Why do we age?

A mutation that causes harm only in old age is barely selected against, because few individuals live to express it. Peter Medawar put that in 1952 and William Hamilton formalised it in 1966: the force of natural selection typically declines with age.

So harmful late-acting mutations can accumulate in a genome. George Williams added the sharper version in 1957, that a gene helping early can be favoured even if it costs later. Ageing on either account is a side effect of the shape of selection rather than a purpose.

Williams also predicted that species safer from outside dangers should evolve longer lives. Formal theory has since shown that only dangers falling unevenly across ages can do that, but the pattern he predicted is there. Across 1,368 birds and mammals, and over and above body mass, the single strongest predictor of a long life is the ability to fly. Among the species that cannot fly, living in trees or underground goes with longer lives.

Repair is expensive: DNA repair, antioxidant defences and the rest of the body’s maintenance all run on energy. An organism has finite resources to split between reproduction and maintenance.

Investing in a body that will be dead within a year is wasted, so selection tunes maintenance to roughly the lifespan the environment allows. Ageing follows from under-investment that was economically correct in the wild. This is Tom Kirkwood’s disposable soma theory, put forward in 1977 and one of the standard evolutionary theories of ageing.

A 2013 review catalogued the processes that change consistently with age: genomic instability, telomere shortening, epigenetic drift, loss of protein quality control, altered nutrient sensing, mitochondrial dysfunction, cellular senescence, stem cell exhaustion and altered intercellular communication.

The same authors revisited it ten years later and the list came back longer, at twelve, with failures of cellular recycling, chronic inflammation and the composition of the gut microbiome added.

A longer catalogue is not a mechanism. Some of these are causes, some are consequences, and some are both, and separating them requires interventions that have mostly been done in short-lived laboratory animals.

Some animals show no increase in mortality with age over the periods they have been studied. The desert tortoise’s risk of dying falls with age right up to the age by which all but five per cent of adults have died. Hydra’s stays so flat that five per cent of a laboratory cohort would still be alive after fourteen hundred years. In a long-studied captive colony, the naked mole rat’s does not rise even at twenty-five times its age at maturity, a claim that was challenged in print, defended with the same data left-censored, and upheld by its authors in 2024 with twice the data.

This is evidence that ageing is not an inevitable consequence of being made of cells. It is also a problem for the theory: the evolutionary account explains rising mortality after maturity, and across forty-six species the trajectories come out increasing, constant, decreasing, humped and bowed. What those species do differently is only partly known, and is one argument that the process is modifiable in principle.

The interventions with the strongest case, a top tier that includes rapamycin, senolytics, metformin, acarbose, spermidine, NAD+ enhancers and lithium, rest mainly on animal studies, with some clinical results. A review that ranks them by the robustness of that evidence then turns to whether clinical trials of them are feasible. An experiment with lifespan as its outcome would take a human lifetime.

The measurement problem is as hard as the biology: there is no agreed way yet to validate a marker of biological age, so a trial has to wait for the frailty, illness and deaths that ageing brings.