Irving Fisher, at the time one of America’s most prominent economists, knew that thief well. Tuberculosis, the disease that killed his father, forced the economist to spend the first three years of his Yale professorship on leave in sanatoria. The illness re-oriented his priorities, turning public health and life extension into a personal cause. As a member of the Committee of 100 on National Health,3 Fisher wrote “National Vitality: Its Wastes and Conservation” for the National Conservation Commission, created by Theodore Roosevelt. There, he gave this project of national vitality conservation an epidemiological target:

Eighty-five per cent of deaths by typhoid fever are unnecessary. […] It would be feasible to prevent at least 75 per cent of cases of tuberculosis of the lungs. […] These seven diseases alone could easily be reduced by these amounts […] simply through insistence by the public on pure milk, pure water, pure air, and reasonable protection from accidents.

Alleviating those diseases, by his estimation, would raise average life expectancy from roughly forty-five years toward sixty, a gain a contemporary digest headlined “Lengthening Human Life In This Country One-Third”:

In December 1913, Fisher went on to co-found the Life Extension Institute, and its supporters showed how well the idea had been received. William Howard Taft, fresh out of his US presidential term, chaired its board. Irving Fisher chaired its Hygiene Reference Board and co-wrote its manual “How to Live,” and its consultant on sanitation was William Crawford Gorgas, the army doctor who eliminated yellow fever and brought malaria under control in Panama. William Welch, the founding dean at Johns Hopkins, lent his name too, as did New York’s health commissioner Hermann Biggs. The Institute called itself a body devoted to “the great war of civilization against needless sickness and premature death.”

Over a few decades, the urge was reduced to practice. We began to chlorinate water, whose safety could be judged by bacterial cultures instead of by our senses, and started pasteurizing milk (https://doi.org/10.2307/2173395). Then came therapeutics that could go after the infectious agents themselves once they had entered the body. Sulfonamide antibiotics, the first effective treatment for several bacterial infections, cut mortality by an estimated 2-3% while adding somewhere between 0.4 and 0.7 years to life expectancy (source). Penicillin followed, as did streptomycin, the first antibiotic against tuberculosis. Although the smallpox vaccine had existed since 1796, the disease killed an estimated 300 million people in the twentieth century alone until sustained vaccination campaigns and global coordination eradicated the disease in 1980.

We have done remarkably well in our war against infectious disease, and the gains in lifespan have far exceeded Fisher’s expectations. In the United States, life expectancy rose from 47.3 years in 1900 to 78.7 years in 2018. The Institute itself bears witness that longevity was already sought as a public good, though in those days the chief enemy was infectious disease; and when that enemy receded, another was revealed behind it: the diseases of aging, and they became the new limit upon longevity.

Subscribe now

Who is our new enemy?

In the US, more than three quarters of deaths now occur after age 65, and almost a third after age 85. Of the top ten causes of death most are chronic diseases, and together they account for somewhat less than two thirds of deaths4. Among the diseases that form the backdrop of later life — neurodegenerative and metabolic syndromes, cardiovascular disease, and the majority of cancers — the greatest risk factor is aging itself.

We’ve responded to these diseases by directing biomedical research and drug development towards new medicines. Statins and antihypertensives have driven a long decline in cardiovascular mortality, and smoking cessation and cancer screening have lowered cancer deaths. Such victories against high-burden, common age-related diseases is obviously very desirable: A treatment for Alzheimer’s, for example, could produce enormous population-wide gains in quality of life. Yet even a win that large would ultimately struggle against the ceiling imposed by late-life chronic disease, accumulating year after year as we age.

By age 65, ~half of people have at least two chronic diseases. By 85+, three quarters do, and the share with four or more rises from minuscule to over a quarter of the 85+ population5. Curing diseases of aging one by one will yield important but unremarkable lifespan returns, because late-life disease is a network of competing risks that becomes increasingly intricate as we age. Take one away, or even two, and you are soon in the grips of the next. If pancreatic cancer vanished, people would still live only as long as their remaining, and newly accumulating, chronic diseases allowed them to6.

This kind of Sisyphean medicine is destined to leave most old people less-but-still sick. Longevity science presents an alternative, and goes far beyond basic research on how aging happens: it aims to discover interventions that may change the course of later life by altering the ways our bodies break over time, and seeks also to broaden the kinds of medicine that can be carried forward into drug development.

A bed for one crop

What kinds of medicines is longevity science working towards, then? As nicely taxonomized by Karl Pfleger, these run from broadly slowing aging by turning its rate down, to divide-and-conquer rejuvenation where accumulated damage is reversed one subpathology at a time. Around those sit smaller sub-fields, like replacing parts of the aging body wholesale.

But coming up with ideas for longevity interventions is only part of the field’s task. We also need to figure out how to test them in humans. This is necessary, but difficult, because these broader approaches to longevity medicine require infrastructure that is less mature, like clinical trial designs suitable for measuring frailty, less legible, or even nonexistent, like measurements that could serve as surrogate endpoints for future mortality. This pushes sponsors to slot longevity drugs into single-disease indications where risk legibility is highest. But trials designed around a single disease are not best suited to show a longevity drug’s broader potential benefits, which handicaps returns for longevity medicine.

These field-specific sources of translation risk sit on top of the already daunting odds that shape investment throughout drug development. This is especially true for non-incumbent sponsors, who have weaker priors about trial design and regulatory perception. Even a drug that works can disappoint commercially if it does not beat cheap, familiar care well enough to justify adoption.

Of all these risks, the most consequential is clinical trials. Trial economics matter enormously: they consume most of the capital a biotech raises, and further distort the search for multimorbidity drugs, because cleaner efficacy signals are easier to produce in narrower patient populations. Sponsors are therefore pushed to slot longevity drugs into increasingly narrow indications7. Once a drug wins approval for one condition, it can usually expand to only a few others before exclusivity expires; but from there, pharma has little incentive to keep sponsoring trials for it.

Longevity drugs, like anything new, therefore carry greater scientific and commercial risk than drugs in more established fields of medicine, and are pushed by biotech and pharma to conform to the narrower indications common in those fields, where uncertainty is easier to contain.

Put differently, the field is being grown in soil prepared for one type of crop. The current seedbed gives regulatory and market support to single-indication candidates. Other interventions need different soil, where benefit is expected to appear in forms our current infrastructural, market and regulatory gates do not provide strong incentives for. A seed can look weak when the soil was never prepared for that plant.

Subscribe now

Building longevity drugs like we built antibiotics

The longevity field is still small and young, without routes of its own for developing drugs that alter age-related decline. It therefore relies on drug-development machinery built around conventional disease indications, creating the conflicts described above. What would longevity-specific infrastructure look like? One that turns every advance into a foundation for enriching the next? And what could the field become once it exists?

Take antibiotics as an example. There was a time when each antibiotic was a single, isolated, and largely serendipitous discovery, in the way that exploratory research often is. Bacitracin emerged from Columbia researchers’ study of the microbial ecology of contaminated wounds; cephalosporin from a University of Cagliari professor’s investigation into why sewage-contaminated seawater was not producing the disease outbreaks he expected. But it became an industry when we figured out how to make each discovery and infrastructural innovation the ground for the next one. Once germ theory took hold in the second half of the nineteenth century, sanitation and medicine acquired a new, legible object of intervention: the pathogen. Then came techniques such as pure culture and viable plate count for bacterial enumeration, which made it possible to attribute an effect to a defined organism and measure its response to intervention. Bacteria had become, at last, tractable.

That tractability gave us methods of discovery that any laboratory could use to take antibiotic research from craft into mass production, as in Waksman’s program, which systematically screened thousands of soil microbes for activity against pathogens and gave us streptomycin and neomycin. Drug companies took note of that success and built soil-screening programmes of their own, helping launch a burst in which we discovered most of the major antibiotic classes derived from microbes. For a while, dirt was an extraordinarily productive pharmaceutical platform.

And then we learned how to make vastly more of these drugs. We figured out how to coax greater yields of penicillin out of better molds and increasingly optimized growth media, and eventually moved the entire enterprise into deep-tank fermentation8. This industrial system became more cumulative still as quality signals at each stage made improvement legible and comparable. We established the Oxford unit, and later the International Unit so laboratories and manufacturers could compare penicillin potency across batches and production sites. We also coordinated our wartime agencies so that improvements developed by one company could spread to others. Then we made the road from the laboratory and the factory to the market and common use, first by randomized clinical trials and later by requiring that a treatment prove its efficacy.

Together, our advances in discovering and producing those drugs compounded into the golden era of antibiotics that ran from the 1940s into the 1960s.

Longevity has not yet found its versions of the soil screen or the penicillin unit, advances that allowed antibiotics to become an industry by letting researchers test interventions at scale and learn from fast, validated readouts. Earlier in discovery, molecular biomarkers of aging could provide those readouts, de-risking new interventions by increasing the speed of discovery and identifying likely responders, once validated for specific uses. But no such measure exists yet, as we’ve argued before. At the clinical end, it needs a path to proof that can formalize and accelerate trials without reducing aging to a single disease. Some routes are emerging here: a multimorbidity composite endpoint like in the aging trial design we proposed asks whether a candidate aging drug can delay the first instance of an age-related disease event. A functional endpoint asks instead whether a candidate drug preserves mobility, frailty, or intrinsic capacity — a composite measure of a person’s total mental and physical abilities — without waiting for a named disease. ARPA-H’s PROSPR program is currently trying to validate intrinsic capacity as an FDA-approved functional endpoint for aging.

For now, the single-disease indication remains longevity’s only established route through drug development, leaving many broader hypotheses about aging without a viable path to human trials. Optimism, meanwhile, is gathering. Silver Linings models how large gains from even modest progress could be, while longer and healthier lives seem to attract more serious attention as a medical frontier among researchers and economists alike. Expanding how we develop and test drugs to match that ambition will be essential to human flourishing.

Longevity Industrialized

News, thoughts, and analyses on longevity, focused on what is driving and limiting progress towards human impact. Writings from Martin Borch Jensen and other members/affiliates of Norn Group.

By Norn Group

I see a lot of optimism for accelerated progress in longevity (and medicine in general), but the optimism is mostly indefinite. “Better and faster drugs" is very much plausible, but delivering years of healthy life is a process of many steps. As one example, skeptics argue that regardless of abundant intelligence, clinical trials will continue to be slow and expensive. Solving this with ‘better drugs that will yield clear answers faster’ requires us to believe that biomedical R&D will be improved in dramatic ways, exceeding, say, every Nobel-winning discovery of the past century. A big ask, but after witnessing LLMs becoming lawyers by reading reddit and solving novel math problems, perhaps we should let indefinite optimism prevail?

I don’t think so, because the conditions that allowed language models to stampede across knowledge work do not exist for ‘curing disease’. You can’t do reinforcement learning without verifiable outcomes. And the corpus of biological data we’ve collected so far mostly doesn’t contain the right information to provide feedback on “what would be a good treatment for Alzheimer’s”. For AI to accelerate cures we need to start deliberately generating data that contains the answers, and we should start now because the intelligence curve is already outpacing longevity data generation.

The potential impact of AI is huge, and whether we end up curing disease and extending life is too important for hope to be our strategy. The rest of this piece will lay out these assumptions and arguments, and describe the types of data we need to generate to take a revolution in longevity off its indefinite hold.

Subscribe now

AI works when it has task-shaped data and verifiable outcomes.

In general, AI has gotten good at tasks when three ingredients exist:

• First, sufficient compute.

• Second, a massive corpus of training data that establishes base capability.

  • Richer data is better, but isn’t sufficient unless it contains task-relevant answers.

• Third, a feedback system where verifiable outcomes of the relevant task allow the model to improve beyond that base.

  • The speed and cost of verification for that task determines how fast capability accrues.

Once compute scaled, the first language models cleared all three: Humanity’s historical writings served as training data, and early test environments had clear reward functions like rule-based games, interactive chats where humans respond to output, and code or math with formal verifiers. In each case the act of using the output reveals whether it serves its purpose, such that feedback signals are built into the product.

Biology is physical, and therefore requires physical verification of hypotheses. The data either already exists in a form that (at least latently) contains the information needed for a task, or it has to be generated via physical experiments. AI ingesting all of PubMed doesn’t substitute, because the literature is a highly incomplete and biased projection of biology that provides no feedback signal on new hypotheses. In the case of AlphaFold the data already existed. Three decades of PDB structures combined with sequence databases contained enough structure-sequence pairs for the mapping to be learnable, and CASP’s held-out structures provided ground truth verification data.

But for “how do we treat this disease”, the data we have collected usually clears neither bar. A disease is itself often a label we apply to a pattern of dysfunction we don’t fully understand, with names like heart failure, Alzheimer’s, and sarcopenia grouping together patients whose physiology has gone wrong in similar-looking ways. The label does not by itself tell us about the underlying causal trajectory. The answers about how to reverse physiological decline are not latent in cross-sectional snapshots of postmortem organs with and without disease, and the verification loop is a clinical trial that runs for years. To understand what data we need to generate, we have to ask where in biology the events that produce these labels live.

Task-shaped data in longevity

Curing age-related disease puts specific demands on each of the three criteria above. The data has to contain answers about how organisms change over the course of decades, it has to support causal inference, and the verification loop has to be fast enough that we can iterate on hypotheses.

Biology spans multiple physical layers of organization, each with different behaviors, measurement tools, and iteration speeds. Knowing which layer your question lives at, and which layer your experiments actually probe, is the most important orientation for applying AI to disease.

Layer 1 is molecular. Drug molecules binding to proteins is how we make most medicines, and this is AlphaFold etc territory. For DNA, RNA, and proteins we’re good at simulating biophysics, and experiments can happen as fast as robotic arms can move samples around. Other challenges at this layer, e.g. chemical synthesis and pharmacology, ADME, and off-target effects, are making progress.

Layer 2 is cellular. What’s the response when a protein gets blocked? Some activity stops, the cell changes gene expression to compensate, and ends up in a new steady state (or maybe dies) minutes or hours later. Tools like CRISPR and single-cell sequencing apply here.

Layer 3 is physiological. What happens to an organ when its cells change? Other cells get involved, and the new steady state might be destructive or ineffective. This unfolds as gradual changes and responses over weeks to years. Many of the diseases we hope to cure live at this layer: heart failure, dementia, aging. Age-related multimorbidity is definitionally physiological, since it concerns whole-body systems failing to maintain coordinated function over time.

Layer 4 is organismal. Is the organism alive and functioning, with organs working in concert? If we’re discussing lifespan, and many other clinical trial outcomes, we’re asking questions here.

Higher layers emerge from lower ones, but the mapping is many-to-many. The same molecular state in a cell can produce different physiological outcomes depending on the surrounding tissue, organ-level signaling, and timing, so a lower-layer measurement on its own doesn't determine what's happening above. The true higher layer readout does not exist - a cell cannot have blood pressure. This produces a critical asymmetry that governs how we apply AI to biology: lower layers cannot provide verifiable outcomes for tasks at higher layers. Lower-layer measurements contain higher-layer answers only when the higher-layer outcome was directly measured and linked back to them. The bridge between layers has to be measured into existence, rather than inferred from one side.

This is why predicting protein structures give us folding and some genetic diseases, but no cure for Alzheimer's or other complex diseases. Why the Human Genome Project alone didn’t lead to a plethora of cures. Why CRISPR hasn’t either, despite being a magnificent genomic tool. The problems longevity seeks to solve live at layers 3 and 4. Finding solutions requires hard-earned data from those layers.

Richer data is better, but isn’t sufficient unless it contains task-relevant answers.

But here biology runs into its own uncertainty principle: We can learn a little about a system while preserving its state, or we can cut its time short to learn more. Take heart failure. We can measure the output of the organ with an echocardiogram or a blood test. In animal models we can stop the system, take it apart, and take images of the tissue and measure molecules within its cells at that instant. Measuring the visual and molecular state of cells after a treatment improves function at layer 3 is a go-to experiment for linking layers 2 and 3, but it captures the new state without the path that got there. Cytokine cascades diffusing between cells, and countless other mechanisms, cannot be measured continuously without destroying the system.

The speed and cost of verification for that task determines how fast capability accrues.

The other challenge is time. Biology is the study of life, and life unfolds in time. Time can show us causal relationships, as in clinical trials. But layers 3 and 4 are time-costly. A drug for heart failure needs years of human follow-up to show whether it changes outcomes, and an aging intervention needs longer still because the endpoints are further away. Speeding this up with biomarkers requires us to run the slow study first to validate the biomarkers. UK Biobank is the most heavily mined longitudinal dataset in human biology, because it spans layers 1 (genetics and proteomics) and 4 (health outcomes). The interesting findings about disease and survival now coming out of it (like this paper) depend on the cohort having been tracked for fifteen-plus years. Even with an unlimited budget you could not generate UK Biobank by next month. Time-course is a non-substitutable input.

Relying on indefinite optimism would create another jagged frontier for AIxBio where cool advances flow to areas where we have or can quickly generate lots of task-shaped data (e.g. regulatory genomics, microbiology, or tracking health metrics and giving recommendations to stay within pre-defined ranges), or tasks that avoid unknowns (e.g. clinical trial documentation). Impact on longevity will remain on hold.

How we can enable AI to impact longevity

We now have principles for generating data that will enable AI to help longevity:

1. Maximize answers per unit time by starting the gating experiments now and avoiding sequential dependencies, since much biology can’t be sped up.

2. Data has to span the layer where we want AI to provide answers (physiological or organismal for longevity) to be task-shaped.

3. Richer data is better, since models can find structure we wouldn’t have known to look for.

4. When we want treatments, the data needs to contain or link to outcomes generated by interventions or by time, to enable causal inference.

Most obviously, we should start things that can’t be hurried. But even with the clock then ticking in our favor, core data generation tradeoffs remain. The good news is that the tradeoffs aren’t fixed. They can be molded by building the right bridges between layers, between systems, or between fast and slow sources of data.

Bridging layers

In humans and aging animals, intervention at scale isn't possible and longitudinal sampling is mostly restricted to scans and blood draws, while higher-layer pathology can't be modeled in isolated cells. But we can build biotechnological bridges between causal data and the layer we need answers from.

A simple example: You might collect blood from a few thousand autoimmune patients and run single-cell transcriptomics on their immune cells. You have a large cross-sectional dataset at the right layer, but attempting to predict disease outcomes purely from gene expression suffers from lack of verifiable outcomes. Adding a functional assay (like T-cell activation) prior to sequencing lets you couple gene expression to function, and function to patient outcomes.

A more ambitious example: pooled in vivo screening can bridge layers 2 and 3 by testing perturbations in cells inside living organs, such that physiological information is encoded in cell states. Many perturbations (causal) can be introduced into individual cells in the animal patient, with effects compared to controls in the same animal. This means that layer 3 information is encoded in perturbed cells, and can be interpreted with the help of independent paired datasets that link cellular signatures to organ-level physiological outcomes. Animal-to-human data paired datasets can further bridge preclinical experiments to human disease.

Bridging time

We can cheat time either by turning historical samples into task-shaped data, or by validating biomarkers to let faster systems stand in for slow ones2.

The US Veterans Affairs system has tracked millions of patients across decades of medical care and stored blood samples from many of them, but most of those samples have only been run through routine clinical chemistry. Adding proteomics, metabolomics, immune profiling, methylation, and cell-free DNA to a chosen subset would yield a longitudinal molecular dataset on a population whose health outcomes are already known, way faster (and cheaper) than starting a new cohort and waiting for outcomes. Stored samples from past clinical trials work the same way, with causal interventions baked in3. Thinking ahead, we can expect measurement technology to evolve (aided by AI), and so should prioritize banking additional samples from every longitudinal human study.

Experiments designed for future intelligence

The bridges described above can be part of any experiment we design today. But if we truly embrace intelligence abundance, we can combine them into experimental designs built to be maximally useful for AI:

Start a middle-aged longitudinal human cohort now, large enough to capture aging trajectories across populations and split off groups later. Every three or six months, collect blood and run high-dimensional proteomics and other molecular assays. Periodically, layer on scans and functional measurements at the physiological layer: echocardiography, MRI, cognitive testing, frailty assessments. This fulfills ‘right layer’, ‘rich data’, and ‘longitudinal for causality’.

Multiple months between hypothesis testing is much slower than AI would like, so we add a fast experimental loop appropriate for specific hypotheses. The AI ingests each new sample alongside the prior timepoints, the published literature, and the participants’ clinical data, then proposes causal hypotheses about what’s driving the trajectories it sees. A hypothesis that a specific inflammatory signal sits upstream of kidney decline can be tested in mice perturbed to match patients, while finding causal drivers among broad molecular changes can be tested with in vivo pooled screens. Because these systems run faster than human aging, each new sampling timepoint arrives with refined hypotheses and tightened predictions for what each participant's sample should show, without needing to reset the aging clock.

When the prediction loop is calibrated well enough to be trusted, the cohort extends into an adaptive prevention trial4. The regulatory and ethical frame changes because you’re now intervening, but the experimental design carries over5 with one critical addition: because the AI already predicts each participant’s individual trajectory from the observational phase, we can compare effects both to the predicted progression of the individual patients, and to their peers in the control group. The AI commits to a specific prediction before each time point, and the measurement either confirms the intervention worked as predicted or shows where the model is still wrong. Compared to a conventional trial that waits years for arm-vs-arm divergence, here every datapoint is a verification event. This fulfills the criteria we laid out: We’re starting the slowest part (human aging) right away, but rather than waiting for one long trial we add multiple data feedback points, spanning physiological layer human data with faster loops of causal intervention data. Starting such a trial does not require new technology (but the trial design can adopt new tech as needed), it only needs foresight and commitment.

Objections to the layer limitation

Before we conclude, I want to address the common arguments for an imminent biological data abundance that would rewrite the rules we’re currently operating under.

1. The most common idea to save AI from physical reality is automated labs. This is a natural idea in part because it’s already happening: biopharma has been using liquid handling robots to test massive drug libraries for years. But whether done by pharma or AI-first biotechs, the automation never goes beyond the cell layer, because today's robots lack the manual dexterity for animal work. Even on Elon timelines, human-equivalent dexterity is a few years out and scaling robotics to ~double current R&D work will take a decade6. This can and should happen, but even that doubling of capacity will leave AI starved for physiological data.

2. If higher-layer behavior is in principle derivable from sufficient lower-layer measurements, scaling lower-layer data collection could yield higher-layer answers without measuring them directly. And layer 2 costs are coming down fast7. However, even our lower-layer measurements only sparsely capture biological reality. Single-cell sequencing captures roughly a third of the mRNAs in a cell, itself a fraction of total RNA. Then there are ~10¹⁰ proteins, along with their splice variants and post-translational modifications. And lipids, and things we don't yet know we should measure. All of this is captured at a single moment in time. We would need dramatically improved methods that fight physical reality to generate data to resolve even weak emergence.

3. If physical experiments are hard, could we just simulate them with virtual experiments?8 The critical question for longevity is whether we would be able to simulate physiological layers without data from those layers. We could discuss how compute-intensive simulations would be under different assumptions9, but the strongest evidence against this possibility is that none of the virtual cell models in development make use of AI breakthroughs from the molecular layer. AlphaFold and its successors have hugely improved protein structure modeling, and genomic models can simulate evolution and predict gene expression. But AlphaFold is not part of any frontier virtual cell model. One leading model, X-Cell, uses protein interaction data extensively, and notably relies on past experimental observations rather than AlphaFold predictions. Rather than build a model from subcomponents that do well at predicting molecular outcomes, every virtual cell model is based on cell-level measurements like imaging or RNA sequencing of whole cells. In other words, data at the layer the model is attempting to predict.

Conclusion

It’s possible that sheer scale of data centers and experiments on isolated cells in plastic wells will end up unlocking biological understanding that fundamentally changes how we approach curing disease. But many billions of dollars are already backing this hypothetical, across frontier AI labs, biotech MegaCos, and plentiful funding for AIxBio startups. The broader thrust of AI for bio is not focused on the problems particular to longevity, and the longitudinal cohorts and bridging datasets that would enable AI-driven longevity research mostly don’t exist yet. The time to start them is now, because the data takes years to mature regardless of how fast intelligence scales. If we want to maximize the probability of longevity benefitting from AI in time to yield medicines in our lifetimes, it seems wise to allocate at least 5-10% of AIxBio funding to the type of enabling work described here.

I’m grateful to Marton Meszaros, Effie Klimi, Hannu Rajaniemi, Julie Vaughn, Henry Miller, Alex Epstein, Michael Reisman, and Oliver Hahn for feedback on early versions of this essay.

Subscribe now

Why?

Unless your company has access to at least a billion mission-aligned dollars, you will find that targeting ‘longevity’ is not a viable path. There is not yet a reimbursable endpoint for aging, no defined approval criteria, and no previous established commercial model for running an aging trial. We believe this will change (and are actively working to accelerate that with e.g. multi-morbidity trial designs for aging and biomarker validation programs). But until that is done, investors will want to see a path to profit that doesn’t involve figuring out new ways to do (long) clinical trials and new ways to pay for medicines in the US healthcare system2. This means picking a recognized disease where you think your drug will work, with plans to expand to additional diseases and/or preventative treatments eventually.

99% of all longevity therapeutic companies end up on this path, even when the founders want to cure aging. BioAge did, NewLimit did, etc. Some do so right away, others push the longevity goal until they’re no longer able to raise funding for that, then adjust plans. We at Norn Group think about the big picture of producing longevity, and care about companies succeeding. So a while back we decided to collect all the most important information for indication selection for every age-related disease we could think of, and serve this up to the community to save everyone else from having to do the same. We’ve now updated the table to add additional diseases and represent current standard of care.

How to use this resource

The spreadsheet covers 47 age-related indications across various organ systems. For each indication, we have included a brief overview of disease mechanism and links to mechanisms of aging, data on human incidence and patient population, current treatment options and clinical endpoints/measurables, relevant animal models of the disease and their relevance to humans, size and duration of clinical trials, as well as market size. This should give you most of the information needed to consider the best fit for your therapeutic candidates. We suggest paying extra attention to the following:

Where does your biology fit?

Use our spreadsheet to orient yourself, then go deep on the papers to build confidence. Don’t stop at a prominent academic or company saying that e.g. senescence is very important in disease X, get into the data. Ultimately drug programs succeed or fail on whether the drug you build is solving the right problem.

Are there compelling preclinical models, or will you have to take things into humans?

When animal models are a poor fit for human disease, as in pulmonary fibrosis or Alzheimer’s, investors will be less impressed with even strong preclinical data. This means you might have to raise enough money to run human trials before things really take off. Can you do that? Look at the key readouts for human trials, does your data suggest you can move those measures?

What will it take to run human trials?

Speaking of human trials, how long will they take? How much money will you burn during those years? And how much will the trials cost? As a rough guideline, assume $50K-200K per patient. If trials are slower and more expensive, you’ll need a more impressive drug to motivate investment.

Is there an unmet need?

How good is the standard of care treatment, and how would your drug differentiate? It’s not enough to have something that works, you need to be either better than or additive with what’s already on the market. Look at the drugs in clinical trials too (not included here). You might also consider whether you’ll be able to pivot to another indication if something really good gets approved while you’re building.

We hope this overview will be useful, but you’ll still have to think deeply about therapeutic strategy. Indication selection is at the intersection of business and science decisions, and having a crisp picture (while being willing to pivot as you learn) will help you go further on what you’ve raised so far.

This resource is free. It was built by Martin Borch Jensen and Longevity Apprentices Madi Ueland and Divya Cohen, with input from Norn Nexus members.

If you find an error or an omission, please reach out.

Subscribe now

1

age1 wrote a nice overview of other key decisions:

Libraries for the Future

Founder’s Playbook: From Target to Development Strategy

Finally, the last piece of the series…

Read more

6 months ago · 5 likes · age1 and Lynetta Wang

This piece was written by Nicolas Di Leo, as a part of the Talent Bridge program.

Introduction

The mechanistic target of rapamycin (mTOR) pathway is central to cell growth and aging, integrating signals from nutrients and stress to regulate metabolism, protein synthesis, and autophagy. Decades of research across various species have identified mTOR inhibition as a powerful strategy for extending lifespan and healthspan. In particular, the mTORC1 inhibitor rapamycin has repeatedly been shown to delay aging in yeast, worms, flies, and mice even when treatment is started late in life [1,2]. These findings position mTOR as one of the most validated targets for translating longevity interventions to humans. Yet turning this laboratory success into clinical practice has proven challenging.

Early human trials of mTOR inhibitors hint at improved immune function, cardiovascular markers, and skin health with little side effects [3]. However, they also expose critical gaps: how to measure “aging” in trials, how to balance efficacy with safety in otherwise healthy people, and how to design studies that regulators and industry can get behind.

This report reviews the clinical translation of mTORC1-targeting interventions for aging and age-related disease, with a focus on rapamycin and related strategies outside of traditional immunosuppressive use. It begins with a brief biological background on mTOR and rapamycin, then outlines why mTORC1 remains one of the most compelling pathways for geroscience translation across major domains of age-related pathology. The report next examines past human clinical trials, followed by a review of ongoing and emerging studies, enriched by exclusive insights from the principal investigators leading these programs. It then synthesizes the main translational gaps that continue to limit progress, including endpoint selection, biomarker validation, safety monitoring, and indication prioritization, and discusses practical directions for improving future trial design. The overall aim is to provide a structured, evidence-based assessment of where the field stands, what has been learned from prior efforts, and what is needed to make mTORC1-targeting therapies more interpretable, testable, and clinically actionable in geroscience and longevity medicine.

mTOR and Rapamycin: A Brief Background

mTOR is a conserved serine/threonine kinase that forms two complexes, mTORC1 and mTORC2, which regulate cell growth and metabolism in response to nutrients, growth factors, and energy status. mTORC1, in particular, promotes anabolic processes (protein/lipid synthesis) and inhibits autophagy. Inhibition of mTORC1 triggers a shift from growth to maintenance: boosting autophagy, improving proteostasis, and mitigating cellular senescence [1]. This mechanism helps explain why chronic dietary restriction, which suppresses mTOR signaling, extends lifespan across many species. Rapamycin, a natural product discovered from Streptomyces bacteria, pharmacologically recapitulates key aspects of this low-nutrient state by binding FKBP12 and preferentially inhibiting mTORC1. Importantly, rapamycin remains one of the most reproducible longevity interventions identified to date: it was the first small molecule shown to robustly extend lifespan in mammals, and it continues to stand out from most candidate “anti-aging” compounds in the strength and consistency of the evidence. In 2009, Harrison et al. from the NIA Intervention Testing Program (a multi-center platform using genetically heterogeneous, outbred mice) demonstrated that rapamycin administration starting as late as 20 months of age increased median lifespan by ~10–16% [4]. Subsequent studies have replicated lifespan extension across multiple cohorts, strains, and both sexes, supporting the conclusion that mTORC1 is not a niche target but a central regulator of mammalian aging biology.

Encouragingly, genetic evidence aligns with pharmacology: mice or flies with reduced mTORC1 activity (or increased upstream inhibitors like TSC) tend to live longer, whereas hyperactive mTOR signaling accelerates aging phenotypes [1]. Meanwhile, chronic hyperactivation of mTOR is implicated in many age-related diseases (e.g. cancer, type 2 diabetes, neurodegeneration) [1]. These insights made mTOR an attractive target to slow aging and prevent multiple chronic diseases at once. Rapamycin and its analogs (“rapalogs” such as everolimus) are already FDA-approved for other indications, including organ transplant rejection (as immunosuppressants), certain cancers, and rare disorders like tuberous sclerosis complex (TSC) [1]. Could lower doses or intermittent schedules of these same drugs improve healthspan in otherwise healthy older adults? This provocative idea has catalyzed numerous trials in the past decade.

However, translating an “anti-aging” therapy into the clinic has proved to be tricky. Aging itself is not recognized as a disease by regulators, so trials must target specific age-related conditions or risk factors. Furthermore, measuring aging is non-trivial – unlike a single disease, aging manifests as a multifaceted decline. Safety is another paramount concern: rapamycin at therapeutic doses can cause immunosuppression, metabolic disruptions, mouth ulcers, and other side effects [1]. For “privileged” use in generally healthy populations, even mild toxicities need to be minimized. Thus, a central challenge has been finding dosing strategies that inhibit mTORC1 enough to slow aging, while avoiding excessive mTORC2 inhibition that may underlie many side effects [2].

In summary, the rationale for targeting mTOR in longevity is compelling: it’s a conserved aging pathway with pharmacological tools already in hand. But the path forward requires careful trial design to demonstrate benefits in humans without unacceptable risks. Below we review what has been learned so far from preclinical studies and human trials, and how this informs current efforts to harness mTOR inhibition to combat chronic diseases of aging.

mTOR’s Clinical and Translational Appeal

A consistent theme in longevity research is that, to gain traction clinically in 2026, given the bureaucratic hurdles in place, therapies should target well-defined chronic diseases or risk factors, thereby treating “ageing” by treating an age-related condition. mTOR inhibitors exemplify this approach. Rather than proposing rapamycin as a general anti-ageing drug (an idea that faces regulatory hurdles), many researchers are testing rapamycin in specific diseases of aging, with the hope that success in those domains will validate its broader geroprotective effects. This section highlights some chronic disease contexts where mTOR targeting is being pursued as a longevity strategy, and what has been learned so far.

In immune senescence, aging is characterized by reduced naïve T-cell function, impaired vaccine responsiveness, weaker antiviral defenses, and a chronic inflammatory baseline that paradoxically coexists with immunodeficiency. mTOR activity is a central regulator of immune-cell activation, differentiation, and memory formation, and excessive or dysregulated signaling with age can contribute to maladaptive immune states. In preclinical systems, partial mTORC1 inhibition often improves aspects of immune function and reduces inflammatory tone, and in humans the most consistent early clinical signals for mTOR modulation have come from immune-related endpoints and infection susceptibility, supporting the idea that immune aging is one of the most tractable entry points for geroprotective translation.

Cancer provides a second, compelling rationale because mTORC1 is a core effector of growth and proliferation downstream of oncogenic pathways, and many tumors exploit mTOR signaling to sustain biomass accumulation, survival, and metabolic reprogramming. Aging increases cancer risk through accumulated mutations, epigenetic drift, and changes in the tissue environment that favor clonal expansion. In preclinical models, dialing down mTORC1 tends to reduce tumor initiation and progression in multiple settings, partly by limiting proliferative signaling and partly by improving cellular housekeeping and reducing inflammatory, pro-tumor microenvironments. Clinically, mTOR inhibitors are already established in oncology and in genetic disorders characterized by hyperactive mTOR signaling, demonstrating that the pathway is druggable in humans. For geroscience, the translational opportunity is not necessarily to treat advanced cancers with low-dose mTOR inhibition, but to test whether carefully dosed, long-term modulation can reduce cancer-promoting conditions that build up with age, potentially delaying incidence or slowing progression of early lesions.

Cardiovascular aging is increasingly understood as a convergence of endothelial dysfunction, vascular stiffness, chronic inflammation, impaired autophagy, mitochondrial decline, and maladaptive remodeling of heart tissue. mTORC1 influences all of these processes, including smooth muscle cell growth, fibrosis pathways, inflammatory signaling, and cellular quality control. In animal models, partial mTORC1 inhibition can improve aspects of cardiac function, reduce age-associated hypertrophy and fibrosis signals, and support better vascular function.

Metabolic disease sits in a similar category. mTOR integrates nutrient availability with anabolic signaling, and chronic overnutrition can lock tissues into a persistent growth mode that contributes to insulin resistance, ectopic lipid deposition, and impaired metabolic flexibility. In preclinical systems, modulating mTORC1 can improve metabolic resilience in some contexts by reducing inflammatory signaling and enhancing autophagy, but sustained inhibition that spills into mTORC2 can impair insulin signaling and glucose control. Human experience mirrors this complexity: metabolic effects depend strongly on dose, schedule, baseline metabolic state, and duration.

Neurodegeneration is another domain where the biology points to mTOR as a plausible upstream lever. Many neurodegenerative diseases involve failures of proteostasis, impaired clearance of misfolded proteins, mitochondrial dysfunction, and chronic neuroinflammation, all processes that mTORC1 can influence through its control of autophagy and cellular stress responses. In preclinical models, reducing mTORC1 signaling often improves clearance pathways and can blunt inflammatory and synaptic dysfunction phenotypes in models of protein aggregation. Translationally, the challenge is that the brain is both difficult to access for pharmacodynamic readouts and sensitive to perturbations in protein synthesis and synaptic plasticity, which may render the therapeutic window impractical. Still, the conceptual appeal is strong: a pathway that enhances cellular clearance may be especially relevant in conditions in which accumulated cellular debris drive progressive loss of function.

Across these domains, the broader point is that mTOR is one of the few targets with a coherent mechanistic link between multiple age-related pathologies, and with real-world drug precedents that reduce risk.

Past Clinical Trials

Across human research to date, mTOR inhibition has been tested in several clinical contexts that go beyond immunosuppression, with the overarching goal of determining whether modulating this pathway can produce measurable improvements in age-related physiology. In general, these trials have used low or intermittent dosing and have focused on outcomes that can shift within weeks to months, including vaccine responsiveness and infection susceptibility, markers of systemic inflammation, cardiometabolic readouts, skin and tissue aging features, and functional measures such as strength, endurance, and aerobic capacity. A consistent theme is that the interventions are usually feasible and generally well tolerated in older adults, while also highlighting substantial heterogeneity in biological response across individuals, sex and endpoints. Collectively, the studies highlighted in the table below support the concept that mTOR modulation can influence human aging-relevant biology in multiple organ systems, but they also emphasize the importance of selecting appropriate populations, defining clinically interpretable endpoints, and incorporating mechanistic biomarkers that can confirm target engagement and help explain variability in outcomes.

Notable past biotech efforts

resTORbio

resTORbio is a useful case study in how promising geroscience biology can fail in late-stage development when trial design, endpoint selection, and commercial incentives become misaligned. The company was formed in 2016 after PureTech Health licensed two Phase II–ready mTORC1-targeting assets from Novartis (RTB101 plus an inhaled companion), with the stated goal of treating age-related decline in immune function and potentially expanding into other age-associated diseases. Early data from Novartis’ program helped set the narrative: a short Phase 2a study in adults over 65 reported meaningful reductions in laboratory-confirmed infections and supported the idea that partial mTORC1 inhibition could improve antiviral immunity in older humans. Backed by PureTech’s initial investment and strong investor enthusiasm, resTORbio moved rapidly: after a Series B raise, the company went public early with a valuation that reflected high expectations that a pivotal respiratory trial could establish the first broadly “anti-immunosenescence” therapeutic indication.

The turning point came during the transition from Phase 2b to Phase 3. In its multicenter, randomized, double-blind Phase 2b trial conducted over a winter period (16 weeks treatment plus follow-up), the company reported positive topline findings for RTB101 10 mg once daily, including a reduction in respiratory tract infections that was treated as sufficient justification to progress into a single large registrational study. However, the Phase 2b dataset also raised concerns: the apparent efficacy signal relied heavily on pooling and subgroup decisions, while higher dosing (and some combination regimens) appeared to increase respiratory infections, indicating a narrow therapeutic window and reinforcing that dose selection was not trivial. Despite these issues, resTORbio advanced into PROTECTOR-1, a Phase 3 trial involving> 1,000 participants, selecting RTB101 as the lead agent rather than maintaining the broader combination strategy that had generated the initial signal. This choice made the program more commercially defensible (because RTB101 was proprietary), but also concentrated the clinical risk into a single compound without the support of a better-established comparator.

Phase 3 ultimately failed to meet its primary endpoint, and the reasons illustrate how sensitive geroscience-style trials can be to operational decisions. Near the end of the Phase 3 planning and discussions, the FDA pushed the program away from a virologic, lab-confirmed endpoint and toward a patient-reported symptomatic composite endpoint (“clinically symptomatic respiratory illness”). At the same time, additional major changes were introduced: the Phase 3 population was healthier and more restricted than in Phase 2b (e.g., excluding smokers and certain chronic lung disease groups that had been included previously), and symptom capture shifted from frequent phone-based assessment to daily eDiary self-reporting. These shifts increased measurement noise and made the endpoint more vulnerable to inconsistent reporting and mild symptoms that may not correspond to clinically meaningful infections. In the final analysis, the symptomatic illness endpoint showed essentially no separation between RTB101 and placebo. Importantly, the program did not completely collapse biologically: across Phase 2b and Phase 3, treatment continued to show upregulation of interferon-stimulated antiviral gene programs, suggesting pathway engagement, yet these molecular effects did not translate into a robust, regulator-acceptable clinical benefit under the Phase 3 conditions.

The aftermath was swift and severe. After the Phase 3 failure was announced in November 2019, the company halted development for the respiratory indication and lost market confidence rapidly. Although resTORbio attempted to continue through smaller follow-on programs (including exploratory work in Parkinson’s disease), the lack of a validated clinical success, combined with the cost of running additional large trials without a major pharmaceutical partner, made recovery difficult. The company eventually concluded with a reverse merger into another biotech. In hindsight, resTORbio’s trajectory highlights several practical lessons for longevity translation: small or fragile Phase 2 signals should ideally be reinforced with an intermediate step (or replicated under closely matched conditions), pivotal endpoints must be both clinically meaningful and operationally robust, and changing multiple variables between phases can erase interpretability.

The Mannick 2021 resTORbio program stands out today as one of the most ambitious mTORC1-focused efforts aimed at regulatory approval of a rapalog for an ageing-related condition.

Navitor Pharmaceutical

Navitor Pharmaceuticals represents a branch of “mTOR translation” in industry, built less around chronic mTORC1 inhibition for aging and more around activation of mTORC1 for defined indications, particularly in the CNS.

The company’s flagship program, NV-5138 (also referred to as SPN-820 by its partner), was an orally bioavailable Sestrin2-binding leucine analog designed to transiently activate mTORC1 in the brain, addressing the practical limitation that dietary leucine is rapidly metabolised and does not reliably engage brain mTORC1.

That strategy translated into a partnership with Supernus in 2020 to develop NV-5138/SPN-820 clinically in neuropsychiatric indications, culminating in a larger Phase 2b trial in treatment-resistant depression that reported topline results in February 2025. In that study, SPN-820 did not meet its primary endpoint, despite the sound mechanism and smaller exploratory signals. The value of the Navitor case is not that it validates or invalidates mTOR as a target, but that it illustrates how difficult it is to convert compelling pathway biology into robust clinical benefit when endpoints are noisy and effect sizes may be modest.

At the same time, Navitor’s broader corporate trajectory shows how mTORC1 assets can still create value through indication choice and partnering: in 2022, Janssen acquired Navitor’s spinout Anakuria Therapeutics, built around an mTOR-pathway program aimed at autosomal dominant polycystic kidney disease.

Ongoing/Future Trials

Thanks to growing evidence and significant philanthropic funding, there is now a surge of new clinical trials testing mTOR-based interventions for longevity. These trials are more ambitious in scope: they involve larger sample sizes, longer durations, multi-omics analyses, and creative endpoints. Many are directly informed by the lessons of earlier studies. Below, we highlight several ongoing or upcoming trials of which PI we have interviewed to obtain the most up-to-date information and exclusive insights on how the trials are going.

RAPID trial (University of Washington)

Chronic periodontitis (gum disease) is an age-associated inflammatory condition that not only causes tooth loss but has systemic links to health (it’s associated with higher risk of cardiovascular disease and frailty). Dr. Jonathan An and colleagues are launching the RAPID trial to test whether rapamycin can reduce oral “inflammaging” and improve periodontal health, motivated in part by earlier pre-clinical work suggesting rapamycin can improve oral health and improve periodontal disease

RAPID is designed as a double-blind, placebo-controlled study intentionally structured as a foundational, first-in-field study to enable subsequent larger-scale trials. Two major bottlenecks shaped the design and the timeline of the study – regulatory approval and academic infrastructure. From a regulatory standpoint, FDA approval required approximately one year. Because systemic rapamycin had not previously been evaluated for an oral health indication, the Division of Dermatology and Dentistry (FDA) lacked precedent and initially raised safety concerns for use in older adults, requesting IND-backed safety data specific to this population. Few published rapamycin-in-aging studies were IND-enabled in a manner that could be directly leveraged for a dental indication, further slowing the approval process. In parallel, substantial academic infrastructure had to be built de novo. This included establishing compliant electronic medical/dental record workflows, secure data servers, and coordinated data capture across the medical and dental schools, which operate on distinct EMR systems and governance structures. Additional effort was required for ongoing IND maintenance, safety reporting, and cross-school coordination, components that are rarely centralized for dental clinical trials within academic settings.

To address these bottlenecks, the team first conducted a small pilot study (n≈15) to establish safety and feasibility at very low dosages (0.5mg/week and 1mg/week) with careful monitoring of systemic biomarkers, gingival crevicular fluid, and periodontal clinical measures, and innate and adaptive immune responses assessed through PBMCs over a 1-year period (each participant was in the trial for up to 1 year). Systemic and local markers remained stable, while periodontal clinical measures shifted in a direction consistent with the study hypothesis. Importantly, this pilot also served as a full end-to-end validation of the newly built regulatory and academic infrastructure. With this infrastructure now tested in practice, the platform can support future dental, medical clinical trials seamlessly without rebuilding foundational systems.

Participants will be randomized to placebo or one of two low-dose sirolimus regimens administered over an 8-week period, followed by standardized cleaning visits with longitudinal data collection over 9 months. RAPID is positioned not as an endpoint, but as a platform to generate larger, more robust, and rigorous efficacy data in fully powered follow-on trials. Building directly on these findings, the team has submitted a large-scale NIH grant proposal to support a comprehensive, definitive efficacy trial informed by RAPID’s safety, feasibility, and preliminary outcome data.

Biomarker Study (University of Wisconsin)

Dudley Lamming is running a biomarker-centric study to evaluate how selective mTORC1 inhibition maps onto measurable hallmarks of ageing, with the explicit goal of identifying which proposed ageing biomarkers are actually responsive to mTORC1 inhibition and therefore useful for designing geroscience-guided Phase 3 trials. The study is collecting blood from roughly 50 self-reported rapamycin users, mostly off-label users and biohackers, who typically take about 6 mg once weekly and have used it anywhere from one month to three years. To make this feasible, mobile phlebotomy operation is used for drawing blood in participants’ homes across the US, and the team has emphasized that this logistics layer, more than the biology, has been the main operational challenge. The samples are being profiled using multi-omics approaches, including metabolomics and lipidomics; the data are already generated but still undergoing full analysis. The early signals they are comfortable sharing include a decrease in hsCRP, consistent with lower systemic inflammation, and an absence of hyperlipidemia despite this being a commonly cited side effect of rapamycin in other contexts. The investigators plan to publish the results early in 2026, and they explicitly note that Impetus funding was instrumental in getting the work off the ground.

VIBRANT Study (Columbia University)

Co-led by Dr. Yousin Suh and Dr. Zev Williams, this is a prospective, randomized, double-blind, placebo-controlled trial testing whether once-weekly, low-dose oral rapamycin (sirolimus; Rapamune) can slow ovarian aging and delay menopause, motivated by the idea that postponing menopause could reduce downstream risks tied to this transition, including cardiovascular disease, dementia, and increased mortality. The study focuses on ovarian aging because the ovaries appear to age faster than other organs in women, with a steep functional decline beginning around ages 30–35 and culminating near age 50, and the team’s multi-omics comparisons across tissues in younger versus older women suggest mTOR is unusually active in the ovary, consistent with strong preclinical evidence that mTOR inhibition can preserve fertility and ovarian reserve. VIBRANT enrolls women in their 30s and uses ovarian reserve as the primary endpoint, assessed through vaginal ultrasound follicle counts together with hormone measures such as FSH and AMH, while also collecting blood for multi-omics profiling. The original VIBRANT-1 design (6 mg/week for 6 months with follow-up) was revised after substantial institutional and regulatory scrutiny to 5 mg/week for 3 months with 9 months of follow-up, and a second practical bottleneck was finding a pharmacy able to manufacture a placebo that closely matched the active pill to maintain blinding. The trial enrolled 50 participants (25 placebo, 25 treatment) who came to the New York clinic weekly for dosing and were monitored monthly; dosing and follow-up are now complete, and while the team still needs to finalize analysis and unblind the dataset, they report early separation in the primary ovarian-reserve trend lines, no disruption of menstrual cyclicity, and an absence of commonly expected side effects, with results anticipated early in 2026.

The team has also announced the submission of a grant proposal for VIBRANT 2, a follow-up study which will involve 8 institutions across the US with the goal of increasing the sample size and extending the treatment duration to better discern clinically meaningful changes.

RAPAEX

A practical translational concern is whether mTORC1 inhibition interferes with muscle adaptation, especially in older adults using resistance training to preserve function. RAPAEX was conceived in 2021 by Dr. Brad Stanfield after he recognized conflicting signals in the human literature: on one hand, acute mechanistic studies indicate that a single rapamycin exposure around exercise can blunt canonical post-exercise anabolic signaling and muscle protein synthesis, raising concern that mTORC1 inhibition might impair training adaptations if timed inappropriately. On the other hand, older adults, particularly those who are sedentary and at risk of sarcopenia, often have hyperactive mTOR1 signaling at the muscle. Stanfield’s working hypothesis was that weekly rapamycin properly timed might “rebalance” muscle mTORC1 activity in older individuals ultimately aiding the anabolic stimulus of training. To test this directly, RAPAEX randomized two groups initially planned as 20 participants per arm (40 total; 9 dropped out over the study) of previously inactive older adults aged 65–85 into a double-blind, placebo-controlled design where both arms completed an exercise program three days per week on Monday, Wednesday, and Friday, while receiving either placebo or sirolimus 6 mg once per week on Saturdays for 13 weeks. The 30-second chair stand test was the primary endpoint, with grip strength, 6-minute walk test, bloodwork, and the SF-36 wellbeing questionnaire as secondary endpoints. A major practical issue was producing a placebo that matched the intervention capsules; the team solved this with a simple but effective blinding strategy by using standard Pfizer sirolimus tablets placed inside a larger white opaque capsule filled with cellulose, while the placebo group received the identical larger capsule filled with cellulose but without the sirolimus tablet inside, and no participant identified this workaround during the trial. From an execution standpoint, Stanfield emphasized that the largest hurdle was funding the study, and that operational success depended heavily on providing clear, detailed instructions to the clinical and hospital staff delivering the protocol to minimize dropouts and other avoidable logistical problems. The study has now concluded, with data collected, analyzed, and submitted to a journal.

Additional Notable Trials

EVERLAST (University of Wisconsin, Madison)

EVERLAST (Everolimus Aging Study) is a Phase 2, randomized, double-blind, placebo-controlled trial designed to test whether pharmacologic mTORC1 inhibition produces measurable changes in aging-relevant physiology and candidate biomarkers in a metabolically enriched older population. The trial has completed enrollment and reached its target sample size of 72 insulin-resistant adults aged 55–80, a design choice that prioritizes clinical detectability by studying individuals with mild-to-moderate metabolic dysfunction rather than exceptionally healthy participants. EVERLAST is funded primarily by the NIH/NIA, with Impetus Grants supporting additional biomarker-focused analyses intended to quantify molecular signatures that may shift under treatment. The investigators anticipate completing data and biospecimen collection in July 2026, after which the primary analyses can be finalized.

RAP PAC (University of Wisconsin, Madison)

In parallel, RAP PAC is an NIH/NIA-funded Phase 1 dose-identification study that is currently ongoing and is intended to map tolerable and biologically active dosing regimens for rapamycin and everolimus in humans.

REACH (University of Texas, San Antonio)

REACH is a Phase 2, randomized, placebo-controlled clinical trial evaluating the safety, tolerability, and feasibility of 12 months of oral rapamycin in older adults with amnestic mild cognitive impairment (aMCI) and early-stage Alzheimer’s disease. The primary objective emphasizes feasibility and safety, but the trial is also positioned to generate outcome data including cognitive measures and Alzheimer’s-related biomarkers.

Marmoset Study (University of Texas, San Antonio)

It’s also worth highlighting marmoset trial led by Adam Salmon at the University of Texas, which tests rapamycin in common marmosets, a new world monkey species, in a controlled trial. The rationale is that marmosets recapitulate several features of human aging more closely than mice, including a broader spectrum of age-related morbidity, whereas laboratory mice often die predominantly from cancer or strain-specific pathology. Nonhuman primates share many of the diseases of age with humans, and thus might inform more on aging effects of rapamycin vs disease specific effects. To address this, his group initiated a long-duration study nearly ten years ago in 66 marmosets split evenly between treatment and control arms. The trial enrolled older animals at baseline (average age ~7 years, with an average lifespan of ~10 years) sourced from multiple facilities across the United States, creating a heterogeneous cohort with diverse backgrounds that better resembles real-world human variability. The intervention uses daily rapamycin at 1 mg/kg, a regimen comparable to doses used in mouse lifespan studies and intended to achieve blood concentrations similar to those seen in transplant medicine, substantially higher than the intermittent, low-dose schedules commonly used off-label for longevity in humans. The primary endpoint is survival, with longitudinal monitoring that includes safety markers, cognitive testing, and selected functional measures; the team also collected blood for multi-omics profiling and for pharmacodynamic assessment of target engagement in circulating cells via S6K phosphorylation, and performed systematic tissue collection at death for pathological evaluation, which remains ongoing. Previous reports from this study have shown that rapamycin was well-tolerated without clear serious side effects in this group, and that mTOR signaling was inhibited though with inter-individual variability. The effects of rapamycin on longevity and healthy aging in this cohort are being prepared for publication in the near future.

TRIAD

Finally, another bridge study is represented by TRIAD, the large, multicenter, randomized rapamycin trial in companion dogs led by Matt Kaeberlein. This isn’t a human trial, but it is one of the most informative translation layers between mice and humans because pet dogs share environmental exposures, spontaneous disease patterns, and clinically meaningful outcomes in a compressed timeframe. The trial is explicitly designed to test whether rapamycin improves healthspan metrics and potentially lifespan in middle-aged dogs, using clinical-trial methodology (randomized, double-masked, placebo-controlled, multicenter). The study launched in 2020 but was interrupted by the pandemic and subsequently faced an 18-month funding gap before securing new support and restarting. The trial is statistically powered to enroll around 600 dogs and detect a 9% change in lifespan over a three-year follow-up period that follows a one-year treatment period during which dogs receive low-dose weekly rapamycin. Enrollment is expected to complete within the next year, after which all enrolled dogs will be followed for three years before a formal primary analysis is conducted. Importantly, this analysis will not yet include lifespan data — the team does not expect to publish survival curves before 2029. Around 200 dogs are currently in some stage of the study, either having completed treatment and follow-up, actively receiving treatment, or in the follow-up phase. The dataset remains blinded, and no interim signal has prompted early unmasking; no major side effects have been detected. One possible complication is that the team is not observing the rate of deaths they originally anticipated, likely due to a healthy cohort bias. Because enrollment selected for otherwise healthy dogs rather than a broadly representative population, the baseline mortality rate may be lower than the power calculation assumed, potentially requiring extended follow-up beyond the initial three years to accumulate sufficient events. TRIAD is among the most anticipated studies in the field: its sample size, duration, and clinical relevance mean that a positive result would carry more weight than almost anything else currently in progress.

VITAL-H and TORnado (ARPA-H PROSPR Program)

In early 2026, the federal government placed a significant bet on the longevity field. ARPA-H, the health-focused cousin of DARPA, created to fund high-risk, high-reward biomedical research, announced the winners of its PROSPR program, distributing up to $144 million across seven teams with the explicit goal of treating aging as something that can be slowed, or even reversed. Two of those seven awards landed in mTOR territory — a meaningful concentration in a single pathway, and a reflection of how central mTORC1 inhibition has become to the field’s translational ambitions in 2026.

The Barshop Institute at UT Health San Antonio received up to $38 million to run VITAL-H (Validation and Intervention Testing for Aging, Longevity and Healthspan), which is positioned as the first Phase 3 clinical trial targeting ageing in adults not yet diagnosed with a specific disease, randomising generally healthy individuals aged 60–65 to rapamycin, dapagliflozin, or semaglutide. The trial’s use of a composite “Intrinsic Capacity” endpoint spanning cognition, mobility, psychological well-being, vitality, and sensory function reflects a deliberate attempt to operationalise ageing as a measurable, multidimensional construct rather than a single disease.

Cambrian Biopharma’s PROSPR award of up to $30.8 million takes a different angle. Rather than working with rapamycin itself, their TORnado platform has produced a new class of compounds designed to hit mTORC1 selectively, without mTORC2 engagement that many researchers believe is responsible for rapamycin’s less desirable metabolic effects.

Notable Current Biotech Efforts

Aeovian Pharmaceuticals

Aeovian Pharmaceuticals is an instructive example of a different translational strategy for mTORC1 biology than the “broad ageing indication” route attempted by some earlier programs. Rather than starting in heterogeneous older populations where endpoints are difficult and signal-to-noise is poor, Aeovian has prioritized genetically defined diseases where mTORC1 hyperactivation is causal, with the goal of de-risking mechanism, dose selection, and target engagement before expanding into broader chronic and age-associated indications. The company was originally incubated at the Buck Institute for Research on Aging and has described itself as building a pipeline of highly selective mTORC1 inhibitors designed to restore “metabolic quality control” while avoiding the tolerability constraints that can occur when less selective mTOR inhibitors impact other nodes of the pathway. This positioning is central to their thesis: in mTOR-driven disorders such as tuberous sclerosis complex (TSC), mTORC1 signaling is pathologically elevated due to loss-of-function mutations in TSC1/TSC2, and existing treatment approaches (including mTOR inhibitors used in related settings) can be limited by adverse effects that restrict chronic use and dose intensity. Aeovian’s lead program is therefore framed as a pharmacology problem as much as a biology problem, building an inhibitor with strong mTORC1 selectivity plus adequate CNS penetration, since epilepsy is among the most common and clinically burdensome manifestations in TSC.

This approach has translated into a clinical development plan centered on AV078, which Aeovian describes as a first-in-class, CNS-penetrant selective mTORC1 inhibitor intended for refractory epilepsy associated with TSC. In March 2024, the company announced dosing of the first cohort in a Phase 1 clinical trial and raised an additional $50 million to support completion of early clinical development and preparation for subsequent Phase 2 testing. Public statements around the 2024 financing repeatedly emphasize two priorities: (i) establishing safety and pharmacology in healthy volunteers and (ii) moving into adult and pediatric TSC epilepsy populations where the clinical endpoints and patient need are clear. In December 2025, Aeovian disclosed a further $55 million Series B to advance AV078 into Phase 2 development in TSC-related refractory epilepsy, again highlighting the intent to deliver improved efficacy and tolerability compared with current options and to achieve sustained seizure control in patients whose seizures remain inadequately controlled. Across these announcements, the company’s messaging is consistent: prioritize a setting where mTORC1 is clearly the driver, where CNS exposure matters, and where efficacy can be read out with established neurological endpoints.

From a translational perspective, Aeovian’s program addresses several limitations that have complicated mTORC1 development in broader aging trials. First, rare genetic disorders such as TSC represent a “high-mechanism-confidence” clinical context: pathway hyperactivation is present by definition, reducing uncertainty about whether an mTORC1 inhibitor is addressing the causal biology. Second, epilepsy provides clinically validated outcomes (e.g., seizure frequency and severity) that are more interpretable than composite symptom endpoints used in respiratory or general healthspan studies. Third, Aeovian’s emphasis on selective mTORC1 inhibition reflects a core hypothesis in the field: that improved selectivity may allow chronic pathway modulation with fewer dose-limiting tradeoffs, which is likely to be essential for any long-term indication. Finally, the company has explicitly positioned AV078 as a platform entry point rather than a single product, arguing that once an mTORC1-selective CNS-active inhibitor is clinically validated in a high-signal disease, the same pharmacology could be explored in other mTORC1-mediated neurological and age-associated conditions.

Gaps and Future Directions

Despite the progress, several critical gaps remain in our knowledge and approach. Addressing these will be vital for the success of mTOR-based longevity therapies and ageing trials in general:

Lack of Validated Surrogate and Biomarkers

Aging trials currently lack an accepted surrogate marker that reliably predicts clinical benefit. Unlike blood pressure for cardiovascular disease or HbA1c for diabetes, we don’t have a single metric for “biological aging.” Epigenetic “aging clocks,” proteomic or metabolomic signatures, and functional indices (like gait speed or grip strength) are all candidates but none are yet proven surrogates. Moreover, many studies failed to measure whether mTOR was actually inhibited. We assume a dose had an effect, but we don’t know if some participants were non-compliant or hyper-metabolizers. What is the best easily accessible biomarker of mTORC1 activity in humans? Muscle biopsies can show phospho-S6 reduction, but those aren’t always available. Blood immune cells are an alternative but phosphorylations are extremely labile and hard to reliably measure in humans. In our interview, Joan Mannick highlighted (and Adam Salmon from the Marmoset study confirmed) how there is a huge variability in baseline mTORC1 activation between individuals and how we don’t really have a way to stratify participants by mTORC1 activation and a way to gauge target engagement via the drug.

These limitations are increasingly recognized in the clinical development community focused on geroscience. A recent Norn Group essay argues that, because validated ageing biomarkers are not yet strong enough for pivotal decision-making, the most feasible path for first-in-class “aging trials” is to rely on reimbursable clinical outcomes rather than clocks alone. Specifically, the authors propose a multi-morbidity composite endpoint, counting time-to-first occurrence across several major age-related events, as a practical way to increase event rates and reduce sample size while staying aligned with outcomes that regulators and payers already accept. This matches the direction taken by composite endpoint designs and provides a concrete rationale for why trials like TAME (Targeting Aging with Metformin) and VITAL-H use multimorbidity endpoints instead of single biomarkers.

Endpoints and Goals

Academic studies tend to cast a wide net on endpoints, for example, measuring dozens of biomarkers and multiple functional outcomes to see where an intervention has an effect. The downside is academics sometimes lack a single strong primary outcome, making it harder to declare “success” in a traditional sense. This is a major concern for the translatability of academic findings, as the FDA cares about a drug only if it improves how a patient feels, functions or survives. Biotechs, in contrast, zero in on one clinically recognized outcome as primary endpoint. This focus is necessary for regulatory approval and investment purposes, but it can be a double-edged sword. As we saw with resTORbio, picking the wrong primary endpoint (or one that’s too ambitious) can sink a trial even if the drug has real biological activity. As mentioned, aging itself isn’t an approved indication. To convince regulators, we may need to do what TAME and VITAL-H are attempting: propose a composite endpoint that essentially captures “aging” (e.g. any of several age-related diseases/events). The FDA has shown some openness but will require strong evidence. Ethically, exposing healthy people to a drug with any risk requires clear rationale.

Optimal Dosing Regimen Unclear

There is still no consensus on how to dose rapamycin for aging. Is it better to give a moderate dose weekly? A higher dose biweekly? Short pulses (a few months on, then off)? The “goldilocks” regimen that maximizes mTORC1 inhibition while minimizing mTORC2 effects is hypothesized but not confirmed. Perhaps different regimens work for different tissues (e.g. immune system might need shorter, frequent dosing whereas metabolic tissues might prefer intermittent). We might consider moving toward a personalized dosing: monitoring biomarkers in each patient and adjusting accordingly (for example, if a patient’s phospho-S6 is still high on current dose, escalate it; if their lipids spike too much, reduce frequency).

Long-term Safety in Healthy Individuals

Long-term safety is one of the main constraints that affects the adoption of mTORC1 inhibitors. While multiple studies support short-term tolerability over weeks to months, there is still limited evidence on multi-year exposure in otherwise healthy older adults, which is the relevant scenario for any preventive aging strategy. Because mTORC1 controls core anabolic and repair programs, systemic inhibition can produce adverse events that are acceptable in high-need disease settings but less acceptable in generally healthy participants. This creates a higher evidentiary burden for trial design, monitoring, and endpoint selection, and helps explain why broad “aging trials” have been more readily proposed for safer, widely used repurposed agents, such as metformin, whereas rapamycin-class drugs have progressed primarily through disease-anchored studies and/or with conservative dosing strategies. Classic high-dose immunosuppression is associated with higher rates of infection and metabolic disturbance, and although geroscience trials use lower or intermittent dosing, rare or cumulative harms may only become detectable over years. In addition, some potential liabilities may not present as acute adverse events but as slowly accumulating deficits in tissue maintenance; for example, rodent studies have reported cataracts and degeneration of thymic or gonadal tissues under lifelong rapamycin exposure in some strains, raising the possibility that long-duration monitoring in humans will be necessary even when short-term studies appear benign. Early observational comparisons of off-label rapamycin users versus non-users have been broadly reassuring over roughly two years, with few differences in self-reported health aside from increased mouth ulcers, but these datasets remain limited by self-selection and reporting bias. For these reasons, preventive trial designs should either anchor efficacy to clearly meaningful clinical outcomes or focus on high-risk subgroups where the benefit-to-risk ratio is more favorable, while building in long-term safety follow-up data.

Conclusion

mTOR inhibition remains one of the most promising and extensively validated strategies in geroscience for slowing biological aging and preventing age-related disease. More than 14 years have passed since rapamycin first extended lifespan in mice, and in that time the field has moved from largely preclinical promise to a growing portfolio of human trials. These studies have shown that mTOR biology can be modulated safely enough to test in humans, and that intervention can produce measurable effects across multiple aging-relevant domains.

At the same time, the evidence reviewed here makes clear that biological plausibility is not enough. The outcomes of mTORC1-targeting trials depend heavily on indication selection, dose and schedule, endpoint design, target-engagement measurement, and operational execution. In other words, the central challenge is no longer only whether mTOR is relevant to aging, but how to test that biology in humans in a way that yields interpretable and clinically meaningful evidence.

The field has, nevertheless, made real progress. Past studies have been essential for establishing feasibility, exploring biomarker panels, and identifying potential signal areas, while ongoing trials are increasingly adopting disease-anchored indications and clearer outcome measures. Important uncertainties remain, especially around long-term safety in healthy individuals, validated surrogate biomarkers of aging, and the optimal strategy for selective mTORC1 modulation.

Taken together, the current evidence supports the conclusion that mTORC1 inhibition is not yet a proven clinical longevity intervention, but it remains one of the most credible entry points for building one. The next phase will depend on rigorous trial design, long-term safety characterization, and reproducible effects across populations and endpoints. If those conditions are met, mTORC1-targeting drugs could become the first inarguable longevity therapies.

This piece was produced as a Talent Bridge Award Project.

References:

1. Roark KM & Iffland PH II (2025). Rapamycin for longevity: the pros, the cons, and future perspectives. Frontiers in Aging, 6: 1628187. DOI: 10.3389/fragi.2025.1628187.

2. Konopka AR & Lamming DW (2023). Blazing a trail for the clinical use of rapamycin as a geroprotecTOR. GeroScience, 45(5): 2769–2783. DOI: 10.1007/s11357-023-00935-x.

3. Lee DJW, Hodzic Kuerec A & Maier AB (2024). Targeting ageing with rapamycin and its derivatives in humans: a systematic review. The Lancet Healthy Longevity, 5(2): e152–e162. DOI: 10.1016/S2666-7568(23)00258-1.

4. Harrison DE, Strong R, Sharp ZD, Nelson JF, Astle CM, Flurkey K, Nadon NL, Wilkinson JE, Frenkel K, Carter CS, Pahor M, Javors MA, Fernandez E, Miller RA, et al. (2009). Rapamycin fed late in life extends lifespan in genetically heterogeneous mice. Nature, 460: 392–395. DOI: 10.1038/nature08221.

5. Mannick JB, Del Giudice G, Lattanzi M, Valiante NM, Praestgaard J, Huang B, Lonetto MA, Maecker HT, Kovarik J, Carson S, Glass DJ, Klickstein LB (2014). mTOR inhibition improves immune function in the elderly. Science Translational Medicine, 6(268): 268ra179. DOI: 10.1126/scitranslmed.3009892.

6. Mannick JB, Morris M, Hockey H-UP, Roma G, Beibel M, Kulmatycki K, Watkins M, Shavlakadze T, Zhou W, Quinn D, Glass DJ, Klickstein LB (2018). TORC1 inhibition enhances immune function and reduces infections in the elderly. Science Translational Medicine, 10(449): eaaq1564. DOI: 10.1126/scitranslmed.aaq1564.

7. Kraig E, Linehan LA, Liang H, Romo TQ, Liu Q, Wu Y, Benavides AD, Curiel TJ, Javors MA, Musi N, Chiodo L, Koek W, Gelfond JAL, Kellogg DL Jr (2018). A randomized control trial to establish the feasibility and safety of rapamycin treatment in an older human cohort: Immunological, physical performance, and cognitive effects. Experimental Gerontology, 105: 53–69. DOI: 10.1016/j.exger.2017.12.026.

8. Chung CL, Lawrence I, Hoffman M, Elgindi D, Nadhan K, Potnis M, Jin A, Sershon C, Binnebose R, Lorenzini A, Sell C (2019). Topical rapamycin reduces markers of senescence and aging in human skin: an exploratory, prospective, randomized trial. GeroScience, 41(6): 861–869. DOI: 10.1007/s11357-019-00113-y.

9. Gonzales MM, Garbarino VR, Kautz TF, Song X, Lopez-Cruzan M, Linehan L, Van Skike CE, De Erausquin GA, Galvan V, Orr ME, Musi N, He Y, Bateman RJ, Wang C-P, Seshadri S, Kraig E, Kellogg D Jr (2025). Rapamycin treatment for Alzheimer’s disease and related dementias: a pilot phase 1 clinical trial. Communications Medicine, 5(1): 189. DOI: 10.1038/s43856-025-00904-9.

10. Moel M, Harinath G, Lee V, Nyquist A, Morgan SL, Isman A, Zalzala S (2025). Influence of rapamycin on safety and healthspan metrics after one year: PEARL trial results. Aging (Albany NY), 17(4): 908–936. DOI: 10.18632/aging.206235.

11. Moody AJ, Wu Y, Romo TQ, Koek W, Li J, Linehan LA, Clarke GD, Yang EY, Espinoza SE, Musi N, Chilton RJ, Kraig E, Kellogg DL Jr (2025). Short-term mTOR inhibition by rapamycin improves cardiac and endothelial function in older men: a proof-of concept pilot study. GeroScience. 2025 Sep 19 (Online ahead of print). DOI: 10.1007/s11357-025-01855-8.

12. Svensson JE, Dörfel RP, Schain M, Bolin M, Sacuiu S, Hagman G, Forsberg Morén A, Kivipelto M, Plavén-Sigray P (2025). Evaluation of rapamycin as a neuroprotective treatment in Alzheimer’s disease: a six-month single-arm open-label clinical pilot trial. medRxiv (Preprint). DOI: 10.64898/2025.12.08.25340853.

13. Mannick JB, Teo G, Bernardo P, Quinn D, Russell K, Klickstein L, Marshall W, Shergill S (2021). Targeting the biology of ageing with mTOR inhibitors to improve immune function in older adults: phase 2b and phase 3 randomised trials. The Lancet Healthy Longevity, 2(5): e250–e262. DOI: 10.1016/S2666-7568(21)00062-3.

Acknowledgements:

Thanks to Soleil Wizman (Norn Nexus member) for her help shaping the resTORbio section and the principal investigators who generously dedicated time for exclusive interviews and clarifications that informed this review: Jonathan An, Yousin Suh, Dudley Lamming, Brad Stanfield, Joan Mannick, Adam Salmon, and Matt Kaeberlein.

Subscribe now

Share this piece with your network:

Share

Plenty of other options if you want extensive coverage of biotech deals. If you want to chuckle your way through a masterful Lovecraftian exposition of JPM, look here.

Healthcare has to change

Many discussion threads felt like loose ends of a fraying fabric: US healthcare can’t continue in its current form. With luck, this will help enable longevity medicines.

People are frustrated by costs and a lack of agency in their own (proactive) health. Biotech companies and investors too are frustrated by the cost of bringing drugs to people, driven in part by the cost of US-based clinical trials. Every part of healthcare is getting more expensive, faster than the medicines we usually complain about. While HHS enthusiastically champions FDA modernization to address these issues, consumers are increasingly exploring ways to sidestep the agency entirely.

Healthcare won’t change overnight, but some things will shift in the next few years.

Subscribe now

The DIY Health Zeitgeist

• Everyone wants GLP-1s. After a brief gold rush for aggressive telemedicine companies, Novo and Lilly obliged by launching direct-to-consumer sales. Once that door was opened and subcutaneous injections become normal, people start asking what else is available. This triggered a big ‘Chinese peptide’ craze, including various experimental drugs still in clinical trials. The ‘peptide’ brand is so strong that some pill mills are labeling non-peptide drugs as peptides…

• The FDA should recognize the latent demand for agency on our health, and adapt accordingly. As it is, the peptide craze has ushered in a wave of longevitainment, normalizing circumvention of a ‘too reactive’ healthcare system in favor of a pre-FDA 'apothecary' model informed by friends and tweets. Much like the apothecaries of old, most consumers won’t benefit, and some will be harmed. For complex diseases, the FDA plays a critical role in enforcing scientific truth-seeking. If medicine ends up like supplements, either from too aggressive deregulation or from lack of adaption leading to consumer actions, we’ll stop getting medicines that actually work because sales costs less than science.

N=1, or ‘DIY with a vengeance’

• Some who want greater medical agency have urgent, life-or-death, reasons1. Some cancers and genetic diseases are so rare that treatments are effectively personalized to a single individual, which our healthcare system isn’t set up for. and organized a great event to discuss what would help expand options for patients who need such therapies.

• Notably, the FDA announced a new ‘Plausible Mechanism’ pathway (which you’d think they would have published open access…) last year. Former commissioner Janet Woodcock published a response bringing in ideas from how we treat surgery, which notably does not need clinical trials before doctors can try new ideas. It’s interesting to think about how doctors could be enabled to try new treatments and combinations in order to support preventative medicine broadly.

• Stepping back, the N=1 model could be a powerful way to accelerate medicines when the readouts are clear, like weight loss, cancer, or viral load. But for prevention and any disease without black-and-white outcomes, scattered N=1 trials won’t give us any clear answers. The solution is to establish thoroughly validated surrogate endpoints, like cholesterol, to give us clean data across small trials.

The Geopolitics of Clinical Trial Abundance

• and coined the excellent term ‘Clinical Trial Abundance’, and hosted an event on this topic. This is largely in response to the rising trend of US biotech running trials overseas where costs can be halved and investigator-led trials can be started months than in the US. Or, of course, rely on China to run faster trials and buy assets with human data. You can imagine that this is not a desirable state of affairs for the FDA/US government.

• Longevity companies are hit harder than most, even if they don’t run actual aging trials, given the costs and durations for age-related diseases. The Alliance for Longevity Initiatives (A4LI) hosted an event where Jim' O’Neill and longevity biotech leaders discussed options for cheaper and faster early human data in the US. A common concept here is a ‘barbell strategy’, where we make sure early trials are robust on safety, then grant accelerated approval based on reasonable evidence, followed by rigorous enforcement of confirmatory evidence. This would move longevity trials out of ‘untenable’ territory by enabling commercial returns within the drug patent life. Notably this mechanism already exists in veterinary medicine (which Loyal is taking advantage of).

These are aspects of the same broader trend: As healthcare costs rise it becomes more expensive to prove that drugs work in humans which kneecaps new drugs and preventative medicine, the rare parts of the healthcare system with the potential to improve care per dollar. This trend has been ongoing for a long time, and consumers increasingly want to take health into their own hands. Something has to shift.

Solutions & progress

ARPA-H National Longevity Strategy Roundtable

• organized a roundtable, where a group of academic and biotech leaders came together with ARPA-H leadership to discuss bottlenecks for longevity, and where ARPA-style projects could push us past the ‘valley of death’ between academic work and market-driven progress.

• Key takeaway is that ARPA-H is keen to do what it was made to do: Take bold bets that could fundamentally change how we keep Americans healthy, and is looking seriously at longevity science for ways to do that. This is great! I’m hopeful for followup to produce impactful new ARPA-H programs.

• There was also discussion of how to incentivize medicines that prevent age-related diseases. This kind of medicine simply isn’t financially viable without some kind of accelerated approval pathway, because the time to make the medicine and show effects in humans would eat up almost all of the 20 year patent life. Surrogate endpoints is one way out of this conundrum.

• One of the most important thing ARPA-H can do for longevity is to create a ‘CASP for measuring aging’ to address the surrogate question, as I wrote about earlier this month:

Pharma keeps tip-toeing in

• This year’s JPM had at least two sessions organized by investment bankers, and I kept seeing longevity companies and investors in the floor Eli Lilly had rented. Other pharmas also made comments about curiosity around longevity (typically alongside questions about endpoints and accelerated approval).

• One topic discussed both at the roundtable and elsewhere is how best to make use of the trove of ‘non-commercial’ drugs that the biopharma industry has created. Drugs whose patent life has expired, or that simply aren’t doing well commercially. This would be a cool way for pharma to dip their toes into longevity trials, without adding risk to the blockbuster drugs that keep them in business.

• Clearly defined trial requirements, with accelerated approval, is likely needed to unlock more companies investing in longevity. There are many ways the HHS could act on this, ranging from new regulatory mechanisms to simply issuing crystal clear guidance documents for diseases of aging.

Subscribe now

1

Check out ‘s excellent story on Sid:

The Century of Biology

Going Founder Mode On Cancer

Welcome to The Century of Biology! This newsletter explores data, companies, and ideas from the frontier of biology. You can subscribe for free to have the next post delivered to your inbox…

Read more

8 months ago · 166 likes · 23 comments · Elliot Hershberg

It’s hard to get people excited about a problem they think is solved. If I pull them into a nuanced discussion of biomarker requirements (that they’re rarely in the mood for), I can sometimes get to “Oh ok makes sense, but sounds kind of boring”. Progress?

is always happy to do boring work that’s required for impact. But we also want to articulate why that work matters, and I’m hoping this post can do that. By reading it I want you to realize that yes biomarkers are important, that yes biomarkers exist, and that there’s work to be done before biomarkers are actually useful.

A large part of the confusion is that the term ‘aging clock’ is currently being applied to a wide range of different tools. Measuring tools with different purposes by the same metrics doesn’t help us improve. So we’ll go through these uses, assess how we’re doing for each, and then call attention to certain actions we should take today if we want a real path to using clocks to find therapies to treat aging in humans. We’ll see that, counterintuitively, some of those require focusing less on human clocks.

(Note that I’m following convention and using ‘clock’ instead of the more accurate ‘biological age test’ throughout this essay, in part to emphasize that what we call clocks are research-grade tools that we can’t yet rely on to test biological age1).

Why do we need clocks?

I’ve previously emphasized iteration speed as a driver of progress, and one that biotech struggles with. Software companies can write and test new products daily and distribute them at the speed of light. Biotech can only iterate at the speed of life. And the goals of the longevity field are particularly slow: Preventative medicine means an inevitable delay between treatment and confirmation, and successfully extending life means that answers arrive even later. The ability to predict the future using measurements today would change these dynamics completely.

Even so, the impact of good biomarkers is probably underestimated: Not only would we cut a year or two off every mouse study on longevity around the world, increasing the pace of research by up to 25%2. Not only would this also reduce the cost of research by another 25% or more, allowing more to happen in parallel. Not only would an early measure convert aging clinical trials from untenable to comparable to what we do for individual diseases, which may be required before the first trials for aging get launched. Shaving years off trials also preserves patent life and reduces the cost of capital, increasing investment in new medicines.

And on top of this, decreasing costs and delays increases momentum for the field: There are many unanswered questions in longevity, like whether a single treatment can reduce biological age across the body or whether we need combinations of treatments, how to design trials with the best signal per cost, and how to dose and deliver different treatments. We won’t have answers to most of these until we try potential solutions and find out what we’re missing. Faster and cheaper iterations means we can put down more puzzle pieces faster, as things get filled in it becomes easier to solve the next piece. So validated biomarkers/clocks will increase not only the rate of progress in longevity, but also the rate of acceleration.

What we talk about when we talk about clocks

The idea of a single number distilling the whole aging process is delectably simplifying, and that apparent simplicity part of why the ‘clocks’ concept has gotten so much attention3. We might be able to do that, but at the moment our clocks actually measure a wide range of things. And you can imagine that different uses with different design choices all have their place: Is surgery the best way to diagnose cancer? It’s the most accurate, but also quite invasive…

So, we currently have many ‘aging clocks’ that don’t do the same thing. One way to distinguish clocks is what they’re physically measuring (proteins, DNA methylation, …), but this only matters for determining the constraints of trying to make the clock do what you want. What you want to do is what drives decision-making and sets the goalposts. So let’s look at the intended uses for aging clocks:

• R&D: You want to do experiments to understand biology and/or find new candidate therapies, and avoid waiting for aging and death as your readouts.

• Consumer health optimization: You want to monitor and probability optimize your health. You’ll do measurements at regular intervals, and change your behavior by whether the clock goes up or down.

• Design and interpret clinical trials: You want to select who goes into your trial, or identify people who respond better or worse to some treatment, purely for your own learning.

• FDA approval: You want to run a clinical trials and get Accelerated Approval from the FDA based on lowering clock scores, ahead of showing improvements in mortality or disease.

• Medical care: You want to run tests know whether prescribed medicines and behaviors are working. Wrong results are a big deal, as they could lead to wrong treatments (and, in the US, lawsuits galore).

It’s telling that today, aging clocks are frequently used for the first two purposes but ~never for the last three (where the cost of being wrong is high). Evidently, people in charge of these costly decisions do not think that aging clocks are ready to use.

Why is that? The original 2012 clocks were strictly correlates to chronological age, which is not a very useful thing to estimate. But later clocks have been trained to predict mortality, frailty, and other important outcomes. So the answer must be that we can’t yet trust their predictions.

Where are the clocks failing us? Let’s look at the most important metrics for clocks:

• Accuracy: Does the clock accurately measure the true biological state with low variance? If two simultaneous measurements can vary by a decade, the clock has limited utility.

• Responsiveness (Precision/Recall): Does the clock only record the past, or does it update if a treatment or intervention modifies what we want it to measure? And for the latter, does it sometimes respond when it shouldn’t?

• Interpretability: Do we know the physical/biological reality that the clock is measuring, or is it a black box? If not, it’ll be hard to convert answers to actions.

• Stability: Does the clock change depending on time of day, or whether you have the flu? If it mostly tells you about your current state, you’ll need multiple measurements to understand trajectories.

• Applicability: Does the clock work across sexes, genetics, or even different species? If not, or if unknown, your answer may be meaningless in some contexts.

• Cost: Maybe everything works, but it’s too expensive to be worth it.

Just optimize everything! But in practice there are tradeoff between improving each metric, especially while also keeping costs down. If we want clocks to help us, we have to think about which parameters are important for our purpose.

Accuracy is straightforward, and of course restricts use cases where precision is needed. One important consideration is that ~every clock is created using data averaged across many individuals, so every result is a probability of living longer, not an exact death year.

Norn Group’s mission depends on new therapies with strong impact on biological aging, for which the Responsiveness metric is important. Good Responsiveness means avoiding false positives and negatives. A biological measure of a subset of what happens with age could predict well under normal conditions, but could fail to respond to an intervention that affects a different aspect of aging (and yields benefits). Conversely, an intervention could act directly on the biological measure without affecting aging itself, e.g. a clock tracking DNA methylation of a specific part of your genome might go to zero if you remove that DNA with CRISPR, but that wouldn’t mean you’re actually younger.

To properly assess Responsiveness, behavior must be characterized across a wide range of interventions; responding correctly to one or two doesn’t tell us how often false positives/negatives occur. For example, consider a frequently used clinical test, simply measures how far patients can walk in six minutes. This is used to assess e.g. heart function, because low oxygenation reduces your capacity. But although it does correctly respond to improved heart function, Responsiveness is moderate at best: giving patients caffeine or other stimulants will increase your walk distance without helping your heart.

High stability means measuring states, low means measuring rates. Rates are more actionable, but more prone to confounding factors. High Stability is important if we want to get robust results from infrequent measurements, e.g. in an annual checkup. But high Stability also means slower responses to intentional interventions. So if you want to know whether a lifestyle change or therapy changed your current aging rate, lower Stability can be better as long as Responsiveness and Accuracy are high.

FDA surrogate biomarkers must be mechanistically plausible, so low Interpretability is a no-go for this purpose. (Notably, ~every clock based on DNA methylation has low Interpretability). Interpretability is also important for R&D: you could use a black box measure to screen treatments, but you couldn’t use the outcomes to guide what interventions to test or what experiments to use to confirm results. Clinical use will also require showing Applicability spanning intended use cases4.

A specific type of Applicability is cross-species, i.e. does your clock work both in (human) patients and in animal models used for early drug discovery (e.g. mice). If not, it’s hard to do controlled experiments to test how your clock behaves or to transfer lessons from early research to e.g. clinical trial design. But cross-species Applicability puts additional constraints on what you can measure.

Going back to our different use cases, we can see that different metrics matter more or less for each one5:

Subscribe now

Are we there yet?

As we suspected from current real-world use, we are not there yet:

Accuracy is OK, but known to vary with time of day and between tests. There are questions as to whether it’s better than simpler measures. Price (hundreds to a thousand dollars) is mostly OK although a magnitude lower would be helpful.

Responsiveness across many interventions remains untested for ~all clocks, because the data to do so properly hasn’t been generated. We don’t have human-validated aging therapies (as in, shown to extend healthy human life), so every human-only clock lacks the ground truth data needed to broadly test Responsiveness. Without this, pharma companies can't make clinical trial decisions and doctors can't use clocks for treatment guidance. For observational studies this isn’t a huge issue, and indeed we see clocks used in observational clinical work and for consumer longevitainment.

Interpretability is near zero for the majority of clocks, and low for the rest6. They might be trained on an interpretable endpoint, like a frailty index or time of death, but what’s being measured is a molecular change (like methylation of a specific part of your chromosome) that doesn’t have any known link to the endpoint. This again makes the clocks less useful for clinical trials and drug development - if you see a clock go down in certain patients but don’t know why, how does that change your actions? Some newer ideas like clocks for specific organs add some interpretability, and I expect the field to shift towards tests for specific biological (dys)function7.

How did we end up with consistent gaps in these two metrics? I believe the main cause is a decision to focus on human clocks. Clocks that work well in humans are obviously most attractive (and splashy), so the decision makes some sense. But human-only makes it much harder to generate new data to validate predictions or fill gaps for clock development.

Very few research labs have funding to run human studies, even observational ones. So we end up making more and more clocks using the available data. Some add nuanced information, like organ-specific aging, some correct for technical noise in a new way. This is useful work, but does not address the key bottlenecks (unknown Responsiveness and Interpretability) required to mature clocks from consumer and academic use into tools for finding therapies and running clinical trials.

Time to move forward

Bold research has given us a proof of concept that the aging process can be tracked with molecular measurements. We should appreciate that. And we should acknowledge that we won’t benefit much from talking about clocks’ potential for practical uses until we’re able to properly benchmark the required performance metrics. So let’s decide on the uses we think are most valuable, and make sure we build the infrastructure to track where we are now and whether we’re improving.

To eventually use some kind of biological age test as a clinical trial endpoint, we have to shift our focus to interpretable measures with mechanistic links to aging and disease. One example of this would be to build predictors based on proteins in blood samples, rather than DNA methylation patterns. Proteins have known functions, so even if you first build your predictor on pure regression you can tease apart the role and value of each component and arrive at something mechanistically interpretable. Building clocks without a path to mechanistic interpretability can still be useful for R&D (and entertainment) purposes, so I’m not saying we should avoid this entirely. But human clinical use is important enough that major efforts ought to be compatible with that goal.

To use clocks for finding powerful therapies for aging, we have to robustly benchmark Responsiveness. There are two ways to do this: We can go directly to human-trained clocks, as is the current trend, which allows optimization towards the final clocks right away. But the optimization will probably take multiple decades: First we need to identify more than one working aging therapy in robust trials, without the benefit of guidance from clocks (since they haven’t yet been validated). It’s unlikely that this happens in less than ten years. Only then do we see whether the current clocks are good enough, and if not start making new ones. Once something fits existing data well, the FDA will likely still ask for a fresh trial for confirmation.

Alternatively, we develop cross-species clocks and take advantage of known ways to modulate aging in mice to test Responsiveness. If our current clocks are inadequate, more work using mouse models can iterate quickly to reach something promising. At that point we can use the clocks in human trials for putative aging therapeutics, while also using them to speed up the search for more therapies. FDA endpoint use won’t be feasible until at least a couple of trials have happened. It will still take years before we’re fully confident in clocks for human use, but not decades. I believe this is the better approach.

Beyond having ground truth data, the best way to benchmark is third-party hosted (for impartiality), standardized (for interoperability), and publicly accessible, similar to ImageNet (which enabled computer vision) or CASP (which enabled AlphaFold etc).

Funding such a clock assessment program is the highest leverage in longevity. Between accelerating R&D and eventually enabling human trials, good clocks are very much worth pursuing. The future we want is public benchmarking of performance for the most valuable uses, norms that all serious work submits to this benchmark, and thus iterative improvement towards better and better predictions of future health.

Appendix 1: Existing efforts & resources

The Biomarkers of Aging Consortium is the most active organization on this topic. The group contains many of the academics working on aging clocks and some biotechs, but notably not people from biopharma developing biomarkers. The consortium is focused almost entirely on human aging clocks. They have published a great set of papers for going deeper on the topics discussed here:

• Biomarkers of Aging for Identification and Evaluation of Longevity Interventions

• Validation of biomarkers of aging

• Challenges and Recommendations for the Translation of Biomarkers of Aging

They also launched a competition for benchmarking, using existing data.

TranslAge & Biolearn have benchmarked current clocks on a set of useful parameters.

The ARPA-H PROSPR program to find aging therapeutics includes a biomarker component, with the intent of using new trials to start validating them. Relying on new therapeutics working makes this opportunity to refine biomarkers rather uncertain.

Their approach involves the ‘Intrinsic Capacity’ concept, which could be an effective way to quantify aging as a medical outcome and is now part of WHO’s ICD-11 codes.

The TAME trial for metformin as an aging therapy similarly has a biomarker component, but again relies on uncertain therapeutic effects to get meaningful signal. The trial has also been seeking funding for a number of years, so far without success.

Mining samples from past, completed, clinical trials is perhaps more likely to succeed, though these trials are in diseased patient populations and did not focus on lifespan.

A number of longitudinal human datasets exist that cover a lot of what would be needed for biomarker developed, but with gaps especially for FDA-recognized surrogate endpoints. We’ve outlined those here.

The best ground-truth data on lifespan effects is the NIA Intervention Testing Program. Unfortunately this program did not collect blood samples to allow, but the results can be used to inform what data to collect for a clock assessment program.

Subscribe now

The above is AI slop (from a real LinkedIn post). You probably noticed writing patterns: the contrasting antithesis, the staccato rhythm, the redemption reveal. But the crime of slop is posting content with zero thought behind it, not using tools to make text or images.

Certainly no thoughtful curator was involved in recommending you avoid drinking water for seven days1. The Shloppoth does not seek to understand what it posts, merely to shape words into the most view-optimized form by hijacking our curiosity and tendency to trust ‘explainers’. It is the triumph of volume over value, content with a polished appearance concealing an gaping void. The output of an inhuman abomination ingesting news and press releases and turning them into engagement.

People have written about the dead internet and brain-rot perils of slop content (and sometimes in pseudo-defense of it). But we are here to warn you about another peril. For the Shloppoth and its marketer acolytes are not content to post about work-life balance and investment advice. They are ever seeking fertile topics to devour, and this year have come upon the sapling forest known as “longevity”.

That’s right, gentle reader. Just as our field has begun to generate solid data that aging and its mechanisms can be targeted to prevent disease and extend lifespan, the channels for learning about its progress are getting jammed by low-quality content: evergreen fitness advice becomes ‘optimizing longevity’, spas add a few (untested) supplements to become ‘longevity clinics’, and Shloppoths furiously expand already-exaggerated news headlines into bullet lists of ‘proven longevity benefits’.

The worst of this is truly “longevislop”: high-volume, ungrounded, and unhelpful. It tries to adopt the aesthetics of science without even a modicum of its rigor and truth-seeking. It speaks with absolute confidence, promising “reversals” and “resets”. Combined with the soothing words of wellness influencers and the ads of supplement vendors, longevity is becoming more of a vibe than a biomedical discipline.

Some serious longevity scientists are brandishing the term ‘geroscience’ to distinguish the productive meaning. It’s a perfectly suitable term, if a bit passive. But it’s also clunky enough that we doubt anyone not wearing a suit or a lab coat will adopt it.

Instead, offers the discerning reader a guide to dissecting content branded as ‘longevity’, to understand the nutritional value of different varieties.

Subscribe now

Axis 1: Is it even longevity?

When you come across “longevity news”, first ask yourself whether this information already exists under a different name? If your grandmother got the same advice from her doctor decades ago, it’s probably not longevity science. Sure, anything that affects your health technically affects longevity, but we don’t need a new term for that. Longevity science as a discipline worth paying attention to delivers new insights and ways to increase longevity.

This gives us the first axis for separating out content: Is it even longevity?

On the right end of the spectrum, you have scientific progress that wouldn’t exist without the longevity field: new science, new therapies, and new understanding of aging itself. You still need to determine whether it’s good science (we’ll get to that), understand whether it’s early stages or ready for human trials, and more. But it’s at least in the class of content that could lead to many more years of health.

In the middle, you have things like GLP-1s or drugs focused on individual diseases. These represent real scientific progress and can have big impacts on diseases of aging, but they’re not uniquely “longevity”. They would have emerged from the fields of metabolism or neurology even if nobody was thinking about aging biology.

And at the left end you have rebranded wellness. Sleep well, exercise, don’t smoke, eat vegetables. All true, all useful, but nothing new. People knew it a year ago, ten years ago, a hundred years ago. It exists independently of any longevity research. Follow it perfectly and you’re optimizing for a healthy 78, not a possible 120. There’s nothing wrong with that, but it’s not what the longevity field is actually trying to do.

Axis 2: Who’s your source?

The second question to ask is whether the source is both capable of and interested in finding and communicating the truth. Noise on the topic of longevity is still noise.

On one end, you have informed sources who think deeply about their topic. These are people who have done the work to digest news and progress, who understand context and relevant events from the past, and who are clear about what do and don’t know. They are genuinely truth-seeking, and bringing you along for the ride. If you asked them a harder question, they could answer it thoughtfully. Following people like this is an amazing way to learn, because they compress hours and years of learning into useful updates on the world. Let’s give them an award:

In the middle, you get people who are excited about what’s happening. They’re likely to take new findings at face value, lacking the time and/or expertise to evaluate. They sometimes like to fit data to commonly held beliefs, and other times are wildly overoptimistic. Compared to the award-winning burger above, consider them a random restaurant with a ‘World Famous’ sign in the window: You will most likely not get what you’re being promised, but it’ll be a burger. Maybe it’ll be awful, maybe it’ll be delicious. No guarantees can be expected from this kind of source.

But there’s a type of source that’s worse than the uninformed/unrealistic: The Slop. The over-hyped burger is at least a burger. Someone tried to give you information, maybe they didn’t do the best vetting but at least you can digest it yourself. Slop has no information: it is headlines run through LLMs, with no thought or craft behind it. It cycles through topics (peptides, fasting, telomeres, epigenetics…) with no context or memory, without relating observations to each other. In the world of would-be burgers, slop is raw ground beef:

Don’t eat it. It could be actively dangerous, and it certainly won’t make you better off.

The Longevity Slop Map

When we put the two axes together, we get a map:

In the bottom-right (legit but not longevity) is what we might call longevitainment. For those who genuinely want longevity, this should be considered entertainment with a longevity aesthetic: activities, purchases, and engagement that makes you feel like you’ll live longer, but that won’t actually make a big difference to your lifespan. Think cold plunges, saunas, red light, and sleep scores. Honest claims of minor health effects, possibly with nice brand. You’ll feel good, stay fit, probably be healthier and at least won’t get hurt. That’s great, as long as you know what you’re getting.

Moving left we get to a wellness hype zone full of sales pitches. ‘Forget going for a run, this new type of exercise will give you unheard of results’ or ‘This faraway group of humans have health secrets you’ve never heard of’. If you believe these claims you’re likely to be disappointed, although you might have a good time. Unless, as typically happens, you were asked to put down a credit card before partaking.

If we move up to uncritical longevity, we get a mix of people who are sure that indefinite lifespans are around the corner, or that the FDA is the only thing standing in your way. Clickbait journalism lives here, giving a false impression of progress. The innocuous enthusiast version is perhaps not very constructive, but at least sincere and well-intentioned. The edgier version wants you to fly to Honduras to inject gene therapies - most likely will do more harm than good given how few medicines made by professionals end up working, but not impossible that something will work. All we’ll say here is that this quadrant has been around for a long time, and we’ve yet to see a breakthrough come from there.

Left of these two hype zones is the terrifying realm of longevislop. This is almost certain to be a grift, either for your money or your attention. Just say no.

Norn Group’s recommendation is a diet primarily consisting of the top right quadrant. Listen to truth-seekers who put lots of effort into understanding the world, explain the science of why we age and probably don’t have to. Celebrate when they get excited about a breakthrough, and pay attention to what they spend time talking about.

Subscribe now

A Field Guide

For readers interested in understanding real progress in longevity (for example towards Norn Group’s mission of enabling a sixty year-old to liver another sixty healthy years) we suggest the following checklist for your precious attention:

First, ask whether the same content exists without the longevity field. If so, is there a unique angle/update? If you’re just being told “sleep and exercise”, you’re on the lower part of the quadrant and should adjust your expectations accordingly.

Second, ask whether this person has done the work to understand what they’re saying. Polish people can pay for, depth not so much. A history of engaging in discussions and tackling hard questions with humility is a good sign. A background in aging science (rather than say marketing) ditto. If big news came from an iffy source, look for a more credible source treating the same topic.

Third, check for financial conflicts of interest. Anyone who is trying to sell you something is automatically less credible. Not necessarily wrong, but look for neutral third-party evaluation. Also ask whether it would be oh so lovely to believe something is true. We all want medical breakthroughs, so pitching them is easy. For anecdotes like “I reversed my age by 12 years”, it’s best to ask ‘how could this claim be wrong’.

Finally, when something is exciting and seems to come from a good source, ask yourself ‘what would normally happen next’, and check back on that later. Drugs with great effects in animals tend to move to human trials, research breakthroughs are picked up by other labs.

The longevity field is growing, but a lot of apparent growth is noise. Follow the data.

Share

Nuclear Marketing

Nucleus Genomic launched a massive consumer marketing campaign, which triggered/resurfaced multiple critiques of their science and integrity [1, 2, 3]. The CEO attempted a rebuttal, not very successfully.

Two points are relevant for the longevity field:

1. The stakes for the discussion are high because the commercial success drivers for polygenic risk embryo cos are brand and sales, with science as a nice-to-have support. This is because most outcomes of what they are selling are not immediately verifiable. “Decreased risk of Alzheimer’s” won’t be relevant for decades, and you’d need to wait years after a kid is born to even run an IQ test. And in both cases, there’s no control group/counterfactual. So even if the products did literally nothing, the companies wouldn’t get market feedback for a long time (but of course would be letting their customers and humanity down).

2. The CEO is actively pitching the product as longevity.

These two points are related:

The need to predict future death and disease is the biggest challenge for longevity (i.e., the field of actually giving people more healthy years of life). If we had a validated measurement that tells you how many healthy years you have left, companies would be rewarded for developing treatments that improve the measure. But we only have the embryo of such a test (in the form of various aging clocks and frailty indices), yet to mature into something we can trust. This means uncertainty, which opens the door to emotional and narrative-based marketing.

As the idea of longevity becomes more broadly accepted, people skilled at marketing vibes are seeing a lot of opportunity. Some of this is ‘longevitainment’, selling a desired identity to people with cash to spare. Some is grift, selling unsubstantiated claims as ‘science’. Some might be the nucleus of real consumer longevity products, by gathering the data needed to validate predictors of future health (fear not, we will help you tell the difference).

Biomarkers - 1

Robust ways to measure future health is the most important missing tool for the longevity field. Several relevant papers came out in the last few weeks:

Elastin-derived extracellular matrix fragments drive aging through innate immune activation

TranslAGE: A Comprehensive Platform for Systematic Validation of Epigenetic Aging Biomarkers

A unified framework for systematic curation and evaluation of aging biomarkers

Organ-specific proteomic aging clocks predict disease and longevity across diverse populations

Biomarkers - 2

Martin’s TEDx talk on the need for third-party, standardized benchmarking of potential aging biomarkers went live on Youtube:

This is a preface for a program is working on.

Subscribe now

We propose using a composite endpoint that counts multiple age-related events, which can greatly increase the effective event rate and cut required sample sizes by half or more. We walk through why this design makes sense, what the endpoint options are, and provide a calculator for estimating trial requirements.

Best practices in clinical development disincentivize therapeutics that directly address aging mechanisms.

For the purposes of this essay, we define aging as:

• The sum of underlying biological processes that happens in every human gradually over long periods of time…

• That makes people more susceptible to diseases associated with high chronological age (e.g., neurodegeneration, cardiometabolic diseases, cancer).

Consequently, for the purposes of this essay, an aging drug is a drug that:

• Works against these underlying natural processes.

• Reduces the risk of multiple diseases with unrelated pathologies associated with high chronological age in virtually every human (vs just patient groups with specific risk factors like high LDL, obesity, etc).

To get a therapy approved and labelled for aging, you eventually need to run your pivotal clinical trial of that drug against ‘aging’, by some definition that relates to how patients feel, function, or survive. In other words, the primary endpoint of the treatment’s Phase 3 trial needs to measure aging, not be limited to any one specific indication, and connect to concrete patient outcomes1.

So far, nobody has run such a trial. This is in part because measuring aging is hard and new: it creates uncertainty around the exact endpoints, approval criteria, and reimbursement pathways.

Instead, the majority of biotech companies that are developing aging drugs follow the conventional path of developing/validating their therapy against a specific disease. That is, they pick an age-related disease as a proxy indication, and plan to get their drug to market against that indication. Then, once the drug is on the market, they plan to expand the drug’s labelling for other indications by running additional trials against other diseases. This is a reasonable approach: we see it working today with e.g. GLP1-R agonists.

However, this approach creates an incentive misalignment: If you aim for approval against a specific indication, you will want to optimize your drug development programme for making a drug that works best against that one indication.

Chances are that the drug that works best for that one indication won’t be the exact same drug that works best against aging as a whole. Thus you will have to pick between optimizing your drug development against your target indication or optimizing your drug development against aging. Considering that making any successful drug is already very hard, and you need your drug to succeed, most will choose to optimize for making a good drug for your initial indication rather than developing a great aging drug.

We believe that developing aging drugs has a higher expected value than traditional drug development because of the much-much-higher upside (similar to what we see for incretin drugs). But the much higher risk of failure will turn many reasonable investors and entrepreneurs away from this path; as we wrote about previously, a large success with known probability is more palatable than a larger success with unknown probability. That is, for most people, having a 5% chance for a $10B upside is more attractive than a 1% chance for a $100B upside.

Being kind-of-forced to chase local maxima is frustrating. It’s therefore worth thinking about how we could align our incentives with what we, as aging researchers and biotechnologists, would want to do in the first place. Besides frustration at the individual level, this incentive-misalignment also leaves a lot of societal and patient benefits on the table. As a society, we should be pursuing the opportunities with the highest expected value.

What endpoints would we use for a trial targeting aging?

Specifically, let’s think about what the primary endpoint of our eventual Phase 3 trial can be. This is the decision that should and will drive the design of the whole trial, and will serve as the north star for all of our earlier-stage drug discovery and development processes.

We know that our primary Phase 3 endpoint needs to be:

• A generally good proxy for aging (vs for a more specific indication), to avoid misaligned incentives.

• Reimbursable.

Reimbursable means that if our Phase 3 trials succeed and a treatment shows improvement on our primary endpoint, that also means that our treatment will be:

• Approved by the FDA (and other regulatory agencies worldwide)

• Reimbursed by payers, for a very wide group of patients, at a reasonably high price tag. Payers generally only pay for therapeutics that produce very obvious and very direct benefit to the patients on their insurance plan.

The main relevant payer types we are talking about here are:

• Private commercial insurance plans (United, Cigna, etc.)

• Medicare

• Governments outside of the US

• Patients themselves (out of pocket)

Private and government plans generally care about the same treatments and evidence. Most important for aging drugs will be Medicare and other government payers, as most of the cost savings and quality-adjusted life years an aging drug will create are going to be on the 65+ year old population.

Good aging drugs will be prescribed earlier in life as well, and provide disease prevention and quality of life improvements for patients on private plans, which is needed for private plans to be incentivized to reimburse2. (In this essay we’ll ignore aging drugs that need to be dosed early in life but only provide medical benefits at age 65+, since trial design for these is even less tenable).

Out of pocket payments may be a bit of a wildcard for aging drugs, as patients themselves may require less efficacy and health economic evidence as long as the upside and safety is established. There are trends in this direction, led by incretin weight loss drugs. We expect that the market-size for this is still small compared to a drug reimbursed by main payers for the foreseeable future, and that aging drugs will have less visible upside, so going forward we’ll assume the need to get reimbursed by government and private payers.

Table 1 shows endpoints used for age-related diseases:

We see that the overwhelming majority of Phase 3 primary endpoints are extremely simple. These endpoints are either:

1. Functional, i.e. measuring how well a patient functions, in the simplest terms possible, or

2. Morbidity or mortality related, i.e. measure the presence/absence/frequency of unwanted medical events (incl. death) or diseases.

The few Phase 3 endpoints in use currently that don’t fit either of the above two categories are very extensively validated surrogate endpoints. We see such endpoints in e.g. cardiovascular disease (LDL cholesterol, blood pressure), type 2 diabetes (HbA1c). The benefit of these endpoints is that we can expect to see change earlier than in functional or morbidity-related endpoints for the same disease.

Currently, surrogate biomarkers for aging are nowhere near the evidence level needed to be used as endpoints in pivotal trials. Long term, we should certainly aspire to develop such surrogate endpoints, and the way such surrogate endpoints should be developed warrant further thinking - but the first aging trials are not going to be against such endpoints.

This guides the options we have for a trial against aging. The primary endpoint of our aging Phase 3 trial will either have to be something very functional, or something morbidity/mortality related.

Subscribe now

Option 1: Function

Functional aging endpoints would function much like today’s frailty indices: they aggregate metrics such as cognition and physical fitness into a single score. A better score indicates a higher quality of life. Importantly, a well-designed frailty index could potentially demonstrate drug efficacy earlier than other Phase 3 endpoints.

However, current frailty indices are not approved for Phase 3 reimbursement. This is partly because combining multiple domains (cognition, mobility, biomarkers) into a single metric makes it difficult to interpret exactly what an “improvement” signifies. Furthermore, these indices have not been extensively validated to predict costly medical events, such as intensive care or multi-day hospitalizations, in patients without specific diagnoses. Within individual diseases, like Parkinson’s and Alzheimer’s, composite endpoints similar to frailty indices are used. So a path is feasible, but for aging we would need to refine and validate a better frailty index than what exists today3. Such effort will not be the focus of this essay.

We would not utilize a frailty index for a first-in-class aging trial today, because using a new frailty index will carry regulatory and reimbursement risk, which means that the uncertainty we are trying to address remains. The risk is significantly higher here than with mortality or morbidity endpoints because the individual components of a frailty index are not independently reimbursable.

Option 2.1: All-cause Mortality

The simplest trial design measures all-cause mortality as the primary endpoint. This involves comparing treatment and control groups to determine if significantly fewer patients die in the treatment arm.

While this endpoint is unquestionably reimbursable, low baseline event rate means low responsiveness. Achieving statistical significance would require extremely large sample sizes or very long follow-up periods to capture enough data points.

Option 2.2: Morbidity prevention

The next logical option is to run trials targeting age-related morbidities. These trials measure the frequency of specific medical events, which can range from acute incidents (like a stroke or myocardial infarction) to new disease diagnoses. (Conversely, trials can also track positive outcomes, such as the resolution of a condition.)

Heart failure studies, such as this finerenone trial, offer a practical example. Investigators compare treatment and control groups to determine if the therapy significantly reduces adverse events, such as cardiovascular death or heart failure complications, over the trial’s duration. (Note: these trials technically measure “time-to-first-event” to allow for more nuanced analysis, but the general principle remains the same.)

Crucially, morbidity endpoints are reimbursable and yield results much faster than all-cause mortality trials. If ten years typically elapse between a diagnosis (the medical event) and death, measuring morbidity confers statistical power a decade sooner. Given the current landscape, this is our preferred trial design.

The first aging trials should focus on multimorbidity prevention

For an aging-specific trial, we expand the morbidity design to track a composite endpoint of multiple distinct medical events, rather than a single disease type. Specifically, we measure “time-to-first-event” across a broad spectrum of age-related pathologies.

A qualifying event might include:

• Cardiovascular events (e.g., heart failure, MI, stroke, or CV death)

• New diagnoses of age-related, lethal cancers

• Neurological or psychiatric disease onset

• Renal disease (e.g., late-stage CKD) onset or progression to specific stage

Note: These examples are illustrative. Actual inclusion criteria would depend on the specific drug and require precise diagnostic definitions.

In the simplest model all events are weighted equally. We count only the first qualifying event per patient. While more complex designs could weight events based on quality-of-life impact, the binary “event vs. no event” approach is standard.

This design is highly likely to be reimbursable because its individual components are already approved endpoints. Precedents already exist in standard practice, such as the 3P-MACE (death, MI, stroke) used in cardiometabolic trials. Notably, the TAME study employs a similar multi-morbidity design, with reports suggesting the FDA considers this approach reasonable.

The problem with prevention trials

It is important to recognize that this design functions fundamentally as a prevention trial. Unlike standard therapeutic trials that aim to slow progression in patients who are already diagnosed, we are recruiting healthy individuals and measuring their conversion to a diseased state.

Structurally, this mirrors vaccine development: healthy participants are treated and monitored to see if they avoid a specific outcome. The inherent challenge with this structure is statistical efficiency. In a prevention trial, participants who would never have developed the disease naturally (even without treatment) do not contribute to the efficacy signal. If the natural event rate for a specific disease is low—for example, if only 1 in 10 participants would typically get sick—the trial requires a massive sample size to achieve the same power as a standard study. This inefficiency is a primary reason why prevention trials for single chronic diseases are the method of last resort, used when biomarkers are not available.

However, the multi-morbidity endpoint helps address this. While the probability of a healthy patient developing a single specific disease in a short timeframe is low, the probability of that same patient developing any one of five or six age-related conditions is significantly higher. By aggregating these distinct pathologies into a single composite endpoint, we dramatically increase the overall event rate. This restores statistical power, allowing us to run a prevention trial with a feasible sample size and timeline.

Aging drugs may unlock prevention trials

The fundamental barrier to prevention trials is that the natural incidence of any single chronic disease in a healthy population is quite low. As shown in the table below, even prevalent age-related conditions often have annual incidence rates of 1–2% or less.

(Note: These figures are approximate estimates for illustration. They are based on published data but from different patient demographics. Precise power calculations would require rigorous adjustment for age groups and disease interdependence, but these numbers demonstrate the conceptual challenge.)

For a traditional drug targeting a single disease these low rates mandate sample sizes so large that the trial becomes economically prohibitive except for large pharma, and even then only ventured for very large potential markets.

Enter aging drugs. Because aging is the primary risk factor for multimorbidity, a true aging drug should prevent not just one, but multiple diseases simultaneously. This transforms the trial mathematics. We no longer need to wait for a specific event (like a heart attack); we can count any qualifying age-related event.

Consider a hypothetical scenario:

• Dementia incidence: 0.02/year

• Cancer incidence: 0.02/year

• Cardiovascular incidence: 0.02/year

By targeting the underlying aging process, we aggregate these risks. Suddenly, our effective incidence rate triples from 0.02 to ~0.06/year. Because sample size is inversely proportional to incidence, this roughly divides our required patient count by three. This shift in the math is what makes a preventative aging trial economically viable.

Understanding the statistics of trial design

With an understanding of endpoint selection and overall trial design, we next look at the statistical design drivers. To do this we must review the basic statistics that drive clinical trial planning. (We will keep this high-level; for a deeper dive into biostatistics, we recommend starting here.)

The primary objective of any trial is to prove that a drug successfully modulates the primary endpoint. To demonstrate this with statistical significance, the most critical decision is determining the sample size. This represents a fundamental trade-off: you need a sample size large enough to detect the drug’s effect with confidence (statistical power), but small enough to ensure the trial remains economically feasible, as cost scales directly with the number of patients.

We calculate the required scope of the trial using the following formula (an approximation of the standard Schoenfeld formula):

Where:

• D is the total number of medical events required to occur during the trial to achieve statistical significance.

• HR (Hazard Ratio) represents the expected risk reduction. It is the ratio of the hazard rates in the treatment arm versus the control arm.

• Alpha and Beta are standard statistical constants:

  • Beta (Type II error): The probability of a false negative (missing a positive result even if the drug works). This is usually set to 10%, which corresponds to 90% Power.

  • Alpha (Type I error): The probability of a false positive (detecting an effect where none exists). This is usually set to 0.05.

  • Z: A standardized normal deviate derived from the chosen alpha and beta values.

Note that D represents the number of events required, not the number of patients you need to enroll. For our proposed multimorbidity trial, we get to the number of patients needed (n) with:

where:

• i is the natural incidence of the event of the patient population in our trial, measured by the number of events per year.

• t is the trial length in years.

It is important to understand the trade-off here. Statistical power depends primarily on accumulating a specific number of events (D). Therefore, sample size (n) and trial duration (t) are inversely related.

We can trade a larger sample size for a shorter trial, or vice versa. For instance, if the formula suggests a specific sample size for a one-year trial, extending the trial to five years would allow us to reduce that sample size by a factor of roughly five (assuming the event rate remains constant). We would likely prefer a longer trial, as aging drugs may require sustained dosing to demonstrate efficacy.

From this relationship, we can see that the two biggest drivers of our required sample size (n) are the hazard ratio, i.e. how good our drug is at preventing the events in question, and the incidence rate.

Estimating the trial size requirements of a drug’s multimorbidity prevention trial

In reality, an aging drug will not work equally well for every disease; it will likely show different effect sizes (Hazard Ratios) for preventing cancer versus dementia or heart failure. Determining the required sample size under these heterogeneous conditions is complex. We used multiple approaches to estimate composite hazard ratios and thus approximate trial size requirements.

Quick and dirty approach: Meta-analysis of events across types

Our first approach to modeling an aging trial with prevention of any one of several age-related events as the endpoint, was to treat it as isomorphic (structurally identical) to the problem of combining multiple trial arms in a large study.

Consider a traditional multi-center trial: a drug might work exceptionally well at one site and less so at another due to local heterogeneity. To find the true global effect size, statisticians calculate a value somewhere in between these extremes. We apply this same logic to an aging drug: if the therapy is highly effective against one morbidity but less effective against another, the overall “aging” effect size should theoretically lie in between.

Consequently, we computed an aggregate Hazard Ratio to represent the drug’s overall efficacy in preventing any contributing aging event. We borrowed statistical methods from meta-analysis, which is typically used to synthesize results from independent trials. You can think of this as a weighted average that accounts for the reliability and incidence of each specific disease endpoint when generating a global estimate.

However, we must acknowledge that this is a “quick and dirty” approximation with several limitations:

1. Dependencies: It ignores the biological reality that comorbidities are often dependent; developing one age-related disease usually increases the risk of developing another, which this model does not account for.

2. Fixed Risk: This model assumes a fixed risk for each morbidity over time, whereas in the real world, biological risk accelerates with age.

3. Underestimation: This approach averages the effects on “time-to-first-event,” which likely underestimates the drug’s true power. For example, if a drug is excellent at preventing cancer—and cancer is the event that typically happens first—the actual delay to the first medical event will be much longer than a simple average of all disease risks would suggest

Nevertheless, we captured this approach in an interactive calculator, for anyone to use in approximating the power of different multimorbidity prevention trial designs.

You can access the calculator on the link below (make a copy of the sheet):

Multimorbidity Trial Size Calculator

How to Use It: The most critical inputs are the Hazard Ratios. These are your estimated effect sizes for the drug’s ability to prevent specific diseases or medical events.

Our hope is that you input the estimates for your lead aging asset and interpret the results as follows:

1. Feasible Numbers: If the estimated sample size is manageable, the trial is economically viable. Proceed to trial planning.

2. Prohibitive Numbers: If the sample size is too large to be affordable, the current asset may not be potent enough. Pivot: You likely need to find or develop a more effective molecule before warranting an aging trial.

Calculator Assumptions:

• Scope: The model assumes the drug is effective only against Cancers, Cardiovascular Diseases, Neurodegenerative Diseases (Dementia), and late-stage Chronic Kidney Disease (CKD).

• Incidence: We utilize population averages for the 65–70 age demographic.

• Independence: We assume different diseases occur independently. While this does not strictly hold in clinical reality (where one age-related disease often predicts another), we maintain this assumption here for simplicity.

Worked example of using the calculator

To illustrate how we determine trial size requirements, let us examine a specific imaginary candidate: rejuvecin.

To calculate feasibility, we first need estimates of rejuvecin’s efficacy against the specific diseases we care about. Estimating clinical effect sizes of any treatment well is very hard, how to do that is maybe the content of another essay. Coming up with the best effect size estimates specifically for rejuvecin’s effect in preventing different age-related diseases is hard again, and maybe the content of another essay.

For this illustration, let us assume that rejuvecin reduces the risk of the following events in a healthy patient:

• Cancer: 40% reduction (HR = 0.6)

• Dementia: 5% reduction (HR = 0.95)

• Cardiovascular Events (Stroke, HF, MI): 15% reduction (HR = 0.85)

• Renal Disease (Stage 4/5 CKD): 10% reduction (HR = 0.9)

(Context: For comparison, widely used drugs like statins typically demonstrate an HR of 0.7–0.8 for major cardiovascular events, while SGLT2 inhibitors often show an HR of 0.8–0.9 for heart failure hospitalizations.)

Putting these numbers into our calculator yields the following estimates:

• Composite Hazard Ratio: 0.87

• Estimated Trial Size: ~9,100 participants

These results demonstrate that a multimorbidity trial for rejuvecin would require substantially fewer patients than running independent trials for each specific arm, except cancer. Ignoring the different cost per patient per indication, launching rejuvecin as a cancer drug only may be cheaper, but offer a much narrower label that, between time versus patent life and total cost of additional trials, would not likely expand to cover the additional indications.

Drugs with ‘balanced’ effects benefit most from multimorbidity trials

The rejuvecin calculation above highlights a critical nuance: the drug’s effects are asymmetric. In our model, the bulk of its benefit comes from a strong anti-cancer signal. While a multimorbidity trial is cheaper than running all independent arms, it is likely not cheaper than running the cancer arm alone.

The economics favor the multimorbidity design when the drug is “balanced”, meaning it has pleiotropic effects across multiple domains rather than a single spike in efficacy.

Consider a drug where the disease-specific Hazard Ratios are more consistent (egalitarian) than in the rejuvecin example. Our calculator indicates that this scenario offers an even stronger economic opportunity. This occurs because the power for an asymmetric drug is effectively limited by the most affected indication, whereas pooling indications for a balanced drug optimally increases the effective incidence rate.

As shown in the table below, if the composite hazard reduction remains similar but applies to more types of events, the required sample size drops and in particular is far lower even the cheapest single-indication trial.

This confirms a key insight for longevity biotech: for an aging trial to be definitively superior to a single-disease trial, the drug must typically possess work broadly across at least several distinct aging pathologies.

Quantitative approaches

To validate the “Meta-Analysis” proxy used above, we sought a more rigorous, first-principles approach. We consulted with biostatisticians on the logic, and with Zane Koch built simulations of outcomes in control and treatment groups (scaling the incidences by the individual HRs we previously estimated) with varying trial sizes, using our previous estimates of yearly morbidity incidence rates.

The composite HR here converged to around 0.80 as the trial became better powered (achieving 90% power at ~5000 participants for a 5 year trial). This is in line with our expectation that our meta-analytic average would underestimate the effect size of an aging drug, which came out to 0.87 with the same assumptions of incidence.

To cross-check our simulation results, Mica Xu Ji proposed a third method: an analytic derivation that approximates the composite Hazard Ratio (HR) without requiring a full simulation loop. This approach calculates the expected value of the hazard ratios across all conditions, weighted by their incidence.

The approximation is defined as:

Where P(c) represents the natural incidence of each condition, and λ(t | c, Z) is the Cox proportional hazards function. The ratio term inside the brackets effectively represents the specific Hazard Ratio for condition c (comparing treatment Z = 1 to control Z = 0).

When we substitute our hypothetical HR estimates into this equation, the result converges to 0.81. This is virtually identical to the 0.80 figure derived from our simulations. This convergence provides high confidence that both the granular simulation and this analytical model are accurately capturing the underlying mechanics of the trial design.

Conclusions

To summarize:

• We believe that understanding the path whereby an aging drug can be tested directly against aging (vs specific proxy-indications) will incentivize the development of true aging drugs that promise a higher upside and a higher expected value for all.

• We explore potential primary endpoints for a potential pivotal trial against aging, and claim that the type of endpoint and trial that is most likely to lead to wide reimbursement is a multimorbidity-prevention trial, measuring time to first of any age-related event.

• We suggest potential ways to calculate sample size requirements of a potential multimorbidity-prevention trial, given a drug’s estimated effect sizes against preventing specific age-related events.

• From these calculations, we conclude that:

  • In all cases, sample sizes needed to run a multimorbidity-prevention trial are substantially lower than running a prevention trial for each disease independently. That is, this trial design offers a stronger drug label for a reduced cost.

  • Aging drugs that have balanced effects across different disease groups provide even better economics: The cost of a multimorbidity-prevention trial becomes cheaper than a prevention trial for any single indication.

Self-evident disclaimer: This is an obviously imperfect analysis of how to run trials, to prompt discussion and further work. We tried to signal where we are less confident in the assumptions we’ve taken, but if that failed: please use the ideas and resources presented here thoughtfully, and use them at most as a starting point in your own careful thinking that precedes e.g. committing resources to a project. We will certainly do the same.

Subscribe now

Acknowledgements

Most of the work was done by Marton Meszaros and Brennan Overhoff. Thanks to the many people who contributed and gave us guidance or feedback. Special thanks to Mica Xu Ji, Zane Koch, Frank David, and everyone in Norn Longevity Nexus.

All remaining errors in the essay are owned by Marton and Brennan only. Feel free to send feedback.

To start off, let’s quickly review why value creation in biotech is harder than most industries. This creates a baseline for the scope of the challenge.

Bioscience is harder than rocket science

The first reason is that ‘science is hard’. Finding the right approach to curing a disease is like finding a specific bug in millions of years of vibe-code. We start with an idea of what we want to do (cure Alzheimer’s, say), but we’re not sure how to make that happen or what the rules of our system are. Even sending rockets to space is a much more predictable kind of engineering.

The science is particularly hard because of the limited tools available to biotechnologists. We can’t take a person with Alzheimer’s and with the press of a button make a change to their brain. In fact we are not allowed to do anything to a human until we’ve done a lot of preliminary work. For both practical and legal reasons, we instead have to go work on something that is not Alzheimer’s but hopefully relevant (individual brain cells, or a mouse engineered to mimic parts of the disease) to be able to try things at all. And even here, we often struggle to manipulate the model system in the ways we want. Cells are crowded and bustling with activity. Testing a biological hypothesis usually means a targeted change to one protein inside the cell. Before we test our idea for (hopefully) treating disease, we often need to engineer a way to make that targeted change. If you’re not a biologist, think of this as trying to add chili powder to a finished burrito, but only on the onions. It can be done, but only with considerable ingenuity and perspiration.

Similar problems can occur in other industries. For example, you know you want a user to get the search results they care most about, but not how. Or you’re building an engine before you have a spaceship to launch. But in general it is much easier to work up an idea and get direct feedback on whether it is good: write and run the code quickly, or test the engine on the ground. Biotech only gets incontestible feedback from human testing. This is akin to not being allowed to test your rocket engine before the final launch [1].

Making the experiments of early biotech faster and cheaper is good and useful. But it won’t speed things up very much as long as real feedback comes from human trials. Predicting human outcomes better would be far more useful, because it would both increase eventual success rates and allow more iteration with strong feedback.

Biotech does not want to become a trillion-dollar company

But the reason there is no trillion-dollar biotech is not just because ‘science is hard’. Rather, the ecosystem we have created to create and capture value in biotech deters and destroys would-be trillion-dollar companies.

Extremely valuable companies tend to have two features:

1. At least one flywheel or network effect where growth makes future growth easier, because more users makes the product more valuable or because multiple offerings boost each other’s success.

2. Near-monopolistic distribution of something that’s valuable to a very large market. This could be the best car or phone, essential tools for a growing industry (like chips or payment systems), or the attention of billions.

The first of these is very rare in biotech. The second is common, but not robust. Let’s explore:

Turning flywheels into firewood

Enough has been said about flywheels in business. When you are good at something and your output makes you even better, growth can take off. Unfortunately, the standard biotech business model actively destroys flywheels as a way to minimize risk.

For example: When a biotech seeks funding to launch clinical trials, it is naturally encouraged to derisk in the most cash-efficient way. Get the key data, and we’ll give you money for big trials. Regardless of AlphaFold, Nobel Prize winners, and whatever else, success rates when drugs first go into humans sit at 10-15%. Investors know this, and so prefer to wait for the human readouts before piling on more cash.

But by that time the company will have ‘molted’, shedding its early discovery team to conserve cash during the years of clinical trials. We are a humane industry, and the scientists are not condemned to death or even exiled. They disseminate out into other biotechs and do good work there. But in most cases they are no longer working together - the superorganism that created a new medicine has died.

This phenomenon occurs because 1) once clinical trials are launched the product and design are locked in by FDA filings, and cannot benefit from new insights or improvements, and 2) successful clinical trials reward successful early discovery and drug development, but not until many years after the work is done. So the stages of value creation and capture are de facto decoupled, and unable to form a flywheel.

Putting a medicine on the market involves many steps that each must succeed, but all rely on clinical trials to capture value. And so molting can happen at multiple points of a biotech’s life: When gleaning biological insights about disease shifts to drug-making medicinal chemistry, when new insights shift focus to a different disease area, when drugs eventually go into clinical trials…

Each time it happens, it makes sense as a way to improve ROI by limiting the cost of failures. And each time it happens, we pull up the tracks behind us and feed them to our train’s furnace: We will get where we’re going, as fast as we can! But no railroad system will be left behind for future efforts.

For my next trick, I’ll make my TAM disappear

At least biotech get monopolies, right? Sort of. When a new drug is approved there’s a period of patent protection (typically ~a decade) where competitors aren’t allowed to copy your product. This what gets investors to fund the science and clinical trials. But it is explicitly temporary. When the protection ends, companies all over the world will start selling the same product and profit margins will fall dramatically.

Because the value we create in biotech is not the physical molecules, but the knowledge that this molecule is an effective way to treat a disease. This is great for society: the knowledge we generate stays with humanity forever. It’s the only part of healthcare where the same treatment gets cheaper over time. (footnote Peter Kolchinsky’s book).

So the innovative company must innovate again, make a new product that’s even better. This too is great. It’s how a free market society makes progress. But companies treating disease have a unique obstacle for how they innovate: If the first product worked, the disease is gone or diminished. Once you’ve brought someone’s cholesterol back to normal, they don’t need another cholesterol drug. Maybe you can reduce some side effects or make it more convenient, but the largest markets will be doing something totally different.

This is very different from providing a way to handle online payments. Or even sending rockets to space (at least with resuable rockets). Companies that keep growing are able to use what they’re good at to make better and better products. As the products get better, more and more people want to use them. It’s good for society that this doesn’t happen when you’re curing disease, but it makes it hard for biotech companies to keep growing.

Lada’s proposal to target age-related diseases touches on this point: the biggest successes in drug development (anti-TNF drugs, GLP1Ras) have targeted very large patient groups. But therapeutic success still shrinks the market for your product, and your patent-protected profits will expire after a decade. To achieve the dynamics that robustly support a trillion-dollar biotech based on solving a single problem, we must go further than targeting a very large market: a product that increases its own demand could help create a flywheel. Within healthcare, we have a terrible example of this playing out: addictive substances including oxycontin. But an effective therapy that delays aging would produce the same dynamic, while making lives better: It starts with a massive TAM, and the product working keeps the users alive and hungry for more.

What can a poor trillion-dollar biotech seeker do?

Perverse flywheel and TAM dynamics in biotech are the result of how we as a society invest in medicine. There are no trillion-dollar biotechs in part because we don’t really want them. We pay for treatments rather than prevention because it solves an urgent problem and we can quickly tell whether we succeeded. Biotech investors are intelligent but operate with constraints: the cost of capital, the rate of progress, the patience and the literal lifespan of everyone involved. They are making rational decisions within those constraints and decide (implicitly or explicitly) to aim for something other than a trillion-dollar company. For example, a good chance of a 3x return when their company with a single promising heart failure drug sells to Novartis in Phase 2 [2].

There’s a lot of focus on technologies that could make it easier, and thus faster/cheaper, to do biotech. This could shift these ecosystem incentives and make investors more likely to double down on platforms. But most of the cost and time happens in clinical trials, where a new CRISPR tool doesn’t make a difference. And without a functional and competitive flywheel, there is no reason why value would accumulate in a single company versus distributed across many.

Successfully setting up a way to sell preventative medicine to a massive market could produce a trillion-dollar biotech, but you’d need to figure out a moat other than patent protection. Especially if you aim to sell directly to consumers, thus competing with people who don’t limit themselves to selling things that actually work.

I suspect that the first trillion-dollar biotech will be one that solves connecting a core technology to repeatable value capture in order to create a flywheel effect. Achieving this will probably mean building towards it from day 1. ‘TechBio’ companies are often pitched as ‘flywheel first’. High-throughput experiments generate data for AI models that will discover new therapies. But a new preclinical candidate is often not enough to capture value. So the companies face a challenge at the clinical stage, where the capital required to take many shots is immense. A flywheel that relies on convincing investors to fund dozens of clinical trials before you show a profit will be hard to get going. If you can link value capture directly to the output of your platform, with a platform that does not start from scratch when you move to a different disease, you might have a shot.

Given how biotech works, what combination of tricks would suffice?

• Targeting a very large market D will pay off if successfuly, but limited duration of protected profits will prevent a proper flywheel from forming.

• A technology that creates a flywheel at just stage B (for example new way to target genes, like Alnylam) could outperform within status quo, but will be no more likely than average to generate a second success.

  • If it opens up a meaningful number of existing insights (e.g. genetic targets), it could still have a very good run!

• There are two designs that seem feasible to create a value-capturing flywheel:

  • A technology platform that can solve both A and B in a self-improving way, combined with either a way to monetize prior to C (challenging given current value curve) or access to funds/profit to take multiple shots at C.

  • A (correct) insight A that fits a market D that does not shrink with success, combined with a self-improving platform for B and access to funds/profit to take multiple shots at C.

[1] Notice that most of the breakthroughs from LLMs and other recent AI are in areas where it’s easy to tell whether the output is good: winning at games, plain language responses, code that runs.

[2] We do sometimes swing for trillion dollar bets. One company that seemed to have a chance a few years ago is Moderna (~$200B at peak). Notably, it had both a platform that could rapidly be redeployed across diseases, had quick clinical readouts (for the areas they eventually broke through with), and in the case of vaccines with boosters avoided the disappearing TAM issue. On the other hand, the most valuable therapeutic companies currently (whether century-old pharmas or the winners of the first wave of ‘founder led biotech’ in 80s-90s) don’t really fit the criteria I describe. Could be I’m wrong, or could be that they’ve grown over time but are not on a trillion-dollar trajectory.

[3] If I’m wrong, it’ll be Eli Lilly solving another huge market like neurodegeneration, after doubling down on innovation using Tirzepatide revenues.

In November of 1944, President Franklin D. Roosevelt wrote a famous letter to Vannevar Bush, director of the Office of Scientific Research and Development. Roosevelt asked him how America’s scientific prowess would guide the world and benefit the public in the post-war world.

Vannevar Bush (Credit: H. WALKER/TIME LIFE PICTURES/GETTY)

Bush’s response was a report detailing “The Endless Frontier” of American innovation, fueled by public investments in scientific research, strong relationships with academic institutions, and resounding support for the early research that would go on to enable countless cures, energy improvements, material development efforts, and even put a man on the moon. This system has driven incredible biotechnology research, leading to improvements in our span of healthy years. The results have ranged from better hygiene to novel therapies for cancer and neurodegenerative diseases.

However, an end to the Endless Frontier is now on the horizon.

The Choice Ahead

As the American population grows older, the economic and personal burden of age-related disease grows more stark, straining Medicare, Medicaid, and other social welfare systems. This is compounded by age-associated ailments taking away late-life productivity. It’s time to revisit the question of how America’s scientific prowess will benefit the public. The original Endless Frontier began at a time when the biology of aging wasn’t a topic of research, and efforts to combat disease were divided on many fronts rather than studying their shared causes. As our understanding of aging biology evolved, this framework did not, and today still only ~1% of the NIH budget is directed towards addressing the largest risk factor for most diseases: aging. Targeting one disease at a time, sometimes referred to as “sickcare”, is seeing diminishing marginal returns in aging populations, as each new medicine offers a short-lived reprieve from disability. Addressing the root causes of aging, on the other hand, could lead to accelerating improvements in health and quality of life. Yet, ongoing cuts to basic scientific research are hindering our ability to understand — and intervene in — these aging processes. Strategic investment is also drying up as biopharma increasingly taps late-stage assets based on past research and relies on Chinese innovation, presaging a future where Americans are sicker for longer. Meanwhile, other countries, such as China and Singapore, are investing heavily in their domestic ability to both innovate and manufacture new interventions for chronic and age-related diseases.

Nations must choose whether to lean into the biology of aging for solutions to public health or stay reactive to a growing burden of disease. Eighty years after initiating The Endless Frontier, America is faced with this choice as well as the consequences it will bring. On one path, our ignorance of aging biology leaves us helpless to prevent the financial, emotional, and physical consequences of an aging population. On another, we ensure that every American can live healthier for longer, free from mounting pill bottles, medical debt, and chronic suffering.

Let’s explore what each of these futures might look like…

The Gerontocene

2025 | A Missed Opportunity

Under the shadow of ongoing debt negotiations, major spending bills, and other budgetary constraints, debates erupt in the U.S. Congress over healthcare and research spending. The rising number of older Americans relying on Medicare/Medicaid and Social Security spurs belt-tightening efforts that slash discretionary spending and grant programs. On top of research funding cuts, increased uncertainty about future drug approvals causes investors to pull capital from early-stage innovation. Pipelines shutter and many smaller companies close their doors, firms are increasingly competing for the smaller pool of later-stage assets within the American and EU markets, and are increasingly turning to Chinese offerings. As institutions slash budgets to account for reduced federal funding, young investigators are lured eastward and to Europe, as both spheres of influence compete to attract talent fleeing U.S. institutions. These consecutive hits deal a lasting blow to the biotechnology industry and the development of therapeutics for diseases of aging.

In the midst of all of this, the rise of GLP-1s after two decades of effort gives the industry a lifeline: a strong and very profitable example of one drug addressing multiple diseases. Experts across the nation spotlight the widespread applicability and apparent safety of GLP-1s – some hailing them as the “first longevity drugs” – emphasizing the broad economic implications of a society less riddled with obesity, diabetes, and other conditions. However, this success depends heavily on decades of prior research. Some scientists view aging biology as a field with the potential to yield the next such drugs. But public fears and doubts from other doctors and scientists scare off politicians from supporting this new approach. Rather than doubling down, the U.S. shies away, with budget deliberations upholding reductions to NIH funding for aging projects and regulatory proposals being ignored.

2028 | Warning Signs

Healthcare disparities make headlines throughout the country: Nearly 70 million people in the US are over the age of 65, and ~80% have two or more chronic diseases, requiring multiple daily medications. As the incidence of age-related disease rises, insurance premiums continue to climb, leaving many middle- and lower-class families to forgo medicines. The situation strains hospital and provider networks while also pushing long-term care infrastructure to its limits. Shortages of professional caretakers, cuts to at-home care eligibility on Medicaid, and subsequently long wait lists all contribute to worsening healthcare standards. The human toll falls on the 70 million Americans who serve as unpaid caregivers, left with little choice but to reduce their working hours or take career breaks entirely to provide for loved ones.

Medicare spending passes $1.5 trillion this year. Politicians are at a loss for options to address the crisis, short of making further cuts or facing tradeoffs with other expenses such as defense, education, and infrastructure. Social Security barrels on track towards insolvency by 2033, and more cuts are made to key services for seniors and retirement benefits. While this plays out, China announces several national moonshot programs intended to target aging and chronic disease, some of which are led by U.S. scientists who left the country following budget cuts in 2025. Investors and innovators alike continue to shift eastward, pouring a growing amount of capital into deals with Chinese companies. This shift drains the American biotechnology sector of cash flow for new startups and scientific spinouts.

Subscribe now

Source: KFF

With all these factors, the engine driving us into The Endless Frontier begins to stutter. Universities report drastically lower numbers of matriculating graduate students, and investors lament over the shriveling early-stage pipeline of new medicines. Much of the scientific community realizes that if action is not taken soon, there will be no reversing the collapse.

2032 | Turning Point

As the nation rings in the New Year, the Bureau of Labor Statistics reports that a record number of individuals left their jobs to care for an older loved one. The special report shows Americans citing unbearably high costs, years-long waitlists, or a lack of a care facility within distance as the main prohibitors. Adding to the expense is the fact that many are still being treated for individual diseases on a reactive basis. As the number and prevalence of these diseases increases, the personal and societal costs grow, putting treatment out of reach for many patients.

Sharp restrictions on Social Security eligibility are put in place by Congress as the program’s ability to make payouts reaches a crisis point. Meanwhile, more stringent working requirements for Medicare and Medicaid are enacted, putting further financial strain on families to afford care. Workforce participation sinks among Americans over 65 for the first time since the pandemic due to the rising number of Americans living longer, but being unable to work due to age-related diseases. Shortages across sectors, including healthcare, education, and the trades, draw comparison to Japan’s trends in the previous two decades. With an old-age dependency ratio of 35% and a total of 72% (i.e. for every 28 working individuals, there are 35 dependent seniors), proposals are circulated throughout the U.S. Congress to raise the retirement age in order to keep Social Security afloat.

Across the Pacific, China’s state-sponsored longevity efforts have begun to pay off, so far resulting in the first aging therapeutic targeting multiple indications and increasing healthy life at a lower cost. However, due to ongoing geopolitical tensions, U.S. access to the drug is limited. With a diminished biotechnology sector and growing reliance on Chinese organizations for biomanufacturing, efforts to produce a similar therapeutic will take years. In the meantime, Americans are left to forgo the benefits. Chinese citizens, however, enjoy a reduced disease burden and less dependence on care or prescriptions. Leaders note marked reductions in hospital admissions for age-related diseases and fragility, while families enjoy the reduced pressures of caretaking responsibilities. All of this is possible despite nearly two decades of a challenging demographic picture resulting from the one-child policy.

2035 | The Gerontocene

For the first time in American history, falling fertility rates have led to the number of individuals over 65 outnumbering those under 18. A decade of insufficient investment in aging therapeutics has produced the “Gerontocene” – a world where social structures, economics, and happiness are dominated by the ills of aging and our inability to combat them.

Successful development of aging therapeutics overseas from adversarial nations has created stark contrasts in quality of life and economic output. However, geopolitical tensions continue to prevent Americans from broadly accessing them. After seeing the possibility demonstrated abroad, American politicians and regulators reverse course, announcing a moonshot program investing billions of dollars into the development of a comparable multi-morbidity therapeutic. Big pharmaceutical companies and startups alike begin spinning up programs to answer the call. However, official predictions do not anticipate the completion of the program or widespread availability to the public for at least another 10 years – a decade of decline under the shadow of the Gerontocene.

How can we avoid this fate? Let’s take a look at another path, of boldness and hope.

2035: The New Endless Frontier

2025 | The Beginning

The Secretary of HHS spends his first months in office focusing heavily on the prevalence of chronic disease, stating, “The chronic disease epidemic is the most urgent and costly health crisis in America today. We can’t afford another decade of delay.” Prioritization of chronic disease at the highest levels of government leads to investment and research into shared causes of many diseases. This momentum leads to several executive orders being signed to support aging therapeutic innovation. These orders build on the progress of several ARPA-H programs launched earlier in the year to target age-related conditions and upstream causes of aging. Simultaneously, the White House announces the creation of a public-private investment partnership, similar to those seen with AI innovation. Leaders see it as an opportunity to capitalize on the enthusiasm from recent blockbusters such as GLP-1s, the expanding market size in an aging population, and rapidly maturing science in the aging field.

China, recognizing the demographic crisis in its own rapidly aging population, follows America’s leadership in this space and begins to accelerate. Expanding its Healthy China 2030 agenda, China invests heavily in aging therapeutics to counter American development and sets its eyes on higher life expectancy goals. However, thanks to moonshot programs such as ARPA-H and strong public-private interest, the U.S. takes the lead, attracting more talent, investment, and popular support than our competitors. The New Endless Frontier is on the horizon, and just as we did with the race to the moon, we are not shying away from the moment; we are bringing the future to fruition.

2028 | A Rising Tide

By 2028, following intense efforts to reform regulatory consideration of aging therapeutics, the first clinical trials for “aging indications" are launched under advanced conditional approval pathways. The earliest Phase 1/2 trials are of existing medications that were previously only approved for single indications. The announcement spurs excitement, generating a wave of investment into new and existing companies to develop drugs specifically for aging. By this point, billions of dollars have been poured into combating age-related diseases. Hundreds of preclinical studies are conducted, innovating on new drugs, attempting to repurpose new ones, and developing novel biologics. Growing understanding of aging processes and their manifestations underlie this upswing in the space.

A new wave of venture capital firms focused on longevity pop up, building on the success of the pioneers in the field. With this momentum, Congress raises the NIH’s investment in aging biology research for the third consecutive year, with FY29 levels double what they were in 2025. American biobanks, such as the All of Us program, continue to unlock troves of data about healthy and super agers, attempting to crack the genetic, environmental, and behavioral code of living healthier, for longer.

2032 | Breakthrough

In 2032, the first aging therapeutic is approved in the United States based on broad benefits across disabilities. A drug that was previously approved for a single indication is now widely available to tens of millions of Americans. After showing a 30% reduction in frailty and significantly improved cognition in a placebo-controlled phase 3 trial, FDA advisory boards are quick to recommend the approval of the drug under the new “aging indication” pathway created in 2028. This marks a milestone in humanity’s progression towards improving our health and beating back the diseases that have plagued our species since its inception. The drug is linked to a reduced risk of cardiovascular and neurodegenerative issues, enhanced physical stamina, and better cognitive capabilities to heightened workforce participation. By the end of the year, the percentage of the population over 55 participating in the workforce crosses 45%, thanks largely to lower disease incidence.

Within just a few years, the economic returns start to be felt, with some Americans reporting fewer hours spent caretaking for older loved ones and lower medical debt as a result of fewer falls, diagnoses, and medications needed. As the availability of these drugs spread, there is a noticeable decline in Medicare and Medicaid claims, despite a growing number of eligible individuals. Realignments in retirement plans seek to provide flexibility to older workers who are gainfully employed, not because they have to, but because they want to and can be. Around the country, waitlists for long-term care facilities drop, while hospitals see declines in the number of visits for age-related diseases and frailty-dependent falls among older individuals.

Subscribe now

This trend is especially true in areas of the country where adoption of these drugs is high. Congress, looking forward to the FY33 budget, sees stabilization of Medicare and Medicaid costs, freeing up the opportunity to pursue payments on debt, investment in infrastructure, and other key discretionary topics. American cities retain the lead in investment and talent attraction compared to other global hubs.

2035 | The New Endless Frontier

By 2035, Americans start to benefit from a system that truly focuses on healthcare, rather than reactive “sick-care”. Years of public funding to understand basic aging biology have created a rich pipeline for continued development of drugs that more efficiently reduce disease burden. 2025 served as a crucial moment when many other nations were seeking to dethrone our leadership in science. Strategic competitors sought to become home to the latest revolution in human health through the development of aging therapeutics. However, the U.S. government rose to the occasion, creating the conditions (regulatory and otherwise) for a whole new class of “longevity drugs” that became first available to the American public.

On the back of strong public-private partnerships, an influx of investment for the space was curated. As the field continues to expand, people worldwide rely less on different drugs to manage multiple chronic diseases and instead take one to combat many if not all. With their strong safety profiles, these drugs are prescribed preventatively for individuals almost as commonly as statins were for cardiovascular risk just 10 years ago. Close attention from the FDA to the value of quality control reduces “snake oil” offerings and uplifts credible interventions. The U.S., partnering with dozens of allied nations, leads a global initiative committed to distributing aging therapeutics, setting the moral, scientific, and economic momentum to tackle chronic diseases.

Expectations of financial ruin from Social Security insolvency never materialize. As a result of fewer Americans needing expensive long-term health care or costly prescriptions, average individual beneficiary claims continue to shrink. Many adults who otherwise would have had to take breaks in their careers to care for aging loved ones or incur crushing medical debt can instead enjoy more time with those in their lives. The economic boost and reduced societal burden of age-related disease secure America’s role as the global leader on health, freeing up investment into other discretionary areas of spending.

A New Endless Frontier has arrived, freeing humanity from the once seemingly inevitable decline that came with aging and the trap we could have fallen into where more years did not mean more health, but more suffering.

Conclusions

Countless matters cry out for our attention. As a nation, we are always forced to choose our battles, and sometimes come to regret our choices. For example, America neglected to invest in industrial semiconductor chip production beginning in the 1960s. Decades of deprioritization and reliance on other nations’ capacity have led to supply chain choke points that are more vulnerable than ever to geopolitical tensions. Starting to reverse course cost ~$280 billion in the form of the CHIPS and Science Act. The same story has played out for mining rare earth minerals and conducting heavy manufacturing. As fertility rates decline, population aging continues, and treating disease becomes more and more costly, it is clear that the battle for healthy aging cannot be ignored.

What we choose to do today defines the next decade, as either an inexorable shrinking of our healthy population or the turning point towards health. Our ancestors looked to the stars, yearning to be among them. It was the United States of America that made it happen. Our ancestors have also looked to their loved ones, yearning to free them from disease and disability. Should the people of the United States of America be the ones to make it happen? If so, the choice we face today is clear.

Subscribe now

Comments on our assumptions

Note on Charts/Figures: Charts and key numbers in this piece are not intended to be a precise analysis of how longevity therapeutics will impact economic and demographic outcomes, which would require proprietary knowledge and extensive modeling. Rather they represent what we think are reasonable estimates (more below), and leave greater precision to economists and other writers. As more information becomes publicly available, we may update this piece with new findings and models that rigorously quantify the transformational impact of longevity therapeutics.

Assumptions for demographic charts: Total populations in each age range are from the U.S. Census Bureau’s International Database. Chronic condition population percentages come from CDC BRFSS 2023. The percent of individuals in each age group reflects linear growth within larger grouping averages for disease burden prevalence percentages (18-34, 35-64, 65+). Numbers for 0-17 are based on US Census ACS Disability Reporting, as no chronic disease data were available. Population prevalence percent is assumed not to change meaningfully from 2025-2035.

Assumptions for impact of new therapeutics: We assume that the percent prevalence of disease is the same for both tracks until 2032 (for example, if multiple chronic condition burden for 65+ is reported by BRFSS is 78%, it is assumed to be 78% in both the Gerontocene and the New Frontier). For the reduction of disease burden starting in the 2032 entry, we assume a 15% reduction in chronic disease prevalence and then a 30% reduction in 2035. These estimates are based on one, then two, medicines from aging research with effects comparable to existing GLP1 agonists (such as Ozempic), roughly 20% on multiple prevalent diseases. We do not attempt to model specific efficacy on individual conditions. These reduction figures do not factor in any form of adoption percentage, discount rate, or other distribution metric.

Science Highlights (skipping most of the published work)

They had parallel tracks this year, so I didn’t catch every presentation.

• Max Yun shared details on aging in axolotls in her colony. They live ~10 years and regenerate throughout life. Many features of human aging are absent: 1) no senescent cells, at least by common markers, 2) no telomere shortening, 3) no loss of stem cells, 4) their thymus doesn’t disappear like in humans, 5) ~flat risk of death by age. So how do they die? She doesn’t know yet. One observation is that their adaptive immunity is very weak, leaving them prone to infections (this is also true of naked mole rats).

• Apropos, Emma Teeling shared findings from large-scale epidemiology and sequencing of different bat species finds that short-lived bats have increased innate immune activity with age, while long-lived bats decrease activity.

• Björn Schumacher and Zane Koch showed early data linking the DREAM protein complex, DNA damage and repair, and neurodegeneration. This is cool both because DREAM is probably druggable, which DNA repair generally isn’t. And because while DNA mutations correlate inversely with lifespan, we don’t know what mechanisms are involved. (Björn’s paper)

• Juan Carlos Belmonte of Altos and James Kirkland of Cedars-Sinai were both using transplant organs as a system for testing putative aging therapies - this is cool since it’s a real aged system, ‘transplant success’ is a fast readout, and success could have a commercial outlet before direct human application.

• More clocks, although still nothing with blinded prospective validation… Andreas Berger is looking to make single-cell transcriptomic signatures of cellular injury to compare with aging, by comparing overlaps between different injuries. Alexander Tyshkovskiy from Gladyshev lab made a cross-species, cross-cell-type transcriptomic clock that captures both in vivo and culture “aging” - to me, this suggests it’s focused on cell intrinsic changes, maybe epigenetic, since that’s what would be shared with cultured cells. Tony Wyss-Coray shared a (published) Alzheimer’s predictors based on YWHAG/NPTX2 protein ratio in CSF or blood.

• There was a refreshing number of presentations from labs new/adjacent to the aging field: Sophia Liu shared spatial transcriptomics exploration of thymic function and changes with age, Uri Alon is working on modeling species lifespan, and argued that heritability of lifespan is actually ~50% when you normalize for extrinsic mortality in twin studies. Very satisfying to see many of these funded by Impetus Grants a couple of years ago, now bearing fruit. Vadim Gladyshev/Bohan Zhang spoke about experiments to study accelerated/reversed aging when organs were transplanted between donors of different ages.

• Nice panels on virtual cells & pharma perspectives.

  Subscribe now

New developments

• Todd White announced the non-profit Thalion Institute, intending to spend several hundred million on research and tools with an emphasis on comparative biology. One major component will be a biobank of samples from species with long lifespan or exceptional disease resistance, through life. Assuming this all happens, it should be a great source of new therapeutic hypotheses.

• Attendance by pharma companies was higher than ever. Eli Lilly had the greatest presence, consistent with their investment in a new longevity-focused group. During a talk they described GLP1s as “the first longevity drugs” based on their broad effects, as did Lotte Bjerre Knudsen of Novo Nordisk. Novartis and Genentech agreed with this description on a panel. AstraZeneca and Lundbeck were also present. Not all of these are investing in longevity, but all were paying attention and looking for opportunities.

• There was a lot of discussion of longevity medicine, and three clinical trials stand out for potential to shift status quo: 1) Eli Lilly’s Alzheimer’s Primary Prevention rial was mentioned (no data yet), which is a rare example of treating for prevention. These trials are expensive but have potential to unlock new surrogate endpoints to speed up later trials. 2) FAST, not strictly a trial but rather using samples from past trials of putative longevity drugs to identify biomarkers of aging. Impetus gave a small amount of early funding, which has now been expanded by Astera and others. No data yet, but excited to see results. 3) Another Impetus-funded trial is VIBRANT-I, testing rapamycin to delay ovarian decline. Youin Suh says they’ll be able to unblind and analyze data by the end of the year. VIBRANT-I one is particularly exciting because the design could be a platform to test other candidate longevity therapies much faster than multimorbidity/morality trials (plus delaying menopause is by itself a huge need). A larger VIBRANT-II study is already bring planned. If this works out it could majorly speed up and validate the field.

Congratulations to organizers Morten Scheibye-Knudsen and Alex Zhavoronkov for a great event. Looking forward to next year.

Subscribe now

This piece explores why biomedicine, and longevity in particular, is not a booming industry and how that might change. We set the stage with a primer on how medicines are made: generally, ideas for how to treat disease come from a university research group or sometimes existing companies. Turning the idea into a way to affect human bodies, e.g. a molecule in a pill, takes a few years and more than ten million dollars. Unlike most businesses, the company is not allowed to sell these pills to patients. First, they have to work with the healthcare system to run controlled tests in patients (‘clinical trials’) for a number of years, which can cost more than a hundred million dollars. If the pill works well in these trials (most don’t), it can be sold (in a complicated way via doctors and pharmacies) to a precisely defined group of patients. Targeting aging to improve health takes all of these dynamics to the extreme: clinical studies take longer, the probability of success is unknown, and markets are theoretically huge but untested in practice.

A feature of every booming industry is productive specialization, where different organizations get good at parts of a larger, valuable production. Improvements from any part benefit the whole. New ideas can also open up new markets, using the tools developed in the industry in new ways. This dynamic happens often when the barrier to entry is low, as for software written by a tiny team and offered for sale to the industry. But it can work for capital-intensive industries too, as long as future profits can be predicted reliably. However, the dynamics of biotech disincentivize specialized collaboration. Let’s contrast space and biotech, both capital-intensive:

Space: Company A makes communications satellites. They sign a $50M deal to provide wifi in rural Alaska. They spend $10M building small satellites and pay rocket company B another $10M to launch them (and get their costs back if the rocket explodes). Assuming their satellite works, Company A gets to fulfill their contract without learning to build rockets, while Company B gets paid right away using revenue from a customer base they weren’t serving. Customers get something they didn’t have before, and both companies become more valuable. The more companies come up with useful reasons to put things in space, the more B’s market expands.

Biotech: Company C has data that a certain protein improves heart failure in mice. They want to raise $50M to run clinical trials to get proof in humans, after which they can raise more money to get the drug to approval. These trials will take >5 years, after which they may have lots of profits. Company D has a way to target proteins to the heart better, which they claim would improve the odds of success for Company C’s trials. They suggest that Company C pays them $20M to license this, but C does not have this money because they don’t have revenue before the drug is approved. They ask investors for $70M instead, but their competitor E is only asking for $50M. C tries offering D royalties on future sales instead, but D needs cash now to pay their employees. Ultimately, C, D, and E are distributing the same pie of revenue if drug gets approved.

This type of problem should be solvable by borrowing on future gains. If the protein targeting technology really improves success rates and thus expected value, it should be worth investment today. Even if the company doesn’t have the money, professional underwriters exist to assume long-term risks with positive expected value. But this model relies on being able to estimate the future expected value. High uncertainty scenarios turn low-margin-but-net-positive underwriting into speculative investment. The core event triggering profits in biotech, clinical trials, is necessarily attempting something new and happen on average once or twice per company, so relevant feedback is scarce. To illustrate how this further disincentivizes productive specialization, let’s revisit our space and biotech companies:

Subscribe now

Space: Company A from the example above doesn’t actually get the $50M contract right away, so they don’t have enough money to build satellites and pay for the launch. But because the future revenue is clear, underwriters are happy to bridge the gap. The remaining risk is that A fails to build a usable satellite or B fails to launch. Usually there is a track record for these, and success rates are very high. The risk of failure can be accounted for in financing terms. This arrangement allows progress that otherwise would not have gotten off the ground.

Bio: For most biotech companies, a successful trial means large profits while failure means the end of the company. The average trial has a ~10% chance of success, and improving that number would be amazing. The problem is that the trial either succeeds or fails based on many factors (C had the right disease understanding, found the right patients, picked the right drug dose, etc). Unlike rockets and satellites, we can’t independently test which part succeeded or failed, so determining the effect of D’s technology would take many such trials, each long and expensive. C and others need to commit to D’s technology before they lock in trial design with the FDA, without clear proof of benefit. Because every company faces this dilemma, it’s difficult for D to generate evidence about their technology’s impact. In most cases, companies like D end up using their technology for making therapies and running their own trials.

Feedback feeds financing. When there are no good signals to gauge ROI, financing comes from discretionary spending rather than core investment funds. This is evident in the current state of the longevity field: A few highly capitalized companies are funded by individuals who believe in a long-term mission. An increasing number of pharmaceutical companies have research groups focused on aging biology, but they are small and exploratory. A large fraction of funding and progress comes from non-profit organizations, compared to biotechnology overall (and to other fields).

What would enable longevity to become a true industry, with an accelerating rate of progress fueled by productive specialization? From the above, we see two options:

• The first is political: Make feedback faster by reducing the burden of running clinical trials, which would mean cheaper and more abundant feedback for companies trying to improve the work of others. This could be done responsibly, e.g. by allowing drugs to go on the market earlier but monitoring their effects more closely when they do. However, there are limits to this option - taken to the extreme, e.g. eliminating clinical trials, would eliminate incentives to make drugs actually work and create an industry of false marketing of ineffective products.

• The second is more technological: Make feedback better by improving our ability to predict eventual outcomes, which would let transactions around future events happen confidently. The threshold that would unlock new financial models isn’t obvious, but even before that happens, individual trials would have higher success rates. And there is no limit to the benefits from improving this metric.

These are not alien ideas - biomedicine in general would like faster time to market and better predictions, and attempts these in various ways. But for longevity there’s a stronger imperative: assumed trial success rates are 0%, and proving otherwise takes more time and money. And there are actionable opportunities: everyone ages, both humans and animals. We can build predictors in short-lived model organisms while incorporating data collected from human blood samples. We can create public benchmarking of predictors, to incentivize what works robustly over what can be published. And anything that really works can be scaled through existing consumer health offerings. It’s possible to work towards the first predictive measures of longevity, and if we want this industry to boom, it’s the most important thing to focus on.

Subscribe now

If you have previously subscribed to updates from Norn Group, we’ve moved your subscription to this newsletter, which will be our main outlet moving forward.

This newsletter aims to be the best source for understanding the state of the longevity field. It is intended for lightly technical audiences who want insider perspectives on new developments, analyses on ways to increase progress, and occasional scientific deep dives. Most content is written by Martin Borch Jensen, with some posts from members of Norn Nexus.

If you're not familiar, Norn Group is a non-profit organization run by scientists focused on progress in the field of longevity. Our work includes Impetus Grants for high-potential research, training programs, and technical analysis and proposals. We also host an overview of the rate of progress in longevity, where we track bottlenecks and opportunities. You can read about the organization on our website, and about our results so far here. To those who have already supported our work, thank you.

Some things you will not find in these notes:

• Endorsement of any consumer products.

• Nutrition or lifestyle advice.

• Blind optimism.

Instead, expect posts that collectively lay out a roadmap for progress towards medicines yielding many more years of healthy life.

MBJ