Abstract
Background: The causes of noncontact anterior cruciate ligament injury remain an enigma.
Purpose: To prospectively evaluate risk factors for noncontact anterior cruciate ligament injuries in a large population of young athletic people.
Study Design: Prospective cohort study.
Methods: In 1995, 1198 new United States Military Academy cadets underwent detailed testing and many parameters were documented. During their 4-year tenure, all anterior cruciate ligament injuries that occurred were identified. Statistical analyses were used to identify the factors that may have predisposed the cadets to noncontact anterior cruciate ligament injuries.
Results: Among the 895 cadets who completed the entire 4-year study, there were 24 noncontact anterior cruciate ligament tears (16 in men, 8 in women). Significant risk factors included small femoral notch width, generalized joint laxity, and, in women, higher than normal body mass index and KT-2000 arthrometer values that were 1 standard deviation or more above the mean. The presence of more than one of these risk factors greatly increased the relative risk of injury. All female cadets who had some combination of risk factors sustained noncontact anterior cruciate ligament injuries, indicating that some combinations of factors are especially perilous to the female knee.
Conclusion: Several risk factors may predispose young athletes to noncontact anterior cruciate ligament injury.
Approximately 100,000 ACL injuries occur in the United States each year.6,8 Most of these injuries are sustained by young athletes participating in sports that require quick changes of direction and jumping. The mechanism of injury is noncontact in about 70% of the cases.4,6 More than half of those sustaining ACL ruptures are treated surgically, at an annual cost of nearly one billion dollars. 6 Anterior cruciate ligament injury is clearly a critical healthcare issue with an enormous impact on the young athletic population.
Over the past 15 years more than 4000 articles have been published related to the ACL and the treatment of ACL injuries, making this one of the most popular topics in the orthopaedic sports medicine literature. Unfortunately, only 3% of these publications (10 of 3572) deal with ACL injury prevention. 6 Recent attention has been directed at solving the enigma of the causes of noncontact ACL injuries. A number of potential risk factors for non-contact ACL injury have been identified and divided into four broad categories: environmental, anatomic, hormonal, and biomechanical. 6
Prospective studies that examine potential risk factors and identify those that may predispose athletes to non-contact ACL injury are a critical step in preventing these injuries. Of particular interest are those factors that can be easily assessed in the typical orthopaedic clinic or sports medicine complex. Some of the factors that fit this profile include anthropomorphic characteristics, such as height, weight, and body mass index (BMI); joint laxity; muscle flexibility; strength; and radiographic indices related to the ACL, such as femoral notch width.5,7,9,15,18,20,21 The purpose of this study was to prospectively examine this set of risk factors in a large population of active young people in an entire United States Military Academy (USMA) class, identify all ACL injuries that occurred in the population, and determine those factors that may have predisposed the athletes to noncontact ACL injuries.
Materials and Methods
Subjects
All incoming cadets in the freshman class of 1995 were asked to participate in this prospective study, which was approved by the Human Subject Research Review Board of Keller Army Community Hospital and the USMA. A total of 1198 cadets (1021 men, 177 women) volunteered to participate in the study. All subjects signed institutionally approved volunteer agreement affidavits. During their first 4 weeks at USMA, the cadet volunteers provided a detailed medical history and underwent physical examination, radiographic studies, and strength testing. All cadets who were identified as having a history of ACL injury by examination or report were excluded from the study.
Physical Examination
The stability of each subject's knees was assessed by a sports medicine fellowship-trained orthopaedic surgeon using the following tests: Lachman, pivot shift, anterior drawer, posterior drawer, and varus/valgus stress tests at 0° and 30°. The surgeon also checked for joint effusions and described patellar mobility by noting the number of patellar quadrants of motion that were present when the patella was translated medially and laterally from its resting position. A physical therapist and an athletic trainer with extensive knee arthrometry experience evaluated the laxity of the subjects' knees with a KT-2000 arthrometer (Medmetric, Inc., San Diego, California) and an X-Y plotter. Trained technicians evaluated height (in centimeters) and weight (in kilograms) using a typical medical examination scale. Height and weight information was normalized by calculating the subjects' BMI with the following formula: BMI = weight (in kilograms)/height (in meters) squared.
The physical therapists and certified athletic trainers also measured joint range of motion with a standard goniometer, tested for generalized joint laxity, and assessed the flexibility of the hamstring muscle-tendon units and lumbar musculature with the sit-and-reach test. 22 The following signs were looked for when assessing generalized joint laxity: small finger metacarpophalangeal hyper-extension (≥90° of extension), elbow hyperextension, knee hyperextension, and the ability to touch the thumb to the volar aspect of the forearm.3,25 Subjects were considered to have generalized joint laxity if they had five or more of these signs (indicating that excessive laxity was observed bilaterally and in at least three joints).
Radiographic Evaluation
Bilateral standardized tunnel radiographs were acquired for all subjects. The radiographs were marked and measured (Fig. 1). First, a line parallel to the tibial plateau surface was drawn from the medial to the lateral border of the tibia; the width of this line was measured to determine tibial width. Next, lines bisecting the peaks of the tibial eminences were drawn perpendicular to the tibial width line. The distance between the lines bisecting the tibial eminences is referred to as eminence width, which is considered to be an estimate of ACL diameter. Another line parallel to the tibial plateau was drawn through the femoral condyles and intercondylar notch at the level of the lateral sulcus. As described by Souryal and coworkers, 21 this line was used to measure the condylar width of the femur and femoral notch width. All radiographic measurements were performed with digital calipers (Mitutoyo Digimatic calipers, Mitumoyal, Japan). The precision of these calipers is ±0.01 mm.
Measurements made on radiographs. AA' = condylar width, BB' = notch width, CC = tibial width, DD' = eminence width.
Femoral notch width indices (notch width divided by condyle width) were calculated for all subjects because researchers have demonstrated an association between small notch width indices and ACL tears.20,21 In addition, we calculated two new indices, the eminence width index and notch width/eminence width index, based on the theory that people with smaller ACLs are more prone to noncontact ACL injuries. 18 The eminence width index is similar to the notch width index but describes eminence width divided by tibial width; this index is intended to describe the estimated size of the ACL in relation to the size of the tibia. The notch width/eminence width index is intended to describe the relationship between the estimated size of the ACL and the area in which it is accommodated (notch width).
Strength Assessment
The strength of the knee extensors and flexors was evaluated by having the subjects perform bilateral concentric and eccentric isokinetic testing at 60 deg/sec with the Kin-Com 500H dynamometer (Chattecx Corp., Chattanooga, Tennessee). After the subjects were properly positioned and given testing instructions, they performed 10 submaximal repetitions to warm up and familiarize themselves with the test. Subjects then performed 10 maximal test repetitions; the 3 repetitions with the greatest peak force were stored for analysis. Gravity correction was used for all tests. The order of knee testing was alternated from subject to subject (subject 1, right then left; subject 2, left then right). The order of testing for the muscle groups was alternated in the same manner. Peak force measurements were normalized to subject body weight. In addition to the traditional concentric and eccentric strength ratios, eccentric hamstring muscles to concentric quadriceps muscles and end-range strength ratios were calculated because these ratios have been suggested to be functional and specific to noncontact ACL injury mechanisms. 1
Participation/Exposure Data
“On the fields of friendly strife are sown the seeds that on other fields on other days will bear the fruits of victory.” This famous quote by General Douglas McArthur has been a hallmark of training at the USMA since his role as superintendent in the early 1920s. The corollary in the 21st century is “Every student is an athlete and every athlete is challenged.” All cadets are required to regularly participate in activities requiring quick changes of direction or cutting. The exposure of men and women to high-risk activities is comparable. All cadets are required to participate in Department of Physical Education classes and activities, which include advanced close-quarters combat, basketball, skiing, a challenging indoor obstacle course, and similar activities. Cadets participate in mandatory military training each summer, in which they perform activities such as war games and repelling or parachuting from helicopters. Cadets must also participate in intramural, competitive club, or varsity sports in all but two semesters of their tenure at the USMA. Common activities performed at the intramural level include tackle football, soccer, rugby, touch-and-go football (ultimate Frisbee with a football), Walleyball, and basketball; the varsity and club athletes compete in more than 25 sports at the NCAA Division IA level. Women are prohibited from participation in intramural football, wrestling, boxing, and rugby; however, they are integrated into all other sports on coeducational teams.
Cadet exposure data were maintained in two separate databases. All time spent participating in physical education classes, intramural sports, and competitive club programs was entered into a database maintained by the Department of Physical Education athletic trainers. Attendance in these events was recorded daily. Athletic trainers in the Office of the Director of Intercollegiate Athletics recorded varsity athlete participation time in practices and games using the Sports Injury Monitoring System database (SIMS, Med Sports System Limited, Iowa City, Iowa). Because all physical education classes and intramural practices or games are approximately 1 hour in duration, we described exposure in terms of hours of participation. Exposure during intercollegiate sports and military training was also calculated to the nearest hour. There was no attempt to monitor subject exposure to risk during the cadets' short summer and winter holiday vacations (2 to 3 weeks) or during leisure time on campus.
Identification of Injuries
Injury data were obtained daily (Monday through Friday) during the 4-year study. The cadets at USMA cannot excuse themselves from their required classes and daily activities without written approval. Permission for excuse because of injury must be approved by a physician or physical therapist. Thus, all noteworthy musculoskeletal injuries are identified, documented, and managed by the orthopaedic or physical therapy staff soon after they occur. All ACL injury diagnoses were based on subject reports and orthopaedic physical examination (including imaging), and were confirmed by arthroscopic examination. The determination of whether the mechanism of injury was noncontact or contact was based on each subject's report of the event and the eyewitness accounts of physicians, athletic trainers, or physical therapists present at the time of injury. Before graduation, all members of the Class of 1999 also underwent a thorough commissioning physical examination as mandated by United States Army regulation. We are confident that we identified all of the ACL injuries that occurred in the study population because of the multiple levels of control within the USMA system.
Statistical Analyses
Descriptive statistics were calculated for each risk factor with SPSS for Windows 11.01 (SPSS Corp., Chicago, Illinois). A significance level of P ≤ 0.05 was set for all hypothesis tests.
A two-factor (ACL injury, sex) multivariate analysis of variance with each risk factor as a dependent variable was used to determine 1) whether the results of subjects who sustained noncontact ACL injuries were different from the results of subjects who did not sustain ACL tears (that is, was there a main effect of ACL injury), 2) whether the results of men and women were different (that is, was there a main effect of sex), and 3) whether there was an interaction between ACL injury and sex for any of the risk factors. One-way analysis of variance was used to determine whether there were significant differences between the results of men who sustained noncontact ACL injuries and men who did not injure their ACL, as well as the results of women who sustained noncontact ACL injuries and women who did not.
Logistic regression analyses were used to determine whether any of the risk factors were predictive of noncontact ACL injuries in 1) all subjects, 2) men, and 3) women. Two logistic regression methods were employed: 1) a hypothesis-driven method and 2) a more general multivariate stepwise regression method. In the hypothesis-driven logistic regression models, risk factors were chosen as inputs based on the results of the hypothesis tests and descriptive statistics. In the multivariate stepwise logistic regression models, all explanatory variables were entered for each of the subject categories. In both logistic regression methods, a representative risk factor was chosen for the radiographic, anthropomorphic, and KT-2000 arthrometer factors to prevent analysis of factors with a high likelihood of covariance. Femoral notch width was used as the representative risk factor for the radiographic findings because this factor was significant in all of the hypothesis tests. Body mass index and KT-2000 arthrometer results at 134 N of pull were used rather than body weight and arthrometer results at other loads, respectively.
Finally, the relative risks associated with subjects having one or more significant risk factors were analyzed by calculating risk ratios. 17 The risk factors included in the analyses were chosen based on the results of the hypothesis tests.
Results
Subjects
Of the 1198 cadets (1021 men, 177 women) who were enrolled in this study, a total of 339 subjects were excluded because they either had a history of ACL injury or they failed to complete the 4-year academic program. According to the USMA Office of Institutional Research, the attrition rate for the Class of 1999 was similar to that observed at USMA over the preceding 25 years. The remaining 859 cadets (739 men, 120 women) were observed for the entire 4-year academic program. The mean age of the 859 subjects at the time of enrollment was 18.4 years (range, 17 to 23). Although a very concerted effort was made to ensure that we obtained a complete data set for each subject, some subjects were not able to have every variable assessed because of strict data collection time limits imposed by USMA or other scheduling conflicts. Consequently, the numbers in Tables 2 through 5 vary. No subject lacked more than three data elements.
Incidence of Injury
A total of 29 complete ACL tears (21 in men, 8 in women) were sustained by the 859 cadets during their tenure at West Point. There were no partial ACL tears in the study population. Five of the ACL injuries (17%) that occurred in men were contact injuries, whereas the remaining 24 ACL tears (16 in men, 8 in women) were noncontact ACL injuries. The overall incidence of ACL rupture in the Class of 1999 was 3.3%. The incidence of noncontact ACL tears was 2.8%; 6.6% in women and 2.1% in men (a female-to-male noncontact ACL injury ratio of about 3:1).
Exposure to Risk
The risk of noncontact ACL injury was 1 in 42,429 hours of exposure during mandatory classes and activities (Table 1). Cadets playing in competitive club and varsity sports had a noncontact ACL injury risk of 1 in 25,782 hours of exposure, whereas those participating in intramural sports had a noncontact ACL injury risk of 1 in 5351 hours of exposure. No ACL injuries occurred during the short cadet vacations; one injury occurred during cadet leisure time on campus, which was not monitored for exposure to risk.
Cadet Exposure to Risk for Noncontact ACL Injury a
Cadets who sustained ACL injuries were excluded from further analysis of exposure risk (hours after injury were not included). One noncontact ACL injury occurred in cadet leisure time on campus, which was not monitored for exposure to risk.
Comparison of Men versus Women
There was a significant main effect for sex for the following variables: height, body weight, condylar width, tibial width, eminence width, notch width/eminence width index, generalized joint laxity, KT-2000 arthrometer values at all loads (45 to 178 N of pull in 22.2-N increments), and strength of the knee extensors and flexors (Table 2). Because the strength results for each of our respective analyses were bilaterally similar, we will present the results for the subjects' right limbs for all analyses.
Comparison of Results by Sex
Statistically significant at the P ≤ 0.05 level.
Percent body weight.
All Subjects: Noncontact ACL Injury versus No ACL Injury
There was a significant main effect for ACL injury for the following risk factors: body weight, BMI, notch width, eminence width, notch width index, notch width/eminence width index, generalized joint laxity, and KT-2000 arthrometer values at 111, 134, 156, and 178 N (Table 3). There were no significant sex-by-injury interactions.
Comparison of Results for All Subjects: Noncontact ACL Injury Versus No ACL Injury
Statistically significant at the P < 0.05 level.
Percent body weight.
Men: Noncontact ACL injury versus No ACL injury
The values of the following risk factors were significantly different when the results of men who sustained noncontact ACL injuries were compared with the results of men who did not tear their ACL: notch width, eminence width, notch width index, tibial width index, and generalized joint laxity (Table 4). Knee extensor and flexor strength ratios, including the eccentric hamstring muscles to concentric quadriceps muscles and end-range of motion ratios (P = 0.353 to 0.961) were not significantly different between the groups.
Comparison of Results for Men: Noncontact ACL Injury Versus No ACL Injury
Statistically significant at the P ≤ 0.05 level.
Percent body weight.
Women: Noncontact ACL injury versus No ACL injury
The values of the following risk factors were significantly different when the results of women who sustained non-contact ACL tears were compared with the results of women who did not sustain ACL tears: body weight, BMI, notch width, eminence width, notch width index, eminence width index, tibial width index, notch width/eminence width index, and generalized joint laxity (Table 5). The results for KT-2000 arthrometer testing with loads of 111 to 178 N of pull were approaching significance but lacked statistical power (observed power ranged from 0.41 to 0.47). The strength ratios evaluating relationships between the quadriceps and hamstring muscle groups as well as strength in the end-range of motion (P = 0.424 to 0.700) were not significantly different between groups.
Comparison of Results for Women: Noncontact ACL Injury Versus No ACL Injury
Statistically significant at the P ≤ 0.05 level.
Noteworthy trend but lacked statistical power.
Percent body weight.
Predictive Models for Noncontact ACL Injuries
The hypothesis-driven and multivariate stepwise logistic regression models yielded nearly the same results in each case analyzed. We will present only the results for the hypothesis-driven models because we believe that this was the more appropriate method of evaluating the data and because presenting the results of both models would be redundant. When all subjects were considered, a model including narrow femoral notch width, higher than average BMI, generalized joint laxity, and KT-2000 arthrometer results at 134 N of pull more than 1 SD above the mean was determined to be the most predictive model for noncontact ACL injuries. This model explained 28% of the variance in noncontact ACL injuries (R 2 = 0.284), was highly specific (the status of all uninjured subjects was successfully predicted), but fair with respect to sensitivity (only 17% of the noncontact ACL injuries were correctly predicted). The mean predictive value of the model was 97.5%, but this finding was influenced substantially by the model's negative predictive value.
In men alone, the most predictive model included narrow femoral notch width and generalized joint laxity. This model had an R 2 value of 0.152 but did not correctly predict any of the noncontact ACL tears. The model was able to successfully predict all of the uninjured mens' status correctly (highly specific); however, the meaningfulness of this finding is questionable when the model's lack of sensitivity is taken into account.
In women alone, a model that included narrow femoral notch width, higher than average BMI, and generalized joint laxity explained 62.5% of the variability in noncontact ACL tears (R 2 = 0.625). This model correctly predicted the status of all of the uninjured subjects and was able to predict 75% (six of eight) of the noncontact ACL injuries; thus, this model was both highly specific and highly sensitive in the female cadet population. The mean predictive value for this model was an impressive 98%.
Relative Risk Associated with Risk Factors
The relative risk of noncontact ACL injury associated with one or more risk factors is presented for 1) all subjects, 2) men, and 3) women (Table 6). Relative risk ratios associated with having one or more risk factors ranged from approximately 1.0 to 37.7, indicating that some combinations of factors led to a risk of noncontact ACL injury that was 40 times higher than was observed when those combinations of factors were not present. The risk ratios for women were considerably higher than those for men and contributed disproportionately to the risk ratios calculated for all subjects. The relative risk associated with women who had the combination of a narrow femoral notch width, BMI 1 SD or more above the mean, and either generalized joint laxity or KT-2000 arthrometer results that were more than 1 SD above the mean could not be calculated because all women with either of these combinations sustained noncontact ACL injuries. There were also some combinations of risk factors that were not observed in any of the cadets who sustained noncontact ACL injuries; consequently, no risk ratios could be presented for these combinations.
Relative Risk Associated with Having One or More Noncontact ACL Injury Risk Factors
Risk factor was not significantly different when injured were compared with uninjured.
Discussion
The purpose of this study was to identify risk factors that may predispose young athletes to noncontact ACL injuries. We prospectively examined an entire USMA class and tracked them during their 4-year tenure at the USMA. We successfully identified all ACL tears that occurred in the population and carefully evaluated the prospective data to determine the factors that may have predisposed the subjects to their noncontact ACL injuries. We identified a group of risk factors that were not only common in the cadets who sustained noncontact ACL injuries but also increased their risk for noncontact ACL injuries. The prospective design, large number of subjects enrolled, and the high level of control within this study (similar activity level/lifestyle of the subjects and ability to identify all injuries) made it unique and significant.
It is now common to divide ACL injury risk factors into modifiable factors (those that can be changed through training or other interventions) and nonmodifiable factors (those that are structural or physiologic and difficult to alter with noninvasive methods). We included both modifiable (strength, flexibility, BMI) and nonmodifiable (femoral notch width, generalized joint laxity) factors in this study; unfortunately, almost all of the risk factors that were significant in our analyses were nonmodifiable factors. Body weight and BMI were the only modifiable risk factors that were significant in our analyses, and these were observed only in the female population. We included the specific risk factors identified in this study because 1) they were some of the most frequently discussed noncontact ACL injury risk factors at the time when this study was designed and 2) they could be assessed in a typical orthopaedic clinic or sports medicine complex. Additional risk factors, such as neuromuscular control,6,23 core stability, 6 and hormonal factors8,11,24 are commonly discussed today, but were not at the forefront of ACL injury research in the early 1990s. These factors are important and should be the focus of future studies.
We included the comparison of results by sex (Table 2) to be thorough. Although there is nothing surprising in the results of this comparison, the data demonstrate the marked differences in the results of the men and women who participated in the study and provide a detailed description of the study population. When the male and female noncontact ACL injury versus no ACL injury results are considered (Tables 4 and 5), it is apparent that a disproportionate number of significant risk factors were observed in women. This finding supports the idea that the cause of noncontact ACL injuries is different in men and women. It appears that the risk factors that we evaluated in this study were more specific to noncontact ACL injuries in women.
The only significant risk factors for noncontact ACL injury observed in the men of this study were the radiographic indices related to the size of the femoral notch (notch width and notch width index) and ACL (eminence width and eminence width index) and generalized joint laxity. The most predictive statistical model in men explained only 15% of the variability in noncontact ACL tears and was unable to predict any of the 16 noncontact ACL ruptures that occurred in the male population. This finding suggests that the cause of noncontact ACL injury in men is multifactorial but remains largely an enigma at the present time. With that said, we believe that the 7.8-fold increased risk for noncontact ACL injury observed in men who had both a narrow femoral notch and generalized joint laxity was a meaningful finding that should be considered in future studies.
Among women, we were able to identify several factors that were significant and when present in combination greatly increased the risk of noncontact ACL injury (Tables 5 and 6). The KT-2000 arthrometer knee laxity measurements at 111 to 178 N of pull approached significance but lacked statistical power. We included all women in the population and identified all ACL tears that occurred during their 4-year tenure at the USMA; consequently, statistical power could not be increased by enrolling more subjects. In these circumstances, a level of significance greater than P < 0.05 may be appropriate. In our opinion, these arthrometer results should also be considered significant results. The following findings provide further support that the arthrometer results are meaningful: 1) women who had laxity values that were 1 SD or more above the mean on testing at 134 N had a relative risk of noncontact ACL tears that was 2.7 times higher than was observed in women with less knee laxity, and 2) including the arthrometer results in the regression models had a noteworthy positive impact on their results. Women who had greater than normal laxity on KT-2000 arthrometer testing (1 SD) and either a narrow femoral notch (≤13 mm) or BMI results more than 1 SD above the mean demonstrated a much greater risk for noncontact ACL injury (risk ratios = 16.8 and 37.7, respectively) than was observed for those having either the narrow femoral notch (risk ratio = 4.0) or higher than normal BMI (risk ratio = 3.5) alone. These findings provide concrete support for our belief that the specified arthrometer results are noteworthy.
The results of this study provided considerable evidence that narrow notch width, generalized joint laxity, knee laxity on KT-2000 arthrometer testing (in women), and increased BMI (in women) are important risk factors for noncontact ACL injuries. Although the presence of one of these factors alone led to a relative risk of noncontact ACL injury that was 2.7 to 4.0 times that observed for cadets without these risk factors, the presence of two or more of these factors significantly increased the risk for injury (Table 6). Most remarkable is the fact that all women in the population who had the combinations of a narrow femoral notch, BMI 1 SD or more above the mean, and either generalized joint laxity or KT-2000 arthrometer (134 N) results that were 1 SD or more above the mean sustained noncontact ACL injuries. Further supporting the importance of these factors is the fact that a regression model that included femoral notch width, BMI, and generalized joint laxity was able to correctly predict 75% (6 of 8) of the noncontact ACL injuries and all of the uninjured cases in the population of women (mean predictive value of 98%). A model including KT-2000 arthrometer results (134 N) rather than generalized joint laxity returned similar, but slightly lower, predictive results. Although the results of these statistical models are impressive, it is noted that the most predictive model explained only 62.5% of the variability in noncontact ACL tears (R 2 = 0.625). This indicates that there are several other contributing risk factors that were not investigated in this study. Future work should attempt to identify these factors.
Several authors have documented a relationship between femoral notch width and ACL injuries.2,7,18,20 Our results confirm that a narrow femoral notch is indeed a common finding in both men and women who sustain noncontact ACL injuries. Shelbourne and coworkers 18 have suggested that it may be the absolute size of the ACL that predisposes people with narrow femoral notches to ACL injuries rather than notch width itself. For this reason, we also evaluated indices that are intended to be a simple method of assessing ACL size and the risk associated with the presence of a small ACL (eminence width, eminence width index, and notch width/eminence width index). These indices are based on the assumption that the diameter of the ACL is approximately the same as tibial eminence width. We have not performed formal testing of the validity of this assumption and readily acknowledge that more accurate determination of ACL size can be obtained by calculating the volume and cross-sectional area of the ligament with MRI studies. With that said, it should be noted that the men and women who sustained noncontact ACL injuries in this study had significantly narrower eminence width and lower eminence width index values than people who did not sustain ACL injuries. It is logical that ACL size would be a risk factor for noncontact ACL injuries because a smaller ACL has less material strength than a larger ACL and should rupture sooner under similar loading conditions. The results of this study do not add any insight into whether narrow femoral notch width is a risk factor for noncontact ACL injuries or merely a noticeable radiographic finding that is associated with ACL size. Our results do, however, suggest that people with narrow femoral notches are at greater risk for injury than those with larger notches, even if it is not the femoral notch itself that contributes to noncontact ACL injuries.
Generalized joint laxity was also significantly more prevalent in the men and women who sustained noncontact ACL injuries than in their uninjured counterparts. More than 30 years ago, Nicholas 15 reported that football players with lax joints were more prone to knee ligament injuries than those who had tight joints. Our finding that generalized joint laxity was associated with a 3.1-fold increased risk for noncontact ACL injury in the men of this study is consistent with Nicholas' findings. More recently, Ostenberg and Roos 16 reported that women who had generalized joint laxity had an odds ratio of lower extremity injury that was 5.3 times higher than that observed in women without generalized joint laxity. Söderman and coworkers 19 reported a similar, but slightly lower, odds ratio (3.1) for generalized joint laxity in their study of risk factors for lower extremity injuries in female outdoor soccer players. The risk ratio associated with generalized joint laxity in the women participating in this study (2.7) was consistent with the findings of these studies. The confluence of evidence from these studies suggests that women with generalized joint laxity have an increased risk for noncontact ACL injuries and lower extremities in general. At the present time, it is unclear whether the increased risk of lower extremity ligament injuries observed in women with generalized joint laxity is related to hormonal factors or is simply a mechanical phenomenon. This topic is beyond the scope of this study, but it should be investigated in detail in future studies.
The female cadets who sustained noncontact ACL injuries in this study had a significantly higher mean BMI than those who did not sustain ACL injuries. Female subjects with a BMI that was 1 SD or more above the mean had a relative risk for noncontact ACL injury that was 3.5 times that of women with a lower BMI. The reason(s) for this increased risk are not clear. It should be noted that the BMI included in this analysis was documented prospectively shortly after the cadets arrived at the USMA, whereas their ACL injuries occurred throughout the 4-year study. Repeat data collection was not performed during the study because of the considerable demands associated with this large-scale collection. Consequently, we do not know whether the subjects' BMI changed between data collection and the time of injury. Regardless, the statistical significance, risk ratios, and importance of the BMI data in the statistical models provide compelling evidence that these BMI data are meaningful.
An explanation of the link between elevated BMI and noncontact ACL injury is difficult and requires speculation. Because an inverse relationship often exists between BMI and physical fitness, the higher BMI observed in the female subjects who sustained noncontact ACL injuries may indicate that these cadets were less fit than their uninjured counterparts on arriving at the USMA. However, it must be acknowledged that BMI is not a direct measure of cardiovascular or skeletal muscle fitness, so associating BMI with fitness may not be valid. A higher than normal BMI can also be present in fit people who have greater muscle mass than is typical for people of their height; however our strength data (Table 5) suggest that it is unlikely that this was the case. Another potential explanation for the link between BMI and noncontact ACL injuries involves activity levels before entering the USMA. It is plausible that those who had a higher BMI on arrival at the USMA were less athletically active in the years before entrance than was the average female cadet. If this were the case, it is possible that these female cadets had not fully developed the motor programs necessary for the sports in which they were injured. In theory, less-developed motor programs may lead to decreased neuromuscular control and increased risk for noncontact ACL injury. 23 We openly acknowledge that these explanations are very speculative. More research is clearly needed to examine and explain the apparent relationship between BMI and noncontact ACL injuries; however, our data suggest that higher than average BMI should be regarded as a risk factor for noncontact ACL injury in women.
Cadets are required to participate in sports at the varsity, club, or intramural level during six of their eight academic semesters at the USMA. As a result, most cadets are forced to participate in sports with which they may be relatively unfamiliar. Approximately 60% of the noncon-tact ACL injuries (14 of 24) occurred during intramural sports, which had a much greater risk of noncontact ACL injury than varsity or club sports (Table 1). This finding could also be the result of insufficiently developed motor programs. Once again, we acknowledge that this is speculative, but current theory related to the causes of non-contact ACL injuries suggests that these ideas are plausible and warrant mentioning. Future research should attempt to define the relationship between preinjury athletic history, motor program development, and the incidence of noncontact ACL injuries. In our prospective data collection we did not question the cadets regarding their previous athletic participation. We recommend that researchers who perform future risk factor studies carefully document when the subjects began playing sports recreationally and competitively, as these data may have implications for motor program development and the causes of noncontact ACL injuries.
Strength imbalances have been suggested as a possible risk factor for both ACL injuries and lower extremity injuries in general.9,19 The hamstring muscles act as protagonists to the ACL by providing resistance to anterior tibial translation, which often results from forces applied by the quadriceps muscles.10,12,13 For this reason, the ratio of hamstring muscle strength to quadriceps muscle strength is often discussed as a potential ACL injury risk factor. Moul 14 found that women had lower eccentric hamstring-to-eccentric quadriceps muscle ratios than men and suggested that this strength imbalance may help to explain higher ACL injury rates in female athletes. We examined several strength ratios in this study, including those thought to be more specific to the ACL injury pattern, but none were determined to be significant factors in our analyses. Despite the fact that there were no statistically significant strength ratios, men who had eccentric quadriceps muscle strength ratios 1 SD or more above the mean did have a relative risk of noncontact ACL tear that was twice that of men who had an average eccentric quadriceps muscle strength ratio. Although it was unclear whether this finding was meaningful, it certainly warrants mentioning. Once again, we are limited by the fact that our data were collected prospectively at a single point in time. Consequently, we cannot rule out that the subjects' strength ratios had changed between the time of data collection and injury; however, the similarity in cadet lifestyle and activity levels leads us to believe that similar strength changes would be observed across the population (by sex), except in cases in which cadets arrived at USMA with much lower than average strength. Although the results of this study suggest that knee strength is not a key ACL injury risk factor, our design prevents us from ruling it out. Future studies should examine these strength ratios in greater detail by using serial studies and should also evaluate core strength/stability, which may be a more likely noncontact ACL injury risk factor. 6
One of the strengths of this study is the fact that we enrolled over 1100 subjects; however, one of its important take-home messages is that much larger multicenter studies are needed. Although we included all cadets in the Class of 1999 and identified every ACL injury that occurred in the population, we still had variables that lacked statistical power. The results of this study suggest that the cause of noncontact ACL injuries will remain unclear until there is a large coordinated effort to examine non-contact ACL injuries with multicenter studies similar to those used in the new drug approval process.
Conclusions
The prospective design, large number of subjects, and high level of control included within this study made it unique and significant. Although several modifiable and non-modifiable factors were studied, most of the factors determined to be significant were nonmodifiable. Our results suggest that radiographic indices associated with notch width and ACL size, as well as generalized joint laxity, are indicators of an increased risk for noncontact ACL injuries in both men and women. Higher than average BMI and KT-2000 arthrometer knee laxity results were also significant factors in the female population. Subjects who had more than one of these risk factors were at much greater risk for noncontact ACL injuries than those with one risk factor or none. All women with the combination of a narrow femoral notch (≤13 mm), a BMI that was 1 SD or more above the mean, and either generalized joint laxity or KT-2000 arthrometer values that were 1 SD or more above the mean sustained noncontact ACL injuries. Logistic regression models did not predict any of the noncontact ACL injuries in men but correctly predicted 75% of the noncontact ACL injuries in the female cadets. Our results indicated that, although the cause of noncontact ACL injuries in men remains somewhat of an enigma, we appear to be making headway on finding the causes of noncontact ACL injury in women. Finally, our results suggest that even studies with over 1000 subjects are far too small for adequate statistical power to be present in all analyses. Much larger coordinated multicenter studies are critical to solving the mystery of the causes of noncontact ACL injuries. This study provides a foundation for future research in the causes and prevention of noncontact ACL injury.