The complaint is specific and it is consistent. A hard training week used to cost a day. Now it costs three. A late flight and two short nights used to be absorbed without comment, and now they show up in mood and output for most of the following week. Most people file this under getting older and stop there.

What has changed is rarely one thing getting worse. It is four things moving at once, each of them modest, all of them pointing the same way.

Sleep is the first. Across adulthood, the proportion of slow-wave sleep falls, total sleep time shortens, and time spent awake after falling asleep rises.¹ Slow-wave sleep is the phase most associated with physical restoration. You can spend the same eight hours in bed at 48 that you did at 35 and get materially less of the sleep that does the repairing. This is the part people most often mistake for a discipline problem, and it is covered in more detail in our piece on the sleep you are not measuring.

Muscle is the second, and it is more forgiving than most people assume. Older muscle still responds to training and to protein. A systematic review comparing young and older adults found the response becomes equivalent once exercise volume and protein dose, including its leucine content, pass a higher threshold. Below that threshold, age-related anabolic resistance becomes apparent.² So the session that used to produce adaptation can now produce mostly damage, for no other reason than that it sits under a bar that has quietly moved.

Then aerobic capacity. In a longitudinal study of relatively fit older adults measured four times across six years, it declined by around 18% per decade in women and 24% in men. Once lung function and maximum exercise heart rate were accounted for, the decline sat near 10% per decade in both sexes.³ A given effort therefore sits closer to your ceiling each year, which means it costs more to produce and more to clear afterwards.

Inflammation is the fourth, and the one that quietly links the other three. Background inflammatory tone tends to rise with age, and is pushed higher by physical inactivity, poor diet, disrupted sleep and sustained psychological load.⁴ Inflammation is part of how tissue repairs itself. Chronically elevated, it is also part of why repair stalls.

None of these is dramatic in isolation. Together they explain the thing people actually notice, which is that the input has stayed the same and the bill has gone up.

Most consumer testing answers a different question from the one worth asking. It returns a biological age: a single number, benchmarked against your birth date. It is memorable and it makes a good headline. It is also a snapshot of where you have arrived, and it says very little about how fast you are still moving.

The more useful quantity is rate. The Dunedin study followed a single birth-year cohort and tracked 19 indicators of organ-system integrity across four time points spanning two decades, which allowed researchers to model how quickly each person was actually ageing rather than simply where they had got to. That work produced DunedinPACE, a blood-based measure expressed as biological years accrued per calendar year.⁵ A result of 1.0 means you are ageing at one year per year. Above 1.0 means faster.

The distinction matters clinically. A biological age sitting two years above your chronological age describes accumulated history you cannot undo. A pace of 1.2 describes what is happening now, which is the only part still open to change.

This is not a novelty metric. In the same research programme, a faster pace was associated with morbidity, disability and mortality, and added predictive information beyond the established GrimAge clock.⁵

This is the finding that turns the measurement into something worth acting on. CALERIE was a randomised controlled trial in which 220 adults without obesity were assigned to either 25% caloric restriction or an unrestricted diet for two years. A post hoc analysis of the trial’s blood samples, meaning one run after the fact rather than the question the trial was built to answer, found that the restricted group showed a slowed pace of ageing on DunedinPACE.⁶

Two details deserve attention, and an honest reading needs both.

First, the same intervention did not significantly shift the biological age clocks at all. PhenoAge and GrimAge were unmoved. The rate changed; the accumulated-state measures did not.⁶ That is a direct argument for measuring rate rather than state, and it came from inside a randomised trial rather than from an observational association.

Second, the effect size was small, and the authors say so plainly. A small change to a rate is not a small thing once it compounds across a decade, but this is not a result that supports dramatic claims and I am not going to make one. What it establishes is the part that matters: pace of ageing is not fixed. It moved in response to an intervention.

The other reason a single number underserves you is that your systems are not travelling together. Using plasma proteins that originate in specific tissues, researchers estimated the age of 11 major organs in 5,676 adults across five independent cohorts. Close to 20% of people showed strongly accelerated ageing in one organ, while only 1.7% were ageing quickly across several at once.⁷

The consequences were organ-specific. Accelerated organ ageing carried a 20 to 50% higher mortality risk overall, and accelerated heart ageing was associated with a 250% increase in heart failure risk.⁷

A 20-year follow-up of 6,235 middle-aged adults in the Whitehall II cohort found the pattern held prospectively. Larger organ age gaps predicted increased risk across 30 diseases, with several conditions tied specifically to accelerated ageing in their own organ.⁸

For someone whose actual complaint is that recovery has slowed, this is the practical part. “You are ageing faster than you should be” is not an answer anyone can act on. Which system, and how fast, is. Autonomic recovery is one of the more accessible entry points here, and we cover its measurement in stress resilience is trainable.

This field is oversold, and the distance between what the research supports and what the marketing implies is wide enough to be worth naming directly.

Epigenetic results carry real technical variability. When researchers ran the same sample twice through six widely used epigenetic clocks, technical noise on its own produced deviations of up to nine years between replicates.⁹ Nine years. A single consumer biological-age result, read as a precise fact about you, can be substantially an artefact of the assay rather than a statement about your biology.

There are two reasonable responses to that, and we use both. Choose measures built for reliability: the DunedinPACE work deliberately excluded low-reliability probes,⁵ and principal-component versions of the older clocks now agree between replicates to within about 1.5 years.⁹ Then read any single result as a starting position rather than a verdict, because the trajectory across repeat testing carries the signal that one reading cannot.

The population-level associations are also not individual predictions. A faster pace is associated with worse outcomes across large cohorts. It does not tell any one person what will happen to them, and anyone presenting it that way has gone past the evidence.