024, roughly eight months after the primary endpoint data already existed.² Retrospective registration forfeits any claim that the analysis was pre-specified.

Randomization was by entry date. That is sequential allocation, not concealed randomization, and the consequence shows up immediately: the paper acknowledges that most of the biologic clocks differ significantly between groups at baseline. The sham arm was eight men and three women, and the sham women averaged 77.0 years against 66.5 in the treatment arm. When your outcome variable is unbalanced before you start, a between-arm difference is not clean evidence of anything.

The functional endpoints vanished. The registry lists grip strength, timed-up-and-go, balance time, and a health survey as secondary outcomes.² None of them appear in the publication, which instead states that no clinically meaningful functional, cognitive, or symptomatic changes were observed or assessed. Those are the endpoints a patient would actually feel. They were planned, and they are not reported.

Three authors are in the company. The disclosure names the co-founder and chief scientific officer of Circulate Inc. along with two colleagues as members of the company. The funding statement, three lines away, reads that the authors received no specific funding for this work. Both statements can be technically true at once. Together they read as more disinterested than the situation supports.

Nine Years of Noise in the Measuring Instrument

Set the trial design aside for a moment, because there is a deeper issue. Even a perfectly run study has to measure something, and what this one measured was epigenetic age.

Epigenetic clocks read DNA methylation patterns and output a number in years. They were built as population-level epidemiological predictors, trained to correlate with chronological age and mortality risk across large cohorts. They are now being used as individual-level clinical outcome measures in trials, which is a different job with different requirements.

In 2022, Higgins-Chen and colleagues at Yale quantified how well they handle that job. Run the same blood sample twice through six prominent clocks and technical noise alone produces deviations of up to nine years.⁵ Nine years of instrument error, on a measurement being used to detect a 2.6-year effect. Their proposed fix was principal-component versions of the clocks, which bring most replicates within about a year and a half.

The plasma exchange trial used those principal-component clocks. It does not cite the paper that created them. I checked the full text: zero occurrences of "Higgins," zero occurrences of any form of the word "reliability."

Then in 2026 the same lab published the finding that lands hardest here. Across eighteen DNA methylation biomarkers, technical reproducibility was mostly excellent, but biological reliability was substantially lower, with most clocks showing only low to moderate stability, and that stability decreased further when adjusting for immune cell composition.⁴ Low-reliability clocks, they noted, yielded variable or misleading results in intervention studies specifically.

Read that alongside what plasma exchange demonstrably does. The trial's own headline mechanistic finding was that the procedure restored age-associated shifts in immune cell composition. The intervention moves the exact variable that destabilizes the measurement.

So: does 2.6 clock-years mean anything for a human being? The honest answer is that nobody knows, for any intervention. No epigenetic clock has been qualified as a surrogate endpoint by any regulator. No intervention of any kind has been shown to move a clock and move a hard outcome in the same study. A useful harmonized database of fifty-one human longevity interventions established that clocks are responsive to interventions, which is a precondition for surrogate status, not proof of it.¹⁴ Responsive and validated are different words.

You can do the arithmetic if you want. The best-calibrated clock, GrimAge, has a published relationship in which roughly 7.5 years of age acceleration corresponds to a doubling of mortality risk. Scale 2.6 years against that and you get something like a twenty percent mortality difference. That calculation requires the causal arrow to run from methylation to death rather than both being downstream of something else, which has never been shown, and it requires the 2.6 years to be real, which uncorrected mid-course data from ten people cannot establish.

Horvath Ran It Again. The Clocks Went the Wrong Way.

In July 2025, a Czech academic group published a randomized crossover trial in Scientific Reports. Forty-one enrolled, thirty-four completed, ages forty to sixty, comparing four versus eight automated plasmaphereses over eighteen weeks. Steve Horvath, who built the first widely used epigenetic clock, is a co-author. The funding was a Charles University grant, not a company.³

The biochemistry moved, and measurably: cholesterol, triglycerides, apolipoprotein A, total protein, albumin and several minerals all fell. The epigenetic age did not improve. The authors write that no significant epigenetic rejuvenation was observed, and that plasmapheresis was instead associated with increases in GrimAge, the Hannum clock, and the Dunedin pace of aging. Their own framing is that it may accelerate epigenetic aging.

I want to be scrupulously fair about what this does and does not refute. This is plasma donation, roughly 600 to 800 millilitres with saline volume replacement. It is not a full one-plasma-volume exchange with five percent albumin. A defender of the Circulate protocol can legitimately say this is a lower dose of a related but non-identical intervention, and they would be right.

What cannot be argued away is the direction. This is the only randomized test of the core premise run by people with no financial stake in the answer, and the clocks moved the wrong way. There is one more detail worth knowing: the Aging Cell paper cites two ongoing trials as corroboration. One of them is this Czech trial, the one that found no rejuvenation.

AMBAR: 347 Patients, One Missed Co-Primary, Zero Approvals

When clinics want to point at something bigger than forty-four people, they point at AMBAR. It is the largest and longest sham-controlled plasma exchange trial ever run, and it is usually oversold.

AMBAR randomized 347 Alzheimer's patients across three albumin and immunoglobulin dosing arms plus sham: six weekly conventional plasma exchanges, then twelve months of monthly low-volume exchange, with a fourteen-month endpoint. It was funded by Grifols, which manufactures both the albumin and the immunoglobulin.⁶

It had two co-primary endpoints. One hit: the activities of daily living scale, with 52 percent less decline, p equals 0.03. One missed: the cognitive subscale, p equals 0.06. Under co-primary rules, missing one is a formal trial failure, full stop. The secondary endpoints were genuinely impressive, with the clinical dementia rating at p equals 0.002 and global impression of change below 0.0001. The mild-Alzheimer's subgroup showed nothing.

Six years later there is no FDA approval, no confirmatory phase 3, and no apheresis society indication for Alzheimer's disease. The only follow-up I could find is a thirty-two-patient real-world series from Argentina comparing against a historical control cohort drawn from 2008 to 2018.¹³ A cohort assembled from a different decade is not a control group in any defensible sense, and that series is nonetheless cited by clinics as independent confirmation.

As for plasma exchange marketed for longevity specifically, the count of trials measuring mortality, incident disease, or physical function is zero. I searched the registries. A frailty trial and the Circulate trial both read status unknown with no results posted. A plasma dilution study in mild cognitive impairment has ten participants.

51,567 Procedures and the One Fluid That Stands Out

Credit where it is due: plasma exchange is reasonably safe in competent hands, and the marketing is not lying about that. It is, however, cherry-picking.

The best real-world data is the World Apheresis Association registry analysis of 51,567 procedures on one widely used apheresis system. Adverse event rates were 1.43 percent mild, 2.81 percent moderate, 0.21 percent severe, roughly 4.5 percent overall, and plasma exchange was among the lowest-risk procedure types in the dataset.⁹ That is a genuinely reassuring number built on a genuinely large denominator.

One detail in that paper matters for this specific use case. The single highest severe adverse event rate associated with any variable in the analysis was albumin as the replacement fluid. Albumin is what longevity plasma exchange uses.

Against 51,567 procedures, the trial's quoted safety figure of one event in 240 procedures, 0.42 percent, is restricted to participants who completed the intervention, and the same abstract reports two adverse events requiring discontinuation, one of them immunoglobulin-related. The Argentine real-world series logged mild-to-moderate adverse events in 18.5 percent of 514 sessions.¹³ These numbers are not in conflict so much as measuring different things with different care.

The actual risk list, drawn across the literature: citrate-induced hypocalcemia, which dominates and reaches nearly half of procedures in series without prophylactic calcium; hypotension; vasovagal reactions; vascular access complications; allergic and occasionally anaphylactic reactions to albumin; depletion of clotting factors; removal of protective antibodies; and removal of whatever medications the patient happens to be taking. Add immunoglobulin and you add thromboembolism, renal injury, aseptic meningitis, and hemolysis.

On regulators: the FDA has never approved plasma or plasma exchange for aging. Its February 2019 safety communication on young-donor plasma infusions ended one $8,000-per-liter operation within a day, and the December 2024 update restates that there is no evidence of clinical benefit for aging.^10,11 The 2024 update contains the line most relevant to this industry, warning that a ClinicalTrials.gov listing, or the fact that an establishment has registered with the agency, does not mean anything has been approved. Both notices address young-plasma infusion rather than albumin-replacement exchange, so in fairness the agency has not specifically addressed this practice, and has taken no enforcement action against it. Silence is not endorsement, and it is not condemnation either.

One more data point, because it tells you what is being optimized. In 2026, a Circulate-affiliated group published a before-and-after series showing plasma exchange lowered circulating microplastic counts in 114 patients.¹⁶ It demonstrates that the machine removes things from blood. It demonstrates nothing about health. That is the shape of this entire evidence base: removal is measurable, benefit is assumed.

The Verdict