The dream that won't die

For most of recorded history, human life expectancy at birth sat somewhere between 20 and 50 years. Then, in the early 20th century, something extraordinary happened. Clean water, sanitation, vaccines, antibiotics, and better nutrition drove death rates down so fast that life expectancy at birth — demographers call it e(0) — began rising by roughly three years per decade. In the span of a single century, high-income nations added about 30 years to average lifespan. Demographers gave this acceleration a name: radical life extension.

That phrase has been doing a lot of work ever since. In popular culture it has migrated from a description of past progress into a prediction of future immortality. Silicon Valley billionaires fund anti-aging startups. Biotech firms promise "healthspan" breakthroughs within a decade. Headlines casually assert that most babies born today will live past 100. The implicit promise is that the curve of the 20th century can be extended indefinitely — that we are on the verge of a second longevity revolution.

A new analysis published in Nature Aging by S. Jay Olshansky, Bradley J. Willcox, Lloyd Demetrius, and Hiram Beltrán-Sánchez asks the unfashionable question: what if it can't?

Two hypotheses, three decades of evidence

The paper is, in effect, a 30-year status check on a bet. In 1990, Olshansky and colleagues published a now-classic paper in Science arguing that humanity was approaching a soft upper limit to life expectancy — somewhere around 85 years on average (about 88 for females, 82 for males). Their reasoning was simple: the easy gains from public health and medicine had largely been banked. Once you've eliminated childhood infectious disease, saved mothers in childbirth, and tamed heart attacks in middle age, what's left to die of is aging itself — and aging is not a disease you can vaccinate against.

A counter-claim arose almost immediately. Other demographers argued that the "limited lifespan" view was too pessimistic — that ongoing medical advances and improvements in behavior (less smoking, better diet, more exercise) would keep pushing the curve upward at the historical pace, that radical life extension had already begun, and that most newborns in wealthy countries would routinely live past 100. One influential paper went further, declaring there was "no convincing evidence (demographic, biological or otherwise) of a lower bound on death rates other than zero."

Three decades later, the authors of the new study had actual data to settle the dispute. They pulled annual age- and sex-specific death rates and period life tables from the Human Mortality Database for the eight longest-lived countries on Earth — Australia, France, Italy, Japan, South Korea, Spain, Sweden, and Switzerland — plus Hong Kong and the United States, from 1990 through 2019. (They stopped at 2019 deliberately, to exclude the distortion of COVID-19.) They asked five specific questions, and the answers, taken together, are remarkably consistent.

Finding 1: The brakes have been applied

The first question was the simplest: has radical life extension — operationally defined as a 0.3-year annual increase in life expectancy at birth, equivalent to 3 years added per decade — actually occurred anywhere since 1990?

The answer: only twice, and only briefly. South Korea and Hong Kong are the only populations that sustained the required pace, and even Hong Kong only did so from 1990 to 2000, largely on the back of economic prosperity and aggressive tobacco control. Everywhere else, the annual rise in life expectancy has decelerated to below 0.2 years — well under the radical-extension threshold. In every population studied, including Hong Kong, the most recent decade of improvement was slower than the previous one.

The average gain in life expectancy from 1990 to 2019 across the eight longest-lived countries plus Hong Kong was 6.5 years over three decades — meaningful, but a far cry from the 9+ years that radical-extension forecasts had implied. And the trajectory is bending the wrong way: more deceleration, not acceleration.

The United States, tellingly, is one of the only countries ever documented to end a decade with lower life expectancy than it began with — a phenomenon historically seen only during extreme events like the 1918 Spanish Flu or wars, but in the U.S. case driven by mid-life mortality from so-called "deaths of despair" (opioids, alcohol, suicide) and chronic disease between 2010 and 2019, then dramatically worsened by COVID-19 in 2020.

Finding 2: The 100-year-old is still rare

What are the odds that a baby born today reaches 100? Using 2019 period life tables, the authors calculated the percentage of each birth cohort expected to survive to age 100. The averages across the longest-lived populations: 5.1% for females and 1.8% for males. Hong Kong leads the world at 12.8% of females and 4.4% of males expected to reach 100.

Crucially, no population on Earth comes anywhere close to the 50% survival-to-100 mark that would be required to support predictions that "most babies born today will live to 100." That claim, however often repeated, is not supported by the demographic record. Centenarians remain a statistical rarity, not the emerging norm.

Even under the most optimistic composite scenario the authors constructed — piecing together the lowest observed age- and sex-specific death rates from any country in the world for 2019 — survival to 100 maxes out at 13.9% for females and 4.5% for males. That composite "best of the best" life expectancy is 88.7 years for women and 83.2 years for men. Eliminating death entirely from birth through age 50 — a hypothetical that would require eradicating every accident, infection, and congenital condition — would add only about 1.0 years for females and 1.5 years for males to that composite ceiling. The diseases that kill young and middle-aged people have already been so thoroughly addressed in long-lived countries that there's almost nothing left to gain there.

Finding 3: Each extra year costs more than the last

Here's where the math gets punishing. The authors asked: how much would total mortality — death rates at every age, from every cause — need to fall to add just one more year to life expectancy at birth? They traced this figure back as far as 1750, where data existed.

The trend is unmistakable and brutal. In every population studied, the percentage reduction in death rates required to add a single year of life expectancy at birth has increased since 1990. The easy years have been harvested. What's left costs progressively more.

Take Japan in 2019 as a representative case. If female life expectancy is to rise from 88 to 89 years, mortality from all causes at all ages would need to fall by 20.3%. For males to rise from 82 to 83 years requires a 9.5% reduction across the board. To put that in perspective: a 20% reduction in all-cause mortality at every age is roughly equivalent to eliminating a major cause of death outright — every cancer death, or every cardiovascular death — and even that might not be enough.

Finding 4: Life table entropy — the hidden brake

This is the most technically interesting part of the paper, and it deserves a moment of attention. Demographers use a metric called life table entropy, denoted H*, which measures how resistant a population's life expectancy is to further mortality reduction. The intuition: when most deaths are concentrated at very old ages (as they now are in long-lived populations), saving any individual life buys very little additional time, because that person is already near the biological end of the lifespan distribution. Low entropy means small mortality reductions yield big life-expectancy gains; high entropy means they don't.

The authors found that H* has increased steadily from 1990 to 2019 in every long-lived population studied, and — strikingly — has converged on a common, higher level across countries. The brake is not only engaged; it's gripping harder than it did three decades ago, and it's gripping everywhere. This is direct empirical evidence that further gains in life expectancy are becoming structurally harder, not easier, regardless of which country you examine.

The mathematical backbone of the paper is a decomposition: the logarithm of life expectancy can be expressed as a function of life table entropy and lifespan inequality. When forecasters naively project life expectancy forward in a straight line based on past gains — which is what most actuarial and government forecasts still do — they implicitly assume entropy must keep improving at the historical pace. The data show it isn't. This is why linear extrapolations of past life expectancy are systematically over-optimistic.

Finding 5: Mortality is compressing, not expanding

The fifth finding concerns lifespan inequality, denoted Φ* — a measure of how spread out the age of death is across a population. If Φ* is declining, deaths are compressing into a narrower window (most people dying at similar old ages); if it's rising, the distribution is spreading out, with more people surviving to extreme old age.

Over the 69 years from 1950 to 2019, Φ* declined uniformly and consistently in every long-lived population, even as life expectancy rose. This is the signature of mortality compression: we are not, as a species, pushing the upper boundary of lifespan outward. We are packing more people into the existing boundary. The Gompertz curve — the exponential rise in death risk with age that has held for two centuries — is not shifting rightward. It's getting steeper at the end.

This single fact demolishes a lot of hopeful rhetoric. A genuine second longevity revolution would require the death distribution to shift toward later ages, not just compress. The data show the opposite is happening.

The 110-year thought experiment

To drive the point home, the authors constructed a hypothetical: what would survival have to look like for life expectancy at birth to reach 110 years — the kind of figure that gets tossed around in anti-aging circles? They used Japanese females in 2019 as the baseline, since that population already has among the world's most favorable mortality profile.

The numbers are sobering. For e(0) to reach 110:

• Death rates at every age, from every cause combined — extended out to age 150 — would need to be 88% lower than the observed death rate at age 109 in Japan in 2019. This is not a marginal improvement; it would require the complete cure or elimination of essentially every major cause of human death.
• About 70% of females would have to survive to age 100.
• More than 24% of females would have to survive to age 122.45 years — the documented maximum human lifespan, held by Jeanne Calment of France, who died in 1997 at 122 years and 164 days.
• About 6% of females would have to survive past age 150 — 28 years beyond the longest verified human life in history.

Read that last bullet again. A life expectancy of 110 isn't a tweak to the current system. It requires a meaningful fraction of the population to outlive every human who has ever been verified to exist. This is not extrapolation. It is science fiction with a spreadsheet.

Where the ceiling actually sits

So where is the real ceiling? The authors constructed a "best practice" composite using the lowest observed age- and sex-specific death rates from any country in the world for 2019. The resulting life expectancy: 88.68 years for females, 83.17 years for males. Other researchers, using different methods, have projected slightly higher figures — Canudas-Romo and colleagues estimated 91.6 years for females and 86.1 years for males by 2039 — but the gap between the two estimates is small, and both are well below 100.

Death rates below age 65, the authors argue, are unlikely to fall meaningfully further in long-lived populations — they're already so low there's little left to remove. Death rates above 65 can still decline, but the pace of decline would have to accelerate dramatically — by a factor that hasn't been seen in three decades — to deliver the 3-years-per-decade gain that defines radical life extension. That acceleration is not happening. It is, in fact, decelerating.

There is, in other words, a mortality floor: a soft limit, not a brick wall, but a limit nonetheless. It can be breached only by interventions that change the underlying biology of aging — not by more of the public-health measures that delivered the 20th-century gains.

The U.S. outlier — and what it teaches

The U.S. data are a useful cautionary tale. America spends more per capita on healthcare than any nation on Earth, yet its life expectancy has been flat or declining for most of the past decade — driven by mid-life mortality from opioids, alcoholism, suicide, obesity-related disease, and unequal access to care. The lesson the authors implicitly draw: even in the wealthiest country in history, behavior, inequality, and access can overwhelm medical technology. The path to longer life is not unidirectional. It can reverse.

Conversely, the authors note that genuine radical life extension could still occur this century — but only in today's low- and middle-income nations that have yet to fully experience the first longevity revolution. Their ceiling is far higher than the wealthy world's because they still have massive preventable mortality at younger and middle ages. The gains there will be real, but they will look like the 20th century's gains, not a new 21st-century leap.

What about gerotherapeutics?

The paper's authors are careful not to dismiss the future of geroscience. Senolytics (drugs that clear senescent "zombie" cells), rapamycin and other mTOR inhibitors, NAD+ precursors, reprogramming therapies, and a growing pipeline of "gerotherapeutics" are now in early clinical trials. The field is real and accelerating. As of late 2024, no gerotherapeutic has been approved by regulators for an anti-aging indication, and aging itself is not classified as a disease by the FDA or EMA — which complicates the regulatory pathway for any drug that tries to target aging as such.

The Olshansky paper's contribution here is a dose of empirical realism about the magnitude of the hill these therapies would have to climb. To move the needle on population-level life expectancy — not just on individual biomarkers or rodent lifespans — a gerotherapeutic would need to deliver mortality reductions comparable to eliminating an entire category of disease. The bar is not "extend mouse lifespan by 25%." The bar is "deliver, at population scale, the equivalent of curing all cardiovascular disease and all cancer and all neurodegenerative disease, simultaneously, sustainably, and cheaply."

The authors explicitly note that their limits "are not brick walls for longevity." They leave conceptual room for medicine that targets the underlying causes of aging, or for behavioral risk-factor improvements, to push mortality at older ages lower than today's best-practice composite. But they insist that any forecast claiming such gains have arrived, or are imminent, must be tested against the demographic data — and the data, so far, do not cooperate.

Why this matters beyond the laboratory

The implications cascade outward from demography into policy, finance, and personal planning.

Insurance and pensions. If actuarial forecasts assume continued linear gains in life expectancy, they will overestimate how long today's 40- and 50-year-olds will live — and underestimate the cost of annuities and life insurance while overestimating pension fund liabilities. The authors explicitly call on insurers and actuarial firms to ground forecasts in projected death-rate changes, not in linear life-expectancy extrapolations, and to require justification for any model that breaks through the documented mortality floor.

Social Security and retirement systems. Government projections of program solvency are highly sensitive to longevity assumptions. If radical life extension is not coming, the picture for Social Security and Medicare trust funds is, paradoxically, somewhat better than doomsayers claim — but only modestly. The bigger story is the number of older people, not the extreme old age of a few. Population aging is real and accelerating; radical individual life extension is not.

Healthspan vs. lifespan. The paper's framing quietly redirects attention from how long we live to how well we live. If the upper bound on lifespan is roughly fixed around 90 for women and 85 for men in long-lived populations, then the highest-leverage medical and behavioral investments are those that compress morbidity — pushing disease and disability into a shorter window at the very end of life, rather than extending the total span. This is the "rectangularization of the survival curve" that demographers have observed for decades, and it remains the most achievable, most ethical, and most economically sensible goal of aging research.

Public health vs. biology. The 20th-century longevity revolution was a public-health story. The 21st century's incremental gains — and they will be incremental — will be a biology story. The two require different investments, different timelines, and different expectations. Conflating them has produced a generation of over-promising.

What the study can't tell us

The authors are candid about limitations. Their data run only to 2019; COVID-19's longer-term demographic impact remains to be fully measured. Their populations are the longest-lived — the analysis is not representative of the global picture, where there is still substantial room for the first longevity revolution to play out. And, by their own admission, they cannot numerically estimate the future impact of as-yet-undeveloped gerotherapeutics. If someone invents a genuine aging-slowing therapy in 2035 and deploys it at population scale by 2045, all of these projections break. The authors concede this. Their argument is not that such a therapy is impossible; it is that, absent such a therapy, the demographic ceiling is real, and that nothing in the current data suggests we are anywhere close to breaching it through ordinary medical progress.

The paper also does not address — and was not designed to address — the question of maximum individual lifespan. Whether any human will ever live past Jeanne Calment's 122 is a separate question from whether life expectancy at birth will reach 110. The two are related, but the authors' focus is on the population-level statistic that drives policy, not on the outliers who capture headlines.

The bottom line

So what does all of this mean if you wanted to act on it today?

If you're an individual planning your life: Plan as though you'll likely live into your mid-to-late 80s if you're a woman in a wealthy country, or early-to-mid 80s if you're a man — with a small but real chance of reaching 100, especially if you're in Hong Kong, Japan, South Korea, or France. Do not plan as though living to 100 is a near certainty, and absolutely do not plan as though living to 110 or 120 is plausible. The math doesn't support it. Fund your retirement for a long but bounded horizon, invest in healthspan over lifespan, and recognize that the highest-impact interventions are the boring ones: don't smoke, maintain a healthy weight, exercise regularly, manage blood pressure and cholesterol, sleep, maintain social connections, and avoid opioids and excessive alcohol. None of these will make you live to 120, but collectively they dramatically improve your odds of reaching your late 80s in good shape rather than your early 70s in poor shape.

If you're a policymaker or planner: Stop using linear life-expectancy extrapolations for long-horizon forecasts. Anchor your mortality assumptions in the demographic data on life table entropy and lifespan inequality, and require explicit justification for any model that assumes death rates below the documented "best practice" floor. Plan for an aging population — meaning more older people — rather than for a population of radically longer-lived individuals. Invest in compressing morbidity: public health measures that extend healthy years are far more cost-effective than speculative anti-aging biotech that hasn't yet cleared Phase II trials.

If you're an investor in longevity biotech: The opportunity is real but the bar is exceptionally high. Population-level life expectancy impact requires interventions that deliver disease-elimination-scale mortality reductions at scale. Be skeptical of any company whose pitch depends on extrapolating rodent or small-human-trial results to population-level lifespan gains. Look instead for measurable, validated compression of morbidity — delayed onset of age-related disease, extended healthspan — which is a more achievable and more valuable target than maximal lifespan.

If you're simply trying to make sense of the noise: The next time you read that "most babies born today will live to 100" or that radical life extension is right around the corner, ask the speaker which population, which decade, and which mortality assumptions they're using. The 2024 Olshansky analysis is now the empirical baseline. Until gerotherapeutics deliver population-scale mortality reductions that rival the elimination of major disease categories — and they have not yet, anywhere — the 21st-century longevity story is one of deceleration, compression, and a ceiling that's a lot closer than the headlines suggest.

Sources

1. Olshansky, S. J., Willcox, B. J., Demetrius, L., & Beltrán-Sánchez, H. "Implausibility of radical life extension in humans in the twenty-first century." Nature Aging, vol. 4, pages 1635–1642 (7 October 2024). https://www.nature.com/articles/s43587-024-00702-3 · PDF: https://www.nature.com/articles/s43587-024-00702-3.pdf
2. Olshansky, S. J., Carnes, B. A., & Cassel, C. "In search of Methuselah: estimating the upper limits to human longevity." Science, 1990 (the original limited lifespan hypothesis paper). https://pubmed.ncbi.nlm.nih.gov/2237414
3. Human Mortality Database, University of California, Berkeley (USA), and Max Planck Institute for Demographic Research (Germany) — the underlying data source for the study. https://www.mortality.org
4. "Human Lifespan Might Be About to Hit a Ceiling, Experts Say." CNN, October 11, 2024 — reporting on the Olshansky et al. paper and its implications.
5. "Despite medical advances, life expectancy gains are slowing." EurekAlert / University of Illinois Chicago press release, October 7, 2024. https://www.eurekalert.org/news-releases/1060125
6. "Human Lifespan Limits: Are We Approaching a Biological Ceiling?" — accompanying commentary on the Olshansky et al. findings, November 2024.
7. "Human Longevity May Have Reached its Upper Limit." Press coverage of the Olshansky et al. paper, October 7, 2024.
8. Calment, J. L. — Guinness World Records verified maximum human lifespan: 122 years, 164 days (born 21 February 1875, died 4 August 1997). https://www.guinnessworldrecords.com/news/2023/2/the-oldest-person-who-ever-lived-and-those-who-came-close
9. Muscedere, J., et al. "Advancing Geroscience Research – A Scoping Review of Regulatory Environments for Gerotherapeutics." Journal of Nutrition, Health & Aging, 2025 — survey of the gerotherapeutics pipeline and regulatory landscape. https://pmc.ncbi.nlm.nih.gov/articles/PMC12310402
10. "The Future of Aging Research." Cedars-Sinai, April 11, 2025 — overview of senolytic and geroscience clinical trial progress as of late 2024. https://www.cedars-sinai.org/stories-and-insights/innovation-and-research/the-future-of-aging-research