Gone After Plasma Exchange #5: The Longevity Signal That Didn’t Last

Here is a sentence from an abstract: TPE "significantly improved biological age markers," with 15 epigenetic clocks showing rejuvenation compared to placebo. [1]

Here is a paraphrased sentence from the Results section of the same paper: no biologic clock changed significantly from baseline, at either follow-up point, once the authors corrected for testing 36 of them. [1]

Both come from the same paper. The distance between them is the whole story, and it is why this paper deserves a slow read.

Why this one matters

Plasma exchange is a procedure I do for a living. I know what it takes to get a patient on the machine, what we replace the plasma with, and what we watch for. So when longevity clinics began offering it to healthy people, I wanted to see the evidence. This is the trial people point to: a single-blind, randomized, placebo-controlled study in healthy adults over 50, using DNA methylation "clocks" to measure biological age. The authors say it is the first human study to report a decrease in biological age with TPE using well-validated epigenetic metrics, and it builds on an earlier eight-person pilot [1, 2].

What they did, and what they deserve credit for

Forty-four people enrolled and 42 finished. They landed in one of four arms: biweekly TPE with IVIG (intravenous immunoglobulin), biweekly TPE alone, monthly TPE, or sham. In the TPE arms, one plasma volume came off and 5% albumin went back. The IVIG arm received 2 g of immunoglobulin afterward. [1]

The sham deserves its own paragraph. Those participants got a peripheral IV with saline. Dark curtains hid the apheresis device. The machine ran against a container of water, lines submerged to mimic blood flow, and a rinseback at the end so everyone believed the procedure was finished. I have a grudging respect for the container of water. Patients, caregivers, and raters were all blinded, and the trial was registered. Plenty of longevity interventions never get this far. [1]

The number everyone will quote

At the first follow-up, the biweekly TPE+IVIG arm showed an average biological age reduction of 2.61 years versus sham. Monthly TPE showed 1.32 years. [1]

A quick definition. An epigenetic clock is an algorithm that reads DNA methylation patterns and estimates how old your biology looks. Age acceleration is the gap between that estimate and what you would expect for your chronological age. The authors ran 36 clocks, averaged the changes, and compared each arm with sham.

Thirty-six clocks is a lot of chances to get a lucky one. That is what false discovery rate (FDR) correction is for: it tightens your threshold to account for how many tests you ran. Against sham, ten clocks passed in the TPE+IVIG arm, and five each in biweekly and monthly TPE. [1]

Now the sentence that should sit next to that number: no arm improved compared with its own baseline. Not one clock, in any arm, changed significantly from that arm's baseline at either follow-up once the correction was applied. The authors show some nominal trends and use them to describe patterns, but they do not survive the correction, and the authors say so. [1]

So what does "2.61 years younger than sham" mean? A difference versus sham needs two groups to move. With roughly ten people per arm, a comparison arm drifting in one direction can produce a gap even when the treated arm itself has not clearly changed. That last part is my reasoning, not a finding in the paper. It is why I would not read 2.61 years as "a person got 2.61 years younger."

And the gap did not last. By the second follow-up, after five sessions, none of the arms differed significantly from sham. The improvement relative to sham was a time-limited signal, visible at one measurement and gone by the next. [1]

Then there is the question of whether any of it matters to a patient. The trial did not measure that. Function, cognition, and symptoms were not predefined endpoints, and the authors state that no clinically meaningful changes were observed or assessed. A biomarker difference that has not been linked to any clinical outcome is, for now, a difference in a biomarker. [1]

The null the authors report themselves

Here is the part I like best. The second follow-up came after five sessions. At that point, no group differed significantly from sham in biological age. The authors say so plainly and suggest "potential compensatory mechanisms." In the biweekly TPE arm, people who dropped more early on tended to rebound more later (nominal p < 0.048), and the authors conclude that most of the rejuvenation effect happened between the first two collection points. [1]

Think about what that implies. Clinics already offer repeated sessions. If the effect shows up after three and is gone after five, then the body pushed back, or the clocks are noisy, or something else is going on. The paper tests none of those. It names one and moves on, which is fair for a pilot. It is also the most interesting result in the manuscript, and it is not in the abstract.

What bothers me

Allocation. The abstract says randomization was based on entry date. The Methods explain: first come, first served, with the shorter biweekly arms filled first and the longer monthly and sham arms filled after. That is a queue, not a randomization. Elsewhere in the Methods the participants are described as randomly allocated 1:1:1:1, and I would like the authors to pick one description. The practical consequence is baseline imbalance, and the authors state that most clocks differed between groups at baseline. [1]

Who ended up in sham. Sham was among the last arms filled. Table 1 shows 8 men and 3 women, with the women averaging 77 years old and the men 64.5. The authors adjust for chronological age; I did not see an adjustment for sex described in the Methods. Three women is not a subgroup. It is a coincidence that happens to sit in the comparator arm. [1]

Regression to the mean. When arms start in different places, changes from baseline are vulnerable to regression to the mean. The Discussion notes that people with poorer baseline health seemed to benefit most. That is also what regression to the mean predicts. I am not saying that is what happened. I am saying the paper cannot tell us. [1]

IVIG. The Discussion says IVIG doubles the effect, from 1.32 to 2.61 years. But 1.32 years belongs to the monthly TPE arm, not the biweekly TPE arm that the IVIG group is built to be compared against. In Table 4, by my reading, the head-to-head comparison of biweekly TPE+IVIG against biweekly TPE passes FDR for two clocks out of 36 and fails for the rest. The authors also note that the cell composition changes appeared only in the IVIG arm, and that it is plausible those changes come from IVIG alone. So this is partly a plasma exchange paper and partly an immunoglobulin paper, and the data cannot cleanly separate them. [1]

Power and endpoints. There was no formal power calculation; sample size was set by the safety evaluation. Function, cognition, and symptoms were not assessed, and the authors state that no clinically meaningful changes were observed or assessed in this short study. In their own introduction, they note that biomarker changes from other interventions have not translated into consistent clinical benefit. I agree with them on that point. [1]

Disclosures. Kiprov is co-founder and chief scientific officer of Circulate and is affiliated with Global Apheresis, which performed the procedures and ran the allocation queue. Two other authors are listed as Circulate members, and Furman co-founded Edifice Health, whose assay produced the iAge measure. The funding statement says no specific funding. None of this invalidates the work. It is disclosed, which is what the system asks of authors. Readers should know it exists. [1]

Where the oversell lives

The Discussion is measured. It calls the findings hypothesis-generating and names sample size as the key constraint. The oversell happens upstream. The abstract says "rejuvenation." It says TPE+IVIG reversed age-related immune decline. It does not mention that the effect was gone by the next draw. Then the institute's press release leads with the 2.6 year figure [3], and that is the version that circulates. I do not think the authors hid anything. I think the abstract is the part everyone reads, and it was written to be quoted.

What I am left with

A well-built pilot, a careful Discussion, an abstract that outruns its own Results section, and a finding that came and went by session five. I would take that over a bigger, cleaner-sounding trial with no sham.

Which leaves a question I cannot answer from the paper. If an intervention moves a biomarker and then stops moving it, what was the biomarker measuring?

References

1. Fuentealba M, Kiprov D, Schneider K, et al. Multi-omics analysis reveals biomarkers that contribute to biological age rejuvenation in response to single-blinded randomized placebo-controlled therapeutic plasma exchange. Aging Cell. 2025;24(8):e70103. doi:10.1111/acel.70103

2. Kim D, Kiprov DD, Luellen C, et al. Old plasma dilution reduces human biological age: a clinical study. GeroScience. 2022;44:2701-2720. doi:10.1007/s11357-022-00645-w

3. Buck Institute for Research on Aging. Clinical trial and multi-omics analysis demonstrates the impact of therapeutic plasma exchange on biological age. May 28, 2025. buckinstitute.org