Robert A.Stern, Boston University School of Medicinebobstern@bu.edu
Julie M. Stamm, Yorghos Tripodis, Boston
Submitted April 29, 2015
We thank the WriteClick submitters for their comments and we are grateful for the opportunity to provide a response. Our study examined the relationship between the age of first exposure to repeated head impacts and later-life cognitive impairment in former National Football League (NFL) players ages 41-65. [1] We observed that those who began playing tackle football before age 12 performed significantly worse on measures of executive function, memory, and estimated verbal IQ compared to those who began playing tackle football at age 12 or older. [1] We will address each author's comments separately.
Maroon et al. primarily focused on the limitations to our study. We agree that, due to the inclusion of only former NFL players, the results of this study may not apply to former football players whose highest level of play was only the youth, high school, or college level. Future research should investigate this important question in these other groups. We acknowledge the limitations of the retrospective cross-sectional study design, including the inability to accurately measure the total head impacts actually incurred by our subjects. In an ideal-world study, all children would wear accurate accelerometers during all athletic involvement in order to measure "actual football incurred head impacts exclusive of other sports played." However, these were not available at the time our subjects participated in football, and they are still not widely used today. Although this is an important limitation, we did use total number of years played as a proxy for total exposure in the statistical analysis in this study, and this did not account for our results. Additionally, participation in other contact sports did not differ between those who began playing football before age 12 and those who began playing at age 12 or older in this study. We are unsure why Maroon and colleagues considered the neuropsychological tests selected for this study "suboptimal." As stated in the methods, these tests were specifically selected for the hypothesized outcomes [1] (references 14, 19-24, and 26 [1]) and are commonly used, highly accepted tests (see references 27-31 [1]). We also reported that the t-scores come from published normative data.
Furthermore, it is unclear what Maroon et al.
are questioning with regard to selection of age 12, given that a large proportion of the introduction and discussion specifically details this critical information.
Maroon and colleagues suggested that "the way football was played by the study subjects in the 1960's to '80's is not representative of the improved, safer practices of modern-day youth football programs." We agree that this may be true. However, while safer practices may have been adopted by some youth football leagues, other leagues have not taken these steps. Furthermore, with increased parental pressure, television programs such as "Friday Night Tykes," [2] and websites that scout football players as early as the sixth grade, [3] it could be argued that the nature of "modern- day youth football programs" may be more intense and violent in some cases today than it was when our participants played youth football. We agree with Maroon et al. that, in an ideal world, a longitudinal study of a large number of children across multiple socioeconomic, cultural, ethnic minority, and geographic locations could be conducted with detailed examinations of the children and their parents, including "parenting styles," prior to decision-making regarding involvement in sports, then examine them again 40 years later. Unfortunately, that logistically (and financially) challenging study has yet to be conducted. We addressed some of the many potential biases and alternate explanations for the results of this study, including socioeconomic status and pre-exposure intelligence. Despite the limitations of this study that we clearly addressed, we believe that, based on our findings, it is appropriate to state that "incurring repeated head impacts during a critical neurodevelopmental period may [italics our emphasis] increase the risk of later-life cognitive impairment." We do not believe it would be appropriate to conclude definitive causation, and we state clearly in the discussion that much more research is needed and should occur before changes are made to rules and policies in youth sports.
Though Dr. Andrikopoulos did not comment on the results of our study, we thank him for his comments on our methods and literature. He began his comments by questioning our modern definition of a concussion [4] and asked:
"Would a neurologist actually record this number of concussions in a patient medical record?" Robbins et al. [4] observed that, after being provided with this modern concussion definition, participants reported a two-fold increase in the number of concussions reported. This, and other evidence, [5] suggests that patients may not understand what a concussion is, and neurologists and other health care providers should provide a definition before asking patients about their concussion history. Furthermore, the clinical manifestation of more mild concussions (i.e., seeing stars), as well as subconcussive impacts are still unclear.
The disease chronic traumatic encephalopathy has been diagnosed in individuals with no reported history of concussions. [6] Therefore, we believe it may be important for health care providers to record this detailed information. We respectfully disagree with Dr. Andrikopoulos' interpretation of our referenced literature on intelligence following concussion. Reference 4 found that academic grade point averages were significantly lower for youth athletes with a history of two or more previous concussions, but without a recent concussion. Grade point average is often used as a proxy for overall intellectual functioning and is correlated with a full scale IQ testing battery. [7] Contrary to the interpretation of Dr. Andrikopoulos, reference five covers a range of severity levels of both focal injuries with "demonstrable lesions" and diffuse brain injury, and reference six was not limited to severe head injury but also included mild and moderate head injury as well. With regard to the control group, this was reported in the methods section as part of the overall DETECT study, from which the data for this study was obtained. As this study compared former NFL players who started playing tackle football prior to age 12 to former NFL players who began playing tackle football at age 12 or older, it was appropriate not to include the control group in the analysis. Furthermore, symptom validity was assessed as part of the larger neuropsychological testing battery, and this did not differ between age of first exposure groups.
Larabee and colleagues commented on our use of the WRAT-4 Reading test as an outcome measure in this study and attempted to use this test as a covariate in order to show that our results are explained by premorbid differences in intelligence. We appreciate their comments yet we believe it is appropriate to use the WRAT-4 Reading test as an outcome measure in this study. As these authors accurately state, the WRAT-4 Reading test "is often used to estimate premorbid cognitive ability due to its resistance to acquired adult cerebral dysfunction in all but aphasic or significantly demented individuals." However, our study does not examine "acquired adult cerebral function." Rather, this study examines the effect of repeated head impacts (irrespective of concussions) occurring during a critical neurodevelopmental period in childhood, with the hypothesis that incurring repeated head impacts during this time could result in disrupted brain development and, therefore, a lower adult IQ. That is, the WRAT-4 Reading test, as with most other estimates of overall intelligence, is a measure of the accumulation of learning and development prior to the age of insult(s) as well as the altered trajectory following the insult(s). The age range reflecting the "premorbid" period in studies that use this measure as a covariate is generally equivalent to middle adulthood. In our study, however, "premorbid" would be equivalent to early to mid-childhood (e.g., ages 5-10). Consequently, we strongly feel that it would be inappropriate to use the WRAT-4 reading test as a covariate in this study. Larabee and colleagues provide a brief description of a somewhat unusual re-examination of our findings by using meta-analysis and subtracting the mean of all WCST and NAB List Learning variables from WRAT-4 Reading within each group. Because of the lack of specific information provided for the analysis, it is difficult to determine exactly what was done. However, it appears that their approach would be equivalent to controlling for WRAT-4 Reading in the linear regression model only if the effect of WRAT-4 Reading were equal across outcomes. Although we do not feel that it should have been included in our study for the reasons stated above, we re-ran our regression models for all outcomes while controlling for WRAT-4 Reading by using it as a covariate. We found that the group differences are still significant for NAB-Long Delay T (group difference=7.4, p-value=0.0226) score and WCST % Perservative Errors (group difference=7.3, p-value=0.0333). Despite their statistical significance, we believe that the results of these regressions are meaningless and biased for the reasons we explained. Even if the measure of pre-morbid intelligence is weakly correlated with exposure to head impacts, group differences will be significantly biased. [8] We do agree with Larabee and colleagues that the possible influences of pre- exposure intellectual differences could be elucidated by examining elementary school achievement tests, but these were not available for this study. As stated above and in the publication, an ideal study would observe these measures in a longitudinal design beginning prior to the onset of sport participation.
We view our publication as the first preliminary study of this specific topic and hope that it will encourage future research on this important issue. We are confident in the interpretation of our results, though, as we clearly state, there are many limitations to this initial work. Much more research is needed before rule and policy changes should be made regarding safety in youth sports.
1. Stamm JM, Bourlas AP, Baugh CM, et al. Age of first exposure to football and later-life cognitive impairment in former NFL players.
Neurology 2015;84:1114-1120.
2. Boucherle D. In: Friday Night Tykes: Esquire Network, 2014.
3. Athletes emerge at NextGen [online]. Available at:
https://footballrecruiting.rivals.com/content.asp?CID=1736635. Accessed April 9, 2015.
4. Robbins CA, Daneshvar DH, Picano JD, et al. Self-reported concussion history: impact of providing a definition of concussion. Open access journal of sports medicine 2014;5:99-103.
5. McCrea M, Hammeke T, Olsen G, Leo P, Guskiewicz K. Unreported concussion in high school football players: implications for prevention.
Clinical journal of sport medicine : official journal of the Canadian Academy of Sport Medicine 2004;14:13-17.
6. McKee AC, Stern RA, Nowinski CJ, et al. The spectrum of disease in chronic traumatic encephalopathy. Brain : a journal of neurology 2013;136:43-64.
7. Kaufman AS, Lichtenberger EO. Assessing adolescent and adult intelligence, 3rd ed. Hoboken, N.J.: Wiley, 2006.
8. Angrist JD, Pischke JS. Mostly Harmless Econometrics: An Empiricist's Companion: Princeton University Press, 2009.
For author disclosures, please contact the editorial office at journal@neurology.org.
We thank the WriteClick submitters for their comments and we are grateful for the opportunity to provide a response. Our study examined the relationship between the age of first exposure to repeated head impacts and later-life cognitive impairment in former National Football League (NFL) players ages 41-65. [1] We observed that those who began playing tackle football before age 12 performed significantly worse on measures of executive function, memory, and estimated verbal IQ compared to those who began playing tackle football at age 12 or older. [1] We will address each author's comments separately.
Maroon et al. primarily focused on the limitations to our study. We agree that, due to the inclusion of only former NFL players, the results of this study may not apply to former football players whose highest level of play was only the youth, high school, or college level. Future research should investigate this important question in these other groups. We acknowledge the limitations of the retrospective cross-sectional study design, including the inability to accurately measure the total head impacts actually incurred by our subjects. In an ideal-world study, all children would wear accurate accelerometers during all athletic involvement in order to measure "actual football incurred head impacts exclusive of other sports played." However, these were not available at the time our subjects participated in football, and they are still not widely used today. Although this is an important limitation, we did use total number of years played as a proxy for total exposure in the statistical analysis in this study, and this did not account for our results. Additionally, participation in other contact sports did not differ between those who began playing football before age 12 and those who began playing at age 12 or older in this study. We are unsure why Maroon and colleagues considered the neuropsychological tests selected for this study "suboptimal." As stated in the methods, these tests were specifically selected for the hypothesized outcomes [1] (references 14, 19-24, and 26 [1]) and are commonly used, highly accepted tests (see references 27-31 [1]). We also reported that the t-scores come from published normative data.
Furthermore, it is unclear what Maroon et al. are questioning with regard to selection of age 12, given that a large proportion of the introduction and discussion specifically details this critical information.
Maroon and colleagues suggested that "the way football was played by the study subjects in the 1960's to '80's is not representative of the improved, safer practices of modern-day youth football programs." We agree that this may be true. However, while safer practices may have been adopted by some youth football leagues, other leagues have not taken these steps. Furthermore, with increased parental pressure, television programs such as "Friday Night Tykes," [2] and websites that scout football players as early as the sixth grade, [3] it could be argued that the nature of "modern- day youth football programs" may be more intense and violent in some cases today than it was when our participants played youth football. We agree with Maroon et al. that, in an ideal world, a longitudinal study of a large number of children across multiple socioeconomic, cultural, ethnic minority, and geographic locations could be conducted with detailed examinations of the children and their parents, including "parenting styles," prior to decision-making regarding involvement in sports, then examine them again 40 years later. Unfortunately, that logistically (and financially) challenging study has yet to be conducted. We addressed some of the many potential biases and alternate explanations for the results of this study, including socioeconomic status and pre-exposure intelligence. Despite the limitations of this study that we clearly addressed, we believe that, based on our findings, it is appropriate to state that "incurring repeated head impacts during a critical neurodevelopmental period may [italics our emphasis] increase the risk of later-life cognitive impairment." We do not believe it would be appropriate to conclude definitive causation, and we state clearly in the discussion that much more research is needed and should occur before changes are made to rules and policies in youth sports.
Though Dr. Andrikopoulos did not comment on the results of our study, we thank him for his comments on our methods and literature. He began his comments by questioning our modern definition of a concussion [4] and asked: "Would a neurologist actually record this number of concussions in a patient medical record?" Robbins et al. [4] observed that, after being provided with this modern concussion definition, participants reported a two-fold increase in the number of concussions reported. This, and other evidence, [5] suggests that patients may not understand what a concussion is, and neurologists and other health care providers should provide a definition before asking patients about their concussion history. Furthermore, the clinical manifestation of more mild concussions (i.e., seeing stars), as well as subconcussive impacts are still unclear.
The disease chronic traumatic encephalopathy has been diagnosed in individuals with no reported history of concussions. [6] Therefore, we believe it may be important for health care providers to record this detailed information. We respectfully disagree with Dr. Andrikopoulos' interpretation of our referenced literature on intelligence following concussion. Reference 4 found that academic grade point averages were significantly lower for youth athletes with a history of two or more previous concussions, but without a recent concussion. Grade point average is often used as a proxy for overall intellectual functioning and is correlated with a full scale IQ testing battery. [7] Contrary to the interpretation of Dr. Andrikopoulos, reference five covers a range of severity levels of both focal injuries with "demonstrable lesions" and diffuse brain injury, and reference six was not limited to severe head injury but also included mild and moderate head injury as well. With regard to the control group, this was reported in the methods section as part of the overall DETECT study, from which the data for this study was obtained. As this study compared former NFL players who started playing tackle football prior to age 12 to former NFL players who began playing tackle football at age 12 or older, it was appropriate not to include the control group in the analysis. Furthermore, symptom validity was assessed as part of the larger neuropsychological testing battery, and this did not differ between age of first exposure groups.
Larabee and colleagues commented on our use of the WRAT-4 Reading test as an outcome measure in this study and attempted to use this test as a covariate in order to show that our results are explained by premorbid differences in intelligence. We appreciate their comments yet we believe it is appropriate to use the WRAT-4 Reading test as an outcome measure in this study. As these authors accurately state, the WRAT-4 Reading test "is often used to estimate premorbid cognitive ability due to its resistance to acquired adult cerebral dysfunction in all but aphasic or significantly demented individuals." However, our study does not examine "acquired adult cerebral function." Rather, this study examines the effect of repeated head impacts (irrespective of concussions) occurring during a critical neurodevelopmental period in childhood, with the hypothesis that incurring repeated head impacts during this time could result in disrupted brain development and, therefore, a lower adult IQ. That is, the WRAT-4 Reading test, as with most other estimates of overall intelligence, is a measure of the accumulation of learning and development prior to the age of insult(s) as well as the altered trajectory following the insult(s). The age range reflecting the "premorbid" period in studies that use this measure as a covariate is generally equivalent to middle adulthood. In our study, however, "premorbid" would be equivalent to early to mid-childhood (e.g., ages 5-10). Consequently, we strongly feel that it would be inappropriate to use the WRAT-4 reading test as a covariate in this study. Larabee and colleagues provide a brief description of a somewhat unusual re-examination of our findings by using meta-analysis and subtracting the mean of all WCST and NAB List Learning variables from WRAT-4 Reading within each group. Because of the lack of specific information provided for the analysis, it is difficult to determine exactly what was done. However, it appears that their approach would be equivalent to controlling for WRAT-4 Reading in the linear regression model only if the effect of WRAT-4 Reading were equal across outcomes. Although we do not feel that it should have been included in our study for the reasons stated above, we re-ran our regression models for all outcomes while controlling for WRAT-4 Reading by using it as a covariate. We found that the group differences are still significant for NAB-Long Delay T (group difference=7.4, p-value=0.0226) score and WCST % Perservative Errors (group difference=7.3, p-value=0.0333). Despite their statistical significance, we believe that the results of these regressions are meaningless and biased for the reasons we explained. Even if the measure of pre-morbid intelligence is weakly correlated with exposure to head impacts, group differences will be significantly biased. [8] We do agree with Larabee and colleagues that the possible influences of pre- exposure intellectual differences could be elucidated by examining elementary school achievement tests, but these were not available for this study. As stated above and in the publication, an ideal study would observe these measures in a longitudinal design beginning prior to the onset of sport participation.
We view our publication as the first preliminary study of this specific topic and hope that it will encourage future research on this important issue. We are confident in the interpretation of our results, though, as we clearly state, there are many limitations to this initial work. Much more research is needed before rule and policy changes should be made regarding safety in youth sports.
1. Stamm JM, Bourlas AP, Baugh CM, et al. Age of first exposure to football and later-life cognitive impairment in former NFL players. Neurology 2015;84:1114-1120.
2. Boucherle D. In: Friday Night Tykes: Esquire Network, 2014.
3. Athletes emerge at NextGen [online]. Available at: https://footballrecruiting.rivals.com/content.asp?CID=1736635. Accessed April 9, 2015.
4. Robbins CA, Daneshvar DH, Picano JD, et al. Self-reported concussion history: impact of providing a definition of concussion. Open access journal of sports medicine 2014;5:99-103.
5. McCrea M, Hammeke T, Olsen G, Leo P, Guskiewicz K. Unreported concussion in high school football players: implications for prevention. Clinical journal of sport medicine : official journal of the Canadian Academy of Sport Medicine 2004;14:13-17.
6. McKee AC, Stern RA, Nowinski CJ, et al. The spectrum of disease in chronic traumatic encephalopathy. Brain : a journal of neurology 2013;136:43-64.
7. Kaufman AS, Lichtenberger EO. Assessing adolescent and adult intelligence, 3rd ed. Hoboken, N.J.: Wiley, 2006.
8. Angrist JD, Pischke JS. Mostly Harmless Econometrics: An Empiricist's Companion: Princeton University Press, 2009.
For author disclosures, please contact the editorial office at journal@neurology.org.