KEY WORDS: Anterior cruciate ligament injury, injury prevention, knee injury, sports biomechanics.

Anterior cruciate ligament (ACL) tear is a debilitating sports injury with an estimated 80,000 ACL occurrences in the United States annually (Griffin et al., 2000; Miyasaka et al., 1991; Pedowitz et al., 2003). The literature almost unanimously suggests that females are substantially more susceptible than males in suffering acute non-contact injury of the ACL (Arendt and Dick, 1995; Delfico and Garrett, 1998; Gray et al., 1985; Griffin, 2001; Hutchinson and Ireland, 1995; Messina et al., 1999; Powell and Barber-Foss, 2000; Stevenson et al., 1998; Tillman et al., 2002). The ACL injury rate in females is higher in a variety of exercise activities such as soccer (2-6 times higher), basketball (4-10 times higher) and military training (10 times higher) (Arendt and Dick, 1995; Gray et al., 1985; Griffin, 2001; Hutchinson and Ireland, 1995; Messina et al., 1999; Powell and Barber-Foss, 2000).

A conference on the prevention of ACL injuries sponsored by the American Orthopaedic Society for Sports Medicine in 1999 issued a consensus statement suggesting that biomechanical and neuromuscular factors appear to be the most important factors associated with ACL injury and the higher incidence of injury in female athletes (Griffin, 2001). A primary recommendation of the conference was that research efforts should focus on the investigation of biomechanical risk factors (Griffin, 2001; Harmon and Ireland, 2000) during activities that can cause knee injury, such as landing from a jump. Two-thirds of ACL injuries occur via a non-contact mechanism, and the majority of these injuries occur at landing from a jump (Arendt and Dick, 1995; Boden et al., 2000; Gray et al., 1985; Griffin et al., 2000; 2001; Hewett et al., 1999; Kirialanis et al., 2003; Kirkendall and Garrett, 2000).

Although the ACL gender bias is likely multifactorial, three main theories have been proposed to explain the higher incidence of female ACL injury: the ligament dominance theory (Hewett et al., 2001), the quadriceps dominance theory (Hewett et al., 2001), and the straight knee landing theory (Huston et al., 2001). The ligament dominance theory suggests that the lower extremity muscles do not adequately absorb the impact of landing, resulting in knee valgus which causes increased loading of the ACL (Ford et al., 2003). The quadriceps dominance theory suggests that females tend to rely on their quadriceps more than their hamstrings creating excessive anterior translation of the tibia (Ford et al., 2003; Hewett et al., 1996; Huston and Wojtys, 1996). The straight knee landing theory suggests that females exhibit less knee flexion at the time of impact that may lead to ACL injury either by hyperextension or by anterior tibial translation (Decker et al., 2003; Huston et al., 2001). The present study investigated whether there is support for these three theories in a controlled laboratory environment by collecting biomechanical data for males and females as they are landing from a drop jump.

Fatigue is another factor that has been linked to athletic injuries. Several epidemiological studies support the notion that fatigue is a predisposing factor responsible for increased number of injuries (Bottini et al., 2000; Gabbett, 2000; 2002; Hawkins et al., 2001; Kersey and Rowan, 1983; Rahnama et al., 2002). Despite this, a very limited number of studies have examined the effect of fatigue on biomechanical variables during drop landing (Fagenbaum and Darling, 2003; Madigan and Pidcoe, 2003; Rozzi et al., 1999a) and only one of them (Madigan and Pidcoe, 2003) has used a functional fatigue protocol consisting of tasks that mimic sports activities such as squat and jumping.

In a study (Rozzi et al., 1999a) of male and female athletes who were fatigued to 25% of their original torque with the use of an isokinetic dynamometer, the researchers found decreased knee proprioception and increased onset of contraction time for the hamstrings and the gastrocnemius as subjects performed a landing task. They found no significant effect of gender or fatigue on balance or anterior tibial translation. However, males exhibited significantly increased mean vastus lateralis normalized EMG (NEMG) after fatigue compared to females (Rozzi et al., 1999a). The authors acknowledged that a significant limitation of their study was the non-functional fatigue protocol (isokinetic) (Rozzi et al., 1999a). Madigan and Pidcoe, 2003 looked at the effects of a functional fatigue protocol on the biomechanics of landing but all subjects were male and the authors did not collect frontal plane knee kinematic data.

Fagenbaum and Darling, 2003 concluded that fatigue had a similar effect on males and females but they used an isokinetic fatigue protocol which may not have replicated the fatigue that athletes experience during sports. Chappell et al., 2005 suggested that a fatigue protocol of vertical jumps and sprints caused subjects to land with increased proximal tibia peak anterior shear forces and decreased knee flexion at the time that peak anterior shear forces occur. Moreover, females landed with an external knee valgus moment that was increased in the post-fatigue condition while males exhibited an external varus moment. Females also exhibited a greater external knee flexion moment that the authors suggested may be due to increased quadriceps contraction, decreased hamstrings contraction or a combination of both conditions. However, NEMG data was not collected to further clarify the mechanism that leads to an increased external knee flexion moment.

The primary aim of the present study was to determine the effect of gender and fatigue on peak NEMG, kinetic, and kinematic variables during bilateral drop landings from a 40 cm platform. We hypothesized that females will exhibit landing biomechanics that predispose them to knee injury [greater knee valgus, greater vertical ground reaction force (VGRF)] and that males and females will land with increased knee valgus, VGRF, and NEMG activity of the rectus femoris and hamstrings after the fatigue protocol.

The study was conducted at the Harkness Dance Center Motion Analysis Laboratory, NYU - Hospital for Joint Diseases. It was a repeated measures pre-fatigue and post-fatigue experimental study using measures of EMG, kinetic, and kinematic data.

Based on a power analysis of the findings of a pilot study (3 males and 3 females) this study required 32 subjects to answer the hypotheses with a power of 0.8 and ? level of 0.05. Power analysis was performed with the use of G Power for knee valgus, VGRF, and rectus femoris NEMG (Buchner et al., 1997).

Thirty-two subjects were recruited from universities and colleges in the New York City area via announcements during classes. Only healthy volunteers between the ages of 20-40 years were recruited because this age group is more susceptible to ACL injury (Griffin, 2001). The inclusion criteria included participation in recreational sports at least twice/week for a minimum of 45 min. per practice session.

Exclusion criteria were: obesity (body mass index greater than 30 kg·m^-2); a history of injuries and/or diseases that would render unsafe the execution of the protocol; and a history of injuries and/or diseases that could affect the biomechanics of landing such as lower extremity fractures. Subjects were excluded if they had received specialized training in jumping and landing techniques such as through participation in gymnastics or dance.

All 32 subjects completed the entire protocol (pre-fatigue data collection, fatigue protocol, and post-fatigue data collection) in a single session. Subjects performed bilateral landings on a force plate from a 40 cm platform. The height was chosen to allow comparisons with the findings of other investigators who used similar protocols ( Decker et al., 2003; Fagenbaum and Darling, 2003; Ford et al., 2003; Madigan and Pidcoe, 2003; Rozzi et al., 1999a; 1999b) and as in these studies all subjects landed from the same height in an effort to minimize variability.

All NEMG and kinematic measurements were in reference to the right lower extremity. The literature supports that athletes injure the dominant and non-dominant extremity with equal frequency (Matava et al., 2002), and therefore comparisons between the dominant and non-dominant limb were not performed.

EMG data were collected with the Noraxon Myosystem 1400 (Noraxon USA, Inc., Scottsdale, AZ). The electrodes were disposable, surface, passive electrodes (Blue Sensor, Ambu, Inc., Linthicum, MD). The force plate was an OR6-5 AMTI biomechanical platform (AMTI, Watertown, MA). Kinematic data were collected with the use of eight Eagle cameras (Motion Analysis Corp. Santa Rosa, CA) and reflective markers were placed as per the "Helen Hayes system" (Richards, 2002). The software for data collection was the EvaRT 4.0 (Motion Analysis Corp. Santa Rosa, CA). Video data were smoothed using a Butterworth fourth order low pass filter with a cut-off frequency of 6 Hz. EMG data were filtered through a low pass 2^nd order Butterworth filter with a 6Hz cut-off frequency.

The force platform was time synchronized to the EMG and the motion analysis system. The kinetic and EMG data were sampled at 1200 Hz and the kinematic data were sampled at 240 Hz as appropriate for fast athletic maneuvers. Before each data collection session the system was calibrated to the manufacturer's recommendations.

Subjects were verbally informed of the study protocol and all risks and possible harms as described in the consent form. All subjects completed a sports activity and medical history questionnaire, signed a consent form approved by the NYU School of Medicine IRB, and were measured for height, weight, knee width, foot width, and foot length.

The skin was prepared and the surface electrodes were placed on the medial gastrocnemius, rectus femoris, biceps femoris, and medial hamstrings as described elsewhere (Fagenbaum and Darling, 2003). These sites of electrode placement are consistent with recent guidelines (Hermens et al., 2000) and are located between the motor point and the distal tendon in order to improve intra and inter-subject comparison reliability (Basmajian and DeLuca, 1985). Two electrodes were placed on each muscle at a 20 mm distance and parallel to fiber orientation (Hermens et al., 2000). Athletic tape was used to fixate the electrodes and decrease movement artifact (Hermens et al., 2000). The reflective markers were placed bilaterally on the second metatarsal, calcaneus, lateral malleolus, fibula, lateral knee joint line, thigh, anterior superior iliac spine, acromion, lateral humeral epicondyle, and distal radioulnar joint. Reflective markers were also placed on the sacrum and the left posterior superior iliac spine (offset) as per the "Helen Hayes" system (Richards, 2002). An initial "neutral" standing position of each subject was used to account for skeletal alignment differences as in similar studies (Hewett et al., 2005; Kernozek et al., 2005). With this technique, zero degree angles were defined as the angles between adjacent segments during the neutral standing trial.

The subjects were allowed two practice jumps and then performed three bilateral drop jumps from the 40 cm platform. They were instructed to drop directly down off the box and land with both legs on the force plate. Subjects did not receive any instructions on the landing technique to avoid a coaching effect. The effect of the arms was minimized by asking the subjects to keep their arms crossed against their chest (Rodacki et al., 2002; Decker et al., 2003). Trials were repeated when they were judged as non-acceptable (such as when subjects lost their balance or did not land with both feet on the force plate). Upon completion of three successful jumps, the wires were disconnected from the electrodes (but the electrodes were not removed). The subjects followed the fatigue protocol that consisted of 100 consecutive jumps over short (5-7 cm) obstacles and 50 maximal vertical jumps. This combination of activities was chosen to simulate activities commonly performed in sports and because an eccentric-concentric fatigue protocol is more effective in producing fatigue than a concentric fatigue protocol (Svantesson et al., 1998). The fatigue protocol was designed in a way that the fatigue-induced pattern is applicable to functional activities outside the laboratory setting. The protocol used in the present study is similar to fatigue protocols used in recent research ( Chappell et al., 2005; Madigan and Pidcoe, 2003). Similar protocols were sufficient in inducing fatigue as measured by quadriceps force output and jump height (Skurvydas et al., 2000; 2002) and mean EMG frequency (Madigan and Pidcoe, 2003). Moreover, the demands of games such as soccer are very similar for males and females in terms of distance covered, sprint duration, and exercise intensity (Davis and Brewer, 1993) suggesting that laboratory fatigue protocols have greater applicability if they fatigue male and female athletes in a similar way as occurs on the athletic field. After the fatigue protocol was completed, the wires were re-connected to the EMG electrodes and the same procedure of landings was repeated for the post-fatigue part of data collection. All subjects completed all post-fatigue trials within six minutes after the completion of the fatigue protocol. The landing cycle was defined as the time between initial contact and peak knee flexion.

The Orthotrak 5.0 (Motion Analysis Corp. Santa Rosa, CA) software was used to derive kinetic, kinematic, and EMG variables. The peak value for all variables during the landing cycle were identified and averaged across the three trials. The VGRF was normalized to body weight as in previous studies (Hewett et al., 1996; 1999). EMG amplitude was normalized to the maximum linear-enveloped EMG of each muscle (Arampatzis et al., 2003; Horita et al., 1999; Rodacki et al., 2002; Viitasalo et al., 1998) exhibited during the landing phase of bilateral landings from a 20 cm platform (mean of three trials).

Statistical analysis
Gender differences for anthropometric and sports participation variables were tested with independent sample t-tests. A repeated measures MANOVA was performed with the use of a statistical software package (SPSS 12.0, SPSS Inc., Chicago, IL, 60606) with gender and fatigue as independent variables., The dependent variables were peak values for knee flexion, knee valgus, foot abduction, VGRF, quadriceps NEMG, lateral hamstrings NEMG, medial hamstrings NEMG, and gastrocnemius NEMG as well as knee flexion angle at impact. The data were inspected and tested to insure that the assumptions for data normality of the univariate and multivariate repeated measures analysis of variance (ANOVA and MANOVA) were not violated, Separate univariate repeated measures ANOVA were performed for each dependent variable when the MANOVA reached statistical significance (p < 0.05) (Bray and Maxwell, 1985; Stevens, 2002). Cohen's d statistic of effect size was calculated in order to give a more complete picture of the effect of the independent variables on the dependent variables. The effect size is defined as trivial if it is <0.2, small 0.2-0.5, medium 0.5-0.8, and large>0.8 (Portney and Watkins, 2000).

Table 1 shows the descriptive statistics of demographic, anthropometric, and sports participation data. There were no gender differences in regards to age and sports activity. As expected, there were differences in the anthropometric variables between genders.

The results of the MANOVA showed that the effects of gender (F_9,22 = 5.763, p < 0.001) and fatigue (F_9,22 = 2.934, p = 0.019) were statistically significant but not the interaction of gender and fatigue (F_9,22 = 1.050, p = 0.434).

Univariate repeated measures ANOVA revealed that females landed with greater peak knee valgus (F_1,30 = 12.242, p = 0.001) and peak VGRF (F_1,30 = 10.548, p = 0.003) but knee flexion at contact was not different between males and females (F_1,30 = 0.003, p = 0.957). After the fatigue protocol subjects landed with increased peak foot abduction (F_1,30 = 4.534, p = 0. 042), peak VGRF (F_1,30 =4.7, p = 0.038), and peak rectus NEMG (F_1,30 = 6.252, p = 0.018). Tables 2 and 3 show the means (95% confidence intervals) and p-values (Cohen's d effect size statistic) for all dependent variables grouped by gender and fatigue condition respectively.

The objective of this study was to investigate the effects of gender and fatigue on the kinetics, kinematics, and NEMG of the lower extremity during landing from a drop jump. Gender had a significant effect on one kinematic variable (knee valgus) and the sole kinetic variable studied (VGRF). The findings regarding knee valgus agree with previous studies (Ford et al., 2003; Hewett et al., 2004; Kernozek et al., 2005) demonstrating that females exhibit greater knee valgus than males during landing from a jump. Increased knee valgus may produce excessive stress on the inert structures and lead to traumatic injury, consistent with the ligament dominance theory (Ford et al., 2003; Hewett et al., 2004).

Studies have shown that the knee is in a position of valgus at the time of ACL injury (Boden et al., 2000; Delfico and Garrett, 1998; Griffin, 2001; Griffin et al., 2000; Olsen et al., 2004) and that females exhibit greater peak knee valgus than males in a variety of athletic activities (Ferber et al., 2003; Horita et al., 1999; Kirkendall and Garrett, 2000; Zeller et al., 2003). Moreover, in a prospective study (Hewett et al., 2005), female athletes who subsequently suffered ACL injury were found to have increased peak knee valgus compared to female athletes who did not injure their ACL. The findings of this study provide further evidence that knee valgus is one of the key gender differences that may explain the increased incidence of ACL injuries in females.

The effect of gender on VGRF was also significant with females landing with higher normalized VGRF than males (5.3 BW vs. 3.9 BW). The VGRF reveals the ability of the athlete to efficiently attenuate the impact of landing. The lower the VGRF the more optimal the landing strategy, while high VGRF can lead to knee injuries (Dufec and Bates, 1991; Hewett et al., 1996; 2005). The VGRF findings of the present study show that females experience greater peak impact forces that may predispose them to non-contact injuries as they transfer to more proximal joints of the kinetic chain, such as the knee joint.

To the authors' knowledge, only one more study (Kernozek et al., 2005) has suggested that females landing from a drop jump have increased normalized VGRF compared to males. A study by Hewett et al., 1996 showed that a plyometric training program is effective in reducing peak VGRF in female athletes. VGRF has been shown to be modifiable with training (Hewett et al., 1996), linked to injury risk (Dufec and Bates, 1991; Hewett et al., 1996; 2005) and a significant gender difference as per the findings of the current study. Injury prevention programs should consider training females to land with a decreased VGRF in order to decrease risk of ACL injury.

Gender did not have a statistically significant effect on the peak NEMG of any of the muscles measured. However, females landed with 22% higher rectus femoris NEMG activity compared to males. It is important to recognize that comparison of NEMG muscle activity depends on the normalization method. In the present study, we normalized to peak EMG activity that was exhibited when each subject performed a similar task (landing from a 20 cm platform). Task specific EMG normalization has been reported to be superior to maximum voluntary contraction (MVC) for cyclical activities (Arendt-Nielsen and Sinkjaer, 1991; Kamen and Caldwell, 1996) and to exhibit less variability between and within subjects (Burden et al., 2003; Yang and Winter, 1984). However, the lack of statistical significance in regards to the effect of gender on rectus femoris NEMG in the present study may be explained by the similarity of the task that was used for normalization to the 40 cm landings. That is, females may have landed with higher rectus femoris activity from the 20 cm platform compared to males; therefore, the normalization method would have minimized the NEMG gender differences during the 40 cm landings. A different normalization method may have resulted in significant gender differences relative to rectus femoris NEMG. Fagenbaum and Darling, 2003 normalized to MVC and reported that females landed with higher quadriceps NEMG activity but did not provide details on the magnitude of the difference which precludes comparisons with the findings of the present study. Zeller et al., 2003 investigated gender differences during a single leg squat task and also normalized to MVC and reported that females exhibited 47% higher mean rectus femoris NEMG than males.

The effect of fatigue on male and female athletes who participated in the present study was significant relative to three variables: peak VGRF, peak foot abduction, and peak rectus femoris NEMG. Although the changes in VGRF were statistically significant, the clinical significance is questionable; males and females landed with increased forces after the fatigue protocol by 0.13 BW (3.4% increase) and 0.34 BW (6.5% increase) respectively. It is unclear to what degree an increase of this magnitude can be said to contribute to injury. These findings, however, are opposite those of Madigan and Pidcoe, 2003 who reported that fatigue caused a 12% decrease in peak VGRF in a group of men. The difference in results may be due to the methodology used by Madigan and Pidcoe, 2003; they did not test female subjects that contributed most of the peak VGRF increase in the present study and they fatigued subjects to exhaustion. Another difference between the two studies is that the level of fatigue was not measured in the present study; therefore some subjects may have been fatigued by the protocol more than other subjects. The VGRF increase in the current study suggests that athletes' ability to attenuate the impact of landing decreases after fatigue.

In the present study, subjects in the post-fatigue condition exhibited a significant increase in foot abduction from 6.6° to 8.3° The role of foot abduction in the occurrence of ACL injury is unclear, however some have suggested that it leads to increased knee valgus (Lin et al., 2001) and subsequent knee injury, although in the present study fatigue did not cause a significant increase in knee valgus. Further research is needed to clarify the role of foot abduction on the mechanism of ACL injury.

Subjects in the post-fatigue condition exhibited a statistically significant increase of rectus femoris NEMG (F_1,30 = 6.252, p = 0.018). This finding, however, does not mean that quadriceps forces were higher after fatigue. Although the NEMG activity was higher, the ability of the muscle fibers to develop tension decreases with fatigue. Dimitrova and Dimitrov, 2003 critically reviewed the literature and reported that fatigue can cause an increase, decrease, or insignificant change on the NEMG amplitude during dynamic activities, however, most studies have shown that fatigue increases the amplitude of quadriceps EMG activity (Bonnard et al., 1994; Nummela et al., 1994; Psek and Cafarelli, 1993; Rodacki et al., 2001; Viitasalo et al., 1993). The findings of the present study also suggest that fatigue causes an increase in muscle activity when measured by dynamic EMG.

Previous studies (Decker et al., 2003; Huston et al., 2001) have provided support to the straight knee landing theory by reporting that females land with their knees closer to extension than males. However, the findings of the present study are in agreement with recent studies (Fagenbaum and Darling, 2003; McLean et al., 2004) that do not support the straight knee landing theory. Males and females landed with very similar knee flexion angles at the time of impact (20.2° vs. 20.0°, F_1,30 = 0.003, p = 0.957).

The effectiveness of sports injury prevention programs has been well documented both in epidemiological studies that showed a decrease of athletic injuries (Cerulli et al., 2001; Hewett et al., 1999; Myklebust et al., 2002) and in laboratory studies that showed a decrease in peak VGRF (Hewett et al., 1996). However, the outcomes of these programs may improve if they focus on variables that have been identified by biomechanical studies as reflecting genuine differences due to gender or fatigue and have been proven to be modifiable. In the present study, females landed with greater peak knee valgus and peak VGRF than males. Fatigued subjects exhibited increased peak VGRF and peak rectus femoris NEMG. The findings of the present study may be used to guide injury prevention programs designed to improve the landing technique of female athletes involved in sports. Control of these variables will likely reduce landing patterns that result in high VGRF, increased stress on the passive structures, and subsequently, ligamentous injuries.

Limitations
The present study was performed in a controlled laboratory environment where subjects knew exactly what to expect. Although this allows accurate comparison between groups, it does not closely simulate the athletic environment. Most ACL injuries are non-contact in nature but there are additional unexpected factors during a game such as decreased friction of the floor surface, motion of the opponent, or miscalculation of ball motion that make landing from a jump more unpredictable and dangerous than in a laboratory setting. Moreover, the landing task itself was not identical to the landing technique observed in the athletic field since we instructed subjects to keep their arms crossed across their chest and jump down from a platform. These modifications decrease the generalizability of the findings to the athletic field.

All subjects were recreational athletes who participated at least twice per week in a variety of sports that involved jumping. There were no differences between males and females in regards to hours of sports participation per week. However, this does not insure equal proficiency in drop landings; some subjects might have been more proficient than others in landing from a jump. We also did not objectively measure the effect of the fatigue protocol on the volunteers' musculoskeletal system. A more homogenous group of subjects such as recreational basketball or volleyball players would make the findings of this study less generalizable but would increase the internal validity of the study.

Finally, a limitation inherent of this study as well as of most sports biomechanics studies is that the values used for statistical analysis were the peak values of the biomechanical variables. Peak values of kinetic, kinematic and NEMG variables would possess greater clinical significance if they were shown to occur at the first part of the landing cycle that coincides with the phase the ACL is loaded (Pflum et al., 2004) and injuries are known to occur (Boden et al., 2000; Griffin, 2001; Olsen et al., 2004).

• Female athletes landed with increased knee valgus and VGRF which may predispose them to ACL injury.
• Fatigue elicited a similar response in male and female athletes.
• The effectiveness of sports injury prevention programs may improve by focusing on teaching females to land softer and with less knee valgus.