Article Text
PDF
Abstract
Aim/objective The objective of this study was to provide a meta-analysis examining the effects of football heading.
Design Meta-analytical review on football heading effects on neurocognitive performance, cognition and symptom reports.
Data sources Combinations of the key terms were entered into the following electronic database search engines: Cochrane Libraries, PyscARTICLE, PyscINFO, PubMed, ProQuest, SPORTDiscus and Web of Science on 7 July 2016.
Eligibility criteria for selecting studies The following inclusion criteria were used to determine eligibility for studies: (1) the study examined and reported on soccer athletes; (2) the population's age, sex and sport position was described; (3) cognitive function, symptoms, balance or other outcomes were quantitatively measured; (4) football heading exposure was quantitatively measured between at least two groups and (5) the study was written in the English language after December 1979.
Results The literature search process identified 467 unique studies. After applying exclusion criteria, 28 studies remained. Included studies had a total of 2288 participants (female participants =933, male participants =1355), aged 13–70 years. The overall results of random effects modelling of football heading were found to be inconclusive across all outcomes, groups and time points. No moderating variables related to methodological, sample or study characteristics were supported in the analysis; age was a potential moderating variable.
Summary/conclusions We provide the first meta-analytical review of football heading effects aggregated from multiple studies and extended findings from a recent systematic review of the effects of football heading. Our analysis indicates no overall effect for heading a football on adverse outcomes.
- Concussion
- Meta-analysis
- Football
- Soccer
Statistics from Altmetric.com
Introduction
Football (soccer) is the most popular sport in the world, attracting more than 265 million participants worldwide.1 Concerns about the potential adverse effects of purposeful heading on cognitive function, postural control and symptoms are present.2–5 Mandatory bans on heading to prevent potential acute and chronic effects have been suggested.6 Recently, heading was reported to cause 25–30% of concussions in football and 62–78% of concussions were from incidental head to head contact.7 Football is unique compared with other contact sports in that players use their unprotected heads to pass, shoot, direct and control the ball. Some high-profile football players have recently started wearing soft, protective headbands to minimise impacts from heading the ball; however, no empirical data supports the effectiveness of these headbands.8
Football heading technique is considered an important part of offensive and defensive play9 and is frequently implemented by players. Yet, heading is hypothesised to negatively affect cognitive function and symptoms. Early studies supported this hypothesis, 10 ,11 but were retrospective, focused on retired professional players, and omitted concussion history. More recently, studies have reported no relationship between heading exposure and cognitive performance in adolescent football players.12–16 In a recent review of the effects of heading and the incidence of concussions in football, the mechanism of injury and the neurocognitive implications of concussions in football indicated that the long-term effects of heading were similar to those effects seen in athletes with a history of concussion.17 Many of the studies examined, however, failed to record concussion history of the participants in the study, thus limiting the accuracy of the assessment. Additionally, many of the studies reported conflicting results, again suggesting the need for further exploration as recent literature continues to be inconclusive.
No single study has provided conclusive evidence for the relationship between heading and cognitive and other deficits. Researchers have yet to conduct a standardised review of the evidence regarding the cognitive and other effects of football heading. The purpose of this study was to perform a meta-analytical review of research on the effects of heading a football on neurocognitive performance, cognition and symptom reports. A secondary purpose of this study was to examine the role of potential moderating variables for the effects of football heading including age, gender, level of play and football heading exposure type and timing (acute vs chronic). We expected that any effect would be stronger in athletes of older age and higher level of play, as these players would have incurred more intense exposure to football heading over a longer period of time.
Methods
Database keywords and search strategy
Our literature search strategy used key words to locate and identify relevant research. Combinations of key terms: concussion, mild traumatic brain injury, mTBI, soccer, European football, soccer heading, header(s), concussion symptoms, cognitive impairment and neurocognitive test/performance were entered into electronic database search engines: Cochrane Libraries, PsycARTICLE, PsycINFO, PubMed, ProQuest, SPORTDiscus and Web of Science on 8 January 2015 and 7 July 2016. Literature search findings were recorded and screened by title and abstract to determine inclusion in or exclusion from the current investigation. During the screening process, if relevant information was insufficient, the article was retrieved to complete the review (see online supplementary appendix 1).
supplementary appendix
Inclusion and exclusion criteria
We set these inclusion criteria: (1) the study examined and reported on football athletes; (2) the population's age, sex and sport position were described; (3) cognitive function, symptoms, balance or other outcomes were quantitatively measured; (4) football heading exposure was quantitatively measured between at least two groups and (5) the study was written in the English language after December 1979. We emailed the authors requesting additional information when studies contained insufficient information or data. We set these exclusion criteria: (1) no full-text version of manuscript available; (2) no response was received from study authors for additional study data and (3) study did not report data regarding football heading.
The University of Pittsburgh Institutional Review Board does not require approval to conduct a meta-analysis, as the research is not considered to include human subjects and the information can be found from publically available databases.
Coding procedures and data extraction
Standard coding forms were developed. Article information was extracted and divided into three categories: (1) methodological characteristics, (2) participant characteristics and (3) study characteristics. Methodological characteristics: study conduct, study design (experimental, descriptive or longitudinal), outcomes measured (physical, cognitive, symptoms or multiple outcomes), test location (laboratory/classroom, field or other), exposure (acute, chronic or not reported), personnel training (athletic trainer, physician, researcher or other), on-field markers (signs/symptoms, Sport Concussion Assessment Tool (SCAT), concussion card, physician/athletic trainer or other), and symptom report (Post-Concussion Symptom Scale (PCSS), Post-Concussion Symptom Inventory (PCSI), SCAT or other). Participant characteristics: sport type; sex; age with range, mean and SD; and country. Study characteristics: reporting methods (self, objective, or combined), grant funded (funded or unfunded) and study status (published or unpublished).
Independent review for bias avoidance
Three independent reviewers coded and rated studies according to methodological, participant and study characteristics. After all studies meeting inclusion criteria had been coded, the independent results were compared for agreement. Disagreements were analysed to determine the type of error and classified as either factual disagreements (transcription errors) or interpretative errors (vague or imprecise study information). A researcher who coded the study reviewed interpretative disagreements and the decision was based on a simple majority.
Data analysis
Outlier and publication bias: Determining outliers and influence of publication bias on overall results was the initial step before data analysis. Outliers were identified by residual values (similar to z-scores) ∼2 SDs (±1.96) above or below the study's mean effect size. Inclusion was based on overall results within the 95% CI and a significant summary effect size. Three procedures identified and controlled for publication bias: (1) review of the funnel plot;18 (2) Fail-safe N calculation;19 (3) ‘Trim & Fill’ method.20
Effect size calculations: All analyses were performed using Comprehensive Meta-Analysis V.2 software.21 The unit of analysis was the study. Studies that contained multiple outcomes, the different metrics were averaged into a summary effect and an inverse weighting method was then applied to improve precision of data analyses.22 Hedges g is used to correct small sample sizes (k<20) and was the effect size metric used.23–25 The rationale for using Hedges g was based on the use of additional analyses (outcomes and moderator) that contained fewer than the recommended number of studies and consistency in reporting results.
Effect size interpretation: Interpretation of statistical information is based on Cohen's criteria for evaluating small (≥0.20), medium (≥0.50) and large (≥0.80) effects using the d metric. To provide a consistent interpretation of results, effect sizes were coded as negative that reflected smaller outcome scores for groups exposed to heading, and positive effect sizes were indicative of increased scores for control groups (those not exposed to heading). For research using pre–post designs (one group), negative effects represented larger pretest score and positive effects signified larger post-test scores. To provide a comprehensive review and direction for future research, data for outcomes and moderators with sample size k≥5 are presented to estimate effect.
Error modelling: Random effects model was prespecified for the current investigation due to the expected variability between studies.
Outcome analyses: To provide consistent reporting of the summary effect for differently named but similar outcomes, all variables were defined: attention/concentration, blood biomarkers, fine motor control, impulse control, IQ, reaction time, personality, processing speed, symptoms, verbal memory and visual memory.
Subgroup analyses: Three statistical approaches were used to evaluate heterogeneity and provide a comprehensive approach to interpreting results: QTotal (QT) value based on χ-square (χ2) distribution; τ) value and I2 value. Significant QT statistic was then categorised into QBetween (QB) and QWithin (QW) values and significant QB values (p<0.05) required statistical techniques (ie, Student's t-test or analysis of variance (ANOVA)) to determine subgroup differences.25 ,26 The τ2 statistic provided an estimate of total variance between studies. Small subgroup sample sizes (k≤5) may have influenced the precision of τ2; therefore, a pooled estimate of variance was used for all calculations.26 The I2 statistic reflected the overlap of CIs and can be interpreted as low (25%), moderate (50%), or high (75%) values of the total variance attributed to covariates.27 I2-values closer to zero indicate random variability, whereas, larger values suggest variance can be explained by subgroup analyses.
Results
Literature search and coding
The literature search process identified 1635 potential studies, from which 467 (29%) unique (ie, not duplicates) studies were identified for the initial screening (see figure 1). After screening of the titles and abstracts for a focus on football heading, 326/467 (70%) were excluded, 34/467 (7%) only contained abstracts and were not available in full text and therefore excluded; leaving 141/467 (30%) studies that progressed to the inclusion screening stage. A total of 120/141 (85%) of these studies failed to meet inclusion criteria, leaving 37/141(26%) studies for analysis. On further review, we identified 8/37 (22%) studies had insufficient data for effect size calculations and we were unable to obtain additional data from study authors. Therefore, these 8 articles were excluded, reducing the total number of studies to 29/37(78%). A total of 14 disagreements were identified during coding, all of which were interpretative in nature. A third reviewer provided feedback on the articles and all 14 were included based on simple majority. Examination of coding and data extraction determined that there was a high inter-rater reliability agreement (Κ=0.99).
PRISMA flow chart.
An overview of the studies included in the current investigation is provided in table 1. In summary, studies included were from the years of 1990 to 2016 with a total of 2288 participants (female participants =933, male participants =1355) from six countries and ages ranging from 13 to 70 years. Twenty-two of the 28 (76%) papers used experimental research designs and 28 of 29 (97%) were published in research journals and involved the study of various performance, cognitive and affective outcomes. Measures used to collect data on football participants included self-report and objective tests exploring acute and chronic incidents of football heading patterns. The quality of studies was low to moderate as a risk of bias assessment determined that only two studies randomised participants and ensured allocation of concealment, no studies used blinding processes for participants, researchers or outcomes and the majority of studies provided information regarding incomplete outcomes data and reported data for all outcomes being studied. Studies not reporting or providing information concerning criteria for risk of bias were considered to not have met the criteria.
Coding for Studies Meeting Inclusion Criteria
Outliers and publication bias
Data were screened for outliers and three studies (Andelinovic et al 29 z=−3.01, Bamac et al 30 z=−2.51, Kawata et al 38 z=−2.61) were identified as having large residual values (>±1.96). A sensitivity analysis was performed in comprehensive meta-analysis (‘one study removed’ procedure) and indicated an increase in effect (g=−0.04, p>0.05) for the largest outlier. Based on these findings, all outliers were retained for analysis as the results were within the 95% CI and non-significant. Publication bias was questionable as there was an equally distributed funnel plot and the Trim and Fill procedure did not add additional studies to the right (positive side) of the mean effect. The Fail-safe N calculation was zero, as the null hypothesis was accepted as a potential explanation of these results.
Effect sizes
The overall results of random effects model for heading a football across all outcomes, groups and time points were found to be inconclusive (k=29, g=−0.09, 95% CI −0.190 to 0.009, p>0.05). Studies included in the current investigation had a significant heterogeneous distribution (QTotal=43.06, p<0.05), however, the variance between studies (τ2) and overlap of CIs (I2) were inconsistent presenting small and large variability in scores. Smaller variability scores suggested additional procedures (ie, moderator analyses) to explain the variance within and between studies were not needed. Figure 2 provides the forest plot and summary statistics for studies included in the current investigation.
Soccer heading studies meeting inclusion criteria.
The largest effect sizes were for reaction time and blood biomarkers.30 However, only three studies included reaction time, thus the results should be cautiously interpreted. Effect size for age was not significant. Seventeen studies sought to determine the impact of football heading on neurocognitive performance, including attention/concentration (k=11), impulse control (k=2), reaction time (k=3), processing speed (k=10), verbal memory (k=16) and visual memory (k=11).8 ,13–16 ,28 ,33 ,34 ,36 ,37 ,42–47 ,49 ,50 Generally, the effect sizes were small, with the largest effect supported for reaction time (g=0.28).
Outcomes analyses: All sets of outcomes were marginal positive and negative non-significant values (Z>0.05, p>0.05) that also included CIs containing zero, which were indicative of accepting the null hypothesis. Outcomes variables ranged from positive values for impulse control (g=0.14), verbal memory (g=0.07) and attention/concentration (g=0.02), to negative values for processing speed (g=−0.08), visual memory (g=−0.02), symptoms (g=−0.13) and fine motor control (g=−0.06) (table 2). Larger negative effect sizes were found for reaction time (k=7, g=−0.41) and blood biomarkers (k=2, g=−0.36).
Football heading outcome analyses
Moderator analyses: A secondary purpose of the current investigation was to analyse moderating influences of outcomes. Results suggest that there were no moderating variables related to methodological, sample, or study characteristics. Studies within the subgroups experimental designs (Z=−1.99, p<0.05), cognitive outcomes (Z=−2.08, p<0.05), male only samples (Z=−2.44, p<0.05) and professional soccer players (Z=−2.76, p<0.05) (table 3) suggested significant effects within each subgroup (similar effect sizes); however, between-subgroup comparisons were inconclusive. Most of the moderating variables identified in the current investigation had a smaller degree of variability as interpreted by the Q, τ2 and I2 statistics. An important point to highlight is that I2 measured inconsistency across study findings, not study dispersion and the current investigation's results suggested a moderate need to explain study variance.
Subgroup analyses
Of the potential moderating variables for the effects of football heading, including age, gender, level of play and football heading exposure type and timing (acute vs chronic), age was the only potential moderator that, though insignificant, trended towards impacting outcomes following football heading.
Discussion
General discussion
The current study was the first meta-analytical review of the empirical evidence for the potential adverse effects of football heading. Our findings, similar to those of Maher et al,17 suggest there is no conclusive evidence that heading is associated with adverse effects, specifically exploring performance on neurocognitive measures and presence of concussion symptoms. We hypothesised any effect would be moderated and therefore stronger in athletes of older age and higher level of play. Age trended towards impacting outcomes following football heading; however, no significant findings to this effect were found. Higher level of play also had a slight trend suggesting professional football players had decreased outcomes; however, this effect was based on only two studies and did not control for concussion history, therefore the interpretation is again limited. Most results of this meta-analysis are not easily interpreted due to the few number of studies in the analysis. In summary, outcomes that have positive and negative CIs suggest inconsistent findings on the effects of heading a football. While no publication bias was noted, if a publication bias did exist, it would lean towards publication of positive findings.
The results of the current meta-analytical review of the literature surrounding the effects of football heading on neurocognitive performance and presence of concussion symptoms gives no conclusive evidence that heading is associated with adverse effects. The findings of the current review suggest the inconclusive results of previous research were largely study and sample specific, failing to consider concussion history.
Strengths and limitations
By combining studies into one large, comprehensive and critical statistical analysis, the current findings can help to clarify the current state of the literature with regard to this topic. However, little research has been conducted on the effects of football heading and much of what has been published lacks scientific rigour. Moreover, recent preliminary findings using advanced neuroimaging techniques suggest that the residual effects of football heading may be measureable,39 ,41 ,51 although it remains unclear whether such imaging results are correlated with any functional changes. As neuroimaging, blood biomarker and more comprehensive and sensitive clinical markers are developed, the results of a future meta-analysis on the effects of football heading may result in different outcomes. The current study was limited to research conducted after 1979. This time period was selected to exclude studies criticised for not accounting for concussion history and other factors known to influence cognitive impairment and symptoms.10 ,11 Another limitation of this study is the small number of studies with sufficient data to be included in a meta-analysis. This problem is not unique to football heading research,52 and further highlights the importance of the need for additional high-quality research and publishing sufficient data to allow for such analyses. We recommend these data be included in appendices or online resource links.
Implications and future directions
Our findings suggest any restrictions on football heading to prevent adverse effects may be premature. The potential effects in professional players suggest closer examination is warranted to determine if intense exposure to heading over a longer period of time needs evaluation. Efforts to monitor heading exposure and related outcomes in this potential at-risk group are recommended. Moving forward, more rigorous (ie, controlling for concussion) and large-scale prospective studies of the effects of football heading are needed. Future research should include and compare sexes, age and skill groups. Researchers should explore the potential effects of new football heading guidelines, such as those recently implemented by US Soccer, that limit heading in youth, on outcomes including heading exposure, effects and incidence of heading-related concussions. Such examination would inform more evidence-based guidelines moving forward.
Conclusion
The current findings indicate that there was no overall effect for heading the football on adverse outcomes. Any effects for heading may be limited to professional players who have had substantial levels of exposure to heading over long careers. However, the evidence is scant even in this at-risk group. As such, additional high-quality research on the effects of exposure to football heading is warranted. In the mean time, policy decisions regarding guidelines for football heading should be informed by the current findings as well as findings from future analyses of multiple studies rather than select findings from a single study.
What is already known?
Heading the football results in 25–30% of concussions in football.
Research on the effects of football heading has resulted in contradictory and inconclusive findings.
No meta-analysis has been performed thus far on research examining potential effects of football heading on symptoms and cognitive function.
What are the findings?
No moderating variables related to methodological, sample or study characteristics.
Age was the only potential moderating variable.
No conclusive evidence that heading is associated with adverse effects in the current meta-analysis sample.
No conclusive finding for the effect of football heading on concussion symptoms.
Acknowledgments
The authors would like to thank Kari Hartbauer, B.S., Hannah Makarevich, B.S., Kishan Shah, B.S. and Michael Scutella for their assistance during the article screening process.
References
Footnotes
Contributors AK designed the work, interpreted data, drafted the Introduction and Discussion, revised the manuscript and approved the final version. RB analysed and interpreted data, drafted the Methods and Results sections, revised the manuscript and approved the final version. SC interpreted data, drafted the Introduction and Discussion sections, revised the manuscript and approved the final version. JMD acquired and interpreted the data, drafted the Methods and Results sections, revised the manuscript and approved the final version. LS acquired and interpreted data, drafted the Introduction section, revised the manuscript and approved the final version. VR acquired and interpreted data, drafted the Abstract, revised the manuscript and approved the final version. MC interpreted data, drafted the Discussion, revised the manuscript and approved the final version.
Funding This research was supported in part by a grant to the University of Pittsburgh from the National Institute on Deafness and Other Communication Disorders (1K01DC012332-01A1).
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.