Anterior Cruciate Ligament Injury Risk in Sport: A Systematic Review and Meta-Analysis of Injury Incidence by Sex and Sport Classification

We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses⁵  (PRISMA) guidelines when conducting and reporting this systematic review and meta-analysis.

A systematic review of the current literature was performed using the electronic databases PubMed (1969–January 20, 2017) and EBSCOhost (CINAHL and SPORTDiscus; 1969–January 20, 2017) and the following search terms: anterior cruciate ligament AND injury AND (incidence OR prevalence OR epidemiology). Results were further limited to peer-reviewed articles written in English.

In addition to the electronic search, we contacted experts in the field for further suggestions and examined references cited in review papers to identify any other relevant articles for potential inclusion. Publication details from all studies identified in the literature search were exported to bibliographic software (Endnote X7; Clarivate Analytics, Philadelphia, PA).

Given the large number of identified studies, a single author (A.M.M.) performed the initial screening of articles for inclusion. Any gray areas were discussed with the second author (D.K.S.), and any disagreements were decided by the senior author (G.D.M.). Articles were screened first by title, second by abstract, and third by full text according to the inclusion and exclusion criteria. We included articles in which the total number of ACL injuries and the total number of athlete-exposures (AEs) were reported by sex and the data were provided in such a way that the number of ACL injuries by sex could be calculated. We excluded articles that included further analyses on previously reported prospective studies, were written in languages other than English, or were review papers. Full texts were retrieved when the title or abstract met the selection criteria or when the status could not be determined from the title and abstract alone.

The primary variables extracted were the sport classification, number of ACL injuries for each sex, and person-time in AEs for each sex. Sports were classified as follows: collision (contact with an opponent or object is inherent), contact (contact with an opponent or object is acceptable), limited contact (contact with an opponent or object is discouraged), and noncontact (contact with an opponent or object is unexpected; Table 1). For each sport classification, we calculated the overall ACL injury rate and separate IRs for men and women. The IR ratio (IRR) between men and women was subsequently calculated using only data from studies in which injury-risk data were reported for both men and women to allow direct comparisons. Additional extracted data included year of publication, sport, sport type, and level of play. One author (A.M.M.) recorded all pertinent data from the included articles, and another author (D.K.S.) independently reviewed those data for accuracy and completeness.

The reported person-time unit was not uniform across studies. Therefore, to establish a common metric, we tabulated AEs. When the number of player-hours was reported, the number of AEs was estimated by dividing player-hours by 2. The assumption for converting player-hours to AEs was that each AE (1 game or 1 practice) on average would last about 2 hours. In addition, not all authors reported the number of ACL injuries by sex; instead, they provided IRs by sex. For these studies, the number of AEs and the reported IRs were used to calculate the number of ACL injuries by sex (number of ACL injuries by sex = total AEs by sex × the rate numerator by sex/the rate denominator by sex).^6–8  For studies in which the number of ACL injuries by sex could not be estimated, we e-mailed the authors to gather those data. If they did not have access to the information or did not respond, the study was excluded from the meta-analysis.^9–15 

Included studies were critically appraised independently by 2 authors (A.M.M., D.K.S.). Given that most included articles described observational cohort studies that did not include an intervention, traditional checklists were not appropriate. After a thorough search for tools to appraise observational cohort studies, we decided that the tool best suited to be used quantitatively was the Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies.¹⁶  This tool, available through the National Institutes of Health (Bethesda, MD), assesses criteria such as participation rate, whether exposure data were collected before the outcome, whether the time frame was sufficient to allow for the outcome to occur, and the number of participants lost to follow-up after baseline. If a criterion was met, the item was scored as 1. If it was absent or not reported, the item was scored as zero. The maximum score possible was 14. Items were scored independently by 2 authors (A.M.M., D.K.S.). These authors discussed any discrepancies in scoring. For discrepancies that could not be resolved, a third author (G.D.M.) was consulted for arbitration. Given that the included studies with interventions were treated as cohort studies in the analyses, they were assessed with the same tool, which allowed for quality comparisons across all included studies.

The number of included studies per analysis varied. For the total IR, any study in which authors reported the rate of either sex was included. For the IR by sex, any study in which the authors reported female or male rates was included for the respective analyses. Only studies that included both female and male athletes were used to calculate ratios. The ACL injury IR in noncontact sports comprised sports with marked differences in ACL injury IRs. Given that several outliers were present, we subdivided the category into sports that did and sports that did not include a fixed-object and high-impact landing. These latter sports were removed from the noncontact category, and a new fixed-object, high-impact rotational-landing (HIRL) category was created. Fixed-object HIRL sports were defined as noncontact sports that included high-impact landings from fixed objects, such as beams, vaults, and obstacles. Injury IRs for the individual studies were summarized in forest plots for the following groups by total, female, and male IRs: collision, contact, limited-contact, noncontact, and fixed-object HIRL sports. These rates were multiplied to calculate ACL injury IRs per 10 000 AEs in each respective group. Incidence rate ratios for women versus men were calculated for each group and summarized in forest plots.

Injury data were analyzed using R (version 3.3.2; R Foundation for Statistical Computing, Vienna, Austria) and the R packages meta and metafor with the functions metarate for IR and metainc for IRR weighted for individual study size. When AEs but no events (ACL injuries) were present, a continuity correction was applied. The default value for the continuity correction, 0.5, was used to calculate individual point estimates and the 95% confidence interval (CI) and to conduct a meta-analysis based on the inverse variance method. We set the α level at .05.

The electronic literature search yielded 3774 abstracts for initial review. After duplicates were removed, a total of 1300 unique titles remained. We screened the titles and abstracts and removed 1155 articles for lack of relevance to the research. The remaining 145 articles were manually cross-referenced, and experts were consulted to identify additional relevant articles, resulting in the inclusion of 17 more articles. Full texts of these 162 articles were obtained and assessed for the inclusion and exclusion criteria. We contacted the corresponding authors of the included articles for additional information as needed. At the end of the search, 36 articles were included in the study.^{1,6–8,17–48}  An outline of the literature review process is presented in Figure 1. The data that were extracted for each analysis and can be used to determine which studies were included in each analysis are shown in Table 2.

In collision sports, the total IR of ACL injury among female and male athletes combined was 1.29/10 000 AEs (95% CI = 1.07, 1.54; P < .01, I² = 95.0%; Figure 2). The injury IR among female athletes was 2.10/10 000 AEs (95% CI = 1.12, 3.96; P < .01, I² = 84.0%; Figure 3) and among male athletes was 1.12/10 000 AEs (95% CI = 0.94, 1.33; P < .01, I² = 93.0%; see Supplemental Figure 1, available online at http://dx.doi.org/10.4085/1062-6050-407-16.S1). We observed no difference between sexes for the ACL injury IR (IRR = 1.14; 95% CI = 0.68, 1.92, P = .63; I² = 0%; see Supplemental Figure 2).

The total IR of ACL injury in contact sports was 1.51/10 000 AEs (95% CI = 1.31, 1.75; P < .01, I² = 90.0%; see Supplemental Figure 3). The injury IR was greater among female (1.88/10 000 AEs; 95% CI = 1.61, 2.20; P < .01, I² = 88.0%; see Supplemental Figure 4) than among male (0.87/10 000 AEs; 95% CI = 0.69, 1.11; P < .01, I² = 84.0%; see Supplemental Figure 5) athletes. We observed a difference between sexes for the ACL injury IR (IRR = 3.00; 95% CI = 2.70, 3.34; P < .001, I² = 4.0%; see Supplemental Figure 6).

In limited-contact sports, the total IR of ACL injury was 0.48/10 000 AEs (95% CI = 0.33, 0.70; P < .01, I² = 91.0%; see Supplemental Figure 7). The injury IR in female athletes was 0.71/10 000 AEs (95% CI = 0.50, 1.01; P < .01, I² = 84.0%; see Supplemental Figure 8) and in male athletes was 0.29/10 000 AEs (95% CI = 0.18, 0.48; P < .01, I² = 63.0%; see Supplemental Figure 9). The IRR was calculated using only data from Mountcastle et al,²⁹  as data comparing injury rates among women and men in this sport type were not available. We observed no difference between sexes for the ACL injury IR (IRR = 1.21; 95% CI = 0.35, 4.20; P = .77, I² = 0%; see Supplemental Figure 10).

The total IR of ACL injury in noncontact sports was 0.25/10 000 AEs (95% CI = 0.10, 0.65; P < .01, I² = 85.0%; see Supplemental Figure 11). The ACL injury IR among female athletes was 0.36/10 000 AEs (95% CI = 0.14, 0.96; P < .01, I² = 74.0%; see Supplemental Figure 12) and among male athletes was 0.21/10 000 AEs (95% CI = 0.07, 0.62; P < .01, I² = 70.0%; see Supplemental Figure 13). We observed no difference between sexes (IRR = 1.49; 95% CI = 0.79, 2.79; P = .22, I² = 0%; see Supplemental Figure 14).

In fixed-object HIRL sports, the total IR of ACL injury was 2.62/10 000 AEs (95% CI = 1.44, 4.75; P < .01, I² = 89.0%; see Supplemental Figure 15). The ACL injury IR among female athletes was 4.80/10 000 AEs (95% CI = 2.37, 9.70; P < .01, I² = 89.0%; see Supplemental Figure 16) and among male athletes was 1.75/10 000 AEs (95% CI = 0.41, 7.48; P < .01, I² = 89.0%; see Supplemental Figure 17). We observed a difference between sexes (IRR = 5.51; 95% CI = 2.80, 10.82; P < .001, I² = 0%; see Supplemental Figure 18).

Most studies were of moderate quality (Table 3). Three studies fulfilled 75% or more of the criteria, and 33 studies fulfilled 50% or more of the criteria. The remaining 3 studies fulfilled fewer than 50% of the criteria and were deemed to be of low quality. Studies may have received lower scores for lack of reporting information, such as the total number of eligible individuals, how outcomes were measured, and attrition. Overall, the risk of bias was deemed to be moderate.

The purpose of our study was to quantify sex differences in ACL injury risk for sports with various amounts of contact. Female athletes participating in contact and fixed-object HIRL sports had greater ACL injury IRs than their male counterparts. In contrast, the ACL injury IRs for collision, limited-contact, and noncontact sports did not differ between sexes. The findings from this meta-analysis support a previous report⁴  indicating that the amount of athlete-to-athlete contact inherent to a sport was correlated with the rate of ACL injury in both male and female athletes. However, adding the fixed-object HIRL category suggested that sports such as gymnastics and obstacle-course races may result in the highest rates of ACL injury.

Identifying the ACL injury IR associated with fixed-object HIRL sports is especially relevant as it pertains to military training and activities. Over a 7-year period, the IR of ACL injury in US military members of all services was 3.09/1000 person-years for men and 2.29/1000 person-years for women.⁴⁹  Investigators⁴⁹  noted that service members were at 10 times greater risk of ACL injury than the general population. This increased risk may be partially explained by participation in fixed-object HIRL activities. In contrast to our findings, Owens et al⁴⁹  did not find women to be at greater risk of ACL injury than men; however, they reported person-years because they did not have exposure information. In addition, men outnumbered women in their study⁴⁹  and, thus, had higher rates of ACL injury. Military service members, especially those participating in regular training that includes fixed-object HIRLs, may benefit from integrative neuromuscular training to mitigate their risk of ACL injury.

In addition to the military application, our findings related to fixed-object HIRL sports are also relevant considering the advent of recreational obstacle-course races (eg, Tough Mudder, Spartan, BattleFrog). These races are based on military training obstacle courses. Currently, no information about the rates of ACL injury associated with these races is available, but our results suggest that participants should exercise caution. For gymnastics, our findings indicated that the unique demands of the sport, including both implement-based activity and high-impact landings after full-body rotation, distinguish the sport from other noncontact sports regarding the ACL injury risk. Hootman et al¹  found that football, a collision sport, resulted in the greatest incidence of ACL injuries in collegiate male athletes. Our findings indicated that fixed-object HIRL sports resulted in ACL injury IRs that were similar to those of collision sports in men (1.75/10 000 versus 1.12/10 000 AEs). The ACL injury IR was more than 3 times greater among women than among men for fixed-object HIRL sports. Considering the likely mechanisms of injury (landing with rotation, stiff-legged landing), this disparity highlights the neuromuscular deficits typically demonstrated by female athletes.⁵⁰  Therefore, female athletes participating in fixed-object HIRL sports may benefit the most from preventive strategies.

We also found that female athletes participating in contact sports sustained ACL injuries at 3 times the rate of male athletes in these same sports (IRR = 3.00). These findings are similar to IRRs previously reported⁴  for male and female collegiate basketball and soccer players, which were approximately 3.6 and 2.8, respectively. However, ACL injury IRs did not differ between women and men for collision sports. The lack of a difference in ACL injury IRs between women and men in collision sports and between women in collision and contact sports may be partially explained by the lack of collision-sport participation by women. When participation was equal among women and men (contact sports), the greater ACL injury IR among women was evident. It is possible that not enough studies were available in which researchers investigated ACL injury incidence among both female and male collision athletes to detect a difference in the rates where one truly exists (ie, low statistical power).

In contrast, the ACL injury IRs for men in collision and contact sports differed (1.12/10 000 and 0.87/10 000 AEs, respectively). The sports included in these categories are similar because they require cutting and pivoting, which are dynamic maneuvers known to contribute to noncontact ACL injury mechanisms. Again, these combined findings further support the idea that neuromuscular deficits may contribute to the greater ACL injury IR among women. Although speculative, it was also possible that the men's decreased IR in collision sports compared with contact sports was due to direct-contact blows to the knee based on the nature of the sports.

Whereas our research may provide a robust estimate of sex differences in ACL injury IRs among sport types, it had limitations. The common metric of AE had to be estimated in some cases when exposure was provided in player-hours. This was necessary to include the maximum amount of data possible. As mentioned, we assumed that 2 player-hours were approximately equal to 1 AE, and we used this assumption to generate estimates of AEs. This assumption may have resulted in the overestimation or underestimation of exposure, depending on the sport. We used broad inclusion criteria to capture the greatest amount of information for generating these estimates. The included articles ranged in study quality, and the estimates are only as strong as the evidence on which they are based. However, we believed it was important to capture a wider range of studies to obtain a truer, more robust picture of ACL injury incidence among athletes. In addition, heterogeneity was relatively high (>75%) for the point estimates, indicating that populations that were grouped together may actually have differed. However, this was expected, as different sports have different demands that change the risk of sustaining an ACL injury. Moreover, heterogeneity for the rate ratios was low, and in some cases was 0%, indicating that the results were consistent and potentially generalizable. Given that female participation in collision sports was less prevalent than male participation, we included relatively few studies in which differences in ACL injury IRs between sexes were investigated. We could not control for variables known to contribute to ACL injury, including surface type, anticipation (anticipated event versus unanticipated event), or mechanism of injury (contact versus noncontact) because of a lack of information. Finally, we did not stratify by age or level of play, as those were not aims of this study.

To address these limitations, future researchers should report their findings in the most accurate units possible (player-hours) or should make both player-hours and AEs available to provide the opportunity for meta-analysis. Given that prospective designs allow for real-time data capture, investigators conducting future research in injury epidemiology should use prospective designs. Developing a standard and comprehensive checklist for criteria that should be met when performing or designing a prospective observational cohort study would provide a guide for researchers to achieve maximum study quality. This meta-analysis should be repeated in the future when more ACL injury data are available to permit comparisons of incidence rates among female and male athletes participating in collision and limited-contact sports. Finally, researchers should establish ACL injury IRs within each sport type while controlling for confounding variables, including age and level of play.

The incidence of ACL injury is associated with the nature of player-to-player contact inherent in the sport. Female athletes had greater ACL injury IRs than male athletes in contact and fixed-object HIRL sports. The latter sports category had the highest ACL injury IRs for both sexes, which might suggest the need for a new sport type to identify athletes at the highest risk of ACL injury. Future strategies aimed at reducing the risk of ACL injury may benefit from considering and integrating sport-related perturbation that mimics contact exposure to better equip athletes with preprepatory and avoidance techniques.

Supplemental Figures. Series of forest plots for incidence rate ratios.

Found at DOI: http://dx.doi.org/10.4085/1062-6050-407-16.S1