Human Movement Science

Transcription

1 Human Movement Science 31 (2012) Contents lists available at SciVerse ScienceDirect Human Movement Science journal homepage: Coordination variability during load carriage walking: Can it contribute to low back pain? Sheng-Che Yen a,b,, Gregory M. Gutierrez b, Wen Ling b, Richard Magill b, Andrew McDonough b,c a Sensory Motor Performance Program, Rehabilitation Institute of Chicago, Chicago, IL 60611, United States b Department of Physical Therapy, New York University, NY 10010, United States c Graduate Programs in Health Sciences, Seton Hall University, NJ 07079, United States article info abstract Article history: Available online 12 June 2012 PsycINFO classifications: 2330 Keywords: Continuous relative phase Gait Mechanical stress Backpack Load carriage walking is frequently associated with low back pain. Mechanical stress is a potential cause of such pain, and a lack of coordination variability may produce mechanical stress. We tested the hypothesis that coordination variability would decrease during load carriage walking. We examined the trunk-thigh coordination variability in the sagittal and frontal planes and the thorax-pelvis coordination variability in the transverse plane. Ten healthy participants were recruited to perform unloaded and load carriage walking. Coordination variability was quantified as the standard deviation of continuous relative phase between two segments across a number of walking trials. During load carriage walking, the coordination variability significantly increased rather than decreased in the sagittal and transverse planes, and it did not change significantly in the frontal plane compared to those during unloaded walking. The findings rejected the hypothesis and suggested that reduced coordination variability may not predict the development of low back pain associate with load carriage walking in healthy people. Ó 2012 Elsevier B.V. All rights reserved. Corresponding author at: Sensory Motor Performance Program (SMPP), Rehabilitation Institute of Chicago, Suite 1406, 345 E. Superior Street, Chicago, IL 60611, United States. Tel.: ; fax: address: syen@ric.org (S.-C. Yen) /$ - see front matter Ó 2012 Elsevier B.V. All rights reserved.

2 S.-C. Yen et al. / Human Movement Science 31 (2012) Introduction Soldiers, students, and recreational hikers usually transport heavy items using a backpack. However, walking with carrying a heavy backpack load (load carriage walking) has been identified as a risk of low back pain (Heuscher, Gilkey, Peel, & Kennedy, 2010; Knapik, Reynolds, & Harman, 2004; Limon, Valinsky, & Ben-Shalom, 2004). Mechanical stress is a potential cause of low back pain associated with load carriage walking (Knapik, Harman, & Reynolds, 1996). For example, individuals may adapt to load carriage walking in manners which increase mechanical stress on the low back, such as increased trunk stiffness (Holt, Wagenaar, LaFiandra, Kubo, & Obusek, 2003) and increased trunk forward inclination (Hong & Cheung, 2003; Ling, Houston, Tsai, Chui, & Kirk, 2004; Martin & Nelson, 1986; Singh & Koh, 2009). Also, a heavy backpack load may not move in synchrony with the trunk, which is another source of mechanical stress (Pierrynowski, Winter, & Norman, 1981). Furthermore, load carriage walking increases the vertical ground reaction force that can be transmitted up the kinetic chain and place additional stress on the low back (Birrell, Hooper, & Haslam, 2007). Another potential factor to increase mechanical stress is a decrease in coordination variability (Hamill, van Emmerik, Heiderscheit, & Li, 1999; Pollard, Heiderscheit, van Emmerik, & Hamill, 2005). Coordination here refers to the temporal-spatial coupling between adjacent body segments, and coordination variability reflects how consistent the coupling pattern is reproduced over time. Low coordination variability indicates that a coupling pattern is constantly repeated, which may create an overuse situation (Hamill et al., 1999; Pollard et al., 2005). Previous studies have shown that people with patellar femoral pain syndrome tend to have lower coordination variability between the thigh and shank as compared to healthy controls (Hamill et al., 1999; Heiderscheit, Hamill, & van Emmerik, 2002). Hamill et al. (1999) suggested that people may limit coordination variability as a strategy to reduce pain, and the decreased coordination variability may produce constant stress on the musculoskeletal structures to further the injury. Grounded upon the above notion, a study was conducted to determine if low coordination variability may predict stress injury (Pollard et al., 2005). Specifically, the researchers were interested in understanding whether lower coordination variability between the leg segments is a reason why women tend to have a higher incidence of having noncontact anterior cruciate ligament (ACL) injury than do men when performing sports like basketball and soccer. Their findings showed that female subjects demonstrated lower coordination variability between the thigh and shank when performing an unanticipated cutting maneuver compared to those of male subjects. While the study was not a prospective study and did not follow up whether more female than male subjects eventually acquire ACL injury, the findings provided initial evidence to support that low coordination variability may contribute to stress related injury. In this current study, we examined whether a decrease in coordination variability is a factor that contributes to the development of low back pain associated with load carriage walking. As the low back serves as a linkage system between the trunk and thigh, we examined the coordination variability between these two segments in the sagittal and frontal planes. We also examined the coordination variability between the thorax and pelvis in the transverse plane as previous studies have reported that load carriage walking affects the coupling pattern between these segments (LaFiandra, Holt, Wagenaar, & Obusek, 2002; LaFiandra, Wagenaar, Holt, & Obusek, 2003). We hypothesized that the trunk-thigh coordination variability in the sagittal and frontal planes and the thorax-pelvis coordination variability in the transverse plane would decrease during load carriage walking. 2. Methods To test the hypothesis, we asked participants to perform overground walking in three conditions: (a) carrying no load (unloaded condition); (b) carrying a backpack containing 10% of the body weight (BW); (c) carrying a backpack containing 20% of the BW. The study design has been described in detail elsewhere (Yen, Ling, Magill, McDonough, & Gutierrez, 2011).

3 1288 S.-C. Yen et al. / Human Movement Science 31 (2012) Participants Ten participants, consisting of 5 males and 5 females, were recruited from the community of New York University (NYU). All participants were young and healthy without neuromuscular or musculoskeletal problems affecting their walking ability. Their mean age was 30.1 ± 3.7 years, mean height was ± 12.5 cm, and mean weight was 67.7 ± 14.2 kg. All participants reported that they do not walk carrying a heavy backpack loads (more than 10% of their BW) on a regular basis. All participants signed the consent form approved by the University Committee on Activities Involving Human Participants of NYU Instrumentation The Modular Light Weight Load Carrying Equipment (MOLLE) was used in this study. A block of foam and a number of sandbags were put into the main rucksack of the MOLLE to form the required loads. The MOLLE was placed on participants upper back and was firmly secured to the body using shoulder straps and a hip belt. A 5-camera Qualisys motion analysis system (Qualisys AB, Göthenborg, Sweden) was used to capture participants walking at a sampling rate of 120 Hz. GAITRite system (CIR Systems, Inc., Clifton, NJ) was used to monitor participants walking speed at a sampling rate of 80 Hz Data collection procedures Data collection was conducted indoors in a laboratory setting. Before data collection, a total of 20 reflective makers were attached to each participant to create a 3-degree-of-freedom, 14-segment model (Hatze, 1980) (Fig. 1A). Markers were placed on the participants forehead and chin, and bilaterally on the acromion process, the lateral elbow joint, the lateral radial styloid process, the mid-dorsum of the hand, the greater trochanter, the lateral femoral epicondyle, the lateral malleolus, the head of the fifth metatarsal bone, and calcaneous. During data collection, participants were asked to walk at a speed of 4.8 km/hr or 3 miles/hr (±10%) without carrying a load and carrying 10% and 20% of the BW using the MOLLE. Each condition included several practice trials and 12 data collection trials. Practice trials were set to help participants become familiarized with the load and walking speed. In each trial, participants were instructed to walk naturally from a starting line, over the GAITRite walkway, and then return to the starting line three times continuously. Data were collected without participants awareness in the second or third pass. The GAITRite software (v3.4) automatically calculated each participant s walking speed. Trials at above or below 10% of the target speed were dropped and repeated until 12 trials were successfully recorded. For safety, the order of conditions was fixed from unloaded, 10%, to 20% of the BW, allowing participants to experience the load in a gradual manner to minimize the risk of back and leg injuries. Due to a high number of test trials (3 conditions 12 trials), a 2-minute break was provided between trials and a 10-minute break was provided between conditions to reduce fatigue Data processing procedures The camera setup allows the motion analysis system to capture all 20 markers in the middle of the walkway for 6-8 steps in each trial. From these steps, we randomly selected 4 consecutive steps (a left cycle and a right cycle) for data analysis. Marker data were smoothed using a 4th-order Butterworth filter at a cut-off frequency of 6 Hz. Based on the smoothed marker data, we measured the positions of trunk, thigh, thorax, and pelvis. The trunk was defined as the segment between the acromion process and the greater trochanter, and the thigh was defined as the segment between the greater trochanter and the lateral femoral epicondyle. The positions of the trunk and thigh in the sagittal plane (Fig. 1B) and in the frontal plane (Fig. 1C) were measured as the angle of intersection between the corresponding segment and a vertical line passing through the greater trochanter. The thorax was represented by the segment between the left and right acromion processes, and the pelvis was defined as the segment between the left and

4 S.-C. Yen et al. / Human Movement Science 31 (2012) Fig. 1. (A) Marker setup. (B) Illustration of the trunk position (a) and thigh position (b) in the sagittal plane. (C) Illustration of the trunk position (a) and thigh position (b) in the frontal plane. (D) Illustration of the thorax position (a) and pelvis position (b) in the transverse plane. right greater trochanters. The positions of the thorax and pelvis in the transverse plane were measured as the angle of intersection between the corresponding segment and a horizontal line passing through the midpoint of that segment (Fig. 1D). Since a 3 degree-of-freedom model was used, the sagittal, frontal, and transverse planes coincide with the lab (global) coordinate system, rather than a local coordinate system within each segment. The coordination pattern between two segments was measured by the continuous relative phase (CRP) (Hamill, Haddad, & McDermott, 2000). To calculate the CRP, we first differentiated the position data of each segment against time to obtain the velocity data. Second, both position and velocity data in a walking cycle were linearly normalized to be between 1 (maximum value) and 1 (minimum value). Next, a phase plane plotting normalized velocity (y-axis) against normalized position (x-axis) over a walking cycle was constructed for each segment. Then, the phase angle of each segment at each time frame was calculated using the following equation: UðtÞ ¼arctan vðtþ sðtþ ð1þ where v(t) is the normalized velocity at given time t, and s(t) is the normalized position at the same time t. The CRP is the difference in phase angle between the two segments tested, which can range from 0 to 360. However, since there is a redundancy in angles (e.g., 0 and 360 are the same), we presented the data from 0 to 180 as suggested by Hamill et al. (2000). The raw CRP data were a series of CRP

5 1290 S.-C. Yen et al. / Human Movement Science 31 (2012) points over each walking cycle (CRP trajectory). The CRP points were first averaged for each of the stance and swing phases, and then averaged between the left and right cycles for each trial. The resultant variable was called the trial CRP index. Coordination variability was measured as the standard deviation of the 12 trial CRP index values for each phase in each condition. The resultant variable was called the CRP variability index. We examined coordination variability during both stance and swing phases of gait. To identify the stance and swing phases, we first determined the timing of heel contact and toe-off using the vertical velocity profile of the foot center (O Connor, Thorpe, O Malley, & Vaughan, 2007). The foot center in the present study was defined as the midpoint between the markers on calcaneous and on the fifth metatarsal head (Fig. 1A). The timing of heel contact was defined as the frame in which the foot center had the lowest vertical velocity among all possible heel contact frames. The timing of toe-off was defined as the frame in which the foot center had the highest vertical velocity during a walking cycle. Visual examination was done to verify the outcome of auto-detection. Stance phase was defined as the period from heel contact to toe-off of the corresponding leg, and swing phase was defined as the period from toe-off to heel contact of the corresponding leg (Perry, 1992). All aforementioned data processing procedures were done using customized programs written in LabView 8.5 Language (National Instruments Corp, Austin, TX) Statistical analysis A linear mixed model (LMM) for repeated measures was conducted at the a level of.05 to examine whether the CRP variability index varied as a function of load, phase, and the interaction between load and phase. A first-order autoregressive (AR(1)) model was used to structure the correlation between repeated observations, assuming that events (conditions by phases) occurring closer to each other in time during data collection should have a stronger correlation (Singer & Willett, 2003). The statistical analysis was conducted using Statistical Package for Social Sciences (SPSS) version 16.0 (SPSS, Chicago, IL). 3. Results 3.1. Coordination variability increased in the sagittal plane The coordination variability between the trunk and thigh increased rather than decreased in the sagittal plane during load carriage walking. We first examined the coordination variability based on the CRP trajectory data (Fig. 2A C). Graphically, the variability was larger in both loaded conditions than in the unloaded condition throughout a walking cycle. We also noticed that the increase in the variability was greater during the swing phase than the stance phase in all conditions. Analyses of the CRP variability index showed similar findings (Fig. 2D). The LMM detected a significant interaction effect between the load and phase on the CRP variability index, F(2, 42.04) = 8.41, p =.001. The CRP variability index tended to increase during load carriage walking, and the increase was larger during the swing phase than the stance phase. The LMM also detected a significant fixed effect of load, F(2, 43.58) = 6.96, p =.002. Based on parameter estimates, the CRP variability indices recorded in the 10% and 20% of the BW backpack load conditions was and 7.69 greater than that recorded in the unloaded condition, respectively, controlling for the interaction effect and the fixed effect of phase. Additionally, the LMM detected a significant fixed effect of phase, F(1, 29.68) = , p < Based on parameter estimates, the CRP variability index recorded during the swing phase was greater than that recorded during the stance phase, controlling for the interaction effect and the fixed effect of load. When comparing each participant s CRP variability index during the swing phase between the unloaded and loaded conditions, we found that 8 participants increased the variability in both loaded conditions, 1 decreased the variability in both loaded conditions, and 1 decreased the variability in the 10% of the BW loaded condition but increased the variability in the 20% of the BW loaded condition. When comparing each participant s CRP variability index during the stance phase between the

6 S.-C. Yen et al. / Human Movement Science 31 (2012) Fig. 2. The group s coordination pattern (solid curve) and coordination variability (the vertical distance between the solid curve and either of the dashed curves) in the sagittal plane in the unloaded condition (A), 10% of the BW loaded condition (B), and 20% of the BW loaded condition (C). The group s coordination pattern resulted from taking the mean of the CRP trajectories between one left and one right cycles in a trial, then across 12 trials in a condition, and finally across all participants. The group s coordination variability resulted from averaging one left and one right CRP trajectories in a trial first, followed by taking the standard deviation of the left-and-right averaged CRP trajectories across the 12 trials in a test condition, and finally taking the mean of the standard deviations across all participants. Before taking the mean and standard deviation, each CRP trajectory recorded in a cycle was normalized to 200 points. (D) The group s mean and standard deviation of the CRP variability index in the sagittal plane during stance and swing phases in each condition. unloaded and loaded conditions, we found that 4 participants increased the variability in both loaded conditions, 5 decreased the variability in both loaded conditions, and 1 decreased the variability in the 10% of the BW loaded condition but increased the variability in the 20% of the BW loaded condition. To understand whether the increase in coordination variability was driven by a particular segment, we examined each segment s phase plane data. The data from a typical participant (case 3) are shown in Fig. 3. The trunk (Fig. 3A) and thigh (Fig. 3B) were moving on the phase plane in the clockwise direction during a walking cycle. The trunk showed more variability during the swing phase than the stance phase in all conditions. Carrying a load increased trunk variability, and the increase was more apparent in the swing phase than in the stance phase. In contrast, the thigh moved on the phase plane in a consistent manner, and load carriage did not affect the consistency. The phase plane data suggest that the increase in the coordination variability between the trunk and thigh in the sagittal plane may be mainly driven by the trunk Coordination variability increased in the transverse plane The coordination variability between the thorax and pelvis in the transverse plane increased slightly during both stance and swing phases in both loaded conditions compared to that of the unloaded condition (Fig. 4A C). The LMM on the CRP variability index (Fig. 4D) detected a significant fixed effect of load, F(2,46.88) = 4.48, p =.02. Based on parameter estimates, the CRP variability indices recorded in the 10% and 20% of the BW backpack load conditions was 2.95 and 4.07 greater than that recorded in the unloaded condition, respectively, controlling for the interaction effect and the fixed

7 1292 S.-C. Yen et al. / Human Movement Science 31 (2012) Fig. 3. The phase plane data for the trunk (A) and thigh (B) in the sagittal plane from a typical participant (case 3). The x-axis represents the normalized position while the y-axis represents the normalized velocity. Each curve represents a walking trial in a condition. The cross sign marks where the swing phase typically begins. The trunk and thigh move on the phase plane in the clockwise direction. effect of phase. The LMM also detected a significant fixed effect of phase, F(1, 42.32) = 5.7, p =.02. Based on parameter estimates, the CRP variability index recorded during the swing phase was 0.19 greater than that recorded during the stance phase, controlling for the interaction effect and the fixed effect of load. However, the LMM detected no significant interaction effect between the load and phase

8 S.-C. Yen et al. / Human Movement Science 31 (2012) Fig. 4. The group s coordination pattern and coordination variability in the transverse plane in the unloaded condition (A), 10% of the BW loaded condition (B), and 20% of the BW loaded condition (C). The figures were constructed in the same way as Fig. 2A C. (D) The group s mean and standard deviation of the CRP variability index in the transverse plane during stance and swing phases in each condition. on the CRP variability index, F(2, 42.22) = 5.7, p =.37, suggesting that the increase in the variability associated with load carriage was similar between the stance and swing phases. When comparing each participant s CRP variability index during the stance phase between the unloaded and loaded conditions, we found that 7 participants increased the variability in both loaded conditions, none decreased the variability in both loaded conditions, and 3 decreased the variability in the 10% of the BW loaded condition but increased the variability in the 20% of the BW loaded condition. When comparing each participant s CRP variability index during the swing phase between the unloaded and loaded conditions, we found that 5 participants increased the variability in both loaded conditions, 2 decreased the variability in both loaded conditions, 2 decreased the variability in the 10% of the BW loaded condition but increased the variability in the 20% of the BW loaded condition, and 1 increased the variability in the 10% of the BW loaded condition but decreased the variability in the 20% of the BW loaded condition. Case 3 s phase plane data in the transverse plane are shown in Fig. 5A (thorax) and Fig. 5B (pelvis). Based on the figures, the variability of both segments increased slightly during both stance and swing phases in load carriage walking. Compared with the thorax, the pelvis seems to move more consistently on the phase plane in load carriage walking. This trend was fairly consistent across participants Coordination variability did not change significantly in the frontal plane The coordination variability between the trunk and thigh in the frontal plane did not have any apparent change during load carriage walking (Fig. 6A C). When examining the CRP variability index (Fig. 6D) using the LLM, we found that the fixed effect of load (F(2,41.87) = 0.76, p =.45), the fixed ef-

9 1294 S.-C. Yen et al. / Human Movement Science 31 (2012) Fig. 5. The phase plane data for the thorax (A) and pelvis (B) in the transverse plane from case 3. Both segments move on the phase plane in the clockwise direction. The figure configuration is the same as Fig. 3. fect of phase (F(1,21.07) = 0.42, p =.58), and the interaction effect between the load and phase (F(2, 41.7) = 0.28, p =.65) were all not significant. When comparing each participant s CRP variability index during the stance phase between the unloaded and loaded conditions, we found that 3 participants increased the variability in both loaded conditions, 2 decreased the variability in both loaded conditions, 4 decreased the variability in the

10 S.-C. Yen et al. / Human Movement Science 31 (2012) Fig. 6. The group s coordination and coordination variability in the frontal plane in the unloaded condition (A), 10% of the BW loaded condition (B), and 20% of the BW loaded condition (C). The figures were constructed in the same way as Fig. 2A C. (D) The group s mean and standard deviation of the CRP variability index in the frontal plane during stance and swing phases in each condition. 10% of the BW loaded condition but increased the variability in the 20% of the BW loaded condition, and 1 increased the variability in the 10% of the BW loaded condition but decreased the variability in the 20% of the BW loaded condition. When comparing each participant s CRP variability index during the swing phase between the unloaded and loaded conditions, we found that 2 participants increased the variability in both loaded conditions, 1 decreased the variability in both loaded conditions, 4 decreased the variability in the 10% of the BW loaded condition but increased the variability in the 20% of the BW loaded condition, and 3 increased the variability in the 10% of the BW loaded condition but decreased the variability in the 20% of the BW loaded condition. The trunk and thigh, in the frontal plane, did not respond consistently to the load on the phase plane across participants. A relatively clear trend is that the thigh moved on the phase plane in a more consistent manner than did the trunk in all conditions. Such a trend can be observed in Fig. 7A (trunk) and 7B (thigh), which shows case 3 s phase plane data in the frontal plane Practice effect During data collection, subjects underwent 12 trials in each test condition. Thus, coordination variability was likely to be affected by practice. To examine potential practice effects, we grouped the 12 trials into 2 blocks in each condition: (a) the early block (containing the first 6 trials) and (b) the late block (containing the last 6 trials). We then calculated the CRP variability index within each block, and compared the results between the early and late blocks using LMM with AR(1) structure, controlling for the test condition. The findings suggested that there may be a practice effect. Specifically, there was a significant fixed effect of block on the thorax-pelvis coordination variability during the swing phase, F(1, 27.96) = 8.19, p =.008. The thorax-pelvis coordination variability during the swing phase was smaller in the late block than in the early block in each test condition (Fig. 8A). The LMM also showed that the interaction

11 1296 S.-C. Yen et al. / Human Movement Science 31 (2012) Fig. 7. The phase plane data for the trunk (A) and thigh (B) in the frontal plane from case 3. Both segments move on the phase plane in the clockwise direction. The figure configuration is the same as Fig. 3. effect between the block and test condition was not significant, F(2, 41.75) = 0.35, p =.71, suggesting that the amount of decrease in coordination variability was similar across the conditions. We also found a significant interaction effect between the block and test condition (F(2, 42.36) = 6.54, p =.003) and a significant fixed effect of block (F(1, 29.78) = 7.77, p =.009) on the trunk-thigh coordination variability in the frontal plane during swing phase. The variability was smaller in the late block than the early block in the 20% of the BW backpack load condition, while it did not

12 S.-C. Yen et al. / Human Movement Science 31 (2012) Fig. 8. Thorax-pelvis coordination variability during swing phase (A) and trunk-thigh coordination variability during swing phase in the frontal plane (B) in the early and late blocks in each condition. have apparent difference between the early and late blocks in the unloaded and 10% of the BW backpack load conditions (Fig. 8B). We did not find significant block effects on the trunk-thigh coordination variability in the sagittal plane during both stance and swing phases, on the trunk-thigh coordination variability in the frontal plane during stance phase, and on the pelvis-thorax coordination variability during stance phase Speed and step frequency In this study, participants were required to walk at a speed of 4.8 km/hr (±10%) in each condition. The GAITRite data indicated that participants average walking speed was 4.83 ± 0.16 km/hr in the unloaded condition, 4.8 ± 0.12 km/hr in the 10% of the BW loaded condition, and 4.81 ± 0.1 km/hr in the 20% of the BW loaded condition. Based on the standard deviation, the between-participant variation in speed was modest in all conditions. Additionally, the between-trial variation in walking speed, as measured by trial-to-trial standard deviation, was also modest. On average, it was 0.17 ± 0.04 km/hr

13 1298 S.-C. Yen et al. / Human Movement Science 31 (2012) in the unloaded condition, 0.13 ± 0.04 km/hr in the 10% of the BW loaded condition, and 0.14 ± 0.04 km/hr in the 20% of the BW loaded condition. Although walking speed was experimentally controlled, step frequency could still vary across load conditions, as one could adjust step length rather than step frequency to achieve the same speed. Since step frequency is associated with the period of leg oscillation and can affect the CRP, it is important to examine whether step frequency could be a potential confounder affecting the relationship between the load and the coordination variability. The GAITRite data indicated that participants average step frequency was reasonably consistent across all conditions. Specifically, step frequency was ± 9.02 steps/min in the unloaded condition, ± 9.89 steps/min in the 10% of the BW backpack load condition, and ± steps/min in the 20% of the BW backpack load condition. We then ran a repeated measures LMM with AR(1) structure to determine whether step frequency varied as a function of load and trial. The results suggested that there was no significant fixed effect of load (F(2, ) = 0.29, p =.75), no significant fixed effect of trial (F(1, ) = 0.59, p =.44), and no significant interaction effect between the load and trial (F(2,343.75) = 0.08, p =.92). Since step frequency did not vary as a function of the load, this variable should not confound the relationship between the load and coordination variability. 4. Discussion Mechanical stress is a potential cause of low back pain associated with load carriage walking (Knapik et al., 2004). A lack of coordination variability has been thought to increase mechanical stress (Hamill et al., 1999; Pollard et al., 2005). We examined whether load carriage walking was associated with a decrease in coordination variability. We found that when participants walked while carrying 10% and 20% of the BW in a backpack, their trunk-thigh coordination variability increased in the sagittal plane, the thorax-pelvis coordination variability increased in the transverse plane, and the trunk-thigh coordination variability did not change significantly in the frontal plane compared with those during unloaded walking. Because coordination variability tended to increase rather than decrease in load carriage walking, our data suggests that reduced coordination variability may not explain how healthy people develop low back pain in load carriage walking Why coordination variability increased in load carriage walking? Coordination variability may increase when the system is perturbed. For example, coordination variability between leg joints increased when people were forced to walk at a speed suitable for running (i.e., the fast speed perturbed the system) (Diedrich & Warren, 1995). The variability may also increase when people perform a task that they are inexperienced in. For instance, participants performing a triple jump tended to have higher coordination variability when the skill had not been well developed (Wilson, Simpson, van Emmerik, & Hamill, 2008). In addition, an increase in coordination variability may result from an inconsistent motor planning (before the movement actually begins). Recent evidence based on an animal model has shown that movement variability is correlated with neural variability during the preparatory stage of the movement (Churchland, Afshar, & Shenoy, 2006). In the present study, the backpack load represents a walking perturbation, as it substantially changes the carriers biomechanical and physiological performance (Knapik et al., 2004). In addition, the participants recruited in this study reported that they were not experienced in heavy (P10% of the BW) load carriage walking. They may have limited ability to formulate motor plans or to produce motor outputs in a consistent manner. In fact, we found evidence to support that practice decreased coordination variability among participants in this study (Fig. 8). Thus, backpack load as a perturbation and participants inexperience with load carriage walking may explain why we observed increased coordination variability in the sagittal and transverse planes during load carriage walking. The increased coordination variability may alternatively be explained from a functional perspective (van Emmerik & van Wegen, 2002). Specifically, variability may offer flexibility for the system to adapt to a perturbation. During load carriage walking, the system has been perturbed by the load. The system may need more flexibility in preparation for additional perturbation (e.g., an unexpected slippery

14 S.-C. Yen et al. / Human Movement Science 31 (2012) floor) and therefore the variability was increased. However, it should be noted that such functional variability is expected to only occur in those who are skilled at the motor task (Wilson et al., 2008). More specifically, both unskilled and skilled people could show an increase in variability during a motor task, but the meaning behind the increase is different for the two groups of people. Unskilled people increase the variability as a result of exploring the motor possibilities through trial-and-error. On the other hand, skilled people have mastered the task, and the increase in variability demonstrates their ability to adapt the motor strategy when there is a task or environmental change. Given that participants recruited in this study were inexperienced with load carriage walking, the alternative explanation (functional variability) is less likely to be the actual cause of the increased variability Increase of coordination variability: is it problematic? We have argued that the coordination variability increased during load carriage walking because participants were not experienced in the task and the backpack load was essentially a walking perturbation. Following this line of thinking, the increased variability may represent error - the system was not able to coordinate the trunk and thigh as well as the thorax and pelvis in a consistent manner due to the additional load. This interpretation is in accordance with the traditional view on variability. Traditionally, variability was thought of as error in motor output, which plays no functional role in a goaldirected movement (Davids, Bennett, & Newell, 2006). Such error has been shown to have some harmful effects on gait. For example, a number of studies have concluded that the variability in temporal and spatial gait parameters (e.g., step length and step time) increases with aging and is a risk factor of falling (Gabell & Nayak, 1984; Hausdorff, Rios, & Edelberg, 2001; Montero-Odasso et al., 2011; Nakamura, Meguro, & Sasaki, 1996). However, we are not aware of any study examining whether increased coordination variability has any harmful effect on the musculoskeletal system. Further study is needed to fill this knowledge gap. In particular, it is essential to determine the optimal range of coordination variability for the low back during load carriage walking. The optimal range of coordination variability may be determined by examining the coordination variability in skilled load carriage walkers who have never suffered low back pain due to the task. Then, a retrospective or prospective study may shed light on whether people whose coordination variability deviates from the optimal range have a higher chance to acquire low back pain An issue with interpreting the trunk variability We found that the trunk was the main component driving the increase in coordination variability in the sagittal plane during load carriage walking (Fig. 3). In this study, the trunk was defined as the segment between the acromion process and the greater trochanter, and thus the trunk movement and the pelvis/hip movement cannot be disassociated. However, since the trunk position reflects the relative position between the acromion process and greater trochanter in the anterior-posterior direction (Fig. 1B), examining the movement of greater trochanter in the anterior-posterior direction may help us understand how much of the trunk variability was coming from the pelvis/hip. The data from the transverse plane provides such information. In this study, the pelvis position actually reflects the position of the greater trochanter in the anterior-posterior direction in relation to a reference horizontal line (Fig. 1D). Based on Fig. 5B, the pelvis moved in a fairly consistent manner during both stance and swing phases in unloaded walking, and load carriage did not drastically change the consistency (Fig. 5B). This suggests that the pelvis/hip may not have a large contribution to the increased trunk variability observed in the sagittal plane in load carriage walking Discrete versus continues measures of coordination variability A major difference between the current study and previous studies on load carriage walking (LaFiandra et al., 2002, 2003) was the walking context. Specifically, the current study examined load carriage walking overground, while the previous studies examined load carriage walking on a treadmill. The advantage of examining load carriage walking overground is that the test condition can better

15 1300 S.-C. Yen et al. / Human Movement Science 31 (2012) reflect the real-life situation; studies have shown that overground walking and treadmill walking are different in many ways (Lee & Hidler, 2008; Warabi, Kato, Kiriyama, Yoshida, & Kobayashi, 2005). The disadvantage, however, is that we were not able to collect continuous walking cycles over time, especially in our limited lab space. As a result, we calculated coordination variability across a number of discrete trials rather than over multiple continuous cycles. Measuring coordination variability based on discrete trials may yield different results from measuring coordination based on continuous cycles, which needs to be clarified in future studies Limitations of the study In this study, we examined the low back from a macroscopic perspective (i.e., as the region between the thorax and thigh). However, low back pain associated with load carriage walking may occur locally within the spine. The coordination variability observed may not reflect the variability between vertebra. This is different from previous work which has examined the relationship between knee injuries and coordination variability around the knee (Hamill et al., 1999; Heiderscheit et al., 2002; Pollard et al., 2005). Those authors measured the variability between the thigh and the lower leg, which are the two segments that articulate at the knee joint. We investigated coordination variability at the behavioral level. The effect of load carriage walking on coordination variability at the muscle level (variability of muscle synergy) was not examined. Investigating this issue may provide further insights into the relationship between coordination variability and stress injury. Additionally, we tested young, healthy adults who were not experienced in load carriage walking in this study. The finding of increased coordination variability in load carriage walking may not be generalizable to other populations (e.g., people already have low back pain or skilled load carriage walker). Also, we controlled for participants walking speed to minimize its potential confounding effect on the variability, as walking speed has been shown to affect coordination variability (van Emmerik & Wagenaar, 1996). Participants may not have felt comfortable walking at the experimental speed, which may affect the results. Furthermore, we intended to minimize the effect of fatigue by giving multiple breaks to participants during data collection. However, fatigue may affect coordination variability and may be associated with stress injury. Lastly, the order of the conditions was not randomized for safety reason, and thus a potential order effect could affect the results of this study. 5. Conclusion We tested the hypothesis that coordination variability would decrease during load carriage walking, which may cause mechanical stress and lead to low back pain. The findings led us to reject the hypothesis and suggested that reduced coordination variability may not predict the development of low back pain associated with load carriage walking in healthy people. Examining whether the increased variability has any harmful effect on the musculoskeletal system may be a new research direction for the development of low back pain associated with load carriage walking. Acknowledgments The authors would like to thank all participants for their time and effort and three anonymous reviewers for their invaluable comments. References Birrell, S.A., Hooper, R.H., Haslam, R.A., The effect of military load carriage on ground reaction forces. Gait & Posture 26, Churchland, M.M., Afshar, A., Shenoy, K.V., A central source of movement variability. Neuron 52, Davids, K., Bennett, S., Newell, K.M., Movement system variability. Human Kinetics, Champaign, IL. Diedrich, F.J., Warren Jr., W.H., Why change gaits? Dynamics of the walk-run transition. Journal of Experimental Psychology: Human Perception & Performance 21, Gabell, A., Nayak, U.S., The effect of age on variability in gait. Journal of Gerontology 39,

16 S.-C. Yen et al. / Human Movement Science 31 (2012) Hamill, J., Haddad, J.M., McDermott, W.J., Issues in quantifying variability from a dynamical systems perspective. Journal of Applied Biomechanics 16, Hamill, J., van Emmerik, R.E., Heiderscheit, B.C., Li, L., A dynamical systems approach to lower extremity running injuries. Clinical Biomechanics 14, Hatze, H., A mathematical model for the computational determination of parameter values of anthropomorphic segments. Journal of Biomechanics 13, Hausdorff, J.M., Rios, D.A., Edelberg, H.K., Gait variability and fall risk in community-living older adults: a 1-year prospective study. Archives of Physical Medicine & Rehabilitation 82, Heiderscheit, B.C., Hamill, J., van Emmerik, R.E.A., Variability of stride characteristics and joint coordination among individuals with unilateral patellofemoral pain. Journal of Applied Biomechanics 18, Heuscher, Z., Gilkey, D.P., Peel, J.L., Kennedy, C.A., The association of self-reported backpack use and backpack weight with low back pain among college students. Journal of Manipulative and Physiological Therapeutics 33, Holt, K.G., Wagenaar, R.C., LaFiandra, M.E., Kubo, M., Obusek, J.P., Increased musculoskeletal stiffness during load carriage at increasing walking speeds maintains constant vertical excursion of the body center of mass. Journal of Biomechanics 36, Hong, Y., Cheung, C.K., Gait and posture responses to backpack load during level walking in children. Gait & Posture 17, Knapik, J., Harman, E., Reynolds, K., Load carriage using packs: A review of physiological, biomechanical and medical aspects. Applied Ergonomics 27, Knapik, J.J., Reynolds, K.L., Harman, E., Soldier load carriage: Historical, physiological, biomechanical, and medical aspects. Military Medicine 169, LaFiandra, M., Holt, K.G., Wagenaar, R.C., Obusek, J.P., Transverse plane kinetics during treadmill walking with and without a load. Clinical Biomechanics 17, LaFiandra, M., Wagenaar, R.C., Holt, K.G., Obusek, J.P., How do load carriage and walking speed influence trunk coordination and stride parameters? Journal of Biomechanics 36, Lee, S.J., Hidler, J., Biomechanics of overground vs. treadmill walking in healthy individuals. Journal of Applied Physiology 104, Limon, S., Valinsky, L.J., Ben-Shalom, Y., Children at risk: Risk factors for low back pain in the elementary school environment. Spine 29, Ling, W., Houston, V., Tsai, Y.S., Chui, K., Kirk, J., Women s load carriage performance using modular lightweight loadcarrying equipment. Military Medicine 169, Martin, P.E., Nelson, R.C., The effect of carried loads on the walking patterns of men and women. Ergonomics 29 (10), Montero-Odasso, M., Muir, S.W., Hall, M., Doherty, T.J., Kloseck, M., Beauchet, O., et al, Gait variability is associated with frailty in community-dwelling older adults. The Journals of Gerontology 66, Nakamura, T., Meguro, K., Sasaki, H., Relationship between falls and stride length variability in senile dementia of the alzheimer type. Gerontology 42, O Connor, C.M., Thorpe, S.K., O Malley, M.J., Vaughan, C.L., Automatic detection of gait events using kinematic data. Gait & Posture 25, Perry, J., Gait analysis: Normal and pathological function. Slack, Thorofare, NJ. Pierrynowski, M.R., Winter, D.A., Norman, R.W., Metabolic measures to ascertain the optimal load to be carried by man. Ergonomics 24, Pollard, C.D., Heiderscheit, B.C., van Emmerik, R.E., Hamill, J., Gender differences in lower extremity coupling variability during an unanticipated cutting maneuver. Journal of Applied Biomechanics 21, Singer, J.D., Willett, J.B., Applied longitudinal data analysis: Modeling change and event occurrence. Oxford University Press, Oxford; New York. Singh, T., Koh, M., Effects of backpack load position on spatiotemporal parameters and trunk forward lean. Gait & Posture 29, van Emmerik, R.E.A., van Wegen, E.E., On the functional aspects of variability in postural control. Exercise and Sport Sciences Reviews 30, van Emmerik, R.E.A., Wagenaar, R.C., Effects of walking velocity on relative phase dynamics in the trunk in human walking. Journal of Biomechanics 29, Warabi, T., Kato, M., Kiriyama, K., Yoshida, T., Kobayashi, N., Treadmill walking and overground walking of human subjects compared by recording sole-floor reaction force. Neuroscience Research 53, Wilson, C., Simpson, S.E., van Emmerik, R.E.A., Hamill, J., Coordination variability and skill development in expert triple jumpers. Sports Biomechanics 7, 2 9. Yen, S.C., Ling, W., Magill, R., McDonough, A., Gutierrez, G.M., Temporal relationship between trunk and thigh contributes to balance control in load carriage walking. Gait & Posture 34,