A Call for Scientific Accuracy in CTE Research
By Chris Nowinski, Ph.D., CEO, Concussion & CTE Foundation
For patients and families, it is critical that we learn how to accurately diagnose CTE during life. Studies across the U.S. are dedicated to this and are getting us closer. In 2021, a criteria known as Traumatic Encephalopathy Syndrome (TES) was published as a proposed research criteria and is actively being investigated. TES combines symptoms with a deep understanding of repetitive head impact (RHI) exposure gathered from families of brain donors. Every two weeks I observe clinicians from Boston University and Harvard review the clinical histories of brain donors to see if TES can help predict who has CTE pathology. The criteria appear to work well, especially in older brain donors who have extensive exposure to RHI.
A new study was published in Nature Medicine yesterday, “Performance of traumatic encephalopathy syndrome criteria in identifying individuals with chronic traumatic encephalopathy,” purporting to test validity of the 2021 TES criteria, and the authors claimed it only had a positive predictive value of 24%.
I’m not sure that TES is the right criteria and I’m happy to follow the science, but this is a rare moment where I feel I need to raise my hand and point out the significant limitations of this study. As noted in their own limitations, this Nature Medicine study did not properly apply the TES criteria. All studies have limitations. However, in this case, if the authors didn’t test the full scope of the TES criteria, how can we conclude that the TES criteria do not work?
That doesn’t mean the study does not have value. It highlights a known weakness of the criteria: that many factors are subjective, they perform differently in different settings, and they can be easily misused.
The Study
Researchers primarily at Penn (where the first active college football player, Owen Thomas, was diagnosed with CTE, and where we just diagnosed another not publicly named former football player with CTE) reviewed clinical records and brain studies of 1,038 donors in their brain bank, which began receiving cases in 1985. They claimed 25 participants met TES criteria, but only 6 (24%) had CTE pathology, and concluded, “We found that most individuals with repetitive head impacts met the current consensus criteria defining TES. However, only a fraction of such cases were found to have CTE-NC on neuropathological examination, resulting in a poor PPV.”
Flaw 1: Years of Exposure was Not Acquired for Subjects
CTE risk is influenced by RHI in a dose response fashion. In contact sports like football, ice hockey, and rugby, studies have shown the odds of developing CTE increases by 15% to 34% per season played, although absolute risk remains unknown. Due to this strong relationship between years of play and pathology, TES requires knowing years of play. Based on studies of football players at Boston University, to be considered for possible TES, one needs 5 years of football. For probable TES, the threshold is 11 years.
The first major flaw in the study is noted in the limitations: “For most of the identified athletes, the exact duration of play was not reported, preventing strict application of TES criterion requiring at least 5 years of play. Accordingly, for the purpose of this study, we considered a threshold of at least high school level of play to satisfy the RHI exposure criterion.”
First, this application of the criteria ignores the history of football participation in America. It is extremely unlikely high school football participation would meet the 5 years of play threshold in a brain bank dating back to 1985, with a median age at death of 74. Assuming a normal distribution, almost the entire cohort played in the age before youth tackle football was a national phenomenon. Pop Warner was not a national organization until 1968, and family interviews of deceased football players from that era show that before 1970, nearly all football players began playing in high school (Alosco 2018). Since high school football has a high attrition rate (in the 90’s at my high school, 80 freshman on A & B teams whittled down to 20 seniors), only a minority of high school football players play all 4 years. Thus, it is unlikely that the high school players (10 of the 14 football players considered to have substantial exposure) in this study played for 5 years.
Second, this application of the criteria does not address the second aspect of exposure, that a minimum 5 years only applies to football. The TES criteria states, “Exposure risk thresholds for other contact or collision sports, or combinations of contact/collision sports, have not yet been established but should be a substantial number of years (e.g., ≥ 5 y) at a level of play involving routine RHIs.” This leaves a lot – and probably too much – wiggle room for clinicians to include rare, unquantified exposures.
Five years of high school football exposure is generally 2,000 to 4,000 head impacts of 10-100g. Other sports likely take much longer to reach that threshold, meaning that the inclusion of a high school soccer player and a wrestler, college and military boxers (who wouldn’t have reached 5 years of exposure at those levels alone, and likely wouldn’t have had high school exposure), was likely inappropriate. They also include a paired skater, and I find it hard to imagine that subject would have been dropped on their head on ice 2,000 times. They also included three victims of intimate partner violence, despite the fact the largest study of victims of intimate partner violence showed 0 of 84 women had CTE. It is becoming clear TES criteria is best applied to older elite athletes because it is nearly impossible to accurately estimate RHI exposure in intimate partner violence.
Flaw 2: Misuse of Probably, Possible, Suggestive
Without knowing years of play for the football players, they could not possibly categorize the subjects into probably, possible, and suggestive, but they did. Researchers shouldn’t equally weight in an analysis of the accuracy of a TES diagnosis, but they did. But even with their misuse, those with “probable” were more likely to have CTE, then “possible”, and zero of “suggestive” had CTE, as the TES criteria authors intended.
Flaw 3: They didn’t Calculate Likelihood Ratios, Which Show TES May Actually Work Quite Well
Scientists use different measures to assess how well diagnostic tests perform. One measure that doctors commonly think about is the one used by the authors – positive predictive value (PPV). This is the probability of actually having the disease among people who test positive. Positive predictive value depends on how common the disease is in the group that you are studying. As a disease becomes more rare, the probability that the disease is present if you test positive goes down. This is one of the reasons that we don’t test everyone in the population for all diseases. In this study, the prevalence of CTE was extremely rare, less than 1%.
Therefore, it is not surprising that the positive predictive value was also low. There are alternative measures that are more helpful in study population where the disease is rare, with the most common being the positive likelihood ratio (LR+). This is the likelihood that the disease is present if you test positive compared to if you test negative. A ratio of 10 or more is considered great, and suggests a test is worthy of being utilized in clinical decision making. The LR+ of TES in this study was more than 24!
LR+=Sensitivity/(1−Specificity)
So:
LR+ = 0.462/(1−0.981) ≈ 24.3
Even though the authors claim that there were too many false positives for the test to be useful, there were actually very few false positives – if you account for the many, many people in the population studied who did not have CTE.
Had they run this analysis, it seems their data supports the value of the TES criteria, even when misused as it was here.
Why didn’t the authors publish the LR+? I can’t answer that.
An Emotional Strawman Argument
This group of authors closed their abstract with, “The poor performance of TES criteria raises substantial concern for its potential negative psychological impact on current and former contact sport athletes, who may be incorrectly diagnosed with a progressive neurodegenerative pathology.”
First, this is a strawman argument, as no one has advocated for use of TES in the clinic. Second, this study actually refutes that claim – there was not a single person in the Penn study who was “incorrectly diagnosed with a progressive neurodegenerative pathology.” 100% of the 25 individuals they (incorrectly) labeled as meeting TES criteria had another progressive neurodegenerative brain disease. No one in this population would have been harmed with an incorrect diagnosis of a progressive brain disease.
That’s the real danger of this paper – it risks making experienced clinicians scared to give patients their best differential diagnosis.
I agree that living 40-year-olds should not be told their cognitive symptoms are caused by CTE or that they have TES. The Concussion & CTE Foundation will continue to beat the drum that every young person concerned about symptoms should seek treatment, because their symptoms could be caused by other factors, such as depression, sleep apnea, or other treatable conditions.
The more difficult question is what to tell the 70-year-old who played 25 years of football or ice hockey who has severe cognitive impairment, clearly has a neurodegenerative disease, and tests negative for Alzheimer’s disease biomarkers and every other known disease with a biomarker. Thus far, more than 90% of this population that has come to autopsy has had CTE. Is it fair to tell this individual “We don’t know” or should they be told they likely have CTE?
The widows and children of our brain donor families diagnosed after death with CTE repeatedly tell us that the worst part of their medical experience was the doctor who refused to consider CTE as a diagnosis. Pretending CTE isn’t there also causes harm – both to patients and families.
Some clinicians working in memory or dementia clinics in that situation are choosing to be honest with their patients and tell them when they suspect CTE, not using TES, but with their decades of training and experience.
While writing this I happen to be at the Tau Global Conference in Washington D.C., with other CTE researchers, so I asked the leading neurologist in Canada who treats CTE patients, Carmela Tartaglia of the University of Toronto, how she approaches telling her patients they may have CTE. She told me,
“CTE has to be part of your differential diagnosis in the setting of repeated head impacts; otherwise, you aren’t practicing 2026 medicine. More importantly, regardless of what you think the cause is, you still have to provide care for their symptoms.”
Neurologists on the front lines have the experience to know they aren’t always correct in their diagnosis, because diagnoses of neurodegenerative diseases (including Alzheimer’s disease) are hard and there is always uncertainty. You navigate that uncertainty with the family, together.
In the Penn study, 100% of the 25 people who were theoretically diagnosed with TES actually had an incurable brain disease (in addition to CTE for the 6). So even if these phantom diagnoses occurred, no harm would have been done, as they all would have been told they have an untreatable progressive neurodegenerative pathology.
As an at-risk individual who annually loses friends to CTE, this is an urgent area of research, and we need to move the field forward toward diagnosis during life. Irrespective of TES, we have come so far in the science and there are certainly situations where confidence for underlying CTE can be high. There are other situations where confidence is much lower but that is how it goes with neurodegenerative disease diagnoses. By no means are the TES criteria perfect and we should not be adopting them in the clinic. But, if we truly want to understand the value of the TES criteria and its components, we need studies that are specifically designed to accurately test the full scope of the criteria.
Last updated: Monday, May 18, 2026