Nearly three-quarters of deceased former NFL players showed signs of chronic traumatic encephalopathy (CTE) in their brains, and the striking prevalence remained high even after researchers accounted for bias in brain donations.
Study: Bias-Adjusted Prevalence of Chronic Traumatic Encephalopathy Among Deceased National Football League Players. Image credit: Luis Rojas Estudio/Shutterstock.com
Chronic traumatic encephalopathy (CTE) is a progressive neurodegenerative disease linked to repetitive head impacts, which can occur through contact sports such as football. Because of its suggested dose-response relationship and the participation of millions in such sports, it poses a public health concern.
However, its prevalence remains difficult to estimate precisely. A recent paper in the journal JAMA Network Open reported that even after adjusting for selection bias, the estimated prevalence of CTE remains very high, warranting further study.
Why NFL brain donation studies need re-examining
CTE can only be diagnosed definitively by examining brain tissue after death. Almost everything known about CTE in American football players therefore comes from donated brains, chiefly those in the Understanding Neurologic Injury and Traumatic Encephalopathy (UNITE) Brain Bank at Boston University.
Brain donation odds change in association with age, sex, race, marital status, depression, and dementia in the general community. Thus, determining the prevalence of a disease based on donated brains is subject to registry bias, a form of selection bias.
In addition, brain donations among football players may be influenced by both their history of repetitive head impacts and whether they experienced cognitive, mood, or behavioral problems while alive. This further exacerbates the risk of overestimation of CTE prevalence when using these brains as the basis of the study.
For instance, a much-cited 2017 report found CTE in 110 of 111 former NFL players, a figure that cannot be applied to all players. An earlier bounding analysis could say only that the true prevalence lay somewhere between about 10% and 100%.
The current study aimed to estimate the prevalence of CTE neuropathology at death among all deceased former NFL players, after correcting for registry bias.
Researchers examine more than 2,300 deceased NFL players
The authors combined UNITE data with National Death Index (NDI) mortality records. The target population included every former player who appeared in at least one NFL regular-season game or a postseason game and died between January 2008 and December 2023.
The study included 2,362 eligible men: 396 were brain donors, and 1,966 were not. Players had spent a mean of 5.1 years in the league, and 41.1% died at age 80 or older.
The authors made an assumption that guides the study methodology: donation is driven mainly by cognitive symptoms observed by the player's family. For this reason, the authors used these symptoms as a proxy for CTE, to arrive at the "selection ratio": the relative probability of brain donation among players with clear cognitive symptoms compared with those without them. This symptom-based ratio was then used to approximate the CTE-specific selection ratio needed to correct for overestimation of CTE prevalence in donated brains.
Cognitive symptoms provide a proxy for CTE selection
To do this, they derived a binary dementia-related cause of death from NDI cause-of-death codes for all players, irrespective of donation status. Notably, many people with dementia or mild cognitive impairment (MCI) have another cause of death. Therefore, the category was enlarged using age-specific ratios of MCI to dementia reported by former players over 50 in the NFL-LONG study, based on self-reported physician-diagnosed conditions, smoothed with a spline. This produced a "clear cognitive symptoms" variable.
The researchers then modeled donation as a function of clear cognitive symptoms, adjusting for age at death, race, position, years played, Pro Bowl appearances, and Hall of Fame status. It was repeated across 1000 simulations to reflect uncertainty in reassigning MCI cases.
The authors used the risk ratio for cognitive symptoms as the selection ratio, and the CTE cases observed among donors were weighted by its inverse. They applied the same approach to all-cause dementia during life.
They performed nine sensitivity analyses, varying the age range, the MCI ratios, and the definition of symptoms.
Differences between donors and nondonors
Donors differed from nondonors in several ways. Donors died at a mean of 67 years, compared with 74.7 years for nondonors. Donors played in the NFL for an average of 7.1 years, compared with 4.7 years for nondonors. As for Pro Bowl appearances, donors appeared once on average, compared with 0.5 for nondonors. While 5.5% of donors were Hall of Fame members, only 2.7% of nondonors were.
A dementia-related cause of death was recorded for 19.3% of donors, compared with 11.2% of nondonors. Among donors with known diagnoses, 93.5% had CTE, and 64.9% had a history of dementia in life.
Adjusted estimates still reveal widespread CTE pathology
The median selection ratio was 4.97, meaning players with clear cognitive symptoms were about five times as likely to have their brains donated. Donation was also more likely with each additional year played (risk ratio 1.13) and among White players (1.47), and less likely with each additional year of age at death (0.95).
The position played, Pro Bowl appearances, and Hall of Fame status were not significant in the adjusted model.
After adjustment, the authors estimated that for the whole population of 2,362 players, CTE neuropathology at death was present in 74.5%, and all-cause dementia history during life in 27.2%.
Alternative assumptions still produce high CTE estimates
In the eight sensitivity analyses that used the MCI-enrichment approach for both groups, CTE estimates ranged from 70.3% to 86.1%, and dementia estimates from 28.2% to 44.2%.
The ninth analysis was designed as a conservative lower bound. It used postmortem consensus-based dementia diagnoses and premortem physician-diagnosed MCI for donors, while retaining the MCI-enrichment approach for nondonors. This increased the selection ratio to 8.27 while lowering the prevalence. Even so, the authors estimated that CTE was 63.7% and the dementia estimate 18.3%.
The authors add that if symptomatic players were ten times as likely to donate as asymptomatic ones, the CTE prevalence would still exceed 60%.
Dementia estimates reveal another concerning pattern
The authors conclude that selection bias has a substantial inflationary impact on estimates of disease prevalence when the estimates are based on brains donated because of the disease in question. Thus, the initial figure of 93.5% fell to approximately 75%.
On the other hand, they also concluded that this bias is unlikely to fully explain the high burden of CTE pathology in deceased former NFL players. They describe these as the first population-level estimates of CTE prevalence at death, and present the method as a general approach for autopsy cohorts where symptoms influence participation.
The authors noted that their estimated dementia prevalence of 27.2% is well above the 8.9% reported among living former players over 50 in NFL-LONG. However, the estimates measure different things: the current study assessed dementia history among deceased players, whereas NFL-LONG examined dementia prevalence among living participants. The earlier study included younger men who might develop dementia in the future.
In addition, people with dementia may be less likely to answer a questionnaire. The authors believe that the true lifetime dementia prevalence is probably even higher than 27.2%, since nondonors lived longer on average and would have more time to develop dementia.
Symptom estimates rely on imperfect clinical proxies
The authors note that their classification of cognitive symptoms among nondonors relied on proxies, including death certificates and estimated MCI rates, rather than clinical assessment. Death certificates often underreport dementia. Moreover, they used a selection ratio, comparing donation by symptom status, while the adjustment requires adjustment by CTE status. Where CTE and symptom status do not coincide, the estimates may be more uncertain.
The authors assumed that the MCI-to-dementia ratios apply equally to donors and nondonors. Unmeasured factors could account for residual confounding in both symptoms and donation.
The study included only deceased NFL players, limiting its generalizability to living former players or to those who played at lower levels. In addition, the impact of CTE on mortality risk is unclear, which prevented adjustment for selection into the deceased group.
The estimated CTE prevalence of 74.5% includes all stages of the disease, which may have been early-stage and even asymptomatic in many cases. Finally, they point out that only all-cause dementia could be estimated, not CTE-specific dementia.
CTE remains widespread after correcting donation bias
This study suggests that selection bias may inflate the estimated prevalence of CTE and all-cause dementia among deceased NFL players, but that it is unlikely to account fully for the high prevalence.
The authors recommend studies to produce similar estimates at the high school and college levels, though they acknowledge the paucity of data on nondonors in these age groups.