A quarter of NFL players who died had CTE. Or almost all of them did.

Every widely quoted figure for how common chronic traumatic encephalopathy is among former professional footballers has come from brain banks, and every brain bank has the same problem: families who suspect something was wrong are the families who donate. A retrospective population-based cohort published in The BMJ on 25 August does not solve that problem. It does something more useful — it puts formal bounds around it [s1].

What was counted

The investigators identified 1,712 former National Football League players, from the era of hard-shell helmets (post-1949), who died during 2008 to 2021. Personal details and causes of death for all of them came from the National Death Index. Of those 1,712, 338 players — 19.7% — had donated their brains to the UNITE brain bank or the University of California, San Francisco Alzheimer's Disease Research Center [s1].

Among those 338 donors, 315 (93.2%) received a neuropathological diagnosis of CTE from neuropathologists masked to clinical and playing histories [s1].

From those two numbers the authors derive the range that is the paper's central result. The minimum possible prevalence at death treats every non-donor as CTE-free — donors with CTE divided by all NFL deaths. The maximum treats every non-donor as a case — 1 minus (donors without CTE / all deaths). The paper reports the resulting bounds as 18.5% and 98.7% [s1]. (Worth noting for anyone checking the arithmetic: 315/1,712 is 18.4%, a rounding-level discrepancy with the 18.5% printed in the abstract; the upper bound reproduces exactly.) Restricted to 2016 to 2021, when brain donation was most frequent and the donor fraction therefore largest, the bounds tighten only slightly, to 24.5% and 97.7% [s1].

Neither end is a plausible estimate on its own. The point of stating both is that the true figure is somewhere inside a range this wide, and no brain-bank series can narrow it further without knowing how donors differ from non-donors.

The dementia finding

The second half of the paper is less about counting and more about consequence. Among the 338 donors, 104 (30.8%) had stage IV CTE — the most advanced stage — and 202 (59.8%) had dementia diagnosed by study clinicians who reviewed medical records and informant-based clinical histories while masked to neuropathological status [s1]. Mean age at dementia onset was 63.4 years (SD 12.5); mean age at death was 73.1 (SD 10.5) [s1].

To account for the selection pressure of donation, the authors used inverse probability weighting. Stage IV CTE was associated with study-clinician-diagnosed dementia at a risk ratio of 1.44 (95% CI 1.16 to 1.78; P<0.001) [s1]. That is an association within the donor group, weighted to reduce donation bias — not a population risk estimate.

One number in the paper is arguably the most consequential and has nothing to do with football: only 40.6% of donors with study-clinician-diagnosed dementia (n=82) had neurodegenerative disease listed as the primary or secondary cause of death on the death certificate [s1]. Separately, 63 donors (18.6%) had neurodegenerative disease listed as the primary cause of death [s1]. Death certificates, in other words, missed most of the dementia this cohort was independently judged to have — which is a caution about every study anywhere that uses cause-of-death coding to count neurodegenerative disease.

How this gets reported

There is a reason to state the bounds carefully. A study published in July examined 61 usable press releases about CTE research retrieved from the EurekAlert! service, scoring each abstract and its corresponding release on a "causal intensity" metric and defining spin as the gap between them [s2]. Thirty-nine of the 61 releases (63.9%) were more deterministic than the abstract they described, and five were classified as severe spin, with z-scores at least 1.5 standard deviations above the mean [s2].

That analysis is descriptive and its scoring metric is novel rather than established, so it should be read as a first measurement rather than a settled one [s2]. But it describes precisely the failure mode this BMJ paper invites: a headline that reports one end of an 18.5%-to-98.7% range as though it were the estimate.

What it does and does not establish

The cohort establishes that at minimum, nearly a quarter of former NFL players who died during 2016 to 2021 had CTE neuropathology at death — a floor, derived without assuming anything favourable about non-donors [s1]. It establishes that among donors, dementia recorded before death was common and was associated with the most advanced CTE stage [s1].

It does not establish a point prevalence, and the authors do not claim one. It does not follow living players, does not measure exposure in games or seasons, and cannot say what fraction of the dementia in this group would have occurred without football. The donors remain a self-selected group; weighting reduces that distortion but does not remove it.

What to watch is whether other national death registries can be linked to other brain banks the same way. The method here — bounding the unknowable rather than estimating past it — is portable to rugby, boxing and military cohorts, all of which currently rely on the same kind of donor series this paper was built to interrogate.

Sources

• [s1] Prevalence of chronic traumatic encephalopathy at death in National Football League players: retrospective population based cohort study, 2008-21. The BMJ, 25 August 2026. https://doi.org/10.1136/bmj-2026-100418
• [s2] Discordance in causal terminology between scientific papers and subsequent press releases: chronic traumatic encephalopathy in the "age of spin". Frontiers in Neurology, 8 July 2026. https://doi.org/10.3389/fneur.2026.1882586