Longevity Lab Report

Analysis

What anti-aging treatments can change

Bryan Johnson named five limits of today’s anti-aging technology. We examine the evidence behind each one.

Saved articles

Ask ChatGPT about this article ↗

Copy the prompt, then paste it into ChatGPT to explore the findings and sources.

Open ChatGPT ↗

Bryan Johnson outdoors against a blue sky
Portrait of Bryan Johnson. Photo: M Robertson · CC BY 4.0

The findings

Bryan Johnson listed five things 2026 anti-aging technology cannot change: face age, hearing, baldness, teeth, and brain structure. Checked one by one, he is mostly right, with one exception: baldness partly reverses inside a window. Everywhere else, interventions slow, improve, or replace without restoring.

On October 23, 2026, Bryan Johnson published an Instagram carousel listing five things he said current anti-aging technology cannot change: face age, hearing, baldness, teeth and brain structure. This article examines those claims separately, including where treatments can help and where evidence remains limited.

The concession in the middle is the part worth keeping, because a list of limits can be checked in a way a percentile cannot.

So we checked it, one limit at a time, against the trial record. The word "change" turns out to hide four different claims, and separating them is most of the answer. An intervention can slow a decline or prevent it.

It can improve function without restoring what was lost. It can partly reverse a change, inside a window, while the underlying structures survive. Or it can restore or replace the thing itself.

Page one of every one of these debates mixes the four meanings together, and products are sold in the blur between them. Held apart, Johnson's list scores well: he is right about four of the five, wrong in an interesting way about one, and the strongest evidence in the whole piece belongs to the limit where he is wrong.

The limitWhat intervention can actually do
Face ageSlow it, and improve the surface. Restore nothing.
HearingImprove the signal. Restore nothing.
BaldnessPartly reverse it, inside a window, for as long as treatment continues.
TeethReplace them. Regrow nothing.
Brain structureImprove the test scores. The structure does not move.

Face age: slowing is real, and it is measured on a hand

The best trial in this family is the Nambour study (Hughes et al., Annals of Internal Medicine, 2013): 903 adults under 55, assigned to daily sunscreen or to their usual discretionary use, followed for four and a half years. The daily group showed no detectable increase in skin aging over the trial, and their measured aging ran 24 percent below the discretionary group's. That is a genuine prevention result, and its boundary matters as much as its size: aging was measured by microtopography, a silicone replica of the skin on the back of the hand.

The trial shows that a daily habit slows a surface measure of skin aging. It does not show that anyone's face looked younger, because perceived face age was never the endpoint.

Topical retinoids go one step further, from slowing to modest improvement. Tretinoin improves the surface measures of photoaging (fine wrinkling, texture, pigmentation) in randomized trials, modestly, over months of use, with irritation as the common cost. That is as far as the human evidence reaches for the face: the surface can be slowed and somewhat improved, and nothing at any price has been shown to restore a younger face.

The library's grades agree with the trial record. Collagen peptides grade C+ on the evidence, vitamin C grades B, niacinamide grades B-, GHK-Cu grades C-, and red light therapy grades C+ for skin appearance and collagen density, and D for the claim that it slows aging systemically. Every one of those grades attaches to a surface measure.

Johnson's first limit stands, with one refinement: the face can be slowed and its surface improved, and the honest verb stops there.

Hearing: the devices work around the loss, not on it

Age-related hearing loss is sensorineural: the hair cells of the inner ear die, and in humans they do not come back. Roughly fifteen thousand of them line each cochlea, and the guidance from the National Institute on Deafness and Other Communication Disorders treats the loss as permanent, managed with amplification and assistive devices rather than reversed by any therapy. Hearing aids amplify what survives.

Cochlear implants bypass the dead hair cells entirely and stimulate the nerve directly, which is engineering going around the damage, not biology repairing it. Since October 2022, an FDA category for over-the-counter hearing aids has put amplification in drugstores for adults with perceived mild-to-moderate loss, and that wider access is a real improvement in function. It changes nothing about the underlying structure.

Regenerating hair cells remains preclinical work: in mice, combinations of transcription factors can coax new hair-cell-like cells from supporting cells, and no human therapy exists.

The one result that complicates the picture is about the brain, not the ear. The ACHIEVE trial (Lin et al., The Lancet, 2023) randomized 977 adults aged 70 to 84 with untreated hearing loss to a hearing intervention or a health-education control, and its primary result was null: three-year cognitive change was nearly identical in the two groups. In a prespecified subgroup of 238 participants drawn from a long-running cohort study, older and at higher risk, the hearing intervention was associated with 48 percent slower cognitive decline, and the interaction was significant.

Both halves belong in the sentence, always: a null primary result, and a subgroup signal strong enough to justify the next trial. What ACHIEVE suggests is that treating hearing may protect cognition in the people most at risk of losing it. It does not suggest that hearing itself can be restored, and Johnson's second limit stands as stated.

Baldness: the exception, with a window and a meter running

Here Johnson is wrong, and the evidence is the strongest in this piece. The European S3 guideline for androgenetic alopecia, the evidence-based standard for the field, recommends topical minoxidil for men and women and oral finasteride 1 mg a day for men, with response assessed at six months. Both drugs act on follicles that are miniaturized but alive: they coax surviving follicles back toward producing full hairs, and a meaningful share of users regrow visible hair.

That is partial reversal of an age-related change, the only clean case on Johnson's list. The window is the boundary. A scalp that has gone smooth has lost the follicles themselves, and no drug regrows hair where the follicle is gone.

The reversal also runs on a meter. The guideline's position is that therapy continues indefinitely to hold the effect, and gains are lost within about six to twelve months of stopping finasteride. The drug's costs are known and should be stated plainly: sexual side effects, including decreased libido and erectile difficulty, run at roughly two percent in the dermatology summaries, and whether those effects can persist after stopping is reported and contested in the literature, with the contest itself part of the record.

Transplant surgery, the expensive end of the market, obeys arithmetic rather than evidence: it redistributes a finite supply of donor follicles from the back of the scalp to the front, creates none, and so caps out wherever the donor supply does. Even the myth end of the shelf behaves here: biotin, sold for hair on almost no basis, grades C on the evidence in our library, a supplement in search of a deficiency most buyers do not have. Baldness, then: reversible in part, inside the follicle-survival window, for as long as treatment continues.

Johnson's weakest claim is the one he stated most flatly.

Teeth: replacement is real, regrowth is a registry entry

A lost adult tooth is the clearest case on the list. Nothing regrows one, and the proven interventions are all replacements: implants, bridges, and dentures, which restore function with manufactured parts. That is the fourth level of change, reached by prosthetics rather than biology, and it works well enough that the biological version has to beat a good incumbent, not a vacuum.

The biological hope has a name and a registry number. TRG035, an antibody against the USAG-1 protein developed by Toregem BioPharma, is the most advanced tooth-regrowth programme in the world, and its human record to date is this: a completed Phase I trial of single intravenous doses in 30 healthy men aged 30 to 64 at Kyoto University Hospital, with safety as the primary endpoint and no efficacy result, registered as jRCT2051240154. A Phase IIa trial in 24 children born missing six or more permanent teeth, the population where the need is most acute, had cleared a regulatory review in August 2026 and had not dosed a single participant at the latest sponsor update.

The laboratory work behind the headlines deserves its species named in the same sentence: the 2026 dental pulp stem cell research from Sichuan University concerns pulp and dentin repair inside existing teeth, its functional regeneration work was done in genetically engineered mice, and human teeth appear in it as comparison tissue. Against that record, treat any page quoting a human success rate for tooth regrowth as what it is: the family's page one currently carries a figure of 68 percent success across 412 participants, and there is no trial behind it. Thirty men, one dose each, safety only, is the whole human record.

Johnson's fourth limit stands, and the distance between the registry and the marketing is the widest on his list.

Brain structure: the scores improve, and the scans do not move

The brain is where the four meanings of "change" do the most commercial work, because cognition can genuinely improve while structure stays exactly where it was. The two flagship multidomain trials show both halves. In FINGER (Ngandu et al., The Lancet, 2015), 1,260 at-risk adults aged 60 to 77 received two years of diet, exercise, cognitive training, and vascular monitoring; their cognitive composite improved more than the control group's, by an estimated 0.022 standard units per year, a small test-score effect with no dementia-prevention endpoint.

In the US POINTER trial, 2,111 adults aged 60 to 79 at elevated risk followed a structured lifestyle programme or a self-guided version, and the structured group improved global cognition more over two years, with both groups improving. Then the imaging ancillary study, POINTER-Imaging, scanned 983 of the participants and reported the result that defines this limit: no between-group difference on any prespecified imaging endpoint at two years. Amyloid burden, entorhinal tau, hippocampal volume, and white matter hyperintensity volume all moved the same in both groups.

The best-supported lifestyle programme in the world made people measurably sharper and left every structural measure it took untouched.

One trial gets quoted against that verdict, and it deserves a fair reading rather than a dismissive one. The INHANCE trial (Attarha et al., JMIR Serious Games, 2025) randomized 92 healthy adults aged 65 and older to ten weeks of computerized speed training, thirty minutes a day, or to non-speeded computer games, in a double-blind design funded by the National Institute on Aging. Its measure was chemical rather than structural: PET imaging of cholinergic terminal binding in the anterior cingulate cortex, a marker of the brain's acetylcholine system.

Binding rose about 2.3 percent in the training group and not in the controls, which the authors frame as reversing roughly a decade of decline, their arithmetic being an estimated loss of about 2.5 percent per decade. Two of the paper's co-authors are scientists at Posit Science, the company that makes the training programme, and company scientists hold shares in it, so the result arrives with its conflict attached.

What the trial shows is a neurochemical marker moving the right way in 92 cognitively healthy people over ten weeks; it measured no cognitive outcome and no structure, and in the grammar of this piece it is bounded movement on a chemistry measure. Chemistry is not architecture, and the verdict line stands.

The library's protocol grades tell the same story from the other side. Zone 2 training grades B+ for cardiovascular and mortality benefit, VO2 max intervals grade B- for raising VO2 max itself, CBT-I grades A for insomnia remission, and sauna grades B- for lower dementia risk on cohort data, and none of those grades is a grade for changing brain structure, because none of that work measured structure moving. Whole-body MRI grades D+ as a screening proposition: it can photograph the structure in fine detail, and detection is not change.

Biological age clocks belong in the same caution, since a clock is a measurement instrument, and a measurement that moves is not a structure restored. The risk side of the ledger is real and worth the effort: the Lancet Commission on dementia attributes about 8 percent of dementia cases to hearing loss in its 2020 model, and about 7 percent in its 2024 update, the largest single modifiable factor in both, which is the ACHIEVE bridge arriving from the other direction. Prevention bends risk.

Improvement is available. Restoration of brain structure, in humans, has no entry in the trial record, and Johnson's fifth limit stands in its strongest form: not merely unproven, but tested at scale and not found.

The boundary is the finding

Four of Johnson's five limits survive checking, and the one that fails, baldness, fails in the instructive direction: the reversal is real, bounded by surviving follicles, and rented rather than owned. Everywhere else, the pattern repeats with different nouns. Interventions slow (sunscreen on a hand), improve (aids, implants, test scores), and replace (implants of the dental kind), and the marketing lives in letting a reader hear "restore" where the evidence said one of the smaller words.

The concession slide can be checked because it names things a stranger can verify, and that is also why the carousel's third slide should be read differently: a percentile is a comparison against a population the reader never sees, from a panel the reader never sees, and a claim built that way cannot lose. Keep the slide that can fail. It is the most useful thing the field's most measured man published this year, and the five verdicts above are what it looks like when someone takes it seriously enough to check.

People mentioned

Supplements mentioned

Peptides mentioned

Protocols mentioned

Biomarkers mentioned

Screening tests mentioned

Grades are research summaries, not medical advice. Decisions about your own health belong with your clinician.

Keep exploring

View all

The weekly newsletter

Keep up with what’s changing.

New research, useful tools, and the people shaping health and longevity. Selected and explained in one email.

See what you’ll receive ↗

Free. Confirm your email to subscribe. Unsubscribe anytime.