Epidemiology and the NFL Collide

“Iron” Mike Webster was one of the greatest centers to ever play professional football. Known for playing sleeveless in freezing temperatures to intimidate his opponents, Webster spent 17 seasons in the NFL and helped lead the Pittsburg Steelers to four Super Bowl Championships in the 1970’s. He was later inducted into the Pro Football Hall of Fame and named to the NFL’s all-time team—a true football legend.[1]

But the glory of football quickly faded after his retirement. Unable to keep a job and losing his marriage, Webster became homeless, depressed, indebted, drug-addicted, and plagued by chronic pain that made sleep almost impossible. At one point Webster became so desperate for rest that he would shoot himself in the leg with a Taser until he lost consciousness. In 2002, Mike Webster died of a heart attack at age 50.[2]

What happened next would change professional football forever. During Webster’s autopsy, forensic pathologist and medical examiner Dr. Bennet Omalu discovered that Webster suffered from Chronic Traumatic Encephalopathy (CTE)—a brain disease never before seen in professional football players.[3] In 2005, Omalu published his findings in the medical journal Neurosurgery. CTE was catapulted into the national spotlight. But was football really the cause of Mike Webster’s rare brain disease? The answer to that question is an epidemiological one.


“Epidemiology is the field of public health and medicine that studies the incidence, distribution, and etiology of disease in human populations.”[4] Epidemiology attempts to understand the cause and prevention of disease[5] and is considered among “the best evidence of causation in the mass torts context.”[6]

Epidemiological evidence is often tendered as evidence of general causation, i.e. whether an agent is capable of causing a particular disease or health outcome. It may also be admitted to prove the safety and efficacy of a product, to explain a defendant’s actions, or as the basis for an expert’s opinion. Given their general nature and inherent bias, epidemiological studies alone do not answer questions of specific causation, i.e. whether an agent caused a particular individual’s disease or health outcome.[7]


There are two categories of epidemiologic studies: experimental and observational. In an experimental study, a researcher selects two groups of individuals from a given population. He then exposes one group to a suspected agent while leaving the second group unexposed. The researcher later evaluates both groups for development of the disease.

Experimental studies include randomized controlled trials and are often double-blinded, making them the “gold standard” of epidemiological evidence. However, experimental studies in humans are ethically prohibited when an agent is known to be potentially harmful. As a result, most epidemiologic studies are observational.

There are two main types of observational studies: cohort and case-control. Both types of studies have a comparison group and determine if there is an association between exposure to an agent and a disease. If an association is present, the study then measures the strength of that association. In a cohort study, a researcher selects a group of individuals who have been exposed to the agent in question and a second group who have not been exposed. The researcher then follows the groups and compares their development of the disease.[8] Behind randomized controlled trials, cohort studies are the strongest form of scientific evidence.[9]

In contrast, case-control studies begin by identifying a group of individuals who actually have the disease and a second group who do not. The researcher then compares the groups’ past exposures to the agent in question. If the agent causes the disease, the researcher should find a higher proportion of past exposures among those who have the disease.[10] Case-control studies fall just below cohort studies in the hierarchy of scientific evidence.[11]

Case reports, like the one regarding Mike Webster, describe clinical events in a single patient and are among the weakest forms of scientific evidence.[12] Alone, case reports cannot establish a causal link between an agent and a disease primarily because there is no comparison group, and they are generally excluded at trial.[13] Even Dr. Omalu conceded that his case report of Mike Webster “by itself [could not] confirm a causal link between professional football and CTE.”[14]


However, Dr. Omalu’s diagnosis of CTE in Mike Webster was just the beginning. In November 2006, Omalu published a second case report[15] after finding CTE in the brain of former NFL player Terry Long, a 45-year-old former Pittsburg Steeler who committed suicide in 2005 by drinking antifreeze.[16] He later found evidence of CTE in the brains of retired NFL players Justin Strzelczyk (age 36) and Andre Waters (age 44).[17] Omalu’s research appeared to show an association between playing professional football and CTE. But were these associations true, or merely the result of random error, bias, or confounding factors?


Epidemiologic studies are often based on relatively small sample groups. As a result, a study may erroneously find an association where one does not actually exist, or not find an association where one does exist, simply due to “chance” or “random error.”[18]

One way to assess the potential for random error is by calculating a p-value. “A p-value represents the probability that an observed positive association could result from random error even if no association were in fact present.”[19] For example, a p-value of 0.05—which is the most common significance level—means that there is a 5% chance that the study will erroneously find an association where no true association exists. Thus, the researcher can be 95% sure that the observed association is true. As long as the observed p-value for the study falls below the preselected significance level, then the relative risk or odds ratio can potentially be “statistically significant.”

A second way to assess random error is by using a confidence interval. A confidence interval is a range of values within which the true value is likely to fall. Suppose a study finds a relative risk (discussed below) of 2.66 with a 95% confidence interval of 2.14 to 3.36. The confidence interval tells researchers that they can be 95% sure that the true relative risk is somewhere between 2.14 and 3.36. The more narrow the confidence interval, the more precise the result. However, if the confidence interval includes 1.0, the result cannot be “statistically significant.”[20]


The overarching objection to the existing body of CTE research is that it is based on inherently biased case reports. Bias refers to anything “that results in a systematic (nonrandom) error in a study result and thereby compromises its validity.”[21] There are two primary types of bias in epidemiological studies: selection bias and information bias.[22]

Selection bias results from the method by which study participants are chosen. CTE research is not based on the random selection of patients. Rather, researchers gain access to brains through donations from families or by direction of the players themselves before their deaths. Because the families of players who exhibited CTE-like symptoms during their lives are more likely to donate a brain for research, the sample of brains received is not representative.

Even Dr. Ann McKee, director of Boston University’s CTE center and now the leading neuropathologist in the study of the disease, admits that “an autopsy series is terribly biased” and by itself is unable to detect incidence and prevalence of the disease.[23] In other words, case reports cannot establish why or how frequently the disease occurs in a given population.

Information bias results from “inaccurate information about either the disease or the exposure status of the study participants or a result of confounding.”[24] Because concussions in football are underreported and largely undocumented, it is difficult to reconstruct accurately a player’s medical history. As a result, researchers must often rely on interviews with family members about the player’s exposures which test memory and are speculative at best. These interviews can also encourage “recall bias,” in which family members are more likely to report past exposures once the disease has been confirmed.


A “confounding factor” is an “extra” factor in a study group which independently increases both the risk of disease and exposure.[25] If not properly accounted for, confounding factors can skew the results of a study by producing an observed association when no true association exists.

Critics of Dr. Omalu’s work initially questioned whether the use of anabolic steroids was a possible confounding factor because both Mike Webster and Terry Long had used steroids during their football careers. After all, the known side effects of steroids include high blood pressure, heart problems, aggression, psychiatric disorders, depression, and drug dependence—symptoms commonly displayed by NFL players later diagnosed with CTE.[26]

While this theory has since been discredited through experimental testing on rats,[27] and by the discovery of CTE in players whose careers predated the use of steroids in the NFL,[28] other potential confounding factors—like age, mental health, and substance abuse—have gone “largely unaccounted for in the published literature.”[29] Without conducting experimental studies that properly control for these factors, the link between football and CTE remains suspect.


Once a true association is determined (understanding that bias and chance can never be ruled out), a researcher can then evaluate the strength of that association. The strength of an association refers to the “degree to which the risk of disease increases when individuals are exposed to an agent.”[30] Epidemiologists commonly measure the strength of an association in terms of relative risk or odds ratio numbers.

Relative risk, which is most commonly used in cohort studies, is calculated by dividing the incidence rate of disease in the exposed group by the incidence rate of disease in the unexposed group. A relative risk of 1.0 means that there is no association between an agent and a disease, and that the risk of contracting the disease is the same in both exposed and unexposed individuals. A relative risk less than 1.0 means there is a negative association between the exposure and disease. A relative risk greater than 1.0 means there is a positive association between an agent and disease, which could be causal. The higher the relative risk, the stronger the association.

For example, suppose as a hypothetical that a researcher wanted to test the safety of modern football helmets compared to the leather helmets worn by early players. After selecting two equally-sized, equally-matched groups, the researcher would give leather helmets to one group (exposed group) and modern helmets to the other (unexposed group). Further suppose that at the end of the season, 66 out of 100 players with leather helmets sustained concussions, compared to only 22 out of 100 players with modern helmets. To calculate the relative risk, the researcher would divide the incidence rate of concussion among players with leather helmets (66/100 = 0.66) by the incidence rate of concussion among players with modern helmets (22/100 = 0.22), equaling a relative risk 3.0 (0.66/0.22 = 3.0). This relative risk not only shows a positive association between leather helmets and concussions (because it is over 1.0), but also implies that players who wear leather helmets are three times more likely to sustain a concussion.[31]

Many jurisdictions will only admit epidemiologic evidence if the relative risk is greater than 2.0—a level that permits an inference that the disease was more likely than not caused by the agent in question. Others courts reject this reasoning and will admit epidemiologic studies with a relative of 2.0 or less as evidence of causation, thereby leaving the sufficiency of the evidence to the jury to decide.[32]


A basic tenet of epidemiology is that an association does not equal causation. Rather, causation may only be inferred after a researcher considers all known evidence in light of scientifically recognized guidelines.

The Bradford Hill criteria provides a number of factors for researchers to consider in assessing causation, including: (1) the existence of a temporal relationship; (2) strength of the association; (3) dose-response relationship; (4) replication of the findings; (5) biological plausibility (coherence with existing knowledge); (6) consideration of alternative explanations; (7) cessation of exposure; (8) specificity of the association; and (9) consistency with other knowledge. While not all factors must be present for a causal relationship to exist, an assessment of causation requires this analysis.[33]

For years the NFL has maintained that there is no scientific evidence directly linking CTE to football-related participation. Nonetheless, since 2002, almost 100 former NFL players have tested positive for CTE, increasing health concerns among players and leading to some early retirements. As the number of players with CTE has continued to grow, so has the chorus of media outlets insisting that football causes CTE and denigrating NFL administrators for failing to acknowledge the same. In March 2016, an NFL executive publically acknowledged for the first time a link between football and CTE.[34]

Despite this acknowledgment, the science behind CTE is far from settled. After all, an association does not equal causation. With only about 200 cases of confirmed CTE across a variety of disciplines, the study of the disease is still in its “infancy.”[35] These limited results are further weakened by inherent selection bias in the CTE’s brain bank, making it nearly impossible to extrapolate the results to the general population. As stated by the Third Circuit Court of Appeals, “[t]he NFL’s recent acknowledgment may very well advance the public discussion of the risks of contact sports, but it did not advance the science.”[36]

What’s more, many of the leading CTE studies are based on incomplete or unreliable information. In two studies that collectively examined the brains of 93 former athletes, researchers were able to reconstruct the medical histories of only about half of the subjects, and those were taken second hand from family members. Moreover, virtually none of the published literature on CTE accounts for potential confounding factors.[37]

Some studies have even failed to confirm the presence of CTE under expected circumstances. In a 2013 study of six retired football players from the Canadian Football League, all six players had a history of repeated concussions and progressive neurocognitive decline prior to their deaths, but only three of the men had neuropathological findings consistent with CTE. The study concluded that “it is difficult to establish a definitive link between a history of multiple concussions and CTE” and that further research is required.[38]


Dr. Omalu’s discovery of CTE has forever changed professional football. Since 2002, the NFL has revised return-to-play guidelines, altered kickoff rules in hopes of reducing collision speeds, instituted new concussion safety measures requiring that an independent neurologist be on the sidelines for every NFL game, banned “crown of the helmet” hits outside of the tackle box (the length of the offensive line), and donated over $100 million dollars to fund brain trauma research and concussion awareness initiatives. Dr. Omalu’s research was also the impetus for the NFL’s concussion injury litigation in which approximately 5,000 retired players sued the NFL for failing to warn them of the risks of concussions, resulting in an approved settlement of nearly $1 billion dollars.[39]

Important and unanswered questions remain about the relationship between football and CTE. Experimental studies of CTE in living subjects need to be conducted, brain banks expanded, and advancements in player safety incorporated at all levels of the game. And at each step along the way, with every study that is conducted, there to guide, interpret, and help researchers better understand America’s favorite sport, will be epidemiology.


A discussion of CTE and professional football may seem a world apart from pharmaceutical and medical device litigation. However, the underlying scientific principles and methods to support or attack epidemiologic evidence are the same in both contexts. So the next time you are presented with epidemiological evidence, remember this discussion on CTE and ask:

  1. Was the type of study appropriate to the research question?
  2. Was an appropriate sample size used?
  3. How were the participants/controls recruited?
  4. Were confounding factors considered and appropriately accounted for?
  5. How strong is the association between exposure and disease?
  6. How wide is the confidence interval?
  7. Does the relative risk meet the jurisdictional requirement for admissibility?
  8. Is the association consistent with other research or scientific literature?
  9.  How many Bradford Hill criteria are satisfied?
  10. Do the numbers suggest causation?

Practitioners should also consider holding a “Science Day” to educate the judge on epidemiology in cases where such evidence plays a crucial role. “Science Days” have been used in New Jersey and Illinois courts to allow parties to explain the history and background of products and to present relevant medical and scientific literature.[40] Among other things, a well-executed “Science Day” lays the groundwork for later motions to bar expert testimony based on unreliable epidemiologic studies.

[1] Frank Litsky, Mike Webster, 50, Dies, N.Y. TIMES (Sept. 25, 2002), http://www.nytimes.com/2002/09/25/sports/mike-webster-50-dies-troubled-football-hall-of-famer.html?_r=0

[2] Greg Garber, A tormented soul, ESPN (Jan. 25, 2005), http://www.espn.com/nfl/news/story?id=1972285

[3] Bennet I. Omalu et al., Chronic Traumatic Encephalopathy in a National Football League Player, 57 Neurosurgery 128-29 (2005).

[4] Michael D. Green et al, Reference Guide on Epidemiology, in REFERENCE MANUAL ON SCIENTIFIC EVIDENCE 549, 551 (3d ed. 2011) (“Reference Guide on Epidemiology”).

[5] Id. at 551.

[6] In re Breast Implant Litig., 11 F. Supp. 2d 1217, 1224 (D. Colo. 1998).

[7] Reference Guide on Epidemiology at 608-09.

[8] Id. at 555-57.

[9] Hassan Murad et al., New evidence pyramid, 0 Evidence Based Medicine 2 (2016), http://ebm.bmj.com/content/early/2016/06/23/ebmed-2016-110401.full

[10] Reference Guide on Epidemiology at 559.

[11] Murad, supra note 9, at 2.

[12] Id.

[13] See, e.g., DeGidio v. Centocor Ortho Biotech, Inc., 3 F. Supp. 3d 674, 684 (N.D. Ohio 2014) (noting the “widespread recognition among the federal courts that case reports alone cannot prove causation.”) (internal quotations omitted).

[14] Omalu, supra note 3, at 132.

[15] Bennet I. Omalu et al., Chronic Traumatic Encephalopathy in a National Football League Player: Part II, 59 Neurosurgery 1086 (2006).

[16] Lauren Ezell, Timeline: The NFL’s Concussion Crisis, PBS (Oct. 8, 2013), http://www.pbs.org/wgbh/pages/frontline/sports/league-of-denial/timeline-the-nfls-concussion-crisis/

[17] Id.

[18] Reference Guide on Epidemiology at 572-73.

[19] Id. at 576.

[20] Id. at 576-581.

[21] Id. at 583.

[22] Id.

[23] Interview by Michael Kirk with Ann McKee, Director of Neuropathology, Dep’t of Veterans Affairs, Bedford, Mass. (May 20, 2013) http://www.pbs.org/wgbh/pages/frontline/sports/league-of-denial/the-frontline-interview-ann-mckee/

[24] Reference Guide on Epidemiology at 585.

[25] Id. at 591.

[26] Mayo Clinic Staff, Performance-enhancing drugs: Know the risks, Mayo Clinic (Oct. 15, 2015), http://www.mayoclinic.org/healthy-lifestyle/fitness/in-depth/performanceenhancing-drugs/art-20046134

[27] James D. Mills et al., Anabolic Steroids and Head Injury, 70 Neurosurgery 205, 209 (2012).

[28] Boston University CTE Center, Member of NFL Hall of Fame Diagnosed with Degenerative Brain Disease (Oct. 29, 2009), http://www.bu.edu/cte/news/press-releases/october-28-2009/

[29] In re Nat. Football League Players’ Concussion Injury Litig., 307 F.R.D. 351, 399 (E.D. Pa. 2015), aff’d sub nom. In re Nat’l Football League Players Concussion Injury Litig., 821 F.3d 410 (3d Cir. 2016), as amended (May 2, 2016) (internal citation omitted).

[30] Reference Guide on Epidemiology at 566.

[31] Interestingly, some researchers believe that wearing leather helmets—or wearing no helmets at all—might actually decrease the number of head injuries among football players. They theorize that the increased safety of modern helmets emboldens players to hit with greater speed and violence and causes them to use the helmet itself as a weapon. If the hard plastic helmets were removed, the players would alter their tackling habits, thereby reducing head injuries.

[32] Id. at 566-67, 612, 616.

[33] Id. at 552, 598-600.

[34] Bill Chappell, In a First, NFL Executive Admits Football is Linked to Brain Damage, NPR (Mar. 15, 2016), http://www.npr.org/sections/thetwo-way/2016/03/15/470513922/in-afirst-nfl-executive-admits-football-is-linked-to-brain-damage

[35] In re Nat. Football League Players’ Concussion Injury Litig., 307 F.R.D. at 398.

[36] See In re Nat’l Football League Players Concussion Injury Litig., 821 F.3d 410, 443 (3d Cir. 2016), as amended (May 2, 2016).

[37] In re Nat. Football League Players’ Concussion Injury Litig., 307 F.R.D. at 398-99.

[38] Lili-Naz Hazrati et al., Absence of chronic traumatic encephalopathy in retired football players with multiple concussions and neurological symptomatology, 7 Frontiers in Human Neuroscience 1 (2013).

[39] See In re Nat’l Football League Players Concussion Injury Litig., 821 F.3d at 447.

[40] See, e.g., In re Depakote, No. 14-CV-847-NJR-SCW, 2015 WL 4775868, at *3 n.2 (S.D. Ill. Feb. 13, 2015).