Weekly Issue — 2025-11-16

Each of the three studies in this piece comes from a different cohort, asks a different question, and uses a different method. What links them is a shared insistence on quantifying the things we usually wave at. Perceived stress becomes odds ratios. Sleep quality becomes machine-learning clusters. The fat woven through your thigh muscle becomes a metabolomic signature tied to how quickly your brain processes a symbol on a page. None of this is destiny. All of it is information — the kind that helps a thoughtful reader, and a thoughtful clinician, ask better questions.

A note on the strength of the evidence before we go further: these are observational studies of older adults. They identify associations, not proof of cause. The findings are consistent enough to take seriously and provisional enough to hold loosely. That is the honest register for this terrain.

The Costa Rican Longevity and Healthy Aging Study, known as CRELES, has been following older adults for years, and a 2025 analysis of 2,743 participants looked specifically at perceived stress — not cortisol on a given morning, but the kind of stress people report living with. The researchers used logistic regression to map stress against chronic disease, and Cox models to map it against mortality. The results are more interesting than a single headline number.

Stress related to the health of close relatives — the worry that comes with watching a spouse, a sibling, or an aging parent navigate illness — was associated with an increased risk of subsequent cardiovascular events and cataracts. Financial stress carried its own signature: in the same analysis, it was linked to roughly twice the risk of developing hypertension. Notably, the study did not find a statistically significant association between perceived stress and overall mortality, which is worth stating plainly. Stress, in this dataset, did not predict death directly. It predicted the diseases that often precede it.

That distinction matters. It pushes back against both the dismissive framing — stress is just in your head — and the catastrophizing framing — stress will kill you. The more useful reading is that chronic worry, especially about money and the health of people you love, appears to nudge the cardiovascular and metabolic systems in directions that accumulate over years.

The second study comes from Taiwan's Gan-Dau Healthy Longevity Plan, which enrolled 810 community-dwelling adults aged 50 and older. The researchers were interested in something the World Health Organization calls intrinsic capacity — a composite of cognitive, locomotor, vitality, psychological, and sensory function that is meant to capture how well an older person is actually doing across domains. They wanted to know how sleep maps onto it.

Rather than treating sleep as a single number, they used the Pittsburgh Sleep Quality Index and then applied unsupervised machine learning — K-means clustering — to let the data sort people into natural groupings. Four sleep patterns emerged. The cluster the researchers labeled the worst sleepers had roughly two and a half times the odds of low intrinsic capacity compared with better-sleeping peers. Higher overall PSQI scores tracked with lower intrinsic capacity as well, with particular hits to the psychological wellbeing and vitality subdomains.

What is useful here is not the implication that bad sleep is bad — we knew that — but the granularity. The analysis suggests sleep difficulty in older adults is not one problem but several, and that the pattern of disturbance may carry different functional consequences. A person who falls asleep easily but wakes at three is not the same case as one who never quite gets there. The clustering approach is a reminder that averages can hide the shape of the problem.

The third study is the most technical and, in some ways, the most provocative. Researchers working with the Health, Aging, and Body Composition Study — Health ABC — looked at intermuscular fat, the fat that infiltrates skeletal muscle and is visible on imaging but not on a bathroom scale. They wanted to understand why higher intermuscular fat keeps showing up as a correlate of slower cognitive processing speed in older adults, as measured by the Digit Symbol Substitution Test.

Working with 2,388 participants with an average age of about 75, the team measured 613 plasma metabolites using liquid chromatography–mass spectrometry. They confirmed the association between higher intermuscular fat and worse processing speed and then asked which circulating metabolites might help explain it. The framing here is biologically sensible: muscle and brain are both metabolically active organs, and the molecules that flow between them are plausible messengers.

The takeaway is not that intermuscular fat causes cognitive decline. The takeaway is that body composition in later life appears to carry metabolic signals that the brain can feel, and that two people of identical weight may have very different aging trajectories depending on what their muscle actually contains. This is part of why strength and protein intake have moved to the center of conversations about aging well — not for vanity, but because the composition of the tissue beneath the skin appears to be doing real work.

It is tempting to braid these three studies into a single tidy thesis. Resist that a little. They come from different populations — Costa Rica, Taiwan, the United States — with different measurement tools and different outcomes. What they share is a methodological seriousness about variables that have historically been treated as lifestyle filler.

The honest synthesis is this: in cohorts of older adults studied with careful statistics, perceived stress is associated with specific disease risks, sleep patterns are associated with overall functional capacity, and the composition of skeletal muscle is associated with how the brain performs a basic cognitive task. None of these associations prove causation. All of them are consistent with a broader picture in which the everyday architecture of life — what worries us, how we sleep, how we move and eat — registers in the body in ways that can be measured.

That is, in the end, the quiet promise of this kind of research. Not a miracle. Not a protocol. Just better questions, asked with better instruments, about the parts of aging we used to wave at.

The study, published in BMC Medicine, drew on 1,940 participants from the Korean Genome and Epidemiology Study and mapped their adherence to the American Heart Association's Life's Essential 8 — diet, sleep, physical activity, nicotine avoidance, BMI, blood lipids, blood glucose, and blood pressure — against five separate epigenetic-age measures, including the widely cited GrimAge2 and the DunedinPACE pace-of-aging metric. The researchers then used a statistical method called quantile-based g-computation to tease apart how much each individual pillar contributed to slower biological aging, rather than treating the eight as an undifferentiated bundle. The headline finding is unsurprising in direction but useful in precision: better cardiovascular health was associated with lower epigenetic age acceleration across every clock tested, with collective effect estimates ranging from roughly −4.29 to −0.79 years depending on which clock you trust (Lee et al., BMC Medicine, 2025).

What is genuinely new — and what makes this paper worth a longevity reader's time — is the ranking. Not every pillar pulled equal weight, and the leaderboard changed depending on the participant's sex.

In male participants, the components contributing most to slower epigenetic aging were nicotine avoidance and better glucose control. That should not be read as a license to ignore the others — the Korean team explicitly frames Life's Essential 8 as a system of interacting factors — but it does suggest that for men in this cohort, the metabolic and tobacco-exposure pillars carried disproportionate weight in the multivariate decomposition (Lee et al., 2025).

The female ranking diverged. The authors note that the dominant contributors among women differed from men's, reinforcing a point that longevity medicine has been slow to internalize: the levers that move biological-age biomarkers are not uniformly distributed across populations. A protocol optimized around a 45-year-old male executive's risk profile may underweight the very factors most predictive for his sister.

One subtlety worth dwelling on: the study measured five different epigenetic clocks, and they did not move in lockstep. Horvath's intrinsic clock, Hannum's extrinsic clock, PhenoAge, GrimAge2, and DunedinPACE each capture slightly different biology — some weight immune-cell composition, others weight mortality-associated proteins, others measure the rate at which aging is currently proceeding rather than how much has accumulated. The collective association between cardiovascular health and epigenetic age acceleration ranged across these clocks from about a fifth of a year to more than four years (Lee et al., 2025).

For readers tracking their own methylation panels, that spread is the story. A test that reports your GrimAge2 acceleration is not interchangeable with one that reports a Horvath number, and an intervention that meaningfully shifts one may barely touch another. The Korean data is a reminder to interpret a single biological-age readout with humility — and to be skeptical of any consumer service claiming a definitive number.

The evidence here is moderate, not definitive. This is a cross-sectional analysis: it captures a snapshot in time and cannot prove that improving glucose control or quitting nicotine causes the epigenetic clock to slow. The cohort is Asian and Korean-specific, which is precisely why the authors undertook it — Asian cohorts have been underrepresented in epigenetic-aging research — but it also means the relative weightings may not transfer cleanly to other populations. And quantile-based g-computation, while a thoughtful tool for decomposing joint exposures, depends on modeling assumptions that thoughtful epidemiologists will debate.

What the paper does deliver is a credible, prospectively-collected, mechanism-agnostic ranking that lets a longevity-literate reader ask sharper questions of their clinician. If you are already meeting most of Life's Essential 8 targets, where is the marginal return on further effort? For men in this cohort, the data points toward metabolic and tobacco-exposure pillars. For women, the picture differs, and clinicians should be reading the supplementary tables rather than the abstract.

The broader arc here is encouraging. Five years ago, the standard advice for slowing biological aging amounted to a vague invocation of "healthy living." Today, researchers are quantifying which components of that bundle do the heaviest lifting, in which populations, against which biomarkers. The Korean analysis is one data point in a rapidly maturing literature — moderate evidence, carefully reported, and refreshingly honest about its own boundaries (Lee et al., 2025). For readers who have spent years optimizing every pillar at once, it offers something more valuable than another protocol: a reason to ask which pillar, for you specifically, deserves the next hour of your attention.

The outward-looking study comes from a group working with UK Biobank data — a deep well of clinical measurements on hundreds of thousands of British adults. The researchers used a tool called PhenoAge, which estimates biological age from routine blood markers (things a physician already orders), then subtracts your chronological age to get what they call PhenoAgeAccel: a positive number means your body is running ahead of the calendar, a negative number means it's running behind. They ran this on roughly 156,000 people and compared each person's score with the air quality and greenspace around their home address.

The pattern was consistent in the direction you'd guess, if not always in the size you'd hope for. Higher exposure to fine particulate pollution (PM2.5 and the slightly coarser PM10) was associated with faster biological aging. More greenspace within roughly a kilometer of home was associated with the opposite. These are correlations, not causal verdicts — the authors are careful about that — but the dataset is large enough, and the markers concrete enough, that the signal is hard to wave away.

A few things are worth holding steady here. First, PhenoAge is a model, not a verdict. It's a reasonable composite of blood chemistry — inflammation markers, kidney and liver values, glucose, a few others — that tracks mortality risk well in large populations. It does not tell any individual man how long he will live. Second, the effect sizes the team report are modest at the individual level. They become interesting because they show up across a very large group and in the directions biology would predict: dirtier air, faster apparent aging; more green around you, slower.

Third — and this is the part I found most useful — the team did subgroup analyses and reported that non-smokers, former smokers, people who drink, those carrying extra weight, and women appeared more sensitive to the air-and-greenspace exposures. That's a counterintuitive finding worth flagging rather than overselling. It may mean that once smoking dominates the signal, environmental exposures get harder to detect on top of it. It does not mean that current smokers are somehow protected from polluted air. The honest read is that the modifiers of this relationship are still being worked out.

The second study takes the opposite vantage point. A team measured three kinds of repeating DNA sequences in blood leukocytes from 535 people aged 5 to 101: ribosomal repeats (the genes that build the cell's protein factories), a stretch called satellite III on chromosome 1q12, and the telomere repeats at the tips of our chromosomes that have become a kind of folk shorthand for aging.

The interesting group in this work is the centenarians — 106 people aged 90 to 101. Compared with younger groups, they stood out in three ways. Their ribosomal repeat content sat in a notably narrower band — less variation, person to person. Their satellite III content was higher. And, somewhat against the folk version of the telomere story, their telomere repeat content was lower. The authors also report a negative correlation between satellite III and telomere content, and propose two composite parameters (S/T and S/(R*T)) that rise with age and reach their highest values in the oldest cohort.

What to make of it. This is a cross-sectional snapshot — different people at different ages, not the same people followed for decades — so it cannot tell us whether the centenarians arrived at this pattern because of how they aged, or because they were built that way from the start. The sample of very old people is modest. And the measurement technique (non-radioactive quantitative hybridization) is well-established but specialized; replication in other labs and other populations is the next thing to watch for. What the authors are proposing, carefully, is that these parameters may turn out to be useful predictors of life expectancy in late life. That is a hypothesis worth tracking, not a clinical tool to ask your doctor about next week.

For years, the longevity field has had a credibility problem: too many single-mechanism stories, too many supplements sold on the back of a mouse experiment. What's quietly changing is the shape of the evidence. PhenoAge is a population-scale tool built from boring blood markers your physician already understands. The repeat-content work is a population-scale tool built from a few specific, measurable features of the genome. Neither asks you to believe in a miracle. Both are trying to give "biological age" a number you could, in principle, audit.

Read together, they sketch a sensible frame. The outside world — the air on your street, the trees in your line of sight — appears to leave a measurable trace on the blood-chemistry version of biological age. The inside world — the architecture of your own DNA — appears to carry signatures that distinguish people who reach very old age from the rest of us. Both signals are real enough to take seriously and modest enough to keep in proportion.

The honest summary, for a man my age and yours, is this. We are getting better at measuring how we are aging, and the measurements are starting to point at two places at once: the environment we choose (or are stuck with) and the genome we were issued. Neither of these papers tells you to buy anything. Both suggest that the smaller, duller decisions — where you walk, how clean the air is on that walk, what your standard bloodwork actually says — are quietly the ones the science keeps circling back to. That is not a thrilling headline. It happens to be a durable one.