Do knee abduction kinematics and kinetics predict future anterior cruciate ligament injury risk? A systematic review and meta-analysis of prospective studies

Anterior cruciate ligament (ACL) injury is a common injury in team sports [1, 2] and often leads to serious consequences for the individual, including pain, functional limitations, reduced quality of life and lower activity levels [3, 4] that may persist several years post injury [4]. There is also an increased risk of developing early-onset osteoarthritis of the knee [5].

Most ACL injuries occur during non-contact episodes [6], typically within 50 milliseconds after foot contact, with the foot planted on the ground with a nearly extended knee together with trunk lean and knee abduction [7, 8]. The main function of the ACL is to provide mechanical stability to the knee during movements by preventing anterior tibial translation and rotational load [9, 10]. Several in vitro studies have also shown that the knee abduction moment is a major contributor to ACL strain and is, thus, suggested to play an important role in the ACL injury mechanism [11‐13]. Added to this, some studies have reported individuals with ACL deficiency to exhibit an increased knee abduction angle compared to non-injured individuals [14‐16]. This, together with the result of one early study establishing a relationship between increased knee abduction angle and knee abduction moment, respectively, and a higher risk of ACL injury in women [17], have given rise to knee abduction being widely accepted as an undesirable movement pattern [6, 8]. That women are reported to perform functional tasks with greater knee abduction than men [18], as well as having a higher risk of sustaining an ACL injury [1] has further perpetuated this hypothesis.

Based on the evidence-based reasoning presented above, numerous studies have been conducted to (1) determine the factors that contribute to knee abduction during weight-bearing activities [19], and (2) to incorporate exercises to reduce knee abduction into ACL injury prevention programs [6, 8]. However, the evidence for an association between knee abduction kinematics and/or kinetics and ACL injury risk seems to be conflicting [17, 20] and to date the findings from all studies investigating knee abduction as a risk factor for future ACL injury have not been synthesized. Thus, the aim of this study was to systematically review knee abduction kinematics and kinetics during weight-bearing activities at baseline as a possible risk factor for future ACL injury development.

A systematic review and meta-analyses were conducted according to the PRISMA guidelines. The study protocol was pre-registered (PROSPERO CRD42017067254; n.b., knee abduction kinetics were added to the protocol after registration).

The result from this systematic review and meta-analysis revealed no association between baseline knee abduction kinematics or kinetics during vertical drop jumps or squats and the risk of sustaining a future ACL injury. No studies were available in other weight-bearing tasks. Our conclusions are based on a large sample (1979 participants across 8 studies), with low to high heterogeneity, and were unaffected in our sensitivity analyses suggesting that our findings hold true irrespective of participant age, sex, or movement task.

A greater knee abduction angle and/or knee abduction moment during weight-bearing activities has commonly been suggested to represent undesirable mechanics and contribute to future ACL injury [6, 8]. Yet, across the 8 studies included in our meta-analyses, we found no difference in 2D knee abduction angle, 3D knee abduction angle, MKD or peak knee abduction moment at baseline between those who sustained a future ACL injury and those who did not. In addition to the possibility that knee abduction kinematics and kinetics are not at all associated with ACL injury risk, one explanation for this apparent contradiction may relate to the magnitude of knee abduction observed in the included studies. The earliest published study to examine the prospective relationship between knee abduction and ACL injury [17], reported that greater knee abduction angle and moment, respectively, were predictive of subsequent ACL injury. In this study, participants that subsequently sustained an ACL injury exhibited ~ 5 degrees of knee abduction at initial contact with the ground, ~ 9 degrees of peak knee abduction and 45 N.m. peak knee abduction moment [17]. Interestingly, all subsequent studies reporting 3D knee abduction mechanics that were included in our meta-analyses report only around 2 degrees of peak abduction for all participants, including those that subsequently sustained an ACL injury [20, 39, 41] and between 21 and 37 N.m. peak abduction moment [20, 39], and found neither measure to be predictive of future ACL injury. Conceivably, the findings of Hewett and colleagues [17], in combination with earlier evidence from cadaver knees [44‐46], lead to the development and adoption of ACL injury prevention training specifically targeting knee abduction in weight-bearing activities; this has subsequently been highlighted in numerous reviews and consensus statements [8, 18, 47, 48]. As a result, the magnitude of knee abduction mechanics observed in the vast majority of studies included in our analyses may not be sufficient to present as a risk factor for ACL injury. Supporting this, the study by Krosshaug et.al., [20] included in our analysis reports that approximately 40% of the included participants in their study “reported to have implemented preventive training as part of their routine during the season”. It is, thus, possible that the results of our meta-analyses are rather a consequence of successful injury prevention training in the last decade, than that excessive knee abduction and/or kinetics are not risk factors for ACL injury. On the other hand, although injury prevention programs may have decreased the amount of knee abduction exhibited during activities, there seem to be no decrease in the incidence of ACL injury during the same time period [49, 50], indicating that knee abduction may play a minor role in ACL injury.

An alternative explanation for our findings could be that instead of a linear relationship between knee abduction and ACL injury risk there may be a non-linear relationship with a certain cut-point beyond which knee abduction is associated with ACL injury risk. None of the studies included in this review have used break-point analysis to investigate if certain thresholds of knee abduction were associated with elevated ACL injury risk. Although greater knee abduction has been postulated to increase the risk of injury, there is no consensus regarding the amount of knee abduction that is considered excessive enough to amplify ACL injury risk. Fox et al., determined normative values for knee abduction angle during a vertical drop jump to 0.30 ± 5.0 degrees for IC and 8.71 ± 9.1 degrees for peak knee abduction [51], implying that the participants in the studies included in this review were all in the normal range of knee abduction, i.e., concurrent with the amount of knee abduction in the general population, which may further mask possible associations between knee abduction and injury risk. Given the lack of an injury risk threshold, it is also not clear if there is an elevated risk of knee injury in individuals presenting with knee abduction at the higher end of the normal range that has been postulated. Furthermore, most studies investigating knee abduction as a risk factor for ACL injury assess knee abduction during a drop vertical jump. The vertical drop jump is a bilateral task and may not reflect movements when injury occurs and does not seem to detect sex differences in knee abduction compared to other tasks [18]. Thus, it is possible that this task is not challenging enough to capture the amount of knee abduction that may be associated with injury. Other more challenging tasks, such as cutting tasks, should, therefore, be considered when evaluating knee abduction as a risk factor for ACL injury in future studies.

Increased knee abduction compared to both non-injured individuals and the contra-lateral leg is reported after ACL injury [14‐16]. Although several video analysis studies report that knee abduction seems to be involved in the ACL injury mechanism in females [52‐54], it is not possible to elucidate the exact time point of the injury on video recordings. Given that the main purpose of the ACL is to provide mechanical stability to the knee [9, 10], it is not clear if the knee abduction (or valgus collapse) observed at the time for injury causes the injury or is due to decreased joint stability as a result of the ACL tear [52‐54]. Although some recent cadaveric studies report an association between increased knee abduction moment and ACL failure [55, 56], in support of the latter, a recent systematic review on bone bruises assessed with MRI after ACL injury [57] concludes that knee abduction occurs after the ACL is ruptured, not before. It should, however, be noted that in the same systematic review a high number (approx. 70%) of bone bruises were located on the lateral side, which could indicate presence of knee abduction at the time of injury. Nevertheless, the conclusion of that meta-analysis is further supported by a study that investigated knee kinematics before and after ACL injury and found participants that sustained an ACL injury to perform a drop vertical jump with significantly greater knee abduction angle 2 years after injury compared to their performance at baseline prior to the injury [41]. Thus, it is possible that persistent deficiencies in motor control after injury cause further risk of sustaining also a second ACL injury [20, 31]. It should be noted that although 3D motion analysis was used in most of the studies the way in which knee abduction is quantified may still vary substantially. Differences in how joint axes are defined [58], the kinematic modelling approach employed (direct versus inverse kinematics) [59], and inertial properties used to determine joint kinetics [60] are all known to result in differences in the magnitude of knee abduction measured during functional activities. Similarly, marker placement locations may be differentially influenced by soft tissue artefact, impacting upon the validity and reliability of the marker model used [61]. While there is recent evidence of good to excellent within and between session reliability for both knee abduction angle and knee abduction moment during double-leg vertical drop jump using 3D analysis [62], this may not hold true for all studies included in our review. Despite these variations in the approach used to quantify knee abduction and the variance in the data that this may produce, mostly low to moderate heterogeneity was observed across our meta-analyses, suggesting that the cumulative effect of these differences upon our findings was minimal.

This review has some limitations. We pooled studies on females alone and those that included both men and women, had different follow-up periods as well as different weight-bearing tasks in some of our analyses. While these primary analyses may have masked associations between knee abduction and injury risk, our sensitivity analyses demonstrate that this is unlikely to be the case. Likewise, we pooled studies including participants of different ages (i.e., ≤15 years or > 15 years) and different activity levels. Neuromuscular and biomechanical differences between males and females during early puberty and through maturation have been suggested to play a role for ACL injury risk in young females [8]. Importantly however, our sensitivity analysis including the only two studies on young females (i.e. ≤15 years) revealed no association between 3D knee abduction at baseline and future ACL injury. Taken together, the result from this review applies across sexes, tasks, age and follow-up period. It was, however, not possible to perform a sensitivity analysis for activity level (elite athletes versus high school athletes), since there were too few studies using the same outcome. The two studies that included high school athletes [17, 40] reported that the participants that sustained an ACL injury had increased 3D knee abduction angles (IC and peak) [17] and increased 2D MKD (IC and peak) [40] at baseline compared to those who did not sustain an injury. Thus, we cannot rule out that factors contributing to knee injury may differ between those on an elite level compared to being active on a lower level. This is worthy of further investigation. Furthermore, the meta-analyses are only able to show if a greater or smaller amount of knee abduction is associated with future ACL injury and not if a certain threshold of knee abduction is related to an elevated injury risk. We included studies that employed differing methodologies to quantify knee joint mechanics. Of note, knee abduction angles were obtained with both 2D and 3D motion analysis systems; knee abduction moments were exclusively obtained with 3D motion analysis. While there is evidence that knee abduction angles measured in 2D are strongly correlated with knee abduction measured in 3D [63‐65], the 2D measure also incorporates components of sagittal and transverse plane rotation and thus our findings with regard to 2D knee abduction kinematics are likely, to a small extent, to reflect the underlying sagittal and transverse plane knee kinematics. In light of these differences we did not pool the results from 2D and 3D studies. Yet, given the strong relationship between 2D and 3D knee abduction, taken together these results both support the absence of a predictive effect of baseline knee abduction on ACL injury development. Moreover, some of the meta-analysis included a relatively low number of individuals with ACL injury, e.g., the analysis on 2D peak knee abduction (n = 8). Performing meta-analysis with a low number of events may increase the risk of overestimating the effect [66]. The 2D peak knee abduction analysis did also include two different tasks, a single-leg squat and a single-leg drop landing with too few studies to perform a sensitivity analysis. Although individuals seem to perform these tasks with a similar amount of knee abduction [67, 68], it is possible that the use of different tasks may have masked findings from individual tasks. Thus, some caution is needed when interpreting the 2D peak knee abduction results. Furthermore, our heterogeneity analysis using I² statistics revealed mostly low to moderate heterogeneity between studies. The analysis for peak knee abduction moment was, however, associated with high heterogeneity. To account for expected heterogeneity, we have performed all analysis under the random effect model that incorporates both within study and between study variance in the analysis. It has also been suggested that the I² statistics may be subject to bias when only a small amount of studies are included in the analysis [69]. Thus, the I² statistics presented in this review should be interpreted with caution. Also, there were too few studies included to be able to explore publication bias. However, since it is more likely that studies reporting no significant results are the studies that are not being published, this is unlikely to have an influence on our result. Finally, this review only included knee abduction kinematics and kinetics as possible risk factors for ACL injury. Several studies highlight that the mechanisms of ACL injury are in fact multifactorial and that several combined factors, such as knee abduction and internal rotation kinematics and kinetics, but also neuromuscular control of the hip and trunk may contribute to the injury mechanism [8, 12, 56, 70‐73]. Even though knee abduction kinematics and kinetics alone cannot predict injury risk, future studies will reveal if knee abduction may contribute to knee injury when combined with other risk factors, such as those described above.

This systematic review and meta-analysis indicates that neither baseline knee abduction kinematics or kinetics during weight-bearing activity may predict future ACL injuries. This is contrary to popular clinical opinion and the findings of the earliest published study examining this relationship. It is possible that as a result of the successful implementation of ACL injury prevention programs in organized sport, emphasizing a knee position in line with the hip and ankle, that knee abduction is not a risk factor for ACL injury development across these cohorts. Future studies are warranted to investigate whether knee abduction during more demanding tasks, in combination with other risk factors and/or in other cohorts, such as recreational athletes, is associated with future primary as well as second ACL injury.