The Trial That Didn’t Exist, Part II: What It Would Cost to Close It

In January 2022, in a press release the longevity industry had been waiting on for months, a biotech called Altos Labs announced it had raised three billion dollars to slow aging. The investors were Jeff Bezos and Yuri Milner. The roster was nearly twenty principal investigators across two continents, with sites in the Bay Area, Cambridge, and San Diego. The mandate was cellular reprogramming, single-cell methylomics, the senescence pathway. The interventions on the table were elegant, expensive, and at the bleeding edge of what biology can do.

The most senior hire was a German-born mathematician named Steve Horvath. Horvath was 54 and had spent more than two decades at UCLA arguing, against the consensus of his field, that aging could be measured: not chronological age, which any wristwatch can do, but the underlying biological process, the slow drift of a body away from youth. He believed the answer was hidden in the chemical marks left on DNA over time. His peers thought he was wasting his time.

In 2013 he proved them wrong. He published an algorithm that read a DNA sample and returned a person’s chronological age within about three years, across cell types and tissues. By 2019 he had a second-generation tool, GrimAge, that moved beyond chronology and predicted mortality risk directly. The algorithm read methylation patterns, methylation-based surrogates of seven plasma proteins, and a methylation-based estimator of smoking pack-years, and returned, in effect, an estimate of a person’s distance from death.

For the first time in the history of the species, a researcher could draw blood from a depressed adult and measure how much closer to dying that depression had moved them. In 2021, Owen Wolkowitz at UCSF did exactly that, and found a two-year gap. Same chronological age. Same lifestyle. Two years closer.

By the time Altos hired Horvath, the most evidenced longevity intervention in modern medicine was, on the existing literature, treatment of major depression.

Altos was not chasing it. Cellular reprogramming, yes. Senolytics, yes. Partial differentiation, yes. The intervention with the most existing human evidence was in a different building, run by a different specialty, funded out of a different pool. Three billion dollars in the room. The thing the room had the most data on, nowhere on the agenda.

Horvath himself has been candid about the field’s vocabulary. In a 2024 podcast interview with the cardiologist Eric Topol, he said: “I typically avoid the terminology biologic age, because to begin with, we don’t have a definition of it. Decades of discussions, nobody really has a precise definition of it.” A working definition was not actually required. The 2021 Wolkowitz paper used GrimAge as a mortality predictor calibrated on time-to-death data, and found that depression accelerates the predictor. The metaphysics could take the day off.

When I wrote Horvath last spring to ask whether anything in the literature linked the social and psychiatric variables to the clocks his lab had built, the answer came back direct. “I am not aware of any epigenetic clock studies of social isolation interventions,” he replied. “To me this is an entirely new frontier. It would be very interesting to carry out such clinical trials.” He added one important caveat. “The FDA does NOT (yet) consider epigenetic clocks as surrogate endpoints.”

There is one more layer of absence underneath that, and a Columbia epidemiologist named Daniel Belsky pointed at it when I sent him the draft. Belsky’s lab built DunedinPACE, the second major aging clock and, most operators agree, the right primary endpoint for a study running over weeks. “There are studies showing depressive symptoms to be correlated with DunedinPACE values,” he wrote. “I’m not aware of any longitudinal studies observing either that recovery from depression is associated with slowing DunedinPACE or that incident depression is accompanied by accelerating DunedinPACE.” That sentence is what keeps both industries in their separate buildings. The cross-sectional evidence is robust but the longitudinal evidence does not yet exist. Without it, an intervention trial cannot be properly powered, because no operator has the prior on how much movement to expect, or how fast, and the missing observational layer is what makes the missing trial possible to keep not running.

The interventional experiment itself is not complicated. Take a few hundred people with major depression. Randomize them. Give half a treatment that works (cognitive behavioral therapy, an SSRI, ketamine, take your pick from what the largest behavioral-health platforms deploy at scale). Give the other half a placebo or a delayed start. After eight weeks, draw blood. Run methylation panels. Does the DunedinPACE of the treated group come down? Does the GrimAge gap close? Forty to sixty million dollars to run. Roughly the marketing budget of a mid-size cosmetics launch.

A search of ClinicalTrials.gov and the European clinical-trial register, run as recently as April 2026, returns one trial that approximates the question. It enrolled twenty patients in Kentucky, gave them six ketamine infusions over two to three weeks, and measured their GrimAge before and after. Twenty patients. No placebo. Single arm. The scientific equivalent of asking your aunt whether she feels younger after the gardening.

A second trial, recruiting in Brescia, Italy, is running 110 depressed adults through a twelve-week exercise program with epigenetic correlates listed as secondary endpoints. The Brescia trial is currently the most rigorous interventional study on this question in the world. It is being run by an Italian hospital that until 1978 did not have the legal right to integrate psychiatric care into general medicine. The Italians had to legislate their way to the question, and they are now the only ones asking it at scale.

The capital, then.

Spring Health, the employer-facing behavioral-health platform founded in 2016, was last valued at three and a third billion dollars. Its February 2025 paper in JAMA Network Open, on almost 14,000 employees and dependents across seven employers, established a clean number: for every dollar spent on enhanced behavioral-health services, the plans saved a dollar ninety in the first year. A separate four-year analysis of nearly 32,000 Spring Health members showed physical-healthcare costs falling 9 percent in year one and 14 percent in year two. Lyra Health, founded in 2015 and last valued at $5.58 billion, published its own large-scale evaluation in Telemedicine and e-Health in 2022 on 6,738 patients in cognitive behavioral therapy. More than four out of five reliably improved, on average within 7.6 weeks.

These two companies, between them, hold the largest unmeasured longevity clinic in human history. They have run more depression-remission protocols on more people than any other private institution alive. The Spring Health data already suggests those protocols are lowering physical-healthcare claims by a measurable double-digit percentage. If those gains showed up on a methylation panel, that 14 percent reduction would also be a measurable signal of biological-age slowing. Nobody has run the panel.

Maya Shahsavari, the UK-trained surgeon and regenerative doctor practicing in Dubai, named what the patient experiences inside that gap. “In practice, patients move between specialties carrying one underlying problem that gets renamed at each door, while responsibility remains fragmented.”

A patient who finishes a successful Spring Health course of cognitive behavioral therapy does not, on graduation, become someone else. The next year, she may walk into a Fountain Life APEX evaluation. The cardiologist she sees about a flutter does not call the therapist. The endocrinologist who flags her A1c does not see the PHQ-9 from two springs ago. Each door renames the patient. Each door bills the patient again. Each door extracts its slice of the longitudinal data nobody is connecting.

This is what I’d call an Operator Gap. A single party would need the regulatory standing of a top-decile pharmaceutical sponsor, the patient access of a Spring Health or a Lyra, the methylation infrastructure of a TruDiagnostic-class lab, the clinical psychiatric depth of a Yale or a UCSF psychiatry department, and the patience of a generation. No such party exists. The fields that would need to combine were separated by design. The party that could combine them has not been built because nothing in the existing capital structure rewards the combination.

Vinod Khosla, who has done as much capital deployment in healthcare as any single investor alive, said it without the philosophy: “I never look at CPT codes, whether something is billable, what the business model is.” Of the kinds of startups he avoids: “Healthcare startups I hate are the ones who help people increase their billing.”

Jennie Byrne has spent two decades treating the kind of patient longevity clinics now compete to serve. She does not appear on a Bloomberg panel about healthspan. She runs a practice. The pattern she sees is not a population of mechanical organisms with measurable wear. “Despite tons of evidence, we still don’t know exactly what depression is,” she wrote me. “The deeper you go into the literature the more you find different types and presentations. Some lean more physical, like inflammation. Some lean more emotional. Some lean more cognitive.”

She does not speak the longevity vocabulary at all. She told me what she asks instead. “What does thriving look like for you?” Most patients cannot answer, or they give an answer that is plainly unrealistic. “What meaning does your life have? What meaning do you want it to have?” Most cannot answer that either. “What are your core values?” She has them name the values, then asks them to describe two or three peak experiences in their lives, and listens. The values from the peak experiences almost never match the values the patient first listed. That gap is where her work begins.

Then she said the thing that has not left me since. “I see people who are no longer depressed but they are not thriving. In fact, sometimes the treatments that target their superficial depression make their malaise worse, because now their emotions are blunted and they are not engaging fully with life.”

The methylation clocks measure the body running ahead. The PHQ-9 measures whether the patient is sad. Neither measures whether the patient is alive in the way they came in hoping to be. It is possible, on the existing instruments, for an intervention to clear the surface depression, lower the GrimAge gap, and still leave the patient further from the life they actually want. None of the trials being run, including the one this essay has spent forty thousand dollars in argument trying to call into existence, would catch that.

Byrne’s hypothesis: “I often wonder if longevity medicine is our new version of how we will heal our souls, now that we aren’t leaning on religion, or culture, or community.”

That is the question underneath the trial. The interventional study would close one loop, the loop the science requires. The deeper loop is whether the longevity industry is selling distance from death or selling distance from a question its customers do not know how to ask out loud. The PHQ-9 cannot ask it. The methylation panel cannot ask it. The CPT codebook never tried.

The cost of the boundary is not an abstraction. It is the trial that should have been run by now and has not. It is the longitudinal study underneath the trial that nobody has funded either. It is two years of life that depression has been adding, quietly, to the cohort that modern medicine serves. It is the twenty-thousand-dollar diagnostic panel that screens for everything the architecture permits and stops short of the variable that drives much of what it detects. It is a 1966 codebook still doing its work in the back of every operator’s mind. And, underneath all of those, it is the question Jennie Byrne asks her patients on the first visit, that nobody at Altos or Spring Health or the FDA has yet found a way to bill for.

Forty-eight years ago, on a Saturday in May, the Italian Parliament voted to dismantle the asylum system.

Somewhere in Naples, Florida, a man in his early fifties drives home with a printout in his lap. The Nespresso machine is still in the corner. The herons have not moved. He will recalibrate. Nobody, in any of the buildings he has paid, will think to ask if he is sad. Or, if he is no longer sad, whether he is yet alive.

Sources & Contributors

This piece is the second of two. Part I, “The Architecture of the Boundary,” is published separately. Quotes from named contributors come from email exchanges with the author in March, April, and May 2026.

Contributors quoted in Part II. Dr Maya Shahsavari, UK-trained surgeon and regenerative doctor practicing in Dubai. Dr Jennie Byrne, psychiatrist, author of Moral Injury: Healing the Healers and Work Smart: Use Your Brain and Behavior to Master the Future of Work. Steve Horvath, Professor of Medicine, David Geffen School of Medicine, UCLA, and creator of the Horvath and GrimAge epigenetic clocks. Daniel Belsky, Associate Professor of Epidemiology at the Robert N. Butler Columbia Aging Center, principal developer of DunedinPACE.

 

• Kusters, C.D.J. and Horvath, S. “Quantification of Epigenetic Aging in Public Health.” Annual Review of Public Health 46: 91–110 (April 2025).
• Teschendorff, A.E. and Horvath, S. “Epigenetic ageing clocks: statistical methods and emerging computational challenges.” Nature Reviews Genetics 26: 350–368 (May 2025).
• Lu, A.T. et al. “DNA methylation GrimAge strongly predicts lifespan and healthspan.” Aging 11(2): 303–327 (January 2019).
• Belsky, D.W. et al. “DunedinPACE, a DNA methylation biomarker of the pace of aging.” eLife 11: e73420 (2022).
• Topol, E. “Steve Horvath: Our Epigenetic Age Clocks.” Ground Truths (Summer 2024).
• Protsenko, E., Yang, R., Nier, B., et al. “GrimAge, an epigenetic predictor of mortality, is accelerated in major depressive disorder.” Translational Psychiatry 11: 193 (April 6, 2021).
• Han, L.K.M. et al. “Lagged effects of childhood depressive symptoms on adult epigenetic aging.” Psychological Medicine 54(12): 3398–3406 (September 2024).
• Liu, C. et al. “Association between depression and epigenetic age acceleration: a co-twin control study.” Depression and Anxiety 39: 741 (2022).
• ClinicalTrials.gov NCT05294835, “Ketamine and Epigenetic Aging,” TruDiagnostic and Wild Health collaboration.
• ClinicalTrials.gov NCT06989944, “Physical Activity-based Intervention in Depressed Patients,” IRCCS Centro San Giovanni di Dio Fatebenefratelli, Brescia.
• Dawson, K.L. et al. “Epigenetic aging and DNA methylation biomarker changes following ketamine treatment in patients with MDD and PTSD.” Translational Psychiatry 15: 452 (2025).
• Newman, J.C. et al. “Endpoints for geroscience clinical trials: health outcomes, biomarkers, and biologic age.” PMCID: PMC9768060.
• Hawrilenko, M. et al. “Return on Investment of Enhanced Behavioral Health Services.” JAMA Network Open 8(2): e2457834 (February 5, 2025).
• Spring Health long-term cost-savings analysis, Validation Institute (n≈32,000 members, four-year follow-up).
• Polizzi, C. et al. “Real-World Evaluation of a Large-Scale Blended Care-Cognitive Behavioral Therapy Program for Symptoms of Anxiety and Depression.” Telemedicine and e-Health (October 2022), n=6,738.
• Fountain Life executive health pricing and ROI release, PR Newswire (December 1, 2025).
• Vinod Khosla, StartUp Health Festival remarks, J.P. Morgan Healthcare Week (January 2017), per MedCity News.
• Altos Labs launch announcement and team page (January 2022; January 2024).