The most arresting result in a vast new strength-training analysis was not the overall longevity figure. It was the number buried in the neurological column.

Researchers followed 147,374 US health professionals across three cohorts for up to 30 years. Those averaging 90 to 119 minutes of resistance training a week had a 27% lower relative hazard of dying from neurological disease than participants reporting none, even after aerobic activity and other measured lifestyle differences were taken into account.

That is a substantial association, and a more specific angle on the study than the broader longevity analysis ScienceBlog published earlier. It is also a result that needs careful handling. This was an observational study, not an experiment, and it measured deaths assigned to a neurological cause. It did not prove that two hours in the gym prevents dementia.

Where the 27% figure came from

The study, published in the British Journal of Sports Medicine, combined 31,540 men from the Health Professionals Follow-up Study with 115,834 women from the Nurses’ Health Study and Nurses’ Health Study II. The median age at the analytical baseline was 54.

The longest-running cohort contributed up to three decades of follow-up. Across all three, researchers documented 35,798 deaths. Of those, 4,635 were classified as deaths from neurological disease.

Exercise was not measured once and left frozen in time. Participants repeatedly reported how long they spent using weight machines or doing resistance training, every two to four years depending on the cohort. The researchers calculated cumulative averages, so the categories were intended to represent long-term habits rather than one unusually energetic month.

At 90 to 119 minutes a week, the hazard ratio for neurological mortality was 0.73, with a 95% confidence interval from 0.58 to 0.92. In plain language, the rate of neurological death during follow-up was estimated to be 27% lower than in the no-training group.

Relative hazard is not personal probability

A 27% lower hazard is easy to hear as a 27-percentage-point reduction in everyone’s chance of dying from neurological disease. That is not what the statistic says.

A hazard ratio compares how quickly an outcome occurs in two groups over time. The absolute difference for any one person depends on age, baseline neurological risk, length of follow-up and competing causes of death. Nor does the number mean that 27% of neurological deaths would disappear if everybody began lifting.

The statistical model adjusted for age, smoking, alcohol intake, diet, family history and aerobic exercise, among other factors. A further model that included body mass index produced a very similar estimate. Adjustment makes the comparison fairer, but it can account only for factors the researchers measured well enough to include.

The people who trained more were different in several visible ways. They tended to be younger, leaner, less likely to smoke, more aerobically active and more likely to report a healthier diet. Unmeasured differences in health, income, healthcare access or daily function could still contribute to the result.

What counted as a neurological death

Deaths were identified through family reports, autopsy records and the National Death Index. Study physicians reviewed the available medical records and death certificates without knowing participants’ exercise histories, then assigned the underlying cause.

The neurological category was built from selected ICD-9 codes covering dementia and several other conditions affecting the nervous system. Most of the neurological deaths in these cohorts were attributed to dementia, according to the authors.

That makes the endpoint important but narrow. The study did not count every new dementia diagnosis, measure changes in memory, or track whether people developed mild cognitive impairment. Someone may live with dementia for years and ultimately have pneumonia, heart disease or another condition recorded as the underlying cause of death.

Death certificates also under-detect dementia and can misclassify neurological causes. The researchers explicitly flagged that limitation. So the 4,635 neurological deaths were not a perfect census of every fatal contribution made by brain disease.

The hardest alternative explanation runs backward

Neurological disease creates an unusually stubborn problem for exercise research: the arrow of cause and effect may point both ways.

Alzheimer’s disease and other neurodegenerative conditions can have long preclinical phases. Movement, motivation and everyday activity may begin declining years before a formal diagnosis. Low resistance training could therefore be an early consequence of disease rather than a cause of later mortality.

The team tried to reduce this reverse-causation problem. People with cancer, heart disease or stroke at baseline were excluded, and exercise measurements were separated from the mortality period by two years. More importantly for neurological disease, the researchers reran the analyses using delays of eight and 12 years. The inverse association remained.

That persistence makes a simple “people stopped lifting because they were already sick” explanation less satisfying. It cannot eliminate reverse causation completely. A decades-long disease process, imperfect diagnosis and unmeasured changes in function are difficult to remove completely from an observational dataset.

Why two hours is a landmark, not a brain prescription

The neurological results formed a J-shaped curve. Relative hazards fell across the first several training categories, reached their lowest published estimate at 90 to 119 minutes, then rose slightly in the group reporting at least 120 minutes. Even that highest group still had a 22% lower neurological mortality hazard than non-lifters, with a hazard ratio of 0.78.

“Little additional benefit” is therefore more accurate than saying “more exercise becomes harmful”. The study did not show that crossing 120 minutes increased neurological mortality above the risk faced by non-lifters.

It also could not distinguish an exact training recipe. The questionnaires captured weekly duration but not load, sets, repetitions, intensity, rest periods or whether participants counted an entire gym visit as training time. Calisthenics and Pilates were not included. Two people in the same 90-minute category may have trained in very different ways.

The plateau applies only to the mortality association observed here. Extra training may affect strength, bone health, mobility, athletic performance and quality of life, none of which this dose-response curve can settle. Current activity guidance recommends muscle strengthening by days per week, not a rigid minute target, as the US Physical Activity Guidelines explain.

A compelling neurological signal, with a question mark

There are plausible reasons to investigate resistance exercise and the ageing brain. Randomised trials using structural neuroimaging have reported brain changes in older adults after resistance training. Strength work can also improve muscular function and glucose metabolism, factors that may help people remain active and independent. But the mortality study did not image participants’ brains or identify a mechanism linking lifting to fewer deaths.

Earlier evidence also places the exact dose in perspective. A 2022 meta-analysis of muscle-strengthening activity and mortality put its lowest overall mortality estimate nearer 30 to 60 minutes a week. Differences in populations, exercise questions and statistical methods can move the bottom of a curve.

The current cohorts were about 78% women, overwhelmingly white and composed largely of health professionals. Their repeated questionnaires and long follow-up are major strengths, but the precise 90-to-119-minute pattern may not transfer unchanged to every population.

The fairest conclusion is both interesting and restrained. Across decades, moderate long-term resistance training tracked with meaningfully fewer deaths assigned to neurological disease, and the signal survived several serious attempts to explain it away. That is a strong reason to study the relationship more directly. It is not yet proof that a particular number of weekly minutes protects the brain.