Validation of force pedals for bilateral strength asymmetry testing

Over 125,000 anterior cruciate ligament reconstruction surgeries are performed each year in the United States. Rates of re-injury to the reconstructed ACL or contralateral ACL up to 49% have been reported in certain populations. Altered biomechanics-including bilateral asymmetry in knee extensor strength-are known predictors of secondary injury. Indeed, ACL reconstruction patients have been reported to produce peak knee extensor moments 20% less in the operated limb as compared to the non-operated limb. Despite the prevalence of bilateral lower limb asymmetries, many physical therapists lack diagnostic capabilities due to the high costs of motion capture systems and force- measuring systems such as force plates or force pedals. The purpose of this study was to access the validity and reliability of a newly released, cost effective commercial force-measuring pedal. Pedals were first tested under static conditions to identify accuracy of force measurements. Regression analyses indicated that measured and applied forces were essentially equal with a coefficient of determination of > 0.999. Dynamic power output testing as compared to a validated power meter showed an average error of 1.19%. However, when used for biomechanical analysis the data displayed a variable delay of approximately two samples. The pedal was not found to be reliable for diagnostic use at this time, but further testing is underway to identify possible solutions to the sample delay.

VALIDATION OF FORCE PEDALS FOR BILATERAL STRENGTH ASYMMETRY TESTING by David Bennion A Senior Honors Thesis Submitted to the Faculty of The University of Utah In Partial Fulfillment of the Requirements for the Honors Degree in Bachelor of Science In Exercise and Sport Science Approved: Jim Martin, PhD Thesis Faculty Supervisor Janet Shaw, PhD Chair, Department of Exercise and Sport Science Kerry Jacques Honors Faculty Advisor Sylvia D. Torti, PhD Dean, Honors College December 2014 Copyright © 2014 All Rights Reserved ABSTRACT Over 125,000 anterior cruciate ligament reconstruction surgeries are performed each year in the United States. Rates of re-injury to the reconstructed ACL or contralateral ACL up to 49% have been reported in certain populations. Altered biomechanics—including bilateral asymmetry in knee extensor strength—are known predictors of secondary injury. Indeed, ACL reconstruction patients have been reported to produce peak knee extensor moments 20% less in the operated limb as compared to the non-operated limb. Despite the prevalence of bilateral lower limb asymmetries, many physical therapists lack diagnostic capabilities due to the high costs of motion capture systems and forcemeasuring systems such as force plates or force pedals. The purpose of this study was to access the validity and reliability of a newly released, cost effective commercial forcemeasuring pedal. Pedals were first tested under static conditions to identify accuracy of force measurements. Regression analyses indicated that measured and applied forces were essentially equal with a coefficient of determination of > 0.999. Dynamic power output testing as compared to a validated power meter showed an average error of 1.19%. However, when used for biomechanical analysis the data displayed a variable delay of approximately two samples. The pedal was not found to be reliable for diagnostic use at this time, but further testing is underway to identify possible solutions to the sample delay. ii TABLE OF CONTENTS ABSTRACT 11 INTRODUCTION 1 METHODS 5 RESULTS 9 DISCUSSION 12 REFERENCES 14 lll INTRODUCTION Over 125,000 anterior cruciate ligament (ACL) reconstruction surgeries are performed each year in the United States (Kim, Bosque, Meehan, Jamali, & Marder 2011). Rates of either re-injury to the reconstructed ACL or rupture of the contralateral ACL range from 3%-49% (Barber-Westin & Noyes, 2011). Injuries to the contralateral knee are twice as likely as an ipsilateral graft rupture in general populations and rates increase for athletic populations participating in higher intensity activities (Wright, Magnussen, Dunn, & Spindler, 2011; Sward, Kostogiannis, Roos, 2010). Risk factors related to re-injury can be separated into two categories: modifiable and non-modifiable (Hewett, Myer, Ford, Paterno, & Quatman, 2012; Di Stasi, Myer, & Hewett, 2013). Non-modifiable risk factors include age of patient, anatomical factors, genetics, and graft type. Individuals under the age of 21 at the time of reconstruction showed significantly greater risk for contralateral ACL injury than those over the age of 21 (Pinczewski et al., 2007). Anatomical factors such as intercondylar notch stenosis (narrow femoral condyle notch) have also been reported as predisposing factors for contralateral injury (Sward et al., 2010). While research in the area is limited, September, Schwellnus, and Collins (2007) suggested support for a genetic component in ACL injury based on a literature review and more recent case-control studies have identified specific genotypes associated with ACL injury (Posthumus et al., 2009; Sward et al., 2010). Posthumus et al. (2009) reported as a secondary finding of their study that the number of participants suffering an ACL rupture who reported a blood relative with a history of any ligament injury was significantly higher than a matching control group without any self­ reported history of tendon or ligament injury, supporting the theory of genetic predisposition. Pinczewski et al. (2007) also found significantly greater risk of contralateral knee injury in patients who underwent ACL reconstruction using a patellar tendon autograft as compared to a hamstring tendon autograft. Modifiable risk factors include returning to higher activity levels, neuromuscular deficits, and altered biomechanics. Individuals returning to activities requiring cutting, pivoting, and jumping are more likely to suffer either an ipsilateral or contralateral ACL injury (Sward et al., 2010; Di Stasi et al., 2013). Examples of common neuromuscular deficits predictive of re-injury are impaired neuromuscular control and deficits in postural stability (Di Stasi et al., 2013). Altered biomechanics such as asymmetric loading of the limbs during landing, high degrees of dynamic valgus motion in landing, and altered lower limb motion are also predictors of ACL injury (Sward et al., 2010; Hewett et al., 2012). Altered biomechanics resulting from strength deficiencies in the quadriceps and biceps femoris in the ACL graft limb are also reported following ACL reconstruction surgeries (Lewek, Rudolph, Axe, Snyder-Mackler, 2002; Hiemstra, Webber, MacDonald, Kriellaars, 2007). These alterations are believed to contribute to the higher incidence rates of ACL injury in the contralateral knee (Salmon, Russell, Musgrove, Pinczewski, Refshauge, 2005). Strength deficits of 25% have been demonstrated when compared to uninjured populations from the same demographic (Hiemstra, Webber, MacDonald, Kriellaars, 2000). Patients less than one year post-surgery showed trends towards greater hip extensor moments in the ipsilateral limb when compared to the contralateral limb (p=0.06) and significant differences between knee extensor moments (p=0.003) in the 2 ipsilateral and contralateral knees during a simple squat exercise. The peak knee extensor moment was 25.5% less in the ACL reconstructed knee than the contralateral knee, while peak ground reaction forces did not differ between limbs (Salem, Salinas, Harding, 2003). These results show that bilateral strength asymmetries can be masked by compensatory strategies during the multiple-joint functionality exercises preferred in some rehabilitation programs allowing the patient to complete therapy before returning to full strength. Another contributing factor to the high re-injury rate is the lack of objective return to athletics criteria among physical therapists and orthopedic surgeons. In a systematic review of 264 studies performed by Westin-Barber and Noyes (2011) to identify return to unrestricted athletics criteria following ACL reconstruction, 40% of studies did not provide return to athletics criteria, 32% reported postoperative time as the only criterion, and 15% reported postoperative time and subjective criteria. Only 13% reported objective criteria such as thigh circumference, single-leg hop test, and Lachman rating. Hiemstra et al. (2007) recommended the modification of initial rehabilitation goals to include the “achievement of bilateral strength normalization of both the knee extensors and the knee flexors” (p. 548) followed by additional strength gains to normalize knee extensors and flexors of both the graft limb and contralateral limb to strength levels consistent with a non-injured population. During submaximal cycling the neuromuscular system is free to select from the various joint actions in order to meet the overall task requirement. Consequently, submaximal cycling represents a multiple degree of freedom, mutli-joint task, which could be used to identify compensatory strategies. Biomechanics during cycling can be 3 assessed from measures of pedal forces and limb kinematics. The use of cycling biomechanics would allow physical therapists and orthopedic surgeons to identify compensatory patterns and strength deficiencies in patients. Historically, force-measuring pedals for biomechanical analyses were custom built and thus not available to most clinicians. Recently, the Garmin Vector force-sensing pedal has become commercially available which may allow for widespread use of biomechanical analysis for evaluation of rehabilitation. Before use in a diagnostic setting, however, it is important to establish the accuracy and consistency of the data provided by these force pedals. The purpose of this study is to establish the reliability and validity of the Garmin Vector force pedals for use in biomechanical assessment. The following static and dynamic evaluations will be performed: static force measurement comparisons, dynamic power output comparisons, force and acceleration time synchronization assessments, and biomechanical analysis comparisons. 4 METHODS The force sensors in these pedals are embedded in the pedal axle and the software used to obtain data from these pedals includes an algorithm to determine the orientation of the axle with respect to the crank. This algorithm requires smooth rotation of the pedals upon initial setup and calibration of the system. Consequently, these pedals can only be used to collect meaningful validation data when mounted on a functional rotating crank system. A mounting system was built to which the crank could be attached to collect static data. After the cranks and pedals were mounted (either to the mounting system for static data collection or the isometric ergometer for dynamic data collection) the system underwent a two-step calibration process that was completed using the Garmin Edge 1000 head unit: first, the pedals are smoothly rotated at a constant rate between 80-90 revolutions per minute to set the pedal angles and second, the cranks are placed horizontal while the system zeros to account for the weight of the pedal bodies on the sensors. Data recorded by the force sensors and accelerometers is saved to pods attached to each pedal and can be retrieved using the Vector Research Tool (VRT) which provides time, force, and acceleration measurements individually for each pedal. Alternatively, the Edge 1000 head unit can be used to record and retrieve overall power output but not individual pedal measurements. Static Force Measurements Static evaluations were performed by loading the pedals with known masses of 5, 10, 25, 45, and 100 pounds. After loading the pedal the crank was set to one of four 5 angles: 0°, 90°, 180°, 270° and the mass was adjusted to ensure it did not contact any surrounding surfaces (Figure 1). Pedal forces were recorded in both the radial and tangential directions. The recorded radial and tangential force values were then compared to the actual applied radial and tangential force values calculated using the known mass and angle measures. Three trials for each mass and angle combination were performed across multiple days and the system was re-calibrated between trials to ensure reliable day-to-day and setup-to-setup data. FIGURE 1—Static Testing Configuration Dynamic Power Output Measurement Comparison Dynamic power output was recorded during 30-second trials while targeting a constant power output at a fixed pedaling rate on an isokinetic ergometer. Multiple trials were recorded across different days and various power outputs were targeted. Power output was recorded simultaneously using both the Garmin Edge 1000 head unit and a validated SRM power meter. The average power output for each trial was calculated and the averages were then compared. Force and Acceleration Time Sync Assessment Force and acceleration data from matching time points are needed to calculate the crank angle and pedal position used during the biomechanical analysis to determine joint 6 force contributions. Interleaved sampling causes the pedal to alternate between sampling force and acceleration data. This causes a misalignment between data points and reduces the effective sampling rate. If the data is interleaved, the missing data points must be interpolated before calculating the crank angle and pedal position. To identify if the data was interleaved a “slide test” was performed during which the crank was locked at a 90° angle on the mounting system and the pedal was place on a level, low friction surface where it was slid forwards and backwards to create changes in force and acceleration data directions. Twenty trials lasting approximately 10-15 seconds were performed and the system was re-calibrated after completing the first ten trials. Biomechanical Analysis Comparison Final diagnostic capabilities were accessed through a biomechanical analysis comparison. Participants performed two 30-second trials targeting a constant, submaximal power output at a fixed pedaling rate on an isometric ergometer. One trial was performed using the Garmin Vector pedals while the other was performed using a validated, custom-built force measuring pedal utilizing Kistler piezoelectric force transducers. The position of the right iliac crest was recorded using an instrumented spatial linkage system described by Martin, Elmer, Horscroft, Brown, and Schultz (2007) and static measurements for kinematic foot length (pedal spindle to lateral malleolus), foot length (heel to toe), leg length (lateral femoral condyle to lateral malleolus), and thigh length (greater trochanter to lateral femoral condyle) were recorded for each trial. From the pedal data, right iliac crest position data, and static measurements, joint positions can be determined. When using the Garmin pedal, the recorded pedal data and instrumented spatial linkage system data must be synced (Kistler force pedal data is 7 collected in conjunction with the instrumented spatial linkage system data). The data is synced using the pedal position calculated from the provided Garmin data and pedal position recorded by the instrumented spatial linkage system. After the two data sources are synced the joint positions are calculated. Once joint positions are established, linear and angular velocities for each joint can be determined through finite differentiation of position data with respect to time. Joint-specific power contributions and percent total power were calculated as average values for each pedal cycle during a three second time period representative of the trial. Average joint power contributions and percent total power from the two trials were then compared. 8 RESULTS Static Force Measurements Accuracy of the static measurements was accessed through linear regression analysis. The force measurements for each of the three trials of a given mass and pedal angle combination were averaged. Radial forces are positive when the crank is compressed (crank angles of 270°-0°-90°) and negative when extended (90°-180°-270°). Positive and negative tangential forces between the left and right pedals occur at inverse crank angles. The right pedal tangential forces are positive when the pedal is forward (0°90°-180°) while the left is positive when back (180°-270°-0°). Negative tangential forces are recorded in the opposite crank angle ranges. A regression analysis was then performed using the average measured values and the values calculated using the known mass and pedal angle. Coefficient of determination values for each regression analysis were greater than 0.999 (Figure 2). FIGURE2—Left and Right Pedal Radial and Tangential Forces. Dynamic Power Output Measurement Comparison The error between the average force value recorded by the Garmin pedals and SRM power meter for each 30-second trial are presented in Table 1. Trials 1-4 were performed on a single day while trials 5-8 were performed on a separate day after resetting the system and completing the calibration process. The average percent error for all trials was 1.19%. A linear regression showed overall power output also does not account for changes in percent error (R2=0.045). TABLE 1—Percent error between average Garmin power and average SRMpower inWatts. Trial Average Garmin Power (W) Average SRMPower (W) 1 92.03 91.71 2 97.88 95.18 3 99.55 98.78 4 106.55 104.86 5 149.49 148.08 6 162.22 161.12 7 274.10 269.41 8 406.25 404.07 Percent Error 0.34% 2.84% 0.78% 1.60% 0.95% 0.68% 1.74% 0.54% Force and Acceleration Time Sync Assessment Force and acceleration data recorded during the slide test were first graphed and visually inspected for alignment of changes in direction (Figure 3). The data was also inspected manually and it was found that the changes in direction for both force and acceleration occurred at simultaneous time points, showing the data was not interleaved. f --------- -S F — rafr" is ------“V FIGURE3—Force and Acceleration Datavs. Time during “Slide Test.” 10 Biomechanical Analysis Comparison Data generated for biomechanical analysis including percent joint force contributions were compared for each participant’s two trials. The Garmin Vector data showed significant differences from the reference data collected using the Kistler pedal. The Garmin data showed a sample delay of approximately two to three samples when time synched with the instrumented spatial linkage system. This delay resulted in the differences seen in percent joint force contributions. The exact time offset required to correct this sample delay varied from trial to trial and a constant offset could not be determined. 11 DISCUSSION The static testing confirmed the pedals recorded accurate and reliable force measurements for both radial and tangential forces. While slight differences were seen between the measured values and calculated values, these can be expected due to the margin of accuracy for the pedal and limitations surrounding the static testing configuration. If allowed to rest in any position for a short period of time, the pedals enter a sleep mode and must be woken up by spinning the pedals. During some trials, slight movement in the hanging masses occurred which could not be completely avoided due to the need to quickly set the pedals at the specified angle and record the data sample. The dynamic power output testing also provided satisfactory results. The Garmin Vector pedals consistently provided average power measurements comparable to the SRM power meter. The highest error in any trial was 2.84% and considered acceptable, while the average error was 1.19%. This further validates the accuracy of the force measurements and as well as the use of the Garmin Vector pedals for other data collection purposes not involving biomechanical analysis. The force and acceleration time sync assessment provides valuable data by showing that force and acceleration data are sampled simultaneously. Changes in direction for force and acceleration occurred at synchronized time points. Lack of a delay in these directional changes shows there is no offset created by the data sampling. This signifies that interpolated values are not needed before the data can be used in crank angle and pedal position calculations. These results also signify that a force and 12 acceleration offset is not responsible for the sampling delay found during the biomechanical analysis. While other tests proved the Garmin pedal to be a viable tool for performing biomechanical assessments, the final comparison shows the pedal cannot be used until the source of the time offset is identified. The data could be corrected by adding an additional time offset when performing the time sync for the pedal and instrumented spatial linkage system data; however, as this additional offset was not constant it can only be determined when reference data is available. In addition to the differences between the Garmin and reference data, the Garmin data also differed from the expected values for joint contributions based on previous research. Subsequent trials were performed using an unweighted cycling method shown to produce high hip extension power and lower kneed extension power during the push phase of the pedal stroke (Rimer, Marshall, Wehmanen, Farley, & Martin, 2012). Data collected during these unweighted trials using the Garmin pedal showed little hip extension power and high knee extension. When the data was adjusted by approximately two samples, the joint contributions become similar to those identified in previous research for unweighted cycling. Due to the aforementioned problems, we do not recommend the Garmin Vector pedal for use in diagnostic settings at this time. Research to identify the source of the sampling delay is still underway with the intent of validating the pedals for future diagnostic use. 13 REFERENCES Barber-Westin, S.D., & Noyes, F.R. (2011). Factors used to determine return to unrestricted sport activities after anterior cruciate ligament reconstruction. Journal o f Arthroscopic and Related Surgery, 27(12), 1697-1705. Di Stassi, S., Myer, G.D., Hewett, T.E. (2014). Neuromuscular training to target deficits associated with second anterior cruciate ligament injury. Journal o f Orthopaedic & Sports Physical Therapy, 43, 777-792. Grassi, A., Zaffagnini, S., Marcheggiani, G., Neri, M. P., Della Villa, S., Marcacci, M. (2015). After revision anterior cruciate ligament reconstruction, who returns to sport? A systematic review and meta-analysis. British Journal o f Sports Medicine, 49(20),1295304. Hewett, T. E., Myer, G. D., Ford, K. R., Paterno, M. V., Quatman, C. E. (2012). The 2012 ABJS Nicolas Andry Award: The sequence of prevention: A systematic approach to prevent anterior cruciate ligament injury. Clinical Orthopaedics and Related Research, 470, 2930-2940. Hiemstra, L.A., Webber, S., MacDonald, P.B., Kriellaars, D.J. (2000). Knee strength deficits after hamstring tendon and patellar tendon anterior cruciate ligament reconstruction. Medicine & Science in Sports & Exercise, 32, 1472-1479. Hiemstra, L.A., Webber, S., MacDonald, P.B., Kriellaars, D.J. (2007). Contralateral limb strength deficits after anterior cruciate ligament reconstruction using a hamstring tendon graft. Clinical Biomechanics, 22, 543-550. Kim, S., Bosque, J., Meehan, J.P., Jamali, A., & Marder, R. (2011). Increase in outpatient knee arthroscopy in the United States: A comparison of national surveys of ambulatory surgery, 1996 and 2006. Journal o f Joint and Bone Surgery, 93, 994-1000. Lewek, M., Rudolph, K., Axe, M., Snyder-Mackler, L. (2002). The effect of insufficient quadriceps strength on gait after anterior cruciate ligament reconstruction. Clinical Biomechanics, 17, 56-63. Martin, J.C., Elmer, S.J., Horscroft, R.D., Brown, N.A., Schultz, B.B. (2007). A low-cost instrumented spatial linkage accurately determines ASIS position during cycle ergometry. Journal o f Applied Biomechanics, 23, 224-229. 14 Pinczewski, L.A., Lyman, J., Salmon, L.J., Russel, V.J., Roe, J., & Linklater, J. (2007). A 10-year comparison of anterior cruciate ligament reconstructions with hamstring tendon and patellar tendon autograft: a controlled, prospective trial. The American Journal o f Sports Medicine, 35, 564-574. Posthumus, M., September, A.V., Keegan, M., O’Cuinneagain, D., Van der Merwe, W., Schwellnus, M.P., Collins, M. (2009). Genetic risk factors for anterior cruciate ligament ruptures: COL1A1 gene variant. British Journal o f Sports Medicine, 43, 352-356. Rimer, E.G., Marshall, C.S., Wehmanen, K.W., Farley, J., & Martin, J.C., (2012). A simple instruction increases relative contribution of hip extension action during concentric cycling. Supplement to Medicine & Science in Sports & Exercise, 44 (5): S309, ACSM 59th Annual Meeting, San Francisco; May, 2013. Salem, G.J., Salinas, R., Harding, F.V. (2003). Bilateral kinematic and kinetic analysis of the squat exercise after anterior cruciate ligament reconstruction. Archives o f Physical Medicine and Rehabilitation, 84, 1211-1216 Salmon, L., Russell, V., Musgrove, T., Pinczewski, L., Refshauge, K. (2005). Incidence and risk factors for graft rupture and contralateral rupture after anterior cruciate ligament reconstruction. Journal o f Arthroscopic and Related Surgery, 21(8), 948-957. September, A.V., Schwellnus, M.P., Collins, M. (2007). Tendon and ligament injuries: the genetic component. British Journal o f Sports Medicine, 41, 241-246. Sward, P., Kostogiannis, I., Roos, H. (2010). Risk factors for a contralateral anterior cruciate ligament injury. Knee Surgery, Sports Traumatology, Arthroscopy, 18, 277-291. Wright, R.W., Magnussen, R.A., Dunn, W.R., & Spindler, K.P. (2011). Ipsilateral graft and contralateral ACL rupture at five years or more following ACL reconstruction. Journal o f Bone and Joint Surgery, 93, 1159-1165. 15