Participation in Collision Sports and Cognitive Aging Among Swedish Twins

Abstract

We examined the association between early-life participation in collision sports and later-life cognitive health over a 28-year period in a population-based sample drawn from the longitudinal Swedish Adoption/Twin Study of Aging (1987–2014). Cognitive measures included the Mini-Mental State Examination and performance across multiple cognitive domains (e.g., global cognition, verbal ability, spatial ability, memory, processing speed). Among a sample of 660 adults (mean age at baseline, 62.8 years (range: 50–88); 58.2% female), who contributed 10,944 person-years of follow-up, there were 450 cases of cognitive impairment (crude rate = 41.1/1,000 person-years). Early-life participation in collision sports was not significantly associated with cognitive impairment at baseline or with its onset over a 28-year period in a time-to-event analysis, which accounted for the semi-competing risk of death. Furthermore, growth curve models revealed no association between early-life participation in collision sports and the level of or change in trajectories of cognition across multiple domains overall or in sex-stratified models. We discuss the long-term implications of adolescent participation in collision sports on cognitive health.

Abbreviations

 
• CI

  confidence interval

 
• csHR

  cause-specific hazard ratio

 
• MMSE

  Mini-Mental State Examination

 
• SATSA

  Swedish Adoption/Twin Study of Aging

Maintaining brain health and function are important aspects of successful aging. Lower cognitive functioning in older age is associated with reduced quality of life (1) as well as increased risks of disability and dependency (2) and mortality (3). Explanatory models of cognitive health implicate demographic, lifestyle, health, and genetic characteristics that span the life course and contribute to cognitive functioning in adulthood (4). In recent decades, there has been increased interest in studying how exposure to head trauma—especially through participation in collision sports—could alter trajectories of cognitive health. This is particularly true of young athletes, for whom head trauma could interrupt neurodevelopmental processes, interfere with brain maturation, and affect the acquisition of new fine-motor and cognitive skills (5, 6).

Numerous studies have documented the immediate and short-term effects of head trauma experienced during adolescence (7–10). The medical community widely recognizes these short-term effects to include headache and blurred vision, as well as deficits in concentration, balance, memory, and visual processing (e.g., 11, 12). However, the full extent of the implications of head trauma might not be realized until later in life, during which time high level social, mental, and cognitive functioning are expected to manifest (13, 14). As such, research implementing longitudinal methodology is needed to understand the long-term implications of adolescent participation in collision sports on later-life cognitive health.

Head trauma resulting from collision sports has been linked cross-sectionally and longitudinally with lower levels of cognitive performance (15–20). However, much of this research has focused on participants who were collegiate or professional athletes as well as selective samples of individuals whose experiences might not be representative of participation in adolescent sports. Few studies have examined the long-term cognitive and mental health implications of adolescent participation in collision sports due, in part, to a lack of survey or administrative data assessing collision sports retrospectively, or prospectively with sufficient follow-up. The studies that have done so report little or no effect (21–23). For example, a recent observational study by Deshpande et al. (21) found that cognitive and mental health outcomes in adulthood were similar for men who played high-school football and a matched sample that did not. Against a backdrop of rising public health concern over the immediate and long-term implications of sports-related head trauma in adolescence, this inconclusive literature might discount the established benefits of physical health (24), social functioning (25), and resilience to adversity (26) seen among individuals who participate in team sports in early life and underscores the importance of studying the long-term effects of adolescent participation in collision sports on cognitive functioning across multiple domains.

The present study examined the association between early-life participation in collision sports and later-life cognitive health spanning a 28-year-period among participants from the Swedish Adoption/Twin Study of Aging (SATSA). Whereas prior work on this topic has relied primarily on selective or clinical samples, participants in SATSA were drawn from population-based administrative registry of twins, which allowed us to estimate these associations net of childhood environment while accounting for genetic confounding and sociodemographic characteristics. Our primary analysis evaluated whether participation in collision sports was associated with cognitive impairment. Secondarily, we examined whether participation in collision sports was associated with the level of and change in functioning across multiple cognitive domains.

METHODS

Data

The SATSA is a longitudinal study composed of all same-sex twin pairs from the population-based Swedish Twin Registry who indicated they were separated from their co-twin prior to age 11 years and reared apart (27). The SATSA also includes a control sample of twins reared together, matched on age, sex, and county of birth in Sweden. Sample ascertainment and study procedures have been described in detail elsewhere (27, 28). Briefly, SATSA researchers collected information about health, lifestyle, and environment from eligible participants by mail questionnaire when the study began in 1984 (response rate 70.7%). Twin pairs aged 50 years or older were invited to complete in-person tests to assess their health and cognitive abilities during a 4-hour visit conducted by a trained and registered nurse practitioner. Questionnaires and in-person visits continued approximately every 3 years with 7 follow-up periods (1987, 1990, 1993, 2004, 2007, 2010, 2014) until the study ended in 2014. SATSA was approved by the Ethics Committee at Karolinska Institutet (27, 28). All participants provided informed consent for participation.

Our primary sample (sample 1) was obtained from a restricted data file with linked mortality records prepared by SATSA investigators for this specific investigation. This sample contained 660 respondents aged 50 years or older at baseline, in 1984, with full covariate information, who completed at least 1 in-person interview over the survey period and responded to the 1993 mail questionnaire (questionnaire 4), wherein participation in collision sports was assessed. Sample 2 was prepared using the public-use SATSA data files with the same exclusion criteria applied, resulting in a sample of 662 respondents. We retained complete twin pairs and singletons who met these criteria. Fisher’s test revealed no statistically significant differences in the missingness of cognitive scores according to participation in collision sports (P = 1).

Outcomes

Cognitive impairment.

We used the Mini-Mental State Examination (MMSE) to examine cognitive impairment. The MMSE is commonly used to assess cognitive functioning, track changes in cognitive function over time, and screen individuals for cognitive impairment (29). The SATSA implemented the MMSE in its baseline assessment in 1984 and in each wave that followed, through 2010. MMSE scores range from 0 to 30, with lower scores indicative of more severe impairment. The traditional MMSE cutoff score for cognitive impairment suggestive of dementia is 24, but higher cutoff scores have been proposed to increase diagnostic accuracy in individuals with higher levels of education and at earlier stages of dementia severity (30). In the absence of a conclusively defined cutoff, we selected a cutoff of 27 as the primary outcome measure for cognitive impairment.

Cognitive domains.

In addition to examining cognitive impairment, we included more fine-grained measures of cognitive functioning that might be more sensitive to subtle changes in underlying cognitive processes. The measures in the SATSA battery capture 4 cognitive domains: verbal, spatial/fluid, memory, and perceptual speed. Principal components analysis has been used in prior work to create component scores indexing each of these domains, which has been described in detail elsewhere (31).

Exposure.

Participation in collision sports was assessed retrospectively via self-report in the wave 4 interview. Specifically, respondents were asked about their childhood: “Have you been involved in any sport (e.g., football, ice hockey, or boxing) that may involve a hit on a head?”

Covariates.

Covariates were selected for their hypothesized association with sport participation and/or cognitive health (4, 32). We accounted for age, sex (either through adjustment or through stratification), and self-reported education. Self-reported education was categorized on the basis of the Swedish school system as primary education (reference category), lower secondary or vocational, upper secondary education, and tertiary education.

Statistical analysis

We developed and posted a protocol before analysis (online at arXiv; identifier: arXiv:1807.10558) as recommended by Rubin (33). Analyses were performed using R, version 4.0.2 for Windows (R Foundation for Statistical Computing, Vienna, Austria) (34), and Stata, version 14 (StataCorp LP, College Station, Texas) (35).

Descriptive analysis.

We characterized the analytical samples at baseline using means (standard deviations) and proportions. We used t tests and Fisher’s exact test to compare differences in sample characteristics by early-life participation in collision sports.

Time-to-event analysis.

We used cause-specific Cox regression models to evaluate the association between early-life participation in collision sports and later life cognitive impairment—as defined by a MMSE score of 27 or below—and obtained cause-specific hazard ratios while accounting for the semi-competing risk of mortality. Competing risks arise when the occurrence of a particular event (e.g., death) precludes the event of interest (e.g., cognitive impairment). We used attained age as the underlying time scale due to its strong association with cognitive impairment. Models adjusted for sex and educational attainment and estimated with robust standard errors clustered at the twin level to account for potential environmental and genetic associations within each twin pair.

Growth curve modeling analysis.

As a complementary analysis, we estimated trajectories of cognitive performance across multiple domains (global cognition, verbal ability, spatial ability, memory, processing speed) using a series of growth curve models to examine the associations of participation in collision sports with mean cognitive performance and change over time. Growth curve models allow for the exploration of both intra-individual change and individual differences in the nature of that change (36, 37). We performed analyses in 2 stages. In stage 1, we included a fixed effect for collision sports participation to investigate the association between collision sports participation and mean cognitive abilities at baseline (i.e., the intercept). In stage 2, we included a fixed effect for collision sports as well as linear and quadratic age interaction terms to investigate whether and how participation in collision sports might affect cognitive abilities over time (i.e., the slope). All models adjusted for sex and education. Age was used as the time scale and centered at 65 years to provide a clear interpretation of the intercept term as the mean cognitive performance at age 65. Analyses also accounted for the hierarchical structuring of repeated measures nested within the individual nested within twin pairs by modeling the latter component as random effects.

Sensitivity analysis.

We performed several sensitivity analyses: 1) a replication of our time-to-event analysis in models stratified by sex; 2) a replication of our time-to-event analysis in the overall sample using 2 additional MMSE cutoffs (27 and 29) that have been used in prior studies (38, 39); and, in accordance with recommendations specified in prior work (40, 41), 3) a parallel analysis in which we used the Fine and Gray model (42, 43) in place of the cause-specific Cox regression model (42). In addition, we replicated our time-to-event analysis among a sample of adults without cognitive impairment in the survey waves prior to and immediately following their retrospective report of early-life sports participation. For this analysis, the survey wave following their retrospective report was used as our baseline (wave 5).

RESULTS

The baseline sample for the time-to-event analysis (sample 1) consisted of 660 individuals composed of 270 twin pairs and 120 singletons. Of the sample, 58.2% was female, with a mean age of 62.8 (standard deviation = 7.9) years at baseline (Table 1). Among the 660 individuals at baseline, 78 (11.8%) reported participating in collision sports. The analytical sample used for estimating the association between participation in collision sports and trajectories of cognitive performance (sample 2) had a similar composition: 58.2% of the 662 individuals (270 twin pairs, 122 singletons) were female with a mean age of 62.6 years (standard deviation = 7.8), and 78 (11.8%) reported participating in collision sports. In both samples, men were more likely than women to report participation in collision sports (sample 1 and sample 2: 77 vs. 1; P < 0.05). Levels of educational attainment for those who did and did not participate in collision sports were comparable.

Characteristics of the Analytical Samples, Swedish Adoption/Twin Study of Aging, 1987–2014

Time-to-event analysis

Over 10,944 person-years of follow-up (median, 18.0 (interquartile range, 7.5–22.6) years), there were 450 cases of cognitive impairment (crude rate: 41.1/1,000 person-years) using an outcome that classified cognitive impairment as a score on the MMSE of 27 or lower. The association between baseline risk factors and cognitive impairment was quantified using cause-specific hazard ratios obtained from cause-specific Cox regression models. Table 2 presents these results in the form of cause-specific hazard ratios and corresponding 95% confidence intervals. Participation in collision sports was associated with a 20% increase in the risk of cognitive impairment using an MMSE score of 27 as the cutoff, which might correspond to a small (not trivial but possibly inconsequential) effect size (44, 45) despite this association not being statistically significant as indicated by the confidence interval containing one (cause-specific hazard ratio (csHR) = 1.20, 95% confidence interval (CI): 0.86, 1.68). In models stratified by sex (Web Table 1, available at https://doi.org/10.1093/aje/kwab177), the association between participation in collision sports and cognitive impairment for men was on par with the complete sample (csHR = 1.23, 95% CI: 0.86, 1.77). Among women, the association between participation in collision sports and cognitive impairment was stronger (csHR = 1.63, 95% CI: 1.22, 2.17); however, among 384 women in the analytical sample, only 1 reported participation in collision sports, and this result is likely an artifact of the model.

Association Between Early-Life Participation in Collision Sports and Adult Cognitive Impairment^a, Swedish Adoption/Twin Study of Aging, 1987–2014

Web Table 2 presents results from our sensitivity analyses, in which we evaluated the association between participation in collisions sports and cognitive impairment using 2 alternative cutoff points for the MMSE. The threshold for cognitive impairment classification from model 1 to model 3 was sequentially reduced; model 1 used a score on the MMSE of 24 or lower, model 2 used a score on the MMSE of 27 or lower (primary analysis; Table 2, main text), and model 3 used a score on the MMSE of 29 or lower. Similar to our primary analysis, in model 1—which used a cutoff of 24 points on the MMSE—participation in collision sports was associated with an increased risk of cognitive impairment, but the association was not statistically significant (csHR = 1.55, 95% CI: 0.85, 2.81). In model 3, which classified participants with cognitive impairment on the basis of an MMSE score of 29 or lower, participation in collision sports was associated with a 9% increased risk of cognitive impairment (csHR = 1.09, 95% CI: 0.65, 1.84). Similar patterns were observed when we used the Fine and Gray model (Web Table 3) in place of cause-specific Cox regression models to evaluate the associations between participation in collision sports and cognitive impairment. Our primary findings were supported by an additional analysis in which we restricted our sample to those without cognitive impairment in the survey waves prior to and immediately following their retrospective report of early-life sports participation (Web Table 4).

Web Figure 1 shows the probability of state occupancy (i.e., cognitive impairment or death) by participation in collision sports. The cross-over and relative agreement of cognitive impairment between adults who did and did not participate in collision sports is evident through age 85. The apparent differences after age 85 were statistically insignificant and likely due to issues related to small sample size and survival effects. Graphically, there appear to be differences in mortality among those who did and did not participate in sports. We evaluated the probability of death obtained from our model against period life tables from Sweden for the year 2000 obtained from the Human Mortality Database (46) and observed strong concordance for adults who did not participate in collision sports.

Growth curve modeling analysis

The results of the growth curve models for the cognitive domains are shown in Table 3. Stage 1 reports fixed effects for female sex, educational attainment, and participation in collision sports. Stage 2 features results from models that build on those shown for stage 1 by including linear and quadratic age interaction terms for participation in collision sports. In all models in stage 1 and stage 2, the intercept corresponds to the mean level of cognitive performance at age 65. Participation in collision sports was not significantly associated with baseline cognitive performance, nor was it associated with change in cognitive performance over time.

Association Between Early-Life Participation in Collision Sports and Adulthood Cognitive Abilities, Swedish Adoption/Twin Study of Aging, 1987–2014

DISCUSSION

We observed that participation in collision sports during adolescence was not significantly associated with cognitive impairment over a nearly 30-year period in a sample of Swedish twins. Further, participation in collision sports was not significantly associated with level of or change in trajectories of cognitive performance across 5 domains of cognition, including general cognitive ability, verbal ability, spatial ability, memory, and processing speed.

Our finding that adolescent participation in collision sports was not associated with cognitive health is consistent with findings from other recent, large epidemiologic studies of youth sports participants (21, 47, 48). These results are counter to an emerging literature reporting long-term neurodegenerative effects of sports-related head trauma among elite career-athletes. For example, 2 studies of retired professional football athletes with a concussion history showed that this group was at greater risk for early onset of Alzheimer’s disease, mild cognitive impairment, and depression (17, 49). Another study of former players in the National Football League found that earlier age at first exposure to football was related to poorer performance on tests of memory, executive function, and verbal intelligence quotient in adulthood (15). The contrast between present findings and conclusions drawn from prior research on selected samples of elite athletes highlights the importance of large epidemiologic studies for elucidating the risk of contact sports participation at all levels of play. Notably, elite athletes might differ from the broader population of amateur athletes in their exposure to head trauma (more prolonged period of exposure, higher impact forces) as well as other aspects of health (e.g., cardiovascular function (50)). It is also important to note that differences in findings across large epidemiologic studies and clinical-based investigations could be driven by the precision with which the exposure is measured. For example, prior clinical investigations used detailed surveys to ascertain information about concussion risk and exposure. In the present study, we were only able to assess whether respondents participated in collision sports. This discrepancy could contribute to the inconsistent findings reported in our study relative to the clinical literature on this topic.

To our knowledge, this is the first observational study to examine the long-term implications of early-life participation in collision sports on adulthood cognitive health among twin pairs. Although sample size limitations precluded us from conducting sufficiently powered analyses utilizing the twin study design (e.g., shared frailty models, discordance analysis), the sample itself was drawn from a large, population-based registry, whereas prior work has relied on small clinical samples in which self-selection into the sample could bias the study results, or studies that are constrained geographically (e.g., the Wisconsin Longitudinal Study). Moreover, whereas much prior work has focused exclusively on American football, utilizing the SATSA data allowed us to assess cognitive health in relation to a measure that spans all collision sports in which there could be a risk of head trauma. Additional strengths of our sample include the rigorous methods used to ascertain and adjudicate cognitive status in the SATSA, the wide scope of cognitive testing, and the near 30-year follow-up with mortality records.

Methodologically, our primary analysis relied on a competing risk framework to test the association between early-life participation in collision sports and cognitive impairment while accounting for mortality. Failure to account for the semi-competing risk of mortality could result in biased estimates if the risk profiles and corresponding mortality among those who did and did not participate in collision sports in early life differ, which they appear to in our study. Our supplementary analysis leveraged growth curve modeling to examine heterogeneity in the level and change in cognition across 5 domains as a function of early-life participation in collision sports. The extant literature examining associations between collision sports participation and cognitive health has mostly relied on cross-sectional studies, longitudinal studies with short follow-up, or studies focused exclusively on cognitive impairment or trajectories of cognitive health but not both. Examining these complementary outcomes, which span a range of cognitive domains and spectrum of impairment, enables us to better understand whether there might be global or domain-specific deficiencies in cognition associated with participation in collision sports.

The results of our study should be viewed in the context of its limitations. Our primary exposure was a self-reported indicator of whether the respondent participated in any sport that could involve a hit on the head. This measure does not inform the age or duration of participation, nor does it address variation in the risk of a hit on the head and its potential severity. Individuals who participated in sports at an earlier age or for a longer period of time as well as those who were at greater risk for more severe head trauma on the basis of the sport they participated in might vary with respect to their exposure and how it is associated with later-life cognition, but this cannot be observed in the present study. Without a direct measure of exposure to head trauma (e.g., medical record or self-report of traumatic brain injury), we were unable to assess the impact of brain trauma on cognitive outcomes. Moreover, given that the exposure was retrospectively reported among a sample of older adults, there is a concern regarding recall bias. We addressed this concern in sensitivity analyses in which we restricted our analysis to respondents without cognitive impairment in the survey waves prior to and immediately following their retrospective report of early-life sports participation. Results from these analyses supported our primary findings. Taken together, these limitations in measurement could contribute to our null findings. In addition, due to limited sample sizes, our study was not sufficiently powered to examine variation across zygosity or discordance. For example, our sample included 99 complete monozygotic twin pairs, 8 of which were discordant for participation in collision sports. Among the 170 dizygotic twin pairs, 21 twin pairs were discordant for sports participation.

The vast majority of youth athletes will not go on to have professional athletic careers, yet most of what is known about long-term risks of contact sports on later-life brain health comes from research on professional athletes. The intention of our investigation was to contribute to the growing body of literature examining the long-term effects on cognition of early-life participation in collision sports. Our null findings do not provide definitive evidence of no association. Instead, they add to the emerging conversations about early-life participation in collision sports using a unique sample and study design yet to be used in this context. Few, if any, studies on this topic have examined the association between early-life participation in collision sports and cognitive impairment while accounting for the semi-competing risk of mortality with extensive follow-up, in conjunction with cognitive domain-specific analyses. Obtaining more precise measurements of exposure to collision sports, perhaps through retrospective population-based surveys, is critical to advance our understanding of the extent to which participation might have long-term health implications as well as the mechanisms through which participation might operate. Finally, considering the established benefits of being physically active at all ages of the life span, it is necessary to view the potential harms of collision sports participation with the documented benefits.

ACKNOWLEDGMENTS

Author affiliations: Department of Demography, University of California, Berkeley, Berkeley, California, United States (Jordan Weiss); Moss Rehabilitation Research Institute, Elkins Park, Pennsylvania, United States (Amanda R. Rabinowitz); Computer Science and Artificial Intelligence Laboratory, Massachusetts Institute of Technology, Cambridge, Massachusetts, United States (Sameer K. Deshpande); and Department of Statistics, Wharton School, University of Pennsylvania, Pennsylvania, United States (Raiden B. Hasegawa, Dylan S. Small).

The Swedish Adoption/Twin Study of Aging has been supported by the National Institute on Aging (grants AG04563, AG10175), the MacArthur Foundation Research Network on Successful Aging, the Swedish Council for Working Life and Social Research (FAS) (award 97:0147:1B, 2009-0795), and the Swedish Research Council (award 825-2007-7460, 825-2009-6141). J.W. was supported by training grants awarded to the University of Pennsylvania (National Institutes of Health grant T32 HD007242) and University of California, Berkeley (National Institutes of Health grant T32 AG000246).

Data availability statement: The data used herein are archived through the National Archive of Computerized Data on Aging (https://www.icpsr.umich.edu/web/NACDA/studies/3843).

Conflict of interest: none declared.

REFERENCES