Journal of Applied Biomechanics, 2014, 30, 649-654 http://dx.doi.org/10.1123/JAB.2013-0261 © 2014 Human Kinetics, Inc.
An Official Journal of ISB www.JAB-Journal.com ORIGINAL RESEARCH
The Role of Running Mileage on Coordination Patterns in Running Katherine A. Boyer, Julia Freedman Silvernail, and Joseph Hamill University of Massachusetts-Amherst Injury rates among runners are high, with the knee injured most frequently. The interaction of running experience and running mechanics is not well understood but may be important for understanding relative injury risk in low vs higher mileage runners. The study aim was to apply a principal component analysis (PCA) to test the hypothesis that differences exist in kinematic waveforms and coordination between higher and low mileage groups. Gait data were collected for 50 subjects running at 3.5 m/s assigned to either a low (< 15 miles/wk) or higher (> 20 miles/wk, 1 year experience) mileage group. A PCA was performed on a matrix of trial vectors of all force, joint kinematic, and center of pressure data. The projection of the subjects’ trial vectors onto the linear combination of PC7, PC10, PC13, and PC19 was significantly different between the higher and lower mileage groups (d = 0.63, P = .012). This resultant PC represented variation in transverse plane pelvic rotation, hip internal rotation, and hip and knee abduction and adduction angles. These results suggest the coordination of lower extremity segment kinematics is different for lower and higher mileage runners. The adopted patterns of coordinated motion may explain the lower incidence of overuse knee injuries for higher mileage runners. Keywords: running, injury, coordination, principal component analysis (PCA), mileage Running has increased in popularity in the last three decades, but with increased participation has come an increase in reported running injuries.1,2 Injury rates among recreational runners are estimated to be as high as 29.4%, with overuse knee injuries such as anterior knee pain (AKP) and iliotibial band syndrome (ITBS) being the most frequency reported.1,3 Overuse injuries typically result from a combination of training errors, as well as anatomical, footwear, running surface, and biomechanical factors. There is evidence to suggest that running experience or training volume may also be a factor, but there are conflicting views on the relative risks for experienced versus novice athletes1,3,4 The interaction of running experience, weekly mileage, and the mechanics of running is not well understood but may be an important factor for understanding relative injury risk in low versus higher mileage runners. Given the multifactorial nature and multitude of types and locations for lower extremity injuries, it is perhaps not surprising that the relationship between training-related factors and running injury risk is not clear. A recent systematic review of risk factors for running injuries suggests there is conflicting evidence for an association between running experience and overall lower extremity injury risk.5 This review reported that inexperience was related to hamstring and knee injuries, while foot injuries were associated with more experienced runners. There is also conflicting evidence for an association between increases in training distance per week (ie, training errors) and overall lower extremity risk. Again, it seems that the risk for knee injuries is different than for thigh or hamstring injuries. There is evidence that an increase in training distance per week is protective for knee injuries,6 while this same increase in Katherine A. Boyer, Julia Freedman Silvernail, and Joseph Hamill are with the Biomechanics Laboratory, Department of Kinesiology, University of Massachusetts-Amherst, Amherst, Massachusetts. Address author correspondence to Katherine A. Boyer at [email protected].
training distance may result in an increase in the risk for thigh or hamstring injuries.7 Together these studies suggest that there may be different risk factors that interact with training distances for the multitude of lower extremity injuries. Running is a multisegmental motion with many degrees of freedom and thus, there exists a high-dimensional space of potential solutions to attain the same movement goal.8 In the sagittal plane, the coupling or coordination of the segment motions is essential to task completion and thus to position the foot properly for the landing phase of the running stride.9 For example, for a given knee angle at heel-strike for a particular running speed, the hip and ankle angles cannot be arbitrarily selected but will be a function of the selected knee angle to successfully complete the stride. Both running experience or skill level and training-induced differences in the neuromuscular patterns or the mechanical properties of the musculoskeletal tissues could influence the coupling or coordination of segment motion. It has been hypothesized that in normal running a disruption or decoupling of these segments’ motions may be deleterious and related to injury development.10 There is substantial evidence linking the frontal plane hip joint motion and strength with knee injury risk and severity.11–14 Thus, the coordination of the secondary motions (frontal and transverse plane) is expected to be more relevant with respect to the discussion on intersegment movement coordination and injury risk. An analysis of a local coupling of the transverse plane thigh and shank motion at heel-strike in both AKP and healthy subjects found there was a decreased variability in the coupling of these motions in the AKP subjects compared with healthy controls.10 This suggests that coordination of segment motions in the secondary planes is an important consideration for injury risk assessment. An increased understanding of the interrelationship between segment motions in running and the role of mileage on this relationship may provide insight into injury mechanisms in running and the relative risks for different groups. 649
Downloaded by New York University on 09/19/16, Volume 30, Article Number 5
650 Boyer, Freedman Silvernail, and Hamill
A traditional analysis of discrete joint angle values provides little information about the pathway of movement or the relative coordination of limb segment motions used to achieve a movement goal. The application of higher-order statistical analysis methods such as principal component (PC) analysis provides a data-driven methodology to analyze correlated deviations from a mean movement pattern.15 The results of a PC analysis that are applied to a data matrix consisting of all variables and time points that represent the lower extremity movement patterns are features of how the relationships between these variables differ between selected groups. Thus, through this application of the PC analysis, segment couplings and differences between groups in the coupling of threedimensional components of joint motion at all lower extremity joints can be examined in a systematic manner. If running mileage is a risk factor for the development of knee injuries then a difference in the multisegment coordination patterns impacting the frontal plane hip and knee motions may be expected. Therefore, the aim of this study was to test the hypothesis that there are differences in the PC weighting coefficients between the higher and low mileage groups indicative of differences in the frontal and transverse plane joint kinematic waveform patterns and altered multisegment coordination.
Methods Participants Fifty healthy adults (26 males and 24 females) completed five running trials at 3.5 m/s while gait analysis data were collected. Participants were enrolled in the study if they had been free from any lower extremity musculoskeletal injury in the past 12 months and had no prior history of major lower limb reconstructive surgery. Participants were assigned to one of two groups: (1) a low mileage and experience group, who ran on average less than 15 miles per week and had no history of higher mileage running (26.3 ± 9.14 y, 22.9 ± 3.0 kg/m2, 7.2 ± 4.6 miles/wk; 13 males); and (2) a higher mileage group with at least 1 year of experience at mileage of greater than 20 miles per week (29.5 ± 10.9 y; 22.2 ± 2.2 kg/m2, 33.8 ± 14.1 miles/wk; 13 males). The protocol was approved by the university institutional review board and all participants provided written informed consent before the completion of any study procedures. Participants were asked to self-report their weekly mileage, the number of months at this training volume, and any previous running experience. Participants wore standard laboratory footwear that could be considered neutral training shoes.
Experimental Set-up Three-dimensional lower extremity joint kinematics and ground reaction force (GRF) data were captured at 240 Hz with a 9-camera motion analysis system (Qualysis Inc., Sweden) and at 1200 Hz infloor AMTI force platform as subjects ran over a 15-m runway. Each subject performed five trials at running speed of 3.5 m/s ± 5%. Running speed was monitored through two sets of photo cells positioned along the runway on either side of the force platform. To capture the position of the limb segments in the laboratory, reflective markers were positioned on the right leg of each participant. The right leg was arbitrarily selected for analysis in this study. The markers included both calibration markers and tracking markers. Calibration markers were placed on the left and right greater trochanters, the right side anterior superior iliac spine (ASIS), iliac crest, L5-S1 interspace, the medial and lateral femoral epicondyles, the medial and lateral
malleoli, the sides of the first and fifth metatarsal heads, and the tip of the shoe. These markers were used to develop the anatomic segment coordinate systems. Tracking marker clusters of four markers each for the shank and thigh and three markers for the heel counter of the shoe affixed to thermoplastic shells were secured around the thigh, shank, and on the heel counter of the shoe. The anatomical markers on the pelvis also served as tracking markers. With all markers affixed, a standing calibration trial was collected with the participant standing in with their feet positioned shoulderwidth apart and with 0° of abduction and adduction; the knees were fully extended and the hips were neutral with 0° of rotation. Following the calibration trial all anatomical markers, with the exception of those on the pelvis, were removed. The joint center positions were located as the midpoint between the medial and lateral malleoli and condyles for the ankle and knee, respectively. The hip joint center was determined using a previously developed regression equation.16
Data Analysis Qualysis Track Manager software (Qualysis, Inc., Sweden) was used to track the marker positions and to export the data into .C3D format. Raw marker data were further processed using Visual 3D software (C-Motion, Inc., Rockville, MD). A fourth order, zero-lag Butterworth digital low-pass filter was used to process the data with a cutoff frequency of 12 Hz. A right-handed local coordinate system was created for each of the pelvis, thigh, shank, and foot segments. The three-dimensional lower extremity joint angles were calculated with respect to the proximal segment using an Xyz Cardan rotation sequence, and a sequence representing flexion/extension, abduction/ adduction, and axial rotation.17 The beginning and end of the stance phase was determined by the instant the vertical GRF exceeded and fell below 15 N, respectively. There were 26 waveforms of interest for this study: the threedimensional motion of the pelvis, hip, knee, ankle, and the mid- to rearfoot, fore- to midfoot, and hallux to forefoot segments; three components of the GRF; and two center of pressure (COP) components. The PC analysis approach used for this study was based on the methods of Boyer et al, Federolf et al, and Nigg.18–21 The data were prepared before the implementation of the PC analysis as follows. The waveforms were interpolated such that 101 points represented the stance phase of the gait cycle and the mean over five trials for each subject was calculated for each waveform. The gait waveforms of interest have different units and scales and the magnitudes of these inputs to the PC analysis are not proportional to the biological relevance of these motions or forces with respect to injury or performance. Thus, it was necessary to standardize the range of all waveforms before the data analysis, as PC analysis is scale sensitive and the aim of this study was to identify group differences in the shape and relative timing of the gait waveforms in running. For each gait waveform, the mean and standard deviation across the stance phase was determined for all participants. Each input waveform was then scaled by subtracting the mean over the stance phase and then dividing the amplitude by the mean standard deviation across the stance phase.22 This preparation of data normalizes the variance of each waveform before the implementation of the PC analysis, and thus, each waveform regardless of scale will have equal importance in the analysis. The determination of PCi depends on the pattern and timing of the different waveforms throughout movement and not on the overall magnitude of a given waveform. The normalized mean waveforms for each participant were then appended together into one column trial vector with 2626 vector
Downloaded by New York University on 09/19/16, Volume 30, Article Number 5
Role of Running Mileage on Coordination Patterns 651
components (waveform × time points). With this representation of the data, each participant’s data can be visualized as a single point in a high dimensional data space. For analysis, the trial vectors for all participants were assembled into a single matrix and the mean of all trial vectors was subtracted from each column vector. The covariance matrix of the matrix of trial vectors was then computed. An eigen decomposition was then performed to determine the characteristic roots (eigenvalues) and the associated orthogonal eigenvectors or principal components (PCi). The eigenvalues (EVi; expressed as a percentage of the sum of all eigenvalues) indicate the variance (%) in the data set explained by a particular PCi. The eigenvectors or principal component vectors represent the primary directions of variance in the data set. With mean centered data, the eigenvectors maximize the variance from the mean of the data set. These vectors are mutually orthogonal and thus created a new coordinate system (or basis) onto which the original data can be transformed. The matrix of trial vectors was projected onto the PCis for further statistical analysis. Each participant’s trial vector could then be represented as a weighted linear combination of the PCis (Equation 1).
n uuur uuuur t rial = trial m + Âci ◊ PC i , (1) i=1
where trial, is an individual participant’s trial vector, trialm is the mean trial vector of all participants, ci is the score or weight coefficient associated for each of the n, PCi. Characteristic differences between the gait patterns of lower and higher mileage runners were identified by comparing the scores (also called weighting coefficients) ci of the first 20 PCi. Unpaired Student’s t test and effect size of difference (Cohen’s d) were used to test for group differences in the PC scores for each component. A threshold of P < .1 or ES > 0.6 was used to select relevant principal component for further data reduction. A resultant vector D was calculated as the weighted linear combination of the selected PCi (Equation 2). A step-wise procedure was used when calculating the resultant vector. The difference in coefficients and the effect size between the higher and lower mileage groups was determined as higher-order PC were incrementally included in the resultant vector determination. This method ensures there was an increase in the effect size with each additional PCi. This vector combined the characteristic deviations from the mean normalized gait pattern that are associated with group assignment. The EVi were used as the weight factors for the resultant vector calculation and the resultant vector was normalized to unit length. To interpret the PC analysis outcomes and visualize the deviations in the mean movement patterns represented by the resultant vector, a trial vector was recomputed for each subject by adding the resultant vector D weighted by the subjects score ci to the mean normalized waveforms and retracing the normalization steps. The specific joint angle or force variables within the trial vector where there was variation evident between participants in the recomputed vectors are those for which there are significant group differences were interpreted.
D = EVi ◊ PCi + EVi+m ◊ PCi+m + EVi+h ◊ PCi+h + …, (2)
where EVi, are the eigenvalues expressed as a percentage of the total data variance corresponding to the PCi. Only those PCi where significant group differences in the scores were found are included in the calculation of D. Each participant’s trial vector was recomputed from the mean trial vector plus the scaled resultant vector for the interpretation of the primary group-based dif-
ferences. The subsections within these computed vectors where there was substantial or visible variation indicated the associated time point and variables in the trial vector that were different based on the group assignment.
Results The projection of the participants’ trial vectors onto the linear combination of PC7, PC10, PC13, and PC19 was significantly different between the higher and lower mileage groups (d = 0.63, P = .012; Table 1). Together these components explained 4.6% of the variance in the data. The examination of the recomputed trial vectors using the resultant vector D indicated that variation related to group assignment was present in the transverse plane kinematics of the hip and pelvis, the frontal plane hip, and the knee and foot segment kinematics (Figure 1). In the transverse plane the internal-external rotation range of motion and pattern of motion of the hip were different between the low and higher mileage runners. The lower mileage runners continued to externally rotate at the hip in midstance and had a relatively more externally rotated hip through terminal stance. There were no differences in the hip internal-external rotation angle for the first 40% of the stance phase. The lower mileage group also showed less pelvic transverse plane rotation toward stance leg throughout ground contact. There was also a systematic offset in the rearfoot internal-external rotation angle throughout stance with the lower mileage group in relatively greater external rotation. In the frontal plane, group differences in the deviation from the mean were found for the knee and hip adduction-abduction angles and for the rearfoot inversion-eversion angles. For the knee, the lower mileage group was more adducted throughout the stance phase and had greater knee adduction range of motion in early stance. There was a systematic offset in the hip adduction position throughout stance with the lower mileage group in a less adducted position. The lower mileage group also has a smaller hip abductionadduction range of motion, particularly in the second half of the stance phase. The primary difference in rearfoot inversion was found in the final two-thirds of the stance phase. The lower mileage group had a smaller rearfoot inversion angle compared with the higher mileage group. There were no differences in the forefoot inversioneversion angle or motion between groups. There were no nonzero components in the resultant PC vector along the axes related to the sagittal plane kinematics for the pelvis, hip, knee, or ankle joint. Thus, no group differences were identified through this analysis for the sagittal plane kinematics with the exception of the forefoot relative to midfoot flexion. The lower mileage group showed less forefoot to midfoot flexion throughout the stance phase. Table 1 The relative variance explained, effect size (Cohen’s d), and p value for the four PCs identified as significantly different between groups PC
Eigenvalue/% Variance Explained
p Value
Effect Size
PC7
2.68
.04
–0.48
PC10
1.12
.09
–0.38
PC13
0.6
.02
0.60
PC19
0.2
.08
–0.38
Resultant PC
4.6
.01
0.63
Downloaded by New York University on 09/19/16, Volume 30, Article Number 5
652 Boyer, Freedman Silvernail, and Hamill
Figure 1 — The group average deviations in joint angles from the mean movement patterns for the lower (black) and higher (gray) mileage groups. The differences related to group assignment in deviations from the mean gait patterns were interpreted by plotting the portion of the weighted resultant vector plus the mean trial vector for each subject corresponding to each of the 26 variables that make up the trial vectors. The mean waveform for both the lower and higher mileage groups for the following variables shown (knee and hip adduction-abduction, hip internal-external rotation, pelvis transverse plane rotation, rearfoot inversion-eversion, transverse plane rotation, and forefoot relative to midfoot flexion) are those for which there was variance among subjects from the mean waveform for at least 20% of the stance phase. An amplification factor of 20 was used for visualization.
Discussion The purpose of this study was to investigate what, if any, role running mileage may have on multisegment movement coordination and running injury risk factors. The results of this study supported the hypothesis that systematic differences between lower and higher mileage runners can be identified using PC analysis to analyze the interrelationship between lower extremity joint kinematics and external forces. A single resultant PC vector was identified for which the PC scores for the higher and lower mileage runners were significantly different. The calculated resultant PC identifies the direction in high-dimensional space along which there were systematic group differences in the deviation from the mean normalized
movement patterns. This direction represented the transverse plane pelvic rotation, hip internal rotation, hip and knee abduction and adduction, and sagittal and frontal plane multisegment foot motion. The group differences identified through this data-driven method are particularly interesting in light of the epidemiological studies suggesting higher mileage runners may be protected to a degree against knee injuries.5 Biomechanical risk factors identified for the development of overuse knee injuries such as AKP and ITBS include the magnitude or excursion of the transverse and frontal plane joint kinematics, more specifically, increased knee abduction and greater hip internal rotation.11,23,24 It has been proposed that weak hip musculature may lead to greater femoral adduction and internal rotation range of motion that will lead to increased knee
Downloaded by New York University on 09/19/16, Volume 30, Article Number 5
Role of Running Mileage on Coordination Patterns 653
valgus, and as a result, increase the stress on the knee and contribute to overuse knee injuries.11,23,24 In the current study, the higher mileage group showed greater transverse plane pelvic rotation toward the stance leg and less hip internal rotation range of motion, but a slightly greater internal rotation angle in the last half of the stance phase and less knee adduction. The biomechanical data from this study seems at odds with the epidemiological data at first glance, however a more internally rotated hip in combination with greater pelvic rotation toward the stance phase suggests that the coupling between the pelvis and the femur in the transverse plane is different and the difference in the distal segment movement between groups may be minimal. This suggests that an analysis of the joint angles in isolation may provide a limited view of injury risk assuming the higher mileage participants in this study are similar to high mileage runners from the literature. While the proximal segment (hip/pelvis) motion and the motion of the knee joint itself have been linked with knee injury presence11,13,25 and risk, other studies have suggested that distal mechanics (ie, foot motion) are also linked with AKP development.11,13,24,26,27 The resultant PC also varied along the axes related to the deviations from the mean normalized inversion-eversion angle for the rearfoot segments. This indicated that for the higher mileage group the rearfoot was in a slightly more inverted position through the later part of the stance phase. There was also a difference in the transverse plane with the higher mileage runners in a more internally rotated position throughout the stance phase. The deviations in the rearfoot inversion angle, but not the forefoot inversion-eversion, suggest that the coupling of the fore- and rearfoot in the frontal plane is different for the higher and lower mileage groups. In conjunction with the proposed biomechanical risk factor and epidemiological literature, this suggests that the neuromuscular adaptations associated with higher mileage and more experience may result in coordination of the lower limb segments in the transverse and frontal planes that lowers the risk for knee injury development. The resultant vector spanned the axes related to the number of joint motions and phases of stance and thus provides some unique insight into the complexity of the multisegment coordination implicit in the mechanics of running. The PCis describe a linear relationship between the variables and time points along which they vary. Therefore, it is possible to say that these variables at specific times within the stance phase are correlated or associated with one another. The relevant and associated variables to describe the group differences were all in the transverse and frontal planes. This supports a functional relationship between the secondary motions of all lower extremity segments and thus suggests that studies focused on injury development in running must not focus on a single pathway (hip or foot motion), but rather, need to consider motion of the entire lower extremity. In running there needs to be a high degree of spatiotemporal coordination between and within the limb segments. Thus, it is of note that the resultant vector did not vary along the axes associated with deviations from the mean for any of the sagittal plane motions. This suggests that, despite having multiple degrees of freedom within the sagittal plane that may vary, there is not a systematically different way in which lower or higher mileage runners organize these degrees of freedom. The sagittal plane motions may be constrained to a large degree by the task or movement goal, a stride of running at 3.5 m/s, and it does not appear that skill level or mileage has a significant impact on an individual’s ability to perform this motion. The PC analysis identified subtle but systematic differences in the joint kinematics for the higher and lower mileage groups. The
differences between groups in the PCi scores were found primarily in the higher order PCi. . In general, the lower PCi will reflect the dominant movements and those that are sensitive to external influences or to training effects will be higher order PCi representing more subtle movements.28,29 While a PC analysis provides a quantitative method for identifying group differences in both the shape and relative coordination of gait waveforms, the interpretation of these differences remains subjective. The presentation of the data as a mean plus a weighted addition of the resultant PC vector illustrates the direction of the variation in the data as a fraction of the overall variance. We examined the variation between participants in the recomputed vectors to identify where and how the gait waveforms differed between groups. The recomputed trial vectors and the mean of these recomputed trial vectors illustrated the portion of the variance at each time point that we can attribute to mileage differences alone. Additional variance in the data may be present but not attributable to the group assignment (ie, other PCis are directed along these axes). One of the challenges when using a PC analysis is that the outcomes of the analysis, while statistically rigorous, are different from that which we get using a traditional discrete time point analysis. Thus, it is difficult to make direct comparisons between the results of these two analyses. However, the additional information related to the overall pattern and timing of gait waveforms that is gained through the application of methods associated with the dynamical system theory may be important in developing potential prevention strategies based on gait-modifying interventions. The results of this study suggest that there are subtle differences in the way that higher and lower mileage runners coordinate movement of the lower extremity and is a first step in understanding the differences in movement coordination between these two groups. A fundamental aim of this study was to apply the concepts and methods from dynamical systems theory to the study of movement mechanics and the potential for differential risk of common lower extremity overuse running injuries in lower and higher mileage runners. A systematic difference between the two groups was identified and this was primarily related to the transverse and frontal plane kinematics of the pelvic, hip, knee, and rear forefoot. The selection of a sagittal plane movement pattern in the higher mileage group that results in less risky frontal and transverse plane joint kinematic pattern may reduce injury risk and improve the efficiency of movement. The athletes in the higher mileage group self-reported no significant history of overuse injury despite some athletes reaching 70 miles/wk. Given the apparent differences in the lower and higher mileage groups, these results raise the question: Do these lower mileage runners change their mechanics as they increase mileage and gain experience or do they get injured before they become higher mileage runners? Acknowledgment This research was supported by a grant from Brooks Sports, Inc.
Reference List 1. Taunton JE, Ryan MB, Clement DB, McKenzie DC, Lloyd-Smith DR, Zumbo BD. A prospective study of running injuries: the Vancouver Sun Run “In Training” clinics. Br J Sports Med. 2003;37:239–244. PubMed doi:10.1136/bjsm.37.3.239 2. Clement DB, Taunton JE, Smart GW, McNicol KL. A survey of overuse running injuries. Phys Sportsmed. 1981;9:47–58. 3. Macintyre J, Taunton J, Clement D et al. Running injuries: a clinical study of 4,173 cases. Clin J Sports Med. 1991;1:81–87.
Downloaded by New York University on 09/19/16, Volume 30, Article Number 5
654 Boyer, Freedman Silvernail, and Hamill 4. van Mechelen W. Running injuries. A review of the epidemiological literature. Sports Med. 1992;14:320–335. PubMed doi:10.2165/00007256-199214050-00004 5. van Gent RN, Siem D. van MM, van Os AG, Bierma-Zeinstra SM, Koes BW. Incidence and determinants of lower extremity running injuries in long distance runners: a systematic review. Br J Sports Med. 2007;41:469–480. PubMed doi:10.1136/bjsm.2006.033548 6. Satterthwaite P, Norton R, Larmer P, Robinson E. Risk factors for injuries and other health problems sustained in a marathon. Br J Sports Med. 1999;33:22–26. PubMed doi:10.1136/bjsm.33.1.22 7. Wen DY, Puffer JC, Schmalzried TP. Injuries in runners: a prospective study of alignment. Clin J Sport Med. 1998;8:187–194. PubMed doi:10.1097/00042752-199807000-00005 8. Bernstein NA. The co-ordination and regulation of movements. London, UK: Pergamon Press; 1967. 9. Winter DA. Foot trajectory in human gait: a precise and multifactorial motor control task. Phys Ther. 1992;72:45–53. PubMed 10. Hamill J, Van Emmerik RE, Heiderscheit BC, Li L. A dynamical systems approach to lower extremity running injuries. Clin Biomech (Bristol, Avon). 1999;14:297–308. PubMed doi:10.1016/S02680033(98)90092-4 11. Ferber R, Noehren B, Hamill J, Davis IS. Competitive female runners with a history of iliotibial band syndrome demonstrate atypical hip and knee kinematics. J Orthop Sports Phys Ther. 2010;40:52–58. PubMed doi:10.2519/jospt.2010.3028 12. Hamill J, Miller R, Noehren B, Davis I. A prospective study of iliotibial band strain in runners. Clin Biomech (Bristol, Avon). 2008;23:1018– 1025. PubMed doi:10.1016/j.clinbiomech.2008.04.017 13. Noehren B, Hamill J, Davis I. Prospective evidence for a hip etiology in patellofemoral pain. Med Sci Sports Exerc. 2013;45:1120–1124. PubMed doi:10.1249/MSS.0b013e31828249d2 14. Noehren B, Davis I, Hamill J. ASB clinical biomechanics award winner 2006 prospective study of the biomechanical factors associated with iliotibial band syndrome. Clin Biomech (Bristol, Avon). 2007;22:951–956. PubMed doi:10.1016/j.clinbiomech.2007.07.001 15. Daffertshofer A, Lamoth CJ, Meijer OG, Beek PJ. PCA in studying coordination and variability: a tutorial. Clin Biomech (Bristol, Avon). 2004;19:415–428. PubMed doi:10.1016/j.clinbiomech.2004.01.005 16. Bell AL, Pedersen DR, Brand RA. A comparison of the accuracy of several hip center location prediction methods. J Biomech. 1990;23:617–621. PubMed doi:10.1016/0021-9290(90)90054-7 17. Cole GK, Nigg BM, Ronsky JL, Yeadon MR. Application of the joint coordinate system to three-dimensional joint attitude and movement representation: a standardization proposal. J Biomech Eng. 1993;115:344–349. PubMed doi:10.1115/1.2895496
18. Boyer KA, Federolf P, Lin C, Nigg BM, Andriacchi TP. Kinematic adaptations to a variable stiffness shoe: mechanisms for reducing joint loading. J Biomech. 2012;45:1619–1624. PubMed doi:10.1016/j. jbiomech.2012.04.010 19. Federolf PA, Boyer KA, Andriacchi TP. Application of principal component analysis in clinical gait research: Identification of systematic differences between healthy and medial knee-osteoarthritic gait. J Biomech. 2013; 46(13):2173–2178. PubMed 20. Nigg B. Biomechanics of sports shoes. Calgary, AB: Topline Printing Inc.; 2010. 21. Federolf P, Tecante K, Nigg B. A holistic approach to study the temporal variability in gait. J Biomech. 2012;45:1127–1132. PubMed doi:10.1016/j.jbiomech.2012.02.008 22. St-Onge N, Duval N, Yahia L, Feldman AG. Interjoint coordination in lower limbs in patients with a rupture of the anterior cruciate ligament of the knee joint. Knee Surg Sports Traumatol Arthrosc. 2004;12:203–216. PubMed doi:10.1007/s00167-003-0420-5 23. Dierks TA, Manal KT, Hamill J, Davis I. Lower extremity kinematics in runners with patellofemoral pain during a prolonged run. Med Sci Sports Exerc. 2011;43:693–700. PubMed doi:10.1249/ MSS.0b013e3181f744f5 24. Ferber R, Kendall KD, Farr L. Changes in knee biomechanics after a hip-abductor strengthening protocol for runners with patellofemoral pain syndrome. J Athl Train. 2011;46:142–149. PubMed doi:10.4085/1062-6050-46.2.142 25. Powers CM. The influence of abnormal hip mechanics on knee injury: a biomechanical perspective. J Orthop Sports Phys Ther. 2010;40:42–51. PubMed doi:10.2519/jospt.2010.3337 26. Dierks TA, Manal KT, Hamill J, Davis IS. Proximal and distal influences on hip and knee kinematics in runners with patellofemoral pain during a prolonged run. J Orthop Sports Phys Ther. 2008;38:448–456. PubMed doi:10.2519/jospt.2008.2490 27. Rodrigues P, TenBroek T, Hamill J. Runners with anterior knee pain use a greater percentage of their available pronation range of motion. J Appl Biomech. 2013;29:141–146. PubMed 28. Maurer C, Federolf P. von T, V, Stirling L, Nigg BM. Discrimination of gender-, speed-, and shoe-dependent movement patterns in runners using full-body kinematics. Gait Posture. 2012;36:40–45. PubMed doi:10.1016/j.gaitpost.2011.12.023 29. von Tscharner V, Enders H, Maurer C. Subspace identification and classification of healthy human gait. PLoS ONE. 2013;8(7):e65063. PubMed doi:10.1371/journal.pone.0065063
An Official Journal of ISB www.JAB-Journal.com ORIGINAL RESEARCH
The Role of Running Mileage on Coordination Patterns in Running Katherine A. Boyer, Julia Freedman Silvernail, and Joseph Hamill University of Massachusetts-Amherst Injury rates among runners are high, with the knee injured most frequently. The interaction of running experience and running mechanics is not well understood but may be important for understanding relative injury risk in low vs higher mileage runners. The study aim was to apply a principal component analysis (PCA) to test the hypothesis that differences exist in kinematic waveforms and coordination between higher and low mileage groups. Gait data were collected for 50 subjects running at 3.5 m/s assigned to either a low (< 15 miles/wk) or higher (> 20 miles/wk, 1 year experience) mileage group. A PCA was performed on a matrix of trial vectors of all force, joint kinematic, and center of pressure data. The projection of the subjects’ trial vectors onto the linear combination of PC7, PC10, PC13, and PC19 was significantly different between the higher and lower mileage groups (d = 0.63, P = .012). This resultant PC represented variation in transverse plane pelvic rotation, hip internal rotation, and hip and knee abduction and adduction angles. These results suggest the coordination of lower extremity segment kinematics is different for lower and higher mileage runners. The adopted patterns of coordinated motion may explain the lower incidence of overuse knee injuries for higher mileage runners. Keywords: running, injury, coordination, principal component analysis (PCA), mileage Running has increased in popularity in the last three decades, but with increased participation has come an increase in reported running injuries.1,2 Injury rates among recreational runners are estimated to be as high as 29.4%, with overuse knee injuries such as anterior knee pain (AKP) and iliotibial band syndrome (ITBS) being the most frequency reported.1,3 Overuse injuries typically result from a combination of training errors, as well as anatomical, footwear, running surface, and biomechanical factors. There is evidence to suggest that running experience or training volume may also be a factor, but there are conflicting views on the relative risks for experienced versus novice athletes1,3,4 The interaction of running experience, weekly mileage, and the mechanics of running is not well understood but may be an important factor for understanding relative injury risk in low versus higher mileage runners. Given the multifactorial nature and multitude of types and locations for lower extremity injuries, it is perhaps not surprising that the relationship between training-related factors and running injury risk is not clear. A recent systematic review of risk factors for running injuries suggests there is conflicting evidence for an association between running experience and overall lower extremity injury risk.5 This review reported that inexperience was related to hamstring and knee injuries, while foot injuries were associated with more experienced runners. There is also conflicting evidence for an association between increases in training distance per week (ie, training errors) and overall lower extremity risk. Again, it seems that the risk for knee injuries is different than for thigh or hamstring injuries. There is evidence that an increase in training distance per week is protective for knee injuries,6 while this same increase in Katherine A. Boyer, Julia Freedman Silvernail, and Joseph Hamill are with the Biomechanics Laboratory, Department of Kinesiology, University of Massachusetts-Amherst, Amherst, Massachusetts. Address author correspondence to Katherine A. Boyer at [email protected].
training distance may result in an increase in the risk for thigh or hamstring injuries.7 Together these studies suggest that there may be different risk factors that interact with training distances for the multitude of lower extremity injuries. Running is a multisegmental motion with many degrees of freedom and thus, there exists a high-dimensional space of potential solutions to attain the same movement goal.8 In the sagittal plane, the coupling or coordination of the segment motions is essential to task completion and thus to position the foot properly for the landing phase of the running stride.9 For example, for a given knee angle at heel-strike for a particular running speed, the hip and ankle angles cannot be arbitrarily selected but will be a function of the selected knee angle to successfully complete the stride. Both running experience or skill level and training-induced differences in the neuromuscular patterns or the mechanical properties of the musculoskeletal tissues could influence the coupling or coordination of segment motion. It has been hypothesized that in normal running a disruption or decoupling of these segments’ motions may be deleterious and related to injury development.10 There is substantial evidence linking the frontal plane hip joint motion and strength with knee injury risk and severity.11–14 Thus, the coordination of the secondary motions (frontal and transverse plane) is expected to be more relevant with respect to the discussion on intersegment movement coordination and injury risk. An analysis of a local coupling of the transverse plane thigh and shank motion at heel-strike in both AKP and healthy subjects found there was a decreased variability in the coupling of these motions in the AKP subjects compared with healthy controls.10 This suggests that coordination of segment motions in the secondary planes is an important consideration for injury risk assessment. An increased understanding of the interrelationship between segment motions in running and the role of mileage on this relationship may provide insight into injury mechanisms in running and the relative risks for different groups. 649
Downloaded by New York University on 09/19/16, Volume 30, Article Number 5
650 Boyer, Freedman Silvernail, and Hamill
A traditional analysis of discrete joint angle values provides little information about the pathway of movement or the relative coordination of limb segment motions used to achieve a movement goal. The application of higher-order statistical analysis methods such as principal component (PC) analysis provides a data-driven methodology to analyze correlated deviations from a mean movement pattern.15 The results of a PC analysis that are applied to a data matrix consisting of all variables and time points that represent the lower extremity movement patterns are features of how the relationships between these variables differ between selected groups. Thus, through this application of the PC analysis, segment couplings and differences between groups in the coupling of threedimensional components of joint motion at all lower extremity joints can be examined in a systematic manner. If running mileage is a risk factor for the development of knee injuries then a difference in the multisegment coordination patterns impacting the frontal plane hip and knee motions may be expected. Therefore, the aim of this study was to test the hypothesis that there are differences in the PC weighting coefficients between the higher and low mileage groups indicative of differences in the frontal and transverse plane joint kinematic waveform patterns and altered multisegment coordination.
Methods Participants Fifty healthy adults (26 males and 24 females) completed five running trials at 3.5 m/s while gait analysis data were collected. Participants were enrolled in the study if they had been free from any lower extremity musculoskeletal injury in the past 12 months and had no prior history of major lower limb reconstructive surgery. Participants were assigned to one of two groups: (1) a low mileage and experience group, who ran on average less than 15 miles per week and had no history of higher mileage running (26.3 ± 9.14 y, 22.9 ± 3.0 kg/m2, 7.2 ± 4.6 miles/wk; 13 males); and (2) a higher mileage group with at least 1 year of experience at mileage of greater than 20 miles per week (29.5 ± 10.9 y; 22.2 ± 2.2 kg/m2, 33.8 ± 14.1 miles/wk; 13 males). The protocol was approved by the university institutional review board and all participants provided written informed consent before the completion of any study procedures. Participants were asked to self-report their weekly mileage, the number of months at this training volume, and any previous running experience. Participants wore standard laboratory footwear that could be considered neutral training shoes.
Experimental Set-up Three-dimensional lower extremity joint kinematics and ground reaction force (GRF) data were captured at 240 Hz with a 9-camera motion analysis system (Qualysis Inc., Sweden) and at 1200 Hz infloor AMTI force platform as subjects ran over a 15-m runway. Each subject performed five trials at running speed of 3.5 m/s ± 5%. Running speed was monitored through two sets of photo cells positioned along the runway on either side of the force platform. To capture the position of the limb segments in the laboratory, reflective markers were positioned on the right leg of each participant. The right leg was arbitrarily selected for analysis in this study. The markers included both calibration markers and tracking markers. Calibration markers were placed on the left and right greater trochanters, the right side anterior superior iliac spine (ASIS), iliac crest, L5-S1 interspace, the medial and lateral femoral epicondyles, the medial and lateral
malleoli, the sides of the first and fifth metatarsal heads, and the tip of the shoe. These markers were used to develop the anatomic segment coordinate systems. Tracking marker clusters of four markers each for the shank and thigh and three markers for the heel counter of the shoe affixed to thermoplastic shells were secured around the thigh, shank, and on the heel counter of the shoe. The anatomical markers on the pelvis also served as tracking markers. With all markers affixed, a standing calibration trial was collected with the participant standing in with their feet positioned shoulderwidth apart and with 0° of abduction and adduction; the knees were fully extended and the hips were neutral with 0° of rotation. Following the calibration trial all anatomical markers, with the exception of those on the pelvis, were removed. The joint center positions were located as the midpoint between the medial and lateral malleoli and condyles for the ankle and knee, respectively. The hip joint center was determined using a previously developed regression equation.16
Data Analysis Qualysis Track Manager software (Qualysis, Inc., Sweden) was used to track the marker positions and to export the data into .C3D format. Raw marker data were further processed using Visual 3D software (C-Motion, Inc., Rockville, MD). A fourth order, zero-lag Butterworth digital low-pass filter was used to process the data with a cutoff frequency of 12 Hz. A right-handed local coordinate system was created for each of the pelvis, thigh, shank, and foot segments. The three-dimensional lower extremity joint angles were calculated with respect to the proximal segment using an Xyz Cardan rotation sequence, and a sequence representing flexion/extension, abduction/ adduction, and axial rotation.17 The beginning and end of the stance phase was determined by the instant the vertical GRF exceeded and fell below 15 N, respectively. There were 26 waveforms of interest for this study: the threedimensional motion of the pelvis, hip, knee, ankle, and the mid- to rearfoot, fore- to midfoot, and hallux to forefoot segments; three components of the GRF; and two center of pressure (COP) components. The PC analysis approach used for this study was based on the methods of Boyer et al, Federolf et al, and Nigg.18–21 The data were prepared before the implementation of the PC analysis as follows. The waveforms were interpolated such that 101 points represented the stance phase of the gait cycle and the mean over five trials for each subject was calculated for each waveform. The gait waveforms of interest have different units and scales and the magnitudes of these inputs to the PC analysis are not proportional to the biological relevance of these motions or forces with respect to injury or performance. Thus, it was necessary to standardize the range of all waveforms before the data analysis, as PC analysis is scale sensitive and the aim of this study was to identify group differences in the shape and relative timing of the gait waveforms in running. For each gait waveform, the mean and standard deviation across the stance phase was determined for all participants. Each input waveform was then scaled by subtracting the mean over the stance phase and then dividing the amplitude by the mean standard deviation across the stance phase.22 This preparation of data normalizes the variance of each waveform before the implementation of the PC analysis, and thus, each waveform regardless of scale will have equal importance in the analysis. The determination of PCi depends on the pattern and timing of the different waveforms throughout movement and not on the overall magnitude of a given waveform. The normalized mean waveforms for each participant were then appended together into one column trial vector with 2626 vector
Downloaded by New York University on 09/19/16, Volume 30, Article Number 5
Role of Running Mileage on Coordination Patterns 651
components (waveform × time points). With this representation of the data, each participant’s data can be visualized as a single point in a high dimensional data space. For analysis, the trial vectors for all participants were assembled into a single matrix and the mean of all trial vectors was subtracted from each column vector. The covariance matrix of the matrix of trial vectors was then computed. An eigen decomposition was then performed to determine the characteristic roots (eigenvalues) and the associated orthogonal eigenvectors or principal components (PCi). The eigenvalues (EVi; expressed as a percentage of the sum of all eigenvalues) indicate the variance (%) in the data set explained by a particular PCi. The eigenvectors or principal component vectors represent the primary directions of variance in the data set. With mean centered data, the eigenvectors maximize the variance from the mean of the data set. These vectors are mutually orthogonal and thus created a new coordinate system (or basis) onto which the original data can be transformed. The matrix of trial vectors was projected onto the PCis for further statistical analysis. Each participant’s trial vector could then be represented as a weighted linear combination of the PCis (Equation 1).
n uuur uuuur t rial = trial m + Âci ◊ PC i , (1) i=1
where trial, is an individual participant’s trial vector, trialm is the mean trial vector of all participants, ci is the score or weight coefficient associated for each of the n, PCi. Characteristic differences between the gait patterns of lower and higher mileage runners were identified by comparing the scores (also called weighting coefficients) ci of the first 20 PCi. Unpaired Student’s t test and effect size of difference (Cohen’s d) were used to test for group differences in the PC scores for each component. A threshold of P < .1 or ES > 0.6 was used to select relevant principal component for further data reduction. A resultant vector D was calculated as the weighted linear combination of the selected PCi (Equation 2). A step-wise procedure was used when calculating the resultant vector. The difference in coefficients and the effect size between the higher and lower mileage groups was determined as higher-order PC were incrementally included in the resultant vector determination. This method ensures there was an increase in the effect size with each additional PCi. This vector combined the characteristic deviations from the mean normalized gait pattern that are associated with group assignment. The EVi were used as the weight factors for the resultant vector calculation and the resultant vector was normalized to unit length. To interpret the PC analysis outcomes and visualize the deviations in the mean movement patterns represented by the resultant vector, a trial vector was recomputed for each subject by adding the resultant vector D weighted by the subjects score ci to the mean normalized waveforms and retracing the normalization steps. The specific joint angle or force variables within the trial vector where there was variation evident between participants in the recomputed vectors are those for which there are significant group differences were interpreted.
D = EVi ◊ PCi + EVi+m ◊ PCi+m + EVi+h ◊ PCi+h + …, (2)
where EVi, are the eigenvalues expressed as a percentage of the total data variance corresponding to the PCi. Only those PCi where significant group differences in the scores were found are included in the calculation of D. Each participant’s trial vector was recomputed from the mean trial vector plus the scaled resultant vector for the interpretation of the primary group-based dif-
ferences. The subsections within these computed vectors where there was substantial or visible variation indicated the associated time point and variables in the trial vector that were different based on the group assignment.
Results The projection of the participants’ trial vectors onto the linear combination of PC7, PC10, PC13, and PC19 was significantly different between the higher and lower mileage groups (d = 0.63, P = .012; Table 1). Together these components explained 4.6% of the variance in the data. The examination of the recomputed trial vectors using the resultant vector D indicated that variation related to group assignment was present in the transverse plane kinematics of the hip and pelvis, the frontal plane hip, and the knee and foot segment kinematics (Figure 1). In the transverse plane the internal-external rotation range of motion and pattern of motion of the hip were different between the low and higher mileage runners. The lower mileage runners continued to externally rotate at the hip in midstance and had a relatively more externally rotated hip through terminal stance. There were no differences in the hip internal-external rotation angle for the first 40% of the stance phase. The lower mileage group also showed less pelvic transverse plane rotation toward stance leg throughout ground contact. There was also a systematic offset in the rearfoot internal-external rotation angle throughout stance with the lower mileage group in relatively greater external rotation. In the frontal plane, group differences in the deviation from the mean were found for the knee and hip adduction-abduction angles and for the rearfoot inversion-eversion angles. For the knee, the lower mileage group was more adducted throughout the stance phase and had greater knee adduction range of motion in early stance. There was a systematic offset in the hip adduction position throughout stance with the lower mileage group in a less adducted position. The lower mileage group also has a smaller hip abductionadduction range of motion, particularly in the second half of the stance phase. The primary difference in rearfoot inversion was found in the final two-thirds of the stance phase. The lower mileage group had a smaller rearfoot inversion angle compared with the higher mileage group. There were no differences in the forefoot inversioneversion angle or motion between groups. There were no nonzero components in the resultant PC vector along the axes related to the sagittal plane kinematics for the pelvis, hip, knee, or ankle joint. Thus, no group differences were identified through this analysis for the sagittal plane kinematics with the exception of the forefoot relative to midfoot flexion. The lower mileage group showed less forefoot to midfoot flexion throughout the stance phase. Table 1 The relative variance explained, effect size (Cohen’s d), and p value for the four PCs identified as significantly different between groups PC
Eigenvalue/% Variance Explained
p Value
Effect Size
PC7
2.68
.04
–0.48
PC10
1.12
.09
–0.38
PC13
0.6
.02
0.60
PC19
0.2
.08
–0.38
Resultant PC
4.6
.01
0.63
Downloaded by New York University on 09/19/16, Volume 30, Article Number 5
652 Boyer, Freedman Silvernail, and Hamill
Figure 1 — The group average deviations in joint angles from the mean movement patterns for the lower (black) and higher (gray) mileage groups. The differences related to group assignment in deviations from the mean gait patterns were interpreted by plotting the portion of the weighted resultant vector plus the mean trial vector for each subject corresponding to each of the 26 variables that make up the trial vectors. The mean waveform for both the lower and higher mileage groups for the following variables shown (knee and hip adduction-abduction, hip internal-external rotation, pelvis transverse plane rotation, rearfoot inversion-eversion, transverse plane rotation, and forefoot relative to midfoot flexion) are those for which there was variance among subjects from the mean waveform for at least 20% of the stance phase. An amplification factor of 20 was used for visualization.
Discussion The purpose of this study was to investigate what, if any, role running mileage may have on multisegment movement coordination and running injury risk factors. The results of this study supported the hypothesis that systematic differences between lower and higher mileage runners can be identified using PC analysis to analyze the interrelationship between lower extremity joint kinematics and external forces. A single resultant PC vector was identified for which the PC scores for the higher and lower mileage runners were significantly different. The calculated resultant PC identifies the direction in high-dimensional space along which there were systematic group differences in the deviation from the mean normalized
movement patterns. This direction represented the transverse plane pelvic rotation, hip internal rotation, hip and knee abduction and adduction, and sagittal and frontal plane multisegment foot motion. The group differences identified through this data-driven method are particularly interesting in light of the epidemiological studies suggesting higher mileage runners may be protected to a degree against knee injuries.5 Biomechanical risk factors identified for the development of overuse knee injuries such as AKP and ITBS include the magnitude or excursion of the transverse and frontal plane joint kinematics, more specifically, increased knee abduction and greater hip internal rotation.11,23,24 It has been proposed that weak hip musculature may lead to greater femoral adduction and internal rotation range of motion that will lead to increased knee
Downloaded by New York University on 09/19/16, Volume 30, Article Number 5
Role of Running Mileage on Coordination Patterns 653
valgus, and as a result, increase the stress on the knee and contribute to overuse knee injuries.11,23,24 In the current study, the higher mileage group showed greater transverse plane pelvic rotation toward the stance leg and less hip internal rotation range of motion, but a slightly greater internal rotation angle in the last half of the stance phase and less knee adduction. The biomechanical data from this study seems at odds with the epidemiological data at first glance, however a more internally rotated hip in combination with greater pelvic rotation toward the stance phase suggests that the coupling between the pelvis and the femur in the transverse plane is different and the difference in the distal segment movement between groups may be minimal. This suggests that an analysis of the joint angles in isolation may provide a limited view of injury risk assuming the higher mileage participants in this study are similar to high mileage runners from the literature. While the proximal segment (hip/pelvis) motion and the motion of the knee joint itself have been linked with knee injury presence11,13,25 and risk, other studies have suggested that distal mechanics (ie, foot motion) are also linked with AKP development.11,13,24,26,27 The resultant PC also varied along the axes related to the deviations from the mean normalized inversion-eversion angle for the rearfoot segments. This indicated that for the higher mileage group the rearfoot was in a slightly more inverted position through the later part of the stance phase. There was also a difference in the transverse plane with the higher mileage runners in a more internally rotated position throughout the stance phase. The deviations in the rearfoot inversion angle, but not the forefoot inversion-eversion, suggest that the coupling of the fore- and rearfoot in the frontal plane is different for the higher and lower mileage groups. In conjunction with the proposed biomechanical risk factor and epidemiological literature, this suggests that the neuromuscular adaptations associated with higher mileage and more experience may result in coordination of the lower limb segments in the transverse and frontal planes that lowers the risk for knee injury development. The resultant vector spanned the axes related to the number of joint motions and phases of stance and thus provides some unique insight into the complexity of the multisegment coordination implicit in the mechanics of running. The PCis describe a linear relationship between the variables and time points along which they vary. Therefore, it is possible to say that these variables at specific times within the stance phase are correlated or associated with one another. The relevant and associated variables to describe the group differences were all in the transverse and frontal planes. This supports a functional relationship between the secondary motions of all lower extremity segments and thus suggests that studies focused on injury development in running must not focus on a single pathway (hip or foot motion), but rather, need to consider motion of the entire lower extremity. In running there needs to be a high degree of spatiotemporal coordination between and within the limb segments. Thus, it is of note that the resultant vector did not vary along the axes associated with deviations from the mean for any of the sagittal plane motions. This suggests that, despite having multiple degrees of freedom within the sagittal plane that may vary, there is not a systematically different way in which lower or higher mileage runners organize these degrees of freedom. The sagittal plane motions may be constrained to a large degree by the task or movement goal, a stride of running at 3.5 m/s, and it does not appear that skill level or mileage has a significant impact on an individual’s ability to perform this motion. The PC analysis identified subtle but systematic differences in the joint kinematics for the higher and lower mileage groups. The
differences between groups in the PCi scores were found primarily in the higher order PCi. . In general, the lower PCi will reflect the dominant movements and those that are sensitive to external influences or to training effects will be higher order PCi representing more subtle movements.28,29 While a PC analysis provides a quantitative method for identifying group differences in both the shape and relative coordination of gait waveforms, the interpretation of these differences remains subjective. The presentation of the data as a mean plus a weighted addition of the resultant PC vector illustrates the direction of the variation in the data as a fraction of the overall variance. We examined the variation between participants in the recomputed vectors to identify where and how the gait waveforms differed between groups. The recomputed trial vectors and the mean of these recomputed trial vectors illustrated the portion of the variance at each time point that we can attribute to mileage differences alone. Additional variance in the data may be present but not attributable to the group assignment (ie, other PCis are directed along these axes). One of the challenges when using a PC analysis is that the outcomes of the analysis, while statistically rigorous, are different from that which we get using a traditional discrete time point analysis. Thus, it is difficult to make direct comparisons between the results of these two analyses. However, the additional information related to the overall pattern and timing of gait waveforms that is gained through the application of methods associated with the dynamical system theory may be important in developing potential prevention strategies based on gait-modifying interventions. The results of this study suggest that there are subtle differences in the way that higher and lower mileage runners coordinate movement of the lower extremity and is a first step in understanding the differences in movement coordination between these two groups. A fundamental aim of this study was to apply the concepts and methods from dynamical systems theory to the study of movement mechanics and the potential for differential risk of common lower extremity overuse running injuries in lower and higher mileage runners. A systematic difference between the two groups was identified and this was primarily related to the transverse and frontal plane kinematics of the pelvic, hip, knee, and rear forefoot. The selection of a sagittal plane movement pattern in the higher mileage group that results in less risky frontal and transverse plane joint kinematic pattern may reduce injury risk and improve the efficiency of movement. The athletes in the higher mileage group self-reported no significant history of overuse injury despite some athletes reaching 70 miles/wk. Given the apparent differences in the lower and higher mileage groups, these results raise the question: Do these lower mileage runners change their mechanics as they increase mileage and gain experience or do they get injured before they become higher mileage runners? Acknowledgment This research was supported by a grant from Brooks Sports, Inc.
Reference List 1. Taunton JE, Ryan MB, Clement DB, McKenzie DC, Lloyd-Smith DR, Zumbo BD. A prospective study of running injuries: the Vancouver Sun Run “In Training” clinics. Br J Sports Med. 2003;37:239–244. PubMed doi:10.1136/bjsm.37.3.239 2. Clement DB, Taunton JE, Smart GW, McNicol KL. A survey of overuse running injuries. Phys Sportsmed. 1981;9:47–58. 3. Macintyre J, Taunton J, Clement D et al. Running injuries: a clinical study of 4,173 cases. Clin J Sports Med. 1991;1:81–87.
Downloaded by New York University on 09/19/16, Volume 30, Article Number 5
654 Boyer, Freedman Silvernail, and Hamill 4. van Mechelen W. Running injuries. A review of the epidemiological literature. Sports Med. 1992;14:320–335. PubMed doi:10.2165/00007256-199214050-00004 5. van Gent RN, Siem D. van MM, van Os AG, Bierma-Zeinstra SM, Koes BW. Incidence and determinants of lower extremity running injuries in long distance runners: a systematic review. Br J Sports Med. 2007;41:469–480. PubMed doi:10.1136/bjsm.2006.033548 6. Satterthwaite P, Norton R, Larmer P, Robinson E. Risk factors for injuries and other health problems sustained in a marathon. Br J Sports Med. 1999;33:22–26. PubMed doi:10.1136/bjsm.33.1.22 7. Wen DY, Puffer JC, Schmalzried TP. Injuries in runners: a prospective study of alignment. Clin J Sport Med. 1998;8:187–194. PubMed doi:10.1097/00042752-199807000-00005 8. Bernstein NA. The co-ordination and regulation of movements. London, UK: Pergamon Press; 1967. 9. Winter DA. Foot trajectory in human gait: a precise and multifactorial motor control task. Phys Ther. 1992;72:45–53. PubMed 10. Hamill J, Van Emmerik RE, Heiderscheit BC, Li L. A dynamical systems approach to lower extremity running injuries. Clin Biomech (Bristol, Avon). 1999;14:297–308. PubMed doi:10.1016/S02680033(98)90092-4 11. Ferber R, Noehren B, Hamill J, Davis IS. Competitive female runners with a history of iliotibial band syndrome demonstrate atypical hip and knee kinematics. J Orthop Sports Phys Ther. 2010;40:52–58. PubMed doi:10.2519/jospt.2010.3028 12. Hamill J, Miller R, Noehren B, Davis I. A prospective study of iliotibial band strain in runners. Clin Biomech (Bristol, Avon). 2008;23:1018– 1025. PubMed doi:10.1016/j.clinbiomech.2008.04.017 13. Noehren B, Hamill J, Davis I. Prospective evidence for a hip etiology in patellofemoral pain. Med Sci Sports Exerc. 2013;45:1120–1124. PubMed doi:10.1249/MSS.0b013e31828249d2 14. Noehren B, Davis I, Hamill J. ASB clinical biomechanics award winner 2006 prospective study of the biomechanical factors associated with iliotibial band syndrome. Clin Biomech (Bristol, Avon). 2007;22:951–956. PubMed doi:10.1016/j.clinbiomech.2007.07.001 15. Daffertshofer A, Lamoth CJ, Meijer OG, Beek PJ. PCA in studying coordination and variability: a tutorial. Clin Biomech (Bristol, Avon). 2004;19:415–428. PubMed doi:10.1016/j.clinbiomech.2004.01.005 16. Bell AL, Pedersen DR, Brand RA. A comparison of the accuracy of several hip center location prediction methods. J Biomech. 1990;23:617–621. PubMed doi:10.1016/0021-9290(90)90054-7 17. Cole GK, Nigg BM, Ronsky JL, Yeadon MR. Application of the joint coordinate system to three-dimensional joint attitude and movement representation: a standardization proposal. J Biomech Eng. 1993;115:344–349. PubMed doi:10.1115/1.2895496
18. Boyer KA, Federolf P, Lin C, Nigg BM, Andriacchi TP. Kinematic adaptations to a variable stiffness shoe: mechanisms for reducing joint loading. J Biomech. 2012;45:1619–1624. PubMed doi:10.1016/j. jbiomech.2012.04.010 19. Federolf PA, Boyer KA, Andriacchi TP. Application of principal component analysis in clinical gait research: Identification of systematic differences between healthy and medial knee-osteoarthritic gait. J Biomech. 2013; 46(13):2173–2178. PubMed 20. Nigg B. Biomechanics of sports shoes. Calgary, AB: Topline Printing Inc.; 2010. 21. Federolf P, Tecante K, Nigg B. A holistic approach to study the temporal variability in gait. J Biomech. 2012;45:1127–1132. PubMed doi:10.1016/j.jbiomech.2012.02.008 22. St-Onge N, Duval N, Yahia L, Feldman AG. Interjoint coordination in lower limbs in patients with a rupture of the anterior cruciate ligament of the knee joint. Knee Surg Sports Traumatol Arthrosc. 2004;12:203–216. PubMed doi:10.1007/s00167-003-0420-5 23. Dierks TA, Manal KT, Hamill J, Davis I. Lower extremity kinematics in runners with patellofemoral pain during a prolonged run. Med Sci Sports Exerc. 2011;43:693–700. PubMed doi:10.1249/ MSS.0b013e3181f744f5 24. Ferber R, Kendall KD, Farr L. Changes in knee biomechanics after a hip-abductor strengthening protocol for runners with patellofemoral pain syndrome. J Athl Train. 2011;46:142–149. PubMed doi:10.4085/1062-6050-46.2.142 25. Powers CM. The influence of abnormal hip mechanics on knee injury: a biomechanical perspective. J Orthop Sports Phys Ther. 2010;40:42–51. PubMed doi:10.2519/jospt.2010.3337 26. Dierks TA, Manal KT, Hamill J, Davis IS. Proximal and distal influences on hip and knee kinematics in runners with patellofemoral pain during a prolonged run. J Orthop Sports Phys Ther. 2008;38:448–456. PubMed doi:10.2519/jospt.2008.2490 27. Rodrigues P, TenBroek T, Hamill J. Runners with anterior knee pain use a greater percentage of their available pronation range of motion. J Appl Biomech. 2013;29:141–146. PubMed 28. Maurer C, Federolf P. von T, V, Stirling L, Nigg BM. Discrimination of gender-, speed-, and shoe-dependent movement patterns in runners using full-body kinematics. Gait Posture. 2012;36:40–45. PubMed doi:10.1016/j.gaitpost.2011.12.023 29. von Tscharner V, Enders H, Maurer C. Subspace identification and classification of healthy human gait. PLoS ONE. 2013;8(7):e65063. PubMed doi:10.1371/journal.pone.0065063
