Nobody Actually Dies From Old Age? Autopsy Studies Reveal What Really Kills Us

The ‘Hallmarks of Aging’ Framework Has A Major Problem That Nobody Talked About Until Now

The next time a death certificate lists “natural causes” or a doctor mentions someone died of “old age,” keep in mind that may not be true after all. Autopsy studies reveal that even centenarians who seemed healthy days before death succumbed to specific, identifiable diseases, not some vague process called aging.

Scientists Maryam Keshavarz and Dan Ehninger from the German Center for Neurodegenerative Diseases argue in a new review that this fundamental misunderstanding has led aging research astray for decades. Their systematic analysis, published in Genomic Psychiatry, suggests that much of what scientists thought they knew about measuring and manipulating aging may be built on flawed assumptions.

When researchers examined 2,410 autopsies, cardiovascular disease emerged as the culprit in the vast majority of deaths, including myocardial infarction (heart attacks, 39%), cardiopulmonary failure (38%), and cerebrovascular lesions (strokes, 17.9%). A study of people over 85 who died unexpectedly outside hospitals found cardiovascular events responsible for 77% of deaths. Even among centenarians perceived as healthy, autopsies revealed that 68% died from cardiovascular causes, 25% from respiratory failure, and smaller percentages from other specific organ failures. Zero died from “old age.”

Autopsy analyses are crucial for correcting misperceptions, the researchers write, noting that relatives and even physicians frequently misjudge causes of death without examining the body.

Why Each Species Dies Differently

The pattern of disease-driven mortality extends across the animal kingdom, but with a crucial twist: each species has its own vulnerable spot. While cardiovascular disease dominates in humans and nonhuman primates, causing over 60% of deaths in aging rhesus macaques, rodents tell an entirely different story.

Cancer killed between 84% and 89% of mice in one major study under normal feeding conditions. Dietary restriction reduced this to 64% in both sexes, while rapamycin treatment lowered it to 74%. Even with interventions widely considered to slow aging, cancer remained the primary cause of death in mice, just occurring at later ages. In rats, tumors accounted for roughly 63% of deaths. Dogs showed a similar pattern, with neoplasia causing nearly half of all deaths in older animals.

Move further down the evolutionary tree and the causes shift again. Fruit flies die primarily from intestinal epithelial failure, a breakdown of the gut lining that leads to bacterial infections. Roundworms succumb to pharyngeal infections or throat muscle atrophy. Each species has a different Achilles’ heel.

This creates a problem for aging research. When a drug extends mouse lifespan by preventing cancer, has it slowed aging or merely delayed one particular disease? And if humans die primarily from cardiovascular problems rather than cancer, why should we expect mouse cancer-prevention strategies to help us live longer?

The Influential Framework Built on Questionable Evidence

The concerns grow when Keshavarz and Ehninger turn their attention to one of aging research’s most influential concepts: the “hallmarks of aging.” Introduced in 2013 as nine hallmarks and expanded to 12 in a 2023 update, this framework identifies cellular and molecular changes (from genomic instability to cellular senescence to mitochondrial dysfunction) that supposedly drive the aging process.

The hallmarks papers have been cited thousands of times and shaped research priorities across the field. Yet when the researchers systematically examined the evidence supporting each hallmark’s causal role in aging, they uncovered a widespread methodological flaw.

Between 57% and 100% of the studies supporting different hallmarks tested interventions only in old animals. This design makes it impossible to tell whether treatments genuinely slow aging or simply improve function at any age.

The distinction is critical. Imagine a drug that enhances memory equally in both young and old mice. That’s valuable, but it’s not evidence the drug slows cognitive aging. It just means the drug improves brain function generally. To claim an intervention slows aging, researchers need to show it changes the rate at which age-dependent decline occurs, not just shifts everyone’s performance upward.

When Keshavarz and Ehninger identified studies that did include young animals, they analyzed 602 age-sensitive traits across those studies. They found that 436 of these traits (72%) showed effects in both young and old age groups. The interventions were producing what the researchers call baseline effects (age-independent physiological changes) rather than genuine alterations to aging trajectories.

What Rapamycin and Fasting Studies Actually Show

The issue extends even to aging research’s success stories. Rapamycin consistently extends lifespan in mice and has become one of the field’s most studied compounds. Intermittent fasting reliably prolongs life in rodents. Genetic manipulations that reduce growth hormone signaling add months to mouse lifespans.

Yet deep phenotyping studies that measured dozens of age-sensitive traits revealed something unexpected. Most effects of these interventions appeared as baseline shifts rather than changes in aging rate. Mice showed similar improvements whether treated young (before age-related changes emerged) or old (after such changes had manifested).

A 2022 study by Xie and colleagues examined three major longevity interventions across 180 age-sensitive traits in mice. They found that 145 traits (roughly 81%) showed intervention effects in young animals equal to or greater than those in old animals. The interventions were altering physiology broadly, not specifically targeting the processes that drive age-dependent change.

Rapamycin-treated mice still died primarily from cancer, just at older ages. Intermittent fasting didn’t eliminate the diseases of aging—it postponed them. The interventions worked more like hitting a snooze button on mortality than fundamentally rewinding the aging clock.

DNA Age Tests May Not Measure What You Think

The problems extend to tools that have become wildly popular both in research and consumer markets: biological age clocks. These algorithms, particularly those based on DNA methylation patterns (chemical modifications to DNA), can predict chronological age with impressive accuracy and have spawned a cottage industry of longevity testing.

But Keshavarz and Ehninger argue these clocks share a fundamental limitation with much of aging research. They’re correlational tools that identify patterns associated with age without revealing causal mechanisms. The comparison to facial aging is apt: wrinkles predict age reliably, but treating wrinkles doesn’t make you biologically younger.

Most DNA methylation clocks rely on a surprisingly small number of genomic sites. Studies show they maintain accuracy even when many sites are removed, suggesting the remaining sites capture redundant information rather than providing comprehensive biological insight. A 2024 study using Mendelian randomization (a technique that can help distinguish correlation from causation) found that traditional aging clocks aren’t significantly enriched for sites with causal roles in aging.

The clocks remain useful for predicting health outcomes and stratifying people by risk. But they may not tell researchers whether an intervention truly slows biological aging or just shifts biomarker values in age-independent ways.

What Aging Research Needs to Change

Keshavarz and Ehninger don’t argue that aging research has been worthless. The field has identified numerous interventions that extend lifespan in model organisms and accumulated vast amounts of data about age-related changes. The problem is interpretation.

A drug that prevents cancer in mice is valuable even if it doesn’t slow aging per se. A biomarker that predicts mortality risk helps even if it doesn’t reveal aging mechanisms. But claiming these as evidence of understanding or manipulating aging itself requires more rigorous proof.

The researchers propose concrete refinements. Studies should measure diverse outcomes across multiple organ systems rather than relying on lifespan or a handful of favorite biomarkers. Interventions should be tested in both young and old subjects to distinguish rate effects from baseline effects. Researchers should explicitly acknowledge species-specific life-limiting pathologies rather than assuming mouse results translate directly to humans.

Most provocatively, they suggest the field needs to confront an uncomfortable possibility: despite decades of research and thousands of papers on aging, scientists may still lack a deep mechanistic understanding of what drives age-dependent decline. They’ve catalogued correlates of aging, identified factors that extend lifespan by postponing specific diseases, and developed predictive biomarkers. But understanding the underlying process that makes organisms more vulnerable over time may remain elusive.

This matters beyond academic debates. As longevity interventions move from mice to humans, the distinction between preventing specific diseases and genuinely slowing aging determines which treatments to pursue, how to test them, and what outcomes to measure. Getting it wrong wastes resources and, more importantly, delays finding interventions that might actually work.

Aging research aims to identify the underlying drivers of biological changes across different levels, the authors write. But first, the field needs to ensure its tools and concepts actually measure what researchers think they’re measuring. Otherwise, aging science risks building an increasingly elaborate structure on a foundation that was never as solid as it seemed.

---