The Elephant Problem

An adult African elephant weighs somewhere around a hundred times what you do. It is built out of cells that are much the same size as yours, so it is carrying something in the order of a hundred times as many of them, and it lives roughly as long as a person does, which means all those cells go on dividing for about as many decades.

Cancer starts when a cell copies its own instructions wrongly and the wrongness is the kind that stops it obeying orders. Every division is another roll of that dice. A hundred times the cells, rolling for sixty or seventy years, ought to mean a great deal more cancer.

In 2015 a team led by Lisa Abegglen and Joshua Schiffman at the University of Utah went through necropsy records for 644 elephants. Lifetime cancer mortality came out at 4.81 per cent. In humans the figure is somewhere between 11 and 25 per cent.

The same survey covered 36 mammal species and found no relationship between cancer mortality and body size or lifespan at all. Rock hyrax, about the size of a rabbit, 1 per cent. Lion, 2 per cent. African wild dog, 8 per cent.

Richard Peto noticed this in 1977 and it has been called Peto's paradox ever since. Half a century later it is still unexplained, and the explanation most people have heard is currently being taken apart.

What cancer actually is

The word gets used as though it named a single disease, and it does not. The mechanism is worth setting out plainly.

Every time a cell divides it copies roughly three billion letters of DNA into the new cell. The copying is extremely good and it is not perfect. Most errors land somewhere harmless, or the cell notices and repairs them, or the cell concludes it is too damaged to continue and kills itself, which is a normal and constant event in a healthy body.

Cancer is what happens when the errors land in the small number of genes that govern that whole process: the ones that tell a cell when to divide, when to stop, and when to die. A cell that has lost the brake and lost the off switch keeps dividing. Its descendants inherit the fault and add their own. Eventually there is a lump of cells growing without permission, taking blood supply, and in the cases that kill people, travelling.

The single most important brake is a gene called TP53. Its job is to sit downstream of DNA damage and make the call: repair this, or destroy the cell. It is the most frequently mutated gene in human cancers, and people born with only one working copy instead of two have Li-Fraumeni syndrome, which carries a lifetime cancer risk above 90 per cent.

One copy short and your odds collapse. That is the fact that makes the elephant finding so striking.

What the elephants have

Humans carry one TP53 gene, which is two alleles, one from each parent.

The African elephant genome contains at least twenty copies, forty alleles. Nineteen of those are retrogenes, copies that got pasted back into the genome from RNA rather than inherited as ordinary duplicates, and the 2015 team found evidence that they were being transcribed rather than sitting inert.

Then the researchers did the obvious experiment. They took white blood cells from elephants, from healthy human volunteers, and from patients with Li-Fraumeni syndrome, hit all three with ionising radiation, and counted how many cells responded by killing themselves.

Li-Fraumeni patients: 2.71 per cent. Healthy humans: 7.17 per cent. Elephants: 14.64 per cent. With doxorubicin, a chemotherapy drug, the gap was wider still, 8.10 per cent against 24.77.

The ladder maps onto the gene dosage exactly, which is the kind of result that makes a finding feel settled. An elephant cell that takes damage is far quicker to give up and die than a human cell is, and a cell that dies cannot become a tumour.

A year later Michael Sulak and Vincent Lynch at Chicago added the evolutionary half. The extra copies were not always there. They accumulated along the lineage as the animals got bigger. Asian elephants carry an estimated twelve to seventeen. Woolly and Columbian mammoths carried around fourteen. American mastodons, three to eight. The gene copies grew as the bodies did.

It is a beautiful story and it has been in every popular account of cancer biology for a decade.

The man who says hold on

In 2022 Leonard Nunney, an evolutionary biologist at Riverside, published a re-evaluation that is considerably less comfortable.

His argument is about what those nineteen retrogenes actually contain. The ancestral copy, he shows, was already broken before the duplication began: a frameshift mutation had cut it from 390 amino acids down to 157. Seventy four per cent of the copies in the modern genome are severely truncated, at 88 amino acids or fewer. These are not twenty working tumour suppressors. They are one working gene and a scatter of fragments.

He also tested how the copies accumulated over time. The pattern fits neutral processes, segmental duplication and genetic drift, at a rate of roughly one per million generations. It does not require natural selection for cancer resistance to explain it. Things that copy themselves into genomes do that anyway.

Nunney is careful about what he is not saying. Elephants almost certainly did evolve better cancer suppression, because something has to account for the numbers. His point is that the retrogenes are not the demonstrated mechanism, and that the presence of multiple TP53 copies "needs to be further justified before being used as a classic example of tumour suppression".

The original authors were more cautious than their coverage was. Their own conclusion said their findings, "if replicated", could represent an evolutionary approach to understanding cancer suppression. That qualifier did not survive contact with the headlines.

So the position in 2026 is that the paradox is real, the elephant numbers are real, the apoptosis result is real, and the tidy explanation everybody repeats is under serious challenge.

Meanwhile, at home

While that argument runs, the thing itself continues at a scale that is difficult to hold in the head.

In 2022 there were close to 20 million new cancer cases worldwide and 9.7 million deaths. Roughly one in five people will develop cancer at some point. Lung cancer was both the most diagnosed, at almost 2.5 million cases, and the biggest killer, at about 1.8 million deaths, ahead of colorectal, liver, breast and stomach.

For 2026 the American Cancer Society projects 2,114,850 new cases and 626,140 deaths in the United States alone, and notes that lung cancer will kill more Americans this year than colorectal and pancreatic cancer combined.

What has actually moved the numbers

The news here is genuinely good, and the thing doing most of the work is not what you would expect.

The American cancer death rate has been falling since 1991 and was still falling through 2023. The society's estimate is that 4.8 million deaths have been averted over that period compared with what would have happened had the rate stayed at its peak. Five-year relative survival across all cancers has reached 70 per cent for diagnoses made between 2015 and 2021, up from 63 per cent in the mid-1990s.

Some of that is treatment, and in specific places the treatment story is extraordinary. Five-year survival for myeloma went from 32 per cent to 62. For liver cancer, 7 to 22. For metastatic melanoma, 16 to 35, which is what checkpoint immunotherapy did to a disease that used to be a death sentence within months. Even metastatic lung cancer, the hardest case in the book, went from 2 per cent to 10.

But the largest single contributor to the fall is none of that. It is that people stopped smoking. In 1965, 42.6 per cent of American adults smoked cigarettes. By 2022 it was 11.6 per cent, a fall of 73 per cent, and since lung cancer is the biggest killer on the list by a wide margin, that one behavioural change did more than any molecule.

Billions have gone into oncology research and it has bought real, measurable life, particularly for people with the worst diagnoses. And it is still being outperformed by a public health campaign about cigarettes that began in the 1960s.

Back at the waterhole

The elephants in the photograph at the top of this piece are at a waterhole in the Eastern Cape. There are perhaps twenty of them, adults and calves. Between them they are carrying something like two thousand times as many cells as any one of the people who might be watching, and by the odds most of them will die of something other than cancer.

Nobody can yet tell you why with any confidence. The gene that looked like the answer has turned out to be mostly broken fragments. Something else in that animal is doing the work, and finding out what it is remains one of the more promising questions in the field, because whatever an elephant is doing, it has been doing it successfully for several million years without a hospital.

Sources

• JAMA, Abegglen and others, "Potential Mechanisms for Cancer Resistance in Elephants and Comparative Cellular Response to DNA Damage in Humans" (the survey of 644 elephant necropsies across 36 mammal species; elephant lifetime cancer mortality of 4.81 per cent against 11 to 25 per cent in humans; rock hyrax at 1 per cent, lion at 2 and African wild dog at 8; at least 20 TP53 copies and 40 alleles in African elephants against 1 copy and 2 alleles in humans, including 19 transcribed retrogenes; the apoptosis figures of 2.71, 7.17 and 14.64 per cent after ionising radiation and 8.10 against 24.77 per cent after doxorubicin; and the authors' own "if replicated" qualifier).
• eLife, Sulak and Lynch and others, "TP53 copy number expansion is associated with the evolution of increased body size and an enhanced DNA damage response in elephants" (one TP53 gene and 19 retrogenes in the African elephant, an estimated 12 to 17 copies in Asian elephants, around 14 in woolly and Columbian mammoths and 3 to 8 in American mastodons, and the association with increasing body size along the lineage).
• Leonard Nunney, "Cancer suppression and the evolution of multiple retrogene copies of TP53 in elephants: A re-evaluation" (the ancestral retrogene already truncated from 390 amino acids to 157 by a frameshift before duplication; 74 per cent of copies at 88 amino acids or fewer; accumulation fitting segmental duplication and drift at about one per million generations; the acknowledgement that elephants likely did evolve enhanced cancer suppression; and the conclusion that the retrogenes need further justification before being treated as a classic example).
• National Human Genome Research Institute, "Base Pair" (one copy of the human genome consisting of approximately 3 billion base pairs of DNA).
• International Agency for Research on Cancer, World Health Organization, "Global cancer statistics 2024: GLOBOCAN estimates of incidence and mortality worldwide" (close to 20 million new cases and 9.7 million deaths in 2022; roughly one in five people developing cancer; lung cancer at almost 2.5 million cases and about 1.8 million deaths, ahead of colorectal, liver, breast and stomach).
• CA: A Cancer Journal for Clinicians, Siegel and others, "Cancer statistics, 2026" (the projection of 2,114,850 new cases and 626,140 deaths in the United States in 2026; the mortality rate still declining through 2023 with 4.8 million deaths averted since 1991, attributed largely to smoking reductions, earlier detection and improved treatment; five-year relative survival reaching 70 per cent for 2015 to 2021 diagnoses against 63 per cent in the mid-1990s; the survival gains for myeloma from 32 to 62 per cent, liver from 7 to 22, metastatic melanoma from 16 to 35 and metastatic lung from 2 to 10; and lung cancer killing more Americans in 2026 than colorectal and pancreatic combined).
• American Lung Association, "Overall Smoking Trends" (adult cigarette smoking falling from 42.6 per cent in 1965 to 11.6 per cent in 2022, a decline of 73 per cent, drawn from the CDC National Health Interview Survey).