Test-Retest Reliability of Discrete Gait Parameters in Children With Cerebral Palsy

ORIGINAL ARTICLE Test-Retest Reliability of Discrete Gait Parameters in Children With Cerebral Palsy Susan Klejman, PEng, Jan Andrysek, PhD, PEng, Annie Dupuis, PhD, Virginia Wright, BSc(PT), MSc, PhD 781 ABSTRACT. Klejman S, Andrysek J, Dupuis A, Wright V. Test-retest reliability of discrete gait parameters in children with cerebral palsy. Arch Phys Med Rehabil 2010;91:781-7. Objectives: To examine the test-retest reliability of discrete gait parameters in children with cerebral palsy (CP) in Gross Motor Function Classification System (GMFCS) levels I, II, and III; to calculate the measurement error between testing sessions of these parameters in the total sample and within GMFCS subgroups using the standard error of measurement; and to evaluate the minimal detectable change (MDC) to identify discrete gait parameters that are most sensitive to change in children with CP. Design: Test-retest reliability study. Setting: Rehabilitation facility with human movement laboratory. Participants: Ambulatory children with CP (N 28). Interventions: Not applicable. Main Outcome Measures: Intraclass correlation coefficients (ICCs), standard error of measurement, and MDC of discrete gait parameters. Results: Parameters measured in the sagittal plane and temporal-spatial parameters were highly reliable across all GMFCS levels (ICC range,.84.97), while test-retest reliability in the frontal and transverse planes varied from poor to excellent (ICC range,.46.91). Using MDC as a guide, hip and pelvis parameters in the transverse and frontal planes were least responsive for GMFCS levels I and III (MDC ranges, 8.3 18.0 and 2.7 23.4, respectively), whereas ankle kinematics were the least responsive for level II (MDC range, 8.2 11.9 ). Reliability was dependent on mobility level, with children in GMFCS level III exhibiting greater test-retest variability overall. Conclusions: Our findings suggest that select discrete gait parameters measured using computerized gait analysis are reliable and potentially responsive measures of performance and can be used as outcome measures in intervention studies. Key Words: Cerebral palsy; Rehabilitation. 2010 by the American Congress of Rehabilitation Medicine COMPUTERIZED GAIT ANALYSIS is commonly used in the assessment of gait deviations in children with CP. 1-4 However, observed changes in CGA measurements may be more attributable to variability associated with the measure than to actual functional change. The 3 primary sources of measurement error that contribute to the day-to-day variability with observational measures such as CGA are variations in performance of the child, measurement error of the instrumentation, and measurement inconsistencies of the examiner administering the test. 5 Prior to use in clinical evaluation, it is essential to understand all sources of variability inherent to CGA to determine whether the data collected are representative of the person s gait pattern and whether the chosen parameters are consistent enough between testing sessions to allow meaningful clinical decision-making and to evaluate clinical change over time. CGA produces large quantities of information, and to simplify the analysis and facilitate interpretation, discrete gait parameters such as maximum knee flexion and hip ROM are typically extracted from the continuous kinematic waveforms. 6,7 Previous work has shown that the intrasession variability of these discrete gait parameters increases inversely with function in children with CP as measured by the GMFCS. 6 Sampling a number of stride repetitions was recommended from this work; specifically, a minimum of 4 strides was specified for children in GMFCS level I and a minimum of 6 strides for children in GMFCS levels II and III. No studies to date document the day-to-day repeatability of discrete gait parameters measured using CGA in children with CP. Instead, previous reliability work in this population has evaluated the measurement of the underlying kinematic waveforms. 8,9 Findings suggested that kinematic variables can be reliably measured using CGA; however, intrasession reliability was generally higher than intersession (test-retest) reliability. 8,9 The intersession repeatability of discrete gait parameters is critical information for outcomes work. Estimates of measurement error facilitate accurate sample size calculations for outcome studies and provide guidance for the choice of gait parameters that are sensitive to change. Previous work investigating the reliability of discrete gait parameters has focused on able-bodied adults. 10,11 Typical results included mean difference between visits of select parameters and the corresponding SDs of the difference scores, which have been informative in identifying repeatable gait parameters in able-bodied adults. In general, reliability was higher for ankle and knee parameters than parameters measured at the hip and pelvis. 10,11 In this context, the objectives of the current study were to (1) examine the test-retest reliability of discrete gait parameters List of Abbreviations From the Bloorview Research Institute, Bloorview Kids Rehab, Toronto, ON, Canada. Supported by the Bloorview Children s Hospital Foundation. No commercial party having a direct financial interest in the results of the research supporting this article has or will confer a benefit on the authors or on any organization with which the authors are associated. Reprint requests to Susan Klejman, PEng, Bloorview Kids Rehab, 150 Kilgour Rd, Toronto, ON, M4G 1R8, Canada, e-mail: sue.redekop@utoronto.ca. 0003-9993/10/9105-00934$36.00/0 doi:10.1016/j.apmr.2010.01.016 CGA CI CP GMFCS ICC MDC MSE ROM SEM computerized gait analysis confidence interval cerebral palsy Gross Motor Function Classification System intraclass correlation coefficient minimal detectable change mean square error range of motion standard error of measurement

782 RELIABILITY OF GAIT PARAMETERS IN CEREBRAL PALSY, Klejman obtained through CGA in children with CP in GMFCS levels I, II, and III; (2) calculate the measurement error between testing sessions of these parameters in the total sample and within GMFCS subgroups using the SEM; and (3) evaluate the MDC to identify discrete gait parameters that are most sensitive to change in children with CP. METHODS A longitudinal single group study was conducted with a baseline and retest session. The retest interval was 1 to 2 weeks and a maximum retest period of 3 weeks was allowed. Actual changes in gait function would not be expected in this period in the absence of intervention. Participants A convenience sample of 28 children with CP was recruited for the study from physiotherapy outpatient caseloads. Parents who had children who met the basic eligibility criteria were given study information letters by their physiotherapist. Those who expressed an interested in learning more about the study then had their names passed to the research assistant on the research team, and the informed consent process followed from there according to the study s Research Ethics Board-approved protocol. The children s characteristics are summarized in table 1. Ten children were in GMFCS level I (9 boys, 1 girl), 10 were in GMFCS level II (4 boys, 6 girls), and 8 were in GMFCS level III (4 boys, 4 girls). Most participants used lower-limb orthoses regularly, but the gait trials were collected in the barefoot walking condition to optimize marker placement and viewing of joint angles. Seven of the children in GMFCS level III walked with a walker, and 1 used bilateral quad canes. Subjects were excluded if they had received botulinum toxin A injections in the lower limbs within the last 3 months or any orthopedic surgery or neurosurgery in the last 6 months. Prior to participation, all procedures were explained to the child, and informed written consent was obtained from the parent or guardian as approved by the research ethics board at our facility. Procedure Details of the data collection procedure are provided elsewhere 6 but are briefly described here. Participants wore spandex shorts and a tank top during data collection to minimize marker movement artifact. Reflective markers were placed on anatomic landmarks on the pelvis and bilaterally on the thighs, shanks, and feet. The markers were applied on all subjects by the same evaluator, who had 5 years of experience evaluating the gait of children with CP. Participants were instructed to walk along a walkway at their self-selected comfortable walking speed. Each child took 2 practice walks prior to the commencement of data collection. A minimum of 3 passes was collected for each subject. Data were collected using a Vicon MX motion a capture system sampling at 120Hz. Reflective markers were manually identified using a Vicon Workstation. a Data were processed to determine spatiotemporal and kinematic parameters using Bodybuilder software. a The same protocol was repeated at retest. Data Analysis Twenty-nine discrete gait parameters were selected for testretest analysis from those commonly used in clinical outcome studies done on this population. 6,12-14 Custom-written Matlab b programs were used to extract relevant features from the kinematic curves. To ensure that a stable intrasession measure was obtained for each discrete gait parameter, 6 repeated measures of each gait parameter were averaged for each subject for each of the 2 visits. 6 Parameters were extracted from 2 consecutive midwalk strides of gait for each of the 3 gait passes. The impact of using this combination of intrapass and interpass strides was investigated in a previous study 6 and did not confound the results. Data were analyzed unilaterally. For children with hemiplegia, data were analyzed for the affected side, whereas the side of the body was arbitrarily chosen for the children with spastic diplegia. Statistical Analysis The between-session reliability was calculated using the ICC measuring agreement (ICC 2,1 ) 15 and associated 95% CI. For the purpose of analysis, mean ICC values of.80 and above reflected excellent reliability, those between.70 and.79 indicated good reliability, and those below.70 reflected poor to moderate reliability. 5 Analyses were completed with all participants and within GMFCS levels I, II, and III. The SEM calculates the total measurement error across repeated measures resulting from performance differences in the child as well as instrument and assessor variability. The SEM was calculated using the MSE from 2-way analysis of variance 16-19 where SEM MSE. This formulation of the SEM has been recommended 17 rather than using calculations based on the ICC statistic because it is unaffected by the range of measurement values (ie, extent of variation in the sample). The MDC was also calculated to estimate the minimal amount of change that is needed to exceed measurement error. The MDC was calculated by multiplying the SEM by 2 and the z score associated with the desired level of confidence. 16,20 An MDC confidence level of 90% was chosen for the current study because this level is often used in clinical outcome studies in children with disabilities. 21-23 Bland-Altman plots were constructed for each parameter to estimate measurement bias. Plots were constructed by plotting the mean difference between visits against the mean of the 2 visits. 24 Plots were examined for the magnitude of the difference between visits and the distribution around the 0 line. The Table 1: Participants Characteristics GMFCS Level Diagnosis Age (y)* Height (cm) Mass (kg) I(n 10) 1 with spastic triplegia 6.6 2.9 122.7 13.9 26.0 8.9 2 with spastic diplegia 7 with hemiplegia (6 left, 1 right) II (n 10) 9 with spastic diplegia 8.1 2.1 126.7 14.7 31.8 11.9 1 with hemiplegia (1 left) III (n 8) 8 with spastic diplegia 7.3 3.0 115.9 12.1 27.0 6.2 NOTE. Values are mean SD. *The ages between the 2 groups did not differ significantly (P.439).

RELIABILITY OF GAIT PARAMETERS IN CEREBRAL PALSY, Klejman 783 mean difference between visits and the SD of the difference scores were used to estimate the limits of agreement. The limits of agreement were used to estimate the magnitude of disagreement between baseline and retest and were calculated by [d] 1.96SD d, where d is the difference score between baseline and retest and SD d is the SD of the difference scores. It is expected that 95% of the differences between baseline and retest for any individual will lie between these limits. Analysis was completed using SAS Version 9.1 for Windows. c RESULTS ICC values and their corresponding 95% CIs are shown in table 2. ICC test-retest estimates were excellent (ICC range,.84.97) for the sample as a whole (ie, across all GMFCS levels) for sagittal plane, transverse plane, and spatio-temporal parameters. With the exception of ankle dorsiflexion at initial contact for children in GMFCS level II, good to high levels of reliability were observed in the sagittal plane and for spatiotemporal parameters within each of the GMFCS levels (ICCs.70). Reliability estimates in the transverse plane were lowest for children in GMFCS level III with ICCs varying from.32 to.93, compared with values of.59 to.93 and.94 to.96 for children in levels I and II, respectively. ICCs reflecting fair to good test-retest reliability were observed across all participants for frontal plane parameters (range,.46.78). Frontal plane ICCs varied from.16 to.48,.48 to.84, and.68 to.91 for children in GMFCS levels I, II, and III, respectively. The measurement variability between testing sessions as reflected by the SEM is shown in table 3. SEM values across kinematic parameters were comparable between GMFCS levels with values varying from 1.0 to 7.8, 2.0 to 5.4, and 1.1 to 9.6 for levels I, II, and III, respectively. MDC values were also similar across GMFCS levels, with values varying from 2.2 to 18.0, 4.7 to 12.5, and 2.7 to 22.2 for children in GMFCS levels I, II, and III, respectively. Bland-Altman plots revealed uniform distribution around the 0 line (ie, uniform variability) for most kinematic parameters analyzed (fig 1A, B, D). However, there were several examples of a nonuniform relationship in difference scores between test Parameter Table 2: Calculated Test-Retest Reliability Subcategorized Based on GMFCS Level All Participants (N 28) GMFCS Level I (n 10) GMFCS Level II (n 10) GMFCS Level III (n 8) ICC 95% CI ICC 95% CI ICC 95% CI ICC 95% CI Sagittal tilt 0.87 (0.71 0.94) 0.96 (0.83 0.99) 0.87 (0.47 0.97) 0.89 (0.49 0.98) Pelvis ROM 0.89 (0.76 0.95) 0.94 (0.75 0.98) 0.83 (0.33 0.96) 0.89 (0.48 0.98) Min hip flexion 0.93 (0.85 0.97) 0.94 (0.72 0.99) 0.95 (0.79 0.99) 0.94 (0.70 0.99) Hip flexion IC 0.90 (0.79 0.95) 0.86 (0.42 0.97) 0.93 (0.75 0.98) 0.84 (0.24 0.97) Hip ROM 0.84 (0.66 0.93) 0.82 (0.26 0.95) 0.86 (0.43 0.97) 0.91 (0.57 0.98) Max knee flexion 0.86 (0.69 0.93) 0.88 (0.55 0.97) 0.91 (0.66 0.98) 0.76 (0.00 0.95) Min knee flexion 0.97 (0.94 0.99) 0.92 (0.69 0.98) 0.98 (0.94 1.0) 0.98 (0.93 1.0) Knee flexion IC 0.96 (0.92 0.98) 0.97 (0.72 0.99) 0.92 (0.67 0.98) 0.93 (0.66 0.98) Min knee flexion stance 0.97 (0.94 0.99) 0.92 (0.68 0.98) 0.99 (0.94 1.0) 0.98 (0.92 1.0) Knee ROM 0.94 (0.86 0.97) 0.93 (0.72 0.98) 0.96 (0.85 0.99) 0.91 (0.57 0.98) Max ankle dorsi 0.88 (0.74 0.94) 0.92 (0.71 0.98) 0.80 (0.25 0.95) 0.89 (0.00 0.98) Min ankle dorsi 0.94 (0.89 0.97) 0.97 (0.89 0.99) 0.87 (0.49 0.97) 0.96 (0.82 0.99) Ankle dorsi IC 0.84 (0.64 0.93) 0.92 (0.67 0.98) 0.55 (0.00 0.88) 0.90 (0.00 0.98) Ankle ROM 0.93 (0.84 0.97) 0.87 (0.50 0.97) 0.94 (0.76 0.98) 0.92 (0.61 0.98) Max ankle plantar swing 0.92 (0.82 0.96) 0.97 (0.88 0.99) 0.89 (0.42 0.97) 0.87 (0.42 0.97) ICC 0.91 0.92 0.88 0.90 Frontal Min up obliquity 0.64 (0.24 0.83) 0.48 (0.00 0.86) 0.56 (0.00 0.89) 0.86 (0.31 0.97) Max up obliquity 0.46 (0.00 0.75) 0.36 (0.00 0.82) 0.48 (0.00 0.87) 0.89 (0.46 0.98) Max up obliquity stance 0.49 (0.00 0.76) 0.36 (0.00 0.82) 0.58 (0.00 0.89) 0.91 (0.58 0.98) Max hip adduction 0.62 (0.17 0.83) 0.31 (0.00 0.80) 0.70 (0.00 0.93) 0.68 (0.00 0.93) Min hip adduction IC 0.78 (0.53 0.90) 0.44 (0.00 0.86) 0.84 (0.00 0.96) 0.79 (0.00 0.96) Min hip adduction swing 0.73 (0.42 0.88) 0.16 (0.00 0.78) 0.66 ( 0.43 0.92) 0.70 (0.00 0.94) ICC 0.62 0.35 0.64 0.81 Transverse Max int rotation 0.84 (0.67 0.93) 0.87 (0.51 0.97) 0.94 (0.76 0.99) 0.54 (0.00 0.91) Min int rotation 0.91 (0.80 0.96) 0.93 (0.71 0.98) 0.95 (0.81 0.99) 0.32 (0.00 0.87) Max hip int rotation 0.84 (0.65 0.93) 0.59 (0.00 0.90) 0.95 (0.79 0.99) 0.83 (0.05 0.97) Min hip int rotation 0.91 (0.81 0.96) 0.75 (0.11 0.94) 0.96 (0.83 0.99) 0.93 (0.70 0.99) ICC 0.88 0.79 0.95 0.66 Temporal Velocity 0.93 (0.85 0.97) 0.78 (0.09 0.94) 0.93 (0.72 0.98) 0.96 (0.79 0.99) Stride length 0.90 (0.79 0.96) 0.67 (0.00 0.92) 0.92 (0.69 0.98) 0.98 (0.90 1.0) Cadence 0.90 (0.78 0.95) 0.79 (0.13 0.95) 0.91 (0.67 0.98) 0.94 (0.71 0.99) Percentage stance 0.87 (0.72 0.94) 0.86 (0.44 0.97) 0.83 (0.31 0.96) 0.74 (0.00 0.95) ICC 0.90 0.78 0.90 0.91 NOTE. Reliability estimates.90 are boldface. Abbreviations: Max, maximum; Min, minimum; Up, upwards; Int, internal; Dorsi, dorsiflexion; Plantar, plantarflexion.

784 RELIABILITY OF GAIT PARAMETERS IN CEREBRAL PALSY, Klejman Parameter Table 3: Difference Scores, SEM, and MDC Values for Kinematic and Spatiotemporal Parameters All Participants (N 28) GMFCS Level I (n 10) GMFCS Level II (n 10) GMFCS Level III (n 8) difference SD SEM MDC difference SD SEM MDC difference SD SEM MDC difference SD SEM MDC Sagittal ( ) tilt 1.5 5.7 4.0 9.3 2.4 3.8 3.6 8.4 0.2 5.1 3.6 8.4 2.6 6.8 4.8 11.1 Pelvis ROM 0.4 2.4 1.7 3.9 0.2 1.4 1.0 2.2 0.6 2.8 2.0 4.7 0.5 2.7 1.9 4.4 Min hip flexion 2.3 5.9 4.2 9.6 4.0 6.5 4.6 10.7 0.2 4.6 3.3 7.6 3.3 5.4 3.8 8.8 Hip flexion IC 1.9 7.3 5.2 12.0 0.1 7.4 5.2 12.1 2.0 7.6 5.4 12.5 4.5 6.2 4.4 10.2 Hip ROM 0.3 7.6 5.4 12.4 2.1 10.3 7.3 16.8 0.6 6.3 4.5 10.3 1.0 5.2 3.7 8.5 Max knee flexion 0.8 7.0 4.9 11.4 2.1 6.7 4.8 11.0 1.7 5.1 3.6 8.4 2.2 9.0 6.4 14.7 Min knee flexion 1.0 5.4 3.8 8.9 1.2 6.1 4.3 10.0 0.9 4.4 3.1 7.2 0.9 5.7 4.0 9.4 Knee flexion IC 0.3 5.0 3.5 8.1 2.8 2.8 2.0 4.6 0.6 5.9 4.2 9.7 1.7 5.0 3.5 8.2 Min knee flexion stance 1.0 5.5 3.9 9.0 1.2 6.3 4.4 10.3 0.9 4.2 3.0 7.0 0.9 5.7 4.0 9.3 Knee ROM 0.3 7.4 5.2 12.1 0.9 7.2 5.1 11.7 0.9 6.5 4.6 10.6 3.1 8.7 6.1 14.2 Max ankle dorsi 1.6 6.6 4.7 10.8 2.1 7.0 5.0 11.5 3.6 6.4 4.5 10.4 4.1 3.1 2.2 5.0 Min ankle dorsi 2.2 6.3 4.4 10.3 0.3 4.6 3.2 7.5 5.6 7.2 5.1 11.9 0.9 4.5 3.2 7.4 Ankle dorsi IC 2.7 6.2 4.4 10.2 0.8 6.2 4.4 10.1 4.6 7.1 5.0 11.6 4.6 3.1 2.2 5.1 Ankle ROM 0.5 5.2 3.7 8.5 2.0 4.7 3.3 7.8 2.0 5.0 3.5 8.2 3.2 4.9 3.4 8.0 Max ankle plantar swing 2.3 6.9 4.8 11.2 0.5 4.8 3.4 7.8 5.6 7.1 5.0 11.6 1.6 7.0 4.9 11.4 Frontal ( ) Min up obliquity 3.3 4.0 2.9 6.6 2.7 5.1 3.6 8.3 1.0 3.6 2.5 5.8 0.7 2.5 1.7 4.0 Max up obliquity 1.3 4.7 3.3 7.7 3.3 5.7 4.0 9.3 1.5 4.5 3.2 7.4 1.3 1.8 1.3 2.9 Max up obliquity stance 1.5 4.5 3.2 7.4 3.5 5.6 4.0 9.2 1.5 4.2 3.0 6.9 1.0 1.7 1.1 2.7 Max hip adduction 0.2 8.0 5.7 13.1 3.5 5.5 3.9 9.0 0.3 4.8 3.4 7.9 5.4 11.1 7.9 18.2 Min hip adduction IC 1.3 7.6 5.4 12.5 3.2 8.4 6.0 13.8 0.6 5.1 3.6 8.4 0.1 9.7 6.9 15.9 Min hip adduction swing 0.6 9.1 6.5 15.0 3.3 7.4 5.3 12.2 1.4 7.5 5.3 12.3 0.4 12.6 8.9 20.7 Transverse ( ) Max int rotation 2.3 8.8 6.2 14.4 2.1 8.2 5.8 13.5 2.2 4.1 2.9 6.7 2.5 13.6 9.6 22.2 Min int rotation 0.4 6.4 4.5 10.5 1.7 6.0 4.3 9.9 0.2 4.5 3.2 7.4 0.5 9.0 6.4 14.8 Max hip int rotation 0.0 9.8 7.0 16.1 0.7 11.0 7.8 18.0 1.0 5.2 3.7 8.5 0.4 13.5 9.6 22.2 Min hip int rotation 0.0 7.7 5.4 12.6 3.5 7.3 5.1 11.9 0.0 4.3 3.0 7.0 4.4 10.4 7.3 17.0 Spatiotemporal Velocity (m/s) 0.0 0.1 0.1 0.2 0.0 0.2 0.1 0.3 0.0 0.2 0.1 0.3 0.0 0.1 0.1 0.2 Stride length (m) 0.0 0.1 0.1 0.2 0.0 0.2 0.1 0.3 0.0 0.2 0.1 0.2 0.0 0.1 0.0 0.1 Cadence (steps/min) 0.7 16.6 11.7 27.1 3.0 23.0 16.2 37.6 3.2 11.5 8.1 18.8 2.6 13.3 9.4 21.8 Percentage stance (%) 0.5 3.7 2.6 6.0 0.5 2.6 1.9 4.3 0.6 3.9 2.8 6.4 1.9 4.4 3.1 7.1 Abbreviations: Max, maximum; Min, minimum; Up, upwards; Int, internal; Dorsi, dorsiflexion; Plantar, plantarflexion. sessions and the magnitude of the mean scores (ie, heteroscedasticity), where greater variability was associated with larger mean values. For example, larger test-retest differences were found in children who had a larger mean ankle ROM (fig 1C). A similar pattern was found for parameters measuring pelvic and hip rotation in the transverse and frontal planes (not shown). There was an outlier associated with a child in GMFCS level I (see fig 1B). On further inspection, this child adopted a dramatically altered gait pattern at retest, which resulted in a considerably longer stride length compared with visit 1. Removing this outlier changed the ICCs calculated for hip ROM and hip flexion at initial contact (ie, recalculated as.96 with 95% CI,.83.99 for both parameters). Stride length was also affected, with an ICC value recalculated as.82 (95% CI,.14.96) with the outlier excluded. There was little effect on other ICC values (ie, less than a 0.1 change in the ICC value). DISCUSSION CGA has become increasingly popular in the assessment of gait patterns in children with CP, and therefore, an understanding of the day-to-day reliability of gait parameters is essential to evaluate therapeutic interventions appropriately in these children. In this study, we investigated the test-retest reliability of discrete gait parameters using the ICC, SEM, MDC, and Bland-Altman methods in children with CP. Our findings suggest that select discrete gait parameters measured using CGA are reliable and repeatable measures of performance and can be used as outcome measures in intervention studies. Specifically, parameters measured in the sagittal plane and temporal-spatial parameters were highly reliable across all GMFCS levels. Lower ICCs in the frontal and transverse planes in all GMFCS levels indicated that caution is needed when interpreting the change of these kinematic parameters. Using MDC estimates as a guide, maximum hip internal rotation was the least reliable kinematic parameter in GMFCS levels I and III (MDC values of 18.0 and 22.2 for GMFCS levels I and III, respectively), while minimum hip adduction during swing was the least reliable parameter for level II, requiring a change of 12.3 between testing sessions to indicate a meaningful clinical change. The variability of the MDC values of individual parameters across GMFCS levels suggests that a GMFCS level specific MDC is needed when using CGA measures for evaluative purposes. From a clinical interpretation standpoint, several of the MDC values calculated in the current study for sagittal plane parameters are within the range of gait change scores typical for therapeutic or surgical intervention in this population. Oun-

RELIABILITY OF GAIT PARAMETERS IN CEREBRAL PALSY, Klejman 785 Fig 1. Bland-Altman plots for selected kinematic variables demonstrating uniform relationships between the difference from session to session (ordinate) and the mean performance between sessions (abscissa) (A, B, D) and nonuniform variability (ie, increasing variability with increasing mean ankle ROM) between the difference from session to session (ordinate) and the mean performance between sessions (abscissa) (C). puu et al 25 evaluated the effect of femoral derotation osteotomies in 20 ambulatory children with CP and showed significant increases in knee flexion at IC (an increase of 11 ), peak ankle dorsiflexion (10 ), and maximum ankle plantarflexion during swing (12 ). The GMFCS level of the participants was not specified, but it is likely that the group included children in GMFCS levels I, II, and III. These difference scores are comparable to the average MDC values calculated here, specifically 8.1, 10.8, and 11.2 for knee flexion at IC, peak ankle dorsiflexion, and maximum ankle plantarflexion during swing, respectively. Galli et al 12 evaluated the effect of botulinum toxin treatment in children in GMFCS levels I and II and found mean changes of 7.0, 7.5, and 4.6 for hip, knee, and ankle ROM, respectively. The MDC values from this study suggest that changes of 16.8, 11.7, and 7.8 are necessary for hip, knee, and ankle ROM, respectively, for children in level I, whereas changes of 10.3, 10.6, and 8.1, respectively, are needed for children in GMFCS level II to indicate a detectable clinical change. In general, larger MDC values were observed for children in GMFCS level III than levels I and II (14 of 29 MDC parameters were largest for children in level III compared with 5 of 29 and 9 of 29 parameters for levels I and II, respectively). A potential reason could be the result of fatigue and spasticity that particularly affect children in level III. McDowell et al 26 have suggested that the larger biarticular muscles in the lower limbs likely contribute to the increased variation associated with random measurements taken on different days, because they tend to be more subject to changes in tone. Across GMFCS levels, high levels of reliability were calculated for spatiotemporal parameters (ICCs.74). Despite this, the MDC values of these parameters indicate that they may not be responsive to change when comparing to gait change scores in the literature. For example, after orthopedic surgery or botulinum toxin injections, children with CP had an increase in velocity of.01m/s 25 and.04m/s, 12 respectively. These scores are well below the MDC values calculated in the current study, in which changes in velocity of.29m/s,.25m/s, and.17m/s for GMFCS levels I, II, and III, respectively, were needed to indicate meaningful change. MDCs calculated in the current study were also considerably larger than actual change scores for stride length, cadence, and percentage stance. 12,25,27 Previous work investigating the test-retest reliability of spatiotemporal parameters in children with CP found that the reliability of these parameters increased with increasing walking distance. 28 The relatively short walking distance in the laboratory (ie, 10m) may help to explain the large change scores found in the current study for these parameters. No studies to date document the repeatability of discrete gait parameters in children with CP using CGA. However, previous studies 8,9 investigating the repeatability of kinematic waveforms have found higher variability in the waveforms of children with higher impairment, suggesting that the day-to-day variability of gait is dependent on the functional level of the

786 RELIABILITY OF GAIT PARAMETERS IN CEREBRAL PALSY, Klejman child. This supports our findings that gait parameters tended to be less reliable for children in GMFCS level III than levels I and II, particularly in the frontal and transverse planes. A direct comparison between studies is difficult, because previous work did not stratify subjects based on GMFCS functional levels and instead focused on children who were independent ambulators (ie, children in GMFCS levels I and II). Furthermore, conclusions based on waveforms may differ from those based on discrete parameters. Thus, the current study expands on previous research by investigating the repeatability of gait kinematics between days calculated separately for GMFCS levels I, II, and III. The factors affecting the repeatability of gait parameters between baseline and retest include variability in the individual s gait pattern and the placement of reflective markers on anatomic landmarks. In order to minimize errors associated with marker placement, the same person applied the markers to all study participants on both visits. It has been reported that small changes in marker placement at the knee can have a large impact on transverse kinematics at the hip and may account for the larger ICCs and MDC values observed in this plane. 9 The present study highlights the need for strategies to improve the reliability of CGA, particularly in the coronal and transverse planes. Models with 6 degrees of freedom have recently been proposed as a method to control errors better than conventional gait models 29 ; however, their reliability between testing sessions has not yet been demonstrated. The use of knee alignment devices may improve reliability by lessening the impact of skin movement artifact during knee flexion. Study Limitations It was hypothesized that over repeated sessions, children may accommodate to the test evaluator, the testing procedure, and the room environment, and this may result in reduced muscle tone over repeated visits. 26,30 This might have potentially affected the results in the current study (ie, the outlier) because some children in the study had had previous gait analysis sessions, whereas others had not. CONCLUSIONS In this study, the test-retest reliability of discrete gait parameters was presented for ambulatory children with CP, grouped according to GMFCS levels. In general, parameters measured in the sagittal and transverse planes displayed higher reliability (ICC.84) than parameters in the frontal plane, for which reliability was no more than fair (ICC range,.46.78). The findings also suggest that the reliability of gait parameters is dependent on mobility level, with children in GMFCS level III exhibiting greater test-retest variability overall. Further study is recommended with a larger sample to confirm the impact of GMFCS level on day-to-day reliability of discrete gait parameters. The MDC estimates provide a preliminary insight into the potential responsiveness of various discrete gait parameters in children with CP across functional levels. However, the MDC estimates may not indicate a clinically important change in an evaluative study. References 1. Sutherland DH, Santi MS, Abel MF. Treatment of stiff-knee gait in cerebral palsy: a comparison by gait analysis of distal rectus femoris transfer versus proximal rectus release. J Pediatr Orthop 1990;10:433-41. 2. Gage JR. Gait analysis: an essential tool in the treatment of cerebral palsy. Clin Orthop 1993;288:126-34. 3. Abel MF, Damiano DL, Gilgannon M, et al. Biomechanical changes in gait following selective dorsal rhizotomy. J Neurosurg 2005;102 (2 Suppl):157-62. 4. DeLuca PA. Gait analysis in the treatment of the ambulatory child with cerebral palsy. Clin Orthop 1991;264:65-75. 5. Streiner DL, Norman GL. Health measurement scales: a practical guide to their development and use. 2nd ed. Oxford: Oxford University Press; 2000. 6. Redekop S, Andrysek J, Wright V. Single-session reliability of discrete gait parameters in ambulatory children with cerebral palsy based on GMFCS level. Gait Posture 2008;28:627-33. 7. Sadeghi H, Allard P, Shafie K, et al. Reduction of gait data variability using curve registration. Gait Posture 2000;12:257-64. 8. Steinwender G, Saraph V, Scheiber S, Zwick EB, Uitz C, Hackl K. Intrasubject repeatability of gait analysis data in normal and spastic children. Clin Biomech 2000;15:134-9. 9. Mackey A, Walt SE, Lobb GA, Stott S. Reliability of upper and lower limb three-dimensional kinematics in children with hemiplegia. Gait Posture 2005;22:1-9. 10. Monaghan K, Delahunt E, Caulfield B. Increasing the number of gait trial recordings maximizes intra-rater reliability of the CODA motion analysis system. Gait Posture 2007;25:303-15. 11. Maynard V, Bakheit AM, Oldham J, Freeman J. Intra-rater and inter-rater reliability of gait measurements with CODA mpx30 motion analysis system. Gait Posture 2003;17:59-67. 12. Galli M, Cimolin V, Valente M, Crivellini M, Ialongo T, Albertini G. Computertized gait analysis of botulinum toxin treatment in children with cerebral palsy. Disabil Rehabil 2007;29:659-64. 13. Damiano DL, Abel MF. Relation of gait analysis to gross motor function in cerebral palsy. Dev Med Child Neurol 1996;38:389-96. 14. Corry IS, Cosgrove AP, Duffy CM, Taylor TC, Graham HK. Botulinum toxin in hamstring spasticity. Gait Posture 1999;10: 206-10. 15. Shrout P, Fleiss JL. Intraclass correlations: uses in assessing rater reliability. Psychol Bull 1979;86:420-8. 16. Eliasziw M, Young SL, Woodbury MG, Fryday-Field K. Statistical methodology for the concurrent assessment of interrater and intrarater reliability: using goniometric measurements as an example. Phys Ther 1994;74:777-88. 17. Atkinson G, Nevill AM. Statistical methods for assessing measurement error (reliability) in variables relevant to sports medicine. Sports Med 1998;26:217-38. 18. Stratford PW, Goldsmith CH. Use of standard error as a reliability index of interest: an applied example using elbow flexor strength data. Phys Ther 1997;77:745-50. 19. Roebroeck ME, Harlaar J, Lankhorst GJ. The application of generalizability theory to reliability assessment: an illustration using isometric force measurements. Phys Ther 1993;73:386-401. 20. Stratford PW. Getting more from the literature: estimating the standard error of measurement from reliability studies. Physiother Can 2004;56:27-30. 21. Schmitt JS, Di Fabio RP. Reliable change and minimum important difference (MID) proportions facilitated group responsiveness comparisons using individual threshold criteria. J Clin Epi 2004; 57:1008-18. 22. Haley SM, Fragala-Pinkham MA. Interpreting change scores of tests and measures used in physical therapy. Phys Ther 2006;86: 735-43. 23. Fragala-Pinkham MA, Haley SM, Goodgold S. Evaluation of a community-based group fitness program for children with disabilities. Pediatr Phys Ther 2006;18:159-67. 24. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986; 1:307-10.

RELIABILITY OF GAIT PARAMETERS IN CEREBRAL PALSY, Klejman 787 25. Ounpuu S, DeLuca P, Davis R, Romness M. Long-term effects of femoral derotation osteotomies: an evaluation using threedimensional gait analysis. J Pediatr Orthop 2002;22:139-45. 26. McDowell BC, Hewitt V, Nurse A, Weston T, Baker R. The variability of goniometric measurements in ambulatory children with spastic cerebral palsy. Gait Posture 2000;12:114-21. 27. Shapiro A, Susak Z, Malkin C Mizrahi J. Preoperative and postoperative gait evaluation in cerebral palsy. Arch Phys Med Rehabil 1990;71:236-40. 28. Thompson P, Beath T, Bell J, et al. Test-retest reliability of the 10-metre fast walk test and 6-minute walk test in ambulatory school-aged children with cerebral palsy. Dev Med Child Neurol 2008;50:370-6. 29. Buczek FL, Rainbow MJ, Cooney KM, Walker MR, Sanders JO. Implications of using hierarchical and six degree-of-freedom models for normal gait analyses. Gait Posture 2010;31:57-63. 30. Ashton BB, Pickles B, Roll JW. Reliability of goniometric measurements of hip motion in spastic cerebral palsy. Dev Med Child Neurol 1978;20:87-94. Suppliers a. Vicon Peak, 14 Minns Business Park, West Way, Oxford, OX2 0JB, United Kingdom. b. The Mathworks Inc, 2 Apple Hill Dr, Natick, MA 01760-2098. c. SAS Institute Inc, 100 SAS Campus Dr, Cary, NC 27513-2414.