Defined population extends prior bounding work
Combining mortality records for 2362 deceased NFL players with donor pathology provides a conditional point estimate where earlier work offered broad bounds.
↳ Methods, NDI Data; Results; Discussion
Abstract
Importance
The majority of knowledge regarding chronic traumatic encephalopathy (CTE) in National Football League (NFL) players is derived from autopsy-based studies on voluntarily donated brains. Because these cohorts are highly vulnerable to selection bias (specifically registry bias), population-level prevalence of CTE among former NFL players is unknown, limiting understanding of the impact of this disease.
Objective
To estimate the selection-bias adjusted prevalence of CTE neuropathology in deceased NFL players.
Design, Setting, and Participants
This population-based cross-sectional study used Understanding Neurologic Injury and Traumatic Encephalopathy (UNITE) brain bank data combined with National Death Index records on former NFL players. Both brain donor and nondonor NFL players who died between January 2008 and December 2023 were included. Data were analyzed from November 2025 to April 2026.
Exposure
Playing in at least 1 regular or postseason NFL game.
Main Outcomes and Measures
The main outcomes of this study were bias-adjusted prevalence of CTE neuropathology and all-cause dementia among the target population, both of which were measured in donors prior to this study. A cognitive symptom–related brain donation selection ratio (probability of donation given CTE divided by the probability of donation given no CTE) was estimated using Poisson regression and was applied to adjust donor-observed prevalence of CTE neuropathology at death and all-cause dementia during life. The prevalence estimates were adjusted for bias using the estimated selection ratio.
Results
The cohort consisted of 2362 deceased former NFL players (mean [SD] years played, 5.1 [3.9]; 971 [41.1%] aged ≥80 years at death), of whom 396 were brain donors and 1966 were nondonors. Of 387 donors, 362 (93.5%) had CTE. The median selection ratio was 4.97 (95% CI, 4.72-5.26). Bias-adjusted population prevalence of CTE at death was 74.5% (95% CI, 69.6%-78.3%), with a conservative lower-bound analysis estimating 63.7% (57.2%-68.9%). Bias-adjusted population prevalence of all-cause dementia history during life was 27.2% (95% CI, 22.6%-31.6%).
Conclusions and Relevance
The findings of this cross-sectional study suggest that selection bias materially inflates donor-based prevalence estimates but is unlikely to fully explain the high burden of CTE neuropathology among this population. Further research is required to understand the clinical impact of this disease.
Key Claims
Primary findingTentative
the estimated bias-adjusted prevalence of CTE neuropathology among all former NFL players who died between 2008 and 2023 was 74.5% (95% CI, 69.6%-78.3%).
Fragile — the claim's weakest requirement is that the effect exists at all. Experimental. Every point below is traceable to a quoted passage, but this forecast has not yet been validated against known replication outcomes.
population inference depends on an unvalidated equivalence between symptom-based and CTE-specific donation ratios
Results, paragraph after Table 2
SecondaryTentative
Bias-adjusted population prevalence of all-cause dementia history during life was 27.2% (95% CI, 22.6%-31.6%).
dementia-history adjustment relies on the same proxy-derived donation ratio and transported MCI inputs
Abstract, Results
SecondaryTentative
Even if individuals with cognitive symptoms were 10 times more likely to donate their brains, the estimated prevalence would still exceed 60%.
the reconstructed adjustment yields approximately 59.1% at a tenfold donation ratio rather than exceeding 60%
Conclusions, paragraph 1
Derived from the full evaluation — not a separate score.
Strengths · Limitations
Strengths
Sensitivity analyses examine multiple adjustment scenarios
Table 3 varies age restrictions, MCI inputs, and symptom definitions, including an asymmetric clinical-benchmark scenario intended to lower estimated prevalence.
↳ Methods, Calculating Bias-Adjusted Prevalence; Table 3
Interpretation separates pathology from clinical burden
The Limitations section distinguishes all-stage neuropathology at death from symptoms in living players and explains why the estimates do not generalise to living players or lower exposure groups.
↳ Limitations
Limitations
Symptom proxy does not identify pathology
The required donation ratio concerns CTE status, but the estimated ratio concerns cognitive symptoms assembled from death certificates and external MCI ratios. The sensitivity analyses do not validate equivalence between those ratios.
↳ Methods, Calculation of the Selection Ratio; Limitations, first two limitations
Numerical reporting requires explicit threshold correction
The Conclusions state that prevalence exceeds 60% at a tenfold donation ratio, whereas the reconstructed adjustment gives approximately 59.1%. The selection ratio of 4.97 also has different confidence intervals in the text and Table 2.
↳ Conclusions; Results, selection ratio paragraph; Table 2
Remaining selection mechanisms are not resolved
The donation model uses 2298 complete cases without a reported missing-data sensitivity analysis. Residual donation mechanisms also remain outside the symptom-based adjustment.
↳ Results, first paragraph; Methods, Poisson Modeling; Limitations
Nabu Commentary
The defined decedent population and explicit donation model advance the comparison with earlier broad bounds. Table 3 examines multiple assumptions, but these scenarios do not establish that symptom-based differential donation equals CTE-specific differential donation. This structural uncertainty, together with complete-case selection and remaining donation mechanisms, weighs more heavily on Methodological Rigour than the number of sensitivity analyses alone. The interval reporting requires reconciliation, and the tenfold-donation statement requires correction to approximately 59.1%, although that correction does not overturn the broader high-burden interpretation.
Check another paper
Author Note / Field Feedback
Nabu’s assessment, alongside the field’s view.
Are you an author of this paper?
Dimension breakdown
Study Quality
Sound3.6
Confidence mediumContribution
3.9
The study advances prior broad bounding work by combining a defined decedent population with donor data and an explicit selection model. The contribution is a conditional population estimate rather than direct ascertainment of CTE in nondonors.
Introduction; Discussion, comparison with Binney et al; Table 3
Methodological Rigour
3.3
Covariate-adjusted donation modelling, simulation, and sensitivity analyses address registry bias, but the required CTE-specific ratio is replaced by a symptom-based ratio using transported MCI inputs. Remaining selection and complete-case handling reduce the score, while conflicting interval reporting and the tenfold-scenario error are also surfaced through the reliability flag.
Methods, Calculation of the Selection Ratio; Results, first paragraph; Table 3; Limitations; Conclusions
Reporting
3.8
The target population and adjustment sequence are understandable, and the main conclusions distinguish neuropathology from symptoms. Conflicting selection-ratio intervals and the inaccurate tenfold threshold impede precise interpretation without making the argument incoherent.
Abstract; Results, selection ratio paragraph; Tables 2–3; Conclusions
Positioning
3.8
The paper positions its estimate against prior bounds and selection-bias work and explains important scope restrictions. Its treatment of symptom–pathology discordance and alternative interpretations is less complete than its discussion of death-certificate underascertainment.
Introduction; Discussion; Limitations
Reliability
Caveats4 of 4 checks
Results checkNoted
- Bias-adjusted population prevalence of CTE at death was 74.5% (95% CI, 69.6%-78.3%): "The median selection ratio was 4.97 (95% CI, 4.72-5.26; P < .001)" (Results, selection ratio paragraph) vs "Clear cognitive symptoms | 4.97 (3.77-6.55)" (Table 2, row 'Clear cognitive symptoms'). The same selection-ratio point estimate has different reported intervals, while an approximate reconstruction links the prevalence interval to Table 3's wider selection-ratio interval. → Conclusion holds
- Even if individuals with cognitive symptoms were 10 times more likely to donate their brains, the estimated prevalence would still exceed 60%: "Even if individuals with cognitive symptoms were 10 times more likely to donate their brains, the estimated prevalence would still exceed 60%." (Conclusions, paragraph 1) vs "Consensus dementia history or MCI for donors | 8.27 (6.54-10.85) | 63.7 (57.2-68.9)" (Table 3, last row). Using the reconstructed adjustment that reproduces the main and lower-bound estimates, a selection ratio of 10 gives (362/25)/[10 + (362/25)] = 59.15%. → Not supported by the paper's own data (approximate check)
Reporting and transparencyClean
Ethics approval, written consent, funding roles, and conflicts of interest are declared. A data-sharing statement is referenced in an unavailable supplement, which is an assessability gap rather than a reliability finding.
Conflict of interest: declared · Ethics approval: declared · Data availability: declared
Reference IntegrityClean
20 references in manuscript 20 of 20 checkable references found in an index
References not verified automatically are resolved manually within 24 hours.
Post-Publication RecordClean
No retraction or expression of concern found in Crossref (includes Retraction Watch).
Last checked: 29 Oct 2026
Sources: Crossref (incl. Retraction Watch) ✓
Readiness
Where this paper’s evidence sits on the path from initial observation to real-world use.
Demonstrated
The findings provide epidemiologic characterisation, not a validated clinical tool or intervention protocol. Population mortality coverage does not validate the proxy assumption underlying the adjusted estimates.
Methods, General Approach and Calculation of the Selection Ratio; Limitations
AI-generated, human-governed. Something look off? Contact us to request a review.