Different tissues run on different clocks. Your processing speed peaks before you can legally rent a car. Your vocabulary keeps climbing into your sixties. Your bones stop gaining at thirty and your aerobic ceiling starts falling the decade before that.
This atlas maps those clocks side by side — and shows, with the actual evidence, which ones you can bend.
Move a single dial and watch six independent biological clocks respond. These are modelled population averages — smoothed curves drawn from reference data, not a prediction about any individual. The point is the shape of each curve, and how little the shapes resemble one another.
DNA polymerase cannot finish the job. Every time a linear chromosome is copied, the lagging strand loses the last stretch it had no room to prime — the end-replication problem. Cells solve it by putting something expendable at the tip.
That expendable thing is the telomere: a tandem repeat of the six-letter sequence TTAGGG, roughly 10,000–15,000 bases long at birth, capped by a six-protein complex called shelterin that folds the end into a loop so the cell does not read it as a double-strand break.10
Each division shaves roughly 20–50 base pairs off the tip. After somewhere around 40–60 divisions the cap gets short enough that shelterin can no longer hide it, the DNA-damage response fires, and the cell exits the cycle permanently. Leonard Hayflick described this ceiling in 1961, decades before anyone knew what was doing the counting.11
Senescent cells do not quietly disappear. They stay metabolically active and secrete a cocktail of inflammatory cytokines, proteases and growth factors — the senescence-associated secretory phenotype — which pushes neighbouring cells toward the same fate. This is one of the main routes by which a cellular event becomes a whole-organism one.12
Longer is not simply better. Mendelian randomisation using inherited telomere-length variants finds that genetically longer telomeres associate with lower risk of coronary disease — and higher risk of several cancers, including glioma, melanoma and lung adenocarcinoma.13 Telomerase, the enzyme that rebuilds caps, is reactivated in roughly 90% of human cancers. Meanwhile, measured leukocyte telomere length is a poor predictor at the level of any individual person: between-person variation swamps the age signal, and direct-to-consumer telomere tests have no validated clinical action attached to them.14 Treat telomeres as a beautiful mechanism, not a dashboard gauge.
The 2023 revision of the hallmarks framework lists twelve interlocking processes. Fixating on the telomere is like blaming a building collapse on one bolt.15
Exercise is the only single intervention that has been shown to touch a majority of these twelve — mitochondrial biogenesis, autophagy, nutrient sensing, inflammation, stem-cell niche and intercellular signalling all respond to training. No pill currently does that.16
The dose–response curve for physical activity and death is not a straight line. It is steep at the start and then flattens, which means the most valuable exercise you will ever do is the exercise you do when you are currently doing none.
Curve fitted to the pooled-cohort dose–response of Arem et al. (n = 661,137).17 Strength offset from a meta-analysis of muscle-strengthening activity and mortality.18 Observational data: association, not proof of causation, though the gradient is steep, consistent and biologically coherent.
Same weekly minutes, different jobs. What training is for changes roughly every twenty years, because the thing most at risk changes.
Ask when cognition peaks and the honest answer is: which cognition? Raw processing speed tops out before most people finish secondary school. The ability to read emotion in a stranger's face peaks around fifty. Vocabulary is still climbing at seventy. These are not a single curve with noise on it — they are genuinely different curves.
An infant brain forms synapses at a rate that would be catastrophic if it continued — estimates run to a million new connections per second in the first years of life. What follows is not more building but competitive elimination: connections that carry traffic are kept, connections that do not are removed. Experience does not add wiring so much as it selects wiring.
Grey-matter volume peaks in late childhood and then declines through adolescence, which sounds like damage and is actually maturation. The prefrontal cortex — the region running inhibition, planning and consequence-modelling — is the last to finish, and it finishes in the mid-twenties. This is the neuroanatomical footnote under every conversation about adolescent judgement.
Two implications sit underneath the whole rest of this section. First, the brain is a use-dependent organ across the entire lifespan, not just in childhood. Second, the enormous developmental reserve built early — often measured crudely as years of education — is the single strongest buffer against symptomatic decline later. Pathology and symptoms are not the same thing.24
Cross-sectional estimates from large online samples and standardised norms. Dots mark the estimated peak; bars show the plateau where performance is within a few percent of peak.8,25
Total brain volume falls roughly 0.2% per year from the mid-thirties, accelerating past 0.5% per year after sixty. The loss is not uniform: prefrontal cortex and hippocampus go first and fastest, sensory and motor cortex are relatively spared, and white-matter integrity degrades in a pattern that mirrors the reverse of the order in which it myelinated — last in, first out.23,26
Functionally that produces a signature: slower processing, more effortful task-switching, more vulnerability to interference, and a harder time encoding new episodic memories. What it does not produce in normal aging is loss of accumulated knowledge, vocabulary, semantic structure or professional expertise. Those keep working, and in many people keep improving.
Here is the genuinely good news that rarely makes headlines: age-specific dementia incidence has been falling in high-income countries — roughly 13% per decade across pooled European and US cohorts, about 20% per decade in Framingham. Absolute case numbers still rise because populations are aging, but an individual's risk at a given age is lower than it was for their grandparents. Something in the last fifty years — better blood pressure control, more education, less smoking — has been working.27
Graded by the strength of human evidence for a cognitive endpoint — not by mechanism, plausibility or enthusiasm.
Intensive systolic targets reduced mild cognitive impairment by 19% in SPRINT MIND. The most reliably brain-protective thing in a primary care clinic is a well-controlled blood pressure.28
One year of moderate aerobic training increased hippocampal volume ~2% in older adults — reversing roughly one to two years of expected loss — with corresponding memory gains. Effects on global cognition in meta-analysis are real but modest.29
FINGER (diet + exercise + cognitive training + vascular care) improved cognition versus control in Finland. The US replication, POINTER, reported improvement in both arms with a modest advantage to the structured one — encouraging, and a reminder that attention itself does something.30
Hearing loss is the largest single modifiable dementia risk factor by population attributable fraction. The ACHIEVE trial found no overall benefit but a ~48% slowing of decline in the higher-risk cohort — a pre-specified subgroup, so promising rather than settled.31
Persistent short sleep (≤6 h) in midlife associates with roughly 30% higher dementia risk over 25 years, and slow-wave sleep drives glymphatic clearance of interstitial solutes. Observational, mechanistically compelling, not yet an RCT.32
Trained tasks improve. Transfer to untrained abilities and to everyday function is small and inconsistent, and the field carries a well-documented publication-bias problem. Learning a genuinely hard new skill is a better bet than a game that gets easier.33
Dietary restriction is the strongest known modifiable predictor of eating-disorder onset, and eating disorders carry among the highest mortality of any psychiatric illness. Longevity culture and a developing restrictive disorder can look identical from the outside — both involve tracking, rules, discipline and an explicitly health-framed rationale. The difference is not in the behaviour. It is in whether the behaviour is still being chosen.34
Signals that restriction has stopped being a health behaviour:
None of these is a diagnosis, and this page cannot make one. If several are recognisable — in yourself or someone you know — that is a reason to talk to a clinician, not a reason to research harder.
National Alliance for Eating Disorders Helpline · 1-866-662-1235 · staffed by licensed clinicians
Nothing below is a protocol. No target intakes, fasting windows or step-by-step instructions appear on this page by design. Restriction interventions used in research are conducted with medical screening, monitoring and nutritional supervision, and that context is part of the intervention rather than an optional extra.
With that in place: what does the human evidence actually show? Less than the popular literature implies, and it is more interesting than the headline.
Grading here reflects the strength of human evidence for a clinical endpoint. Not mechanism. Not mouse lifespan. Not biomarker movement. A compound can be mechanistically beautiful, extend life in three model organisms, and still sit in the bottom tier — because that is where the human data puts it.
Context worth holding: under the 1994 US Dietary Supplement Health and Education Act, supplements need no pre-market demonstration of efficacy, and manufacturers — not the FDA — are responsible for verifying identity and purity. Herbal and dietary supplements now account for a substantial and rising share of drug-induced liver injury in US referral centres.39 "Natural" is a marketing category, not a safety category.
Two things move at once across a life: fat mass rises and lean mass falls. Because they partly cancel, a person's weight can stay flat for thirty years while their body composition changes completely. This is why the scale is such a poor instrument and why the two sexes need entirely separate charts.
The gap is not cultural. Essential fat — structural lipid in nerve tissue, marrow, organ membranes and, in females, sex-specific reproductive stores — runs about 2–5% of body mass in men and roughly 10–13% in women. Everything above that is storage fat, and the two curves are offset by that essential floor from puberty onward. A woman at 25% body fat and a man at 15% are in comparable physiological positions.
Then the curves both drift upward with age, and the drift understates what is happening. Fat migrates from subcutaneous to visceral and intramuscular depots, and visceral fat is the metabolically consequential kind. A person can hold the same weight and the same body fat percentage from forty to sixty while their visceral fat and their liver fat both rise substantially.
The shaded bands are published healthy reference ranges; the lines are what people actually measure. Note that the typical female curve sits above the healthy band for most of adult life and the typical male curve sits above it too. That is a population-level observation about the last fifty years of the food and movement environment, not a comment on any individual.6
Body fat percentage carries substantial measurement error, and the error is bigger than the differences people chase. Bioimpedance scales vary by several percentage points with hydration alone. Skinfold calipers depend heavily on the operator. Even DEXA — the practical reference standard — has a precision error of roughly 1–2 percentage points and gives systematically different numbers between manufacturers, which means a change of two points between two different machines is not a change.
More importantly: these are population distributions, not goals. There is no body fat percentage that is correct for a person, no threshold at which health switches on, and no evidence that pursuing a specific number produces better outcomes than pursuing fitness, strength and dietary quality directly. The measurable health signal in this whole domain comes overwhelmingly from cardiorespiratory fitness, muscle mass and metabolic markers, all of which can be improved without ever weighing anything.
If tracking body composition tends to become preoccupying rather than informative, that is a good reason not to track it. The training works the same either way.
| Descriptive band | Female | Male | Note |
|---|---|---|---|
| Essential | 10–13% | 2–5% | Structural minimum. Below this is pathology, not leanness. |
| Athletic range | 14–20% | 6–13% | Sport-specific; sustained low values disrupt hormonal function in both sexes. |
| Fitness range | 21–24% | 14–17% | Typical of consistently active adults. |
| Healthy reference | 21–33% (20–39 y) 24–36% (60–79 y) | 8–20% (20–39 y) 13–25% (60–79 y) | Age-adjusted ranges from cross-sectional reference data. |
| Above reference | >39% | >26% | Associated with metabolic risk at population level; a single number is not a diagnosis. |
Bands vary between organisations and reference populations, and none of them were derived to be used as personal targets. Ranges shown after Gallagher et al. and ACE descriptive categories.6
Dual-energy X-ray absorptiometry fires two X-ray beams at different energies through the body. Bone and soft tissue attenuate the two energies differently, and that difference is solved algebraically to give areal bone mineral density in grams per square centimetre. The radiation dose is roughly a tenth of a chest X-ray — less than a transatlantic flight's worth of cosmic exposure.
Standard deviations away from the mean peak bone mass of a healthy young-adult reference population. This is the score used for diagnosis in postmenopausal women and men over 50, because the clinical question is how far you have fallen from peak.
Standard deviations from the mean for people of the same age, sex and often ethnicity. Used in premenopausal women, men under 50 and children. A Z-score of −2.0 or below is reported as "below the expected range for age" and should prompt a search for a secondary cause rather than a diagnosis of osteoporosis.
Lumbar spine (L1–L4), femoral neck and total hip, with the forearm substituted when the others are unusable. The lowest of the valid sites drives the diagnosis. Spine readings are frequently falsely elevated in older adults by degenerative facet disease, aortic calcification, scoliosis or old compression fractures — which is why the hip usually carries more weight after seventy.
It measures density, not architecture. Two people with identical T-scores can have very different trabecular structure and very different bone strength. Trabecular bone score adds some of this back from the same scan; high-resolution peripheral QCT does it properly but is a research tool. This is a real limitation, not a footnote — most fragility fractures occur in people whose T-scores are not in the osteoporotic range, simply because far more people sit in the osteopenic range.53
US Preventive Services Task Force: all women 65 and older, and younger postmenopausal women whose fracture risk equals that of a 65-year-old. For men, the USPSTF finds evidence insufficient; specialty societies commonly suggest 70, or earlier with risk factors. Anyone with a fragility fracture after 50, long-term glucocorticoids, or a relevant endocrine or malabsorptive condition should be scanned regardless of age.54
The FRAX tool combines BMD with age, sex, prior fracture, parental hip fracture, smoking, glucocorticoids, rheumatoid arthritis, alcohol and BMI to give a ten-year probability of major osteoporotic and hip fracture. US treatment thresholds commonly sit at 20% for major osteoporotic fracture or 3% for hip. A T-score alone is an input, not a decision.55
Roughly one in two women and one in four to five men over 50 will sustain an osteoporotic fracture in their remaining lifetime. Incidence rises approximately exponentially with age, and hip fracture is the one that changes a life: excess mortality in the first year runs around 20–30%, and a substantial minority of those who survive never return to their previous level of independence.56
The federal guideline is 150–300 minutes of moderate aerobic activity per week plus muscle-strengthening on two or more days. In 2024, 26.4% of US adults met both. The aerobic half alone is met by roughly half the population; it is the strength half that most people miss.2
Self-reported leisure-time activity, National Health Interview Survey. Self-report inflates activity relative to accelerometry — device-measured adherence is meaningfully lower — so read these as an optimistic ceiling rather than a floor.58
The trend is genuinely improving. Strength-training participation in particular has roughly doubled since the early 2000s, and the increase shows up across age groups including adults over 65. That is a real public health gain and it rarely gets reported, because "slowly getting better" is not a headline.
But the distribution is the story. Adherence tracks education and income steeply — from roughly 12% among adults with a high school education or less to about 34% among those with a bachelor's degree or higher — and it tracks geography, with rural adults substantially less likely to meet either guideline than adults in large metropolitan areas.9,59 Physical activity has become another axis along which health advantage compounds.
And the flip side of the exercise-mortality curve applies here as a population fact: with roughly three-quarters of adults not meeting both guidelines, the largest single lever on American healthspan is not a drug, a supplement or a diet. It is moving a substantial fraction of that three-quarters into the steepest part of the dose–response curve — the part between nothing and something.
Five things you can measure with a chair, a wall and a clock. Each is here because it has published age norms and independent links to function or survival — not because it looks impressive. Compare yourself to typical values for your age, note the lowest one, and train that.
US life expectancy at birth was about 47 years in 1900 and about 79 in 2024 — the highest figure ever recorded, reached only after a pandemic knocked it back to 1996 levels.1,64 Thirty-two added years sounds like the triumph of medicine. Most of it was not medicine.
There is no such thing as American life expectancy. Partitioning the population by race, geography, income and urbanicity yields ten distinct "Americas" whose life expectancies differed by 12.6 years in 2000 and 20.4 years by 2021 — a gap that widened steadily through two decades of economic growth and then accelerated during the pandemic.3
At county level the range in 2019 was 27.2 years. Break it down by county and race simultaneously and it reaches 36.3 years — the distance between contemporary Switzerland and a country in the middle of a humanitarian emergency, inside one national border.68
Internationally the US is the anomaly its spending does not predict. Japan, Switzerland, Spain, Italy, Australia and South Korea all sit near or above 83–85 years. The US spends more per capita on health care than any of them and ranks well outside the top thirty. The gap is not explained by hospitals: it is smoking history, obesity and the food environment, drug overdose, firearms, road deaths, and the absence of the income and social supports that buffer all of the above.
Lifespan is how long you live. Healthspan is how long you live in good health. They have not been growing at the same rate — globally the gap runs close to a decade, and the United States has the widest gap of any large country, on the order of twelve years of life lived with significant morbidity.69
James Fries proposed in 1980 that the goal should be compression of morbidity — pushing the onset of disability closer to the end of life rather than simply extending the end of life.70 Forty-five years later, in the US, the opposite has been happening. Which reframes the whole project: the interesting target was never the last number on the chart. It was the shape of the years before it.
The demographic case for caution is strong. Analysis of the eight longest-lived populations plus Hong Kong and the US found that improvements in life expectancy have been decelerating since 1990, and that on current trajectories fewer than 15% of women and 5% of men in those populations will reach 100. Radical life extension this century, the authors argue, is implausible without an intervention that alters the biology of aging itself rather than treating diseases one at a time.71
That last clause is exactly the geroscience wager: that aging is the shared upstream driver of cancer, dementia, cardiovascular and metabolic disease, and that slowing it would compress all of them at once — the "longevity dividend". It is a coherent hypothesis with real preclinical support and, as of today, no human trial that has demonstrated it.
The near-term realistic gains are less romantic and much larger. Tobacco, obesity and metabolic disease, overdose, firearms, road safety and physical inactivity account for most of the distance between the US and its peer nations. Moving those does not require new biology — it requires policy, environment and behaviour at population scale, which is harder than biology in a different way.
The plausible exception on the horizon is metabolic: incretin-based therapies are producing weight and cardiovascular effect sizes not previously seen from a pill, with signals across kidney disease, heart failure, sleep apnoea and possibly addiction. If anything shifts the US curve in the next twenty years it is more likely to be that than a senolytic.
And underneath all of it, the least glamorous finding in this entire atlas keeps holding: the intervention with the largest effect size, the widest range of affected systems, the best safety profile and the lowest price is still the one that requires you to be slightly out of breath a few times a week.
Primary literature where possible. Where a claim rests on observational data it is labelled as such in the text. Where the evidence is contested, both sides are noted rather than the convenient one.