DEVELOPMENTAL MEDICINE & CHILD NEUROLOGY
ORIGINAL ARTICLE
Reliability and validity of the Duncan-Ely test for assessing rectus femoris spasticity in patients with cerebral palsy SEUNG YEOL LEE 1 *
| KI HYUK SUNG 2 * | CHIN YOUB CHUNG 3 | KYOUNG MIN LEE 3 | SOON-SUN KWON4 | TAE GYUN KIM 5 | SANG HYEONG LEE 6 | IN HYEOK LEE 7 | MOON SEOK PARK 3 1 Department of Orthopaedic Surgery, Ewha Womans University Mokdong Hospital, Seoul; 2 Department of Orthopaedic Surgery, Myongji Hospital, Kyungki; 3 Department of Orthopaedic Surgery, Seoul National University Bundang Hospital, Kyungki; 4 Biomedical Research Institute, Seoul National University Bundang Hospital, Kyungki; 5 Department of Orthopaedic Surgery, Konyang University Hospital, Daejon; 6 Department of Orthopaedic Surgery, Dongguk University Ilsan Hospital, Kyungki; 7 Department of Orthopaedic Surgery, Sungkyunkwan University Samsung Changwon Hospital, Gyeongnam, Korea. Correspondence to Moon Seok Park at Department of Orthopedic Surgery, Seoul National University Bundang Hospital, 300 Gumi-Dong, Bundang-Gu, Sungnam, Kyungki 463-707, Korea. E-mail: [email protected] *These authors contributed equally to this work. This article is commented on by Stott on pages 895–896 of this issue.
PUBLICATION DATA
Accepted for publication 27th February 2015. Published online 6th April 2015. ABBREVIATIONS
3DGA ICC RFT ROC
Three-dimensional gait analysis Intraclass correlation coefficient Rectus femoris transfer Receiver operator characteristics
AIM The aim of this study was to clarify the method of the Duncan-Ely test and to estimate its interobserver reliability and validity by comparing it with three-dimensional gait analysis (3DGA). METHOD This study included 36 consecutive ambulatory patients with cerebral palsy (CP) who underwent preoperative 3DGA. The Duncan-Ely test was performed during three different velocities (slow, gravity, and fast). The interobserver reliability was assessed by three examiners. The results of the test were compared with kinematic variables derived from the gait analysis to assess the sensitivity and specificity of the test. The cut-off value was determined at the point of trade-off between the highest sensitivity and specificity. RESULTS The intraclass correlation coefficient measuring interobserver reliability of the Duncan-Ely test was greatest during fast velocity (0.819). The sensitivity and specificity of the test during gravity velocity for knee range of motion total were 63.0% and 100% respectively, with a cut-off value of 78.3°. The sensitivity and specificity of the test during fast velocity for knee range of motion total were 66.7% and 100% respectively, with a cut-off value of 65°. INTERPRETATION The Duncan-Ely test shows excellent reliability in fast knee-flexion velocity, and good sensitivity and specificity compared with 3DGA during physical examination as a preoperative assessment of rectus femoris spasticity in patients with CP.
Stiff-knee gait, which is defined as a restricted arc in knee motion during swing,1 is a common gait disorder in ambulatory patients with cerebral palsy (CP), regardless of the anatomical type of CP.2 Stiff-knee gait results in foot clearance problems and reduces gait velocity and step length. Some patients complain of their shoes wearing out rapidly. Although there is still controversy,3 it is now widely accepted that one of the causes of stiff-knee gait in patients with CP is spasticity of the rectus femoris.4–7 Since rectus femoris transfer (RFT) has been recommended for restoring knee motion during the swing phase, it is now the standard surgical treatment for stiff-knee gait in patients with CP and is frequently performed as part of a single-event multilevel surgery.1,5,8,9 Before single-event multilevel surgery, a physical examination is essential to assess the patient’s physical status. The Duncan-Ely test, among others, has been recommended as a clinical tool for assessing rectus femoris spasticity.10 In the classic Duncan-Ely test, the examiner passively flexes the knee rapidly while the patient lies prone © 2015 Mac Keith Press
in a relaxed state. The test is considered positive if the patient simultaneously flexes their ipsilateral hip or if resistance is felt by the examiner during passive knee flexion.10,11 A positive test result reflects the velocitydependent spasticity of the rectus femoris muscle during the rapid movement test as well as muscle contracture during the slow movement test.12 Although a previous study reported that the hip rise was not necessarily specific for rectus femoris spasticity and that it might be due to the electromyographic response in both the rectus femoris and the iliopsoas,11 the Duncan-Ely test is a helpful predictor of rectus femoris dysfunction during gait and of outcomes in children being considered for RFT surgery.12,13 The test is easy to perform; however, the test method is vague in terms of the knee flexion velocity, and the results are based on the examiners’ judgements and are therefore subjective. In addition, the test cannot reflect the severity of spasticity of the rectus femoris. Therefore, physicians might be unsure of whether surgery is required when the patient’s hip rise is felt by the examiner at the end of knee flexion. DOI: 10.1111/dmcn.12761
963
Currently, three-dimensional gait analysis (3DGA) is commonly used as a tool in guiding the treatment of gait disorders.14 In this analysis, rectus femoris spasticity or contracture may result in a decrease in knee range of motion, a decrease in peak knee flexion, and a decrease in its flexion velocity during the swing phase. This analysis produces objective values for the kinematics and kinetics of the joint using optical tracking. Although 3DGA is relatively reliable and accurate,14 it might not always be available before surgery. The Duncan-Ely test showed only moderate levels of reliability in one study,15 but the method of the test was not clarified. Therefore, we performed this study to clarify the method of the Duncan-Ely test and to estimate the interobserver reliability and validity by comparing it with 3DGA.
METHOD The present prospective study was approved by the institutional review board at Seoul National University Bundang Hospital, a tertiary referral centre for CP. Informed consent was obtained from all the patients or their legal guardians. Between March 2011 and October 2013, we screened consecutive ambulatory patients with CP (Gross Motor Function Classification System [GMFCS] levels I–III) who showed positive results in the classic Duncan-Ely tests and underwent preoperative 3DGA for surgical planning purposes. Preoperative 3DGA was performed within 1 month before surgery, and the Duncan-Ely test (i.e. the reliability test) was performed 1 day before surgery. Patients with a history of orthopaedic surgical procedures or with neuromuscular diseases other than CP were excluded. Thirty-six consecutive patients with CP were enrolled. The median age of the patients at the time of examination was 10 years (interquartile range 8–19y; Table I). Consensus building and the test protocol Six orthopaedic surgeons with a median of 12 years’ orthopaedic experience (interquartile range 8y 8mo–17y 4mo)
Table I: Patient demographics Parameter
Value
Number of patients (male/female) Median age (y)
36 (25/11) 10 (interquartile range 8–19)
Involvement of CP
Number of patients (male/female)
Unilateral Bilateral
8 (5/3) 28 (20/8)
GMFCS level
Number of patients (male/female)
I II III
8 (4/4) 19 (16/3) 9 (5/4)
CP, cerebral palsy; GMFCS, Gross Motor Function Classification System.
964 Developmental Medicine & Child Neurology 2015, 57: 963–968
• •
What this paper adds The interobserver reliability of the Duncan-Ely test was greatest during fast velocity. The sensitivity and specificity of the Duncan-Ely test were greatest during fast velocity.
held a consensus-building session before the patients’ physical examination measurements were taken. We established the details of the test method, including the velocity. Three orthopaedic surgeons (KML, KHS, and SYL) with 11, 9, and 8 years’ orthopaedic experience respectively, were selected as examiners for the session. All decisions required unanimity.16 Then the test was performed on the patients. First, with the patient in the prone position, one hip was stabilized in extension while the lower leg was brought into flexion. Next, the lower leg was placed in flexion to test velocity, which was divided into three levels according to the Tardieu scale (V1, V2, and V3).17 V1 is as slow as possible for the examiner. It is slower than the rate of the natural drop of the limb segment under gravity. V2 is the velocity of the limb segment naturally falling under gravity. We tried to simulate the similar velocity with gravity, because it was not possible to evaluate the natural fall into knee flexion in the prone position. V3 is as fast as possible, and it is faster than the rate of the natural drop of the limb segment under gravity. The slow velocity test assesses the shortening of the rectus femoris, and the fast velocity test assesses the velocity-dependent spasticity of the muscle. V1 served as a comparison for V2 and V3. V2 was used as a reference velocity. Because the measurement of the angular velocity of the knee joint was difficult to obtain during the test in a general clinical setting, the gravity velocity was selected as a reference velocity for reliability, according to the Tardieu scale. Finally, the knee flexion angle, at which the hip rise was felt by the examiner, was measured and recorded (Fig. 1).
The reliability of the Duncan-Ely test The reliability of the Duncan-Ely test was assessed by calculating the interobserver reliability for the three examiners who had 13, 10, and 9 years’ experience in orthopaedics. The tests were conducted during a single day before single-event multilevel surgery for all patients. Examiners were blinded to the measurements of the other examiners as well as to the patients’ clinical information. Examinations were performed by three teams; each team was headed by one of the three orthopaedic surgeons and included physician assistants with more than 4 years’ experience. An orthopaedic surgeon positioned the patients’ limbs and made measurements using a goniometer, while a physician assistant held the limb in place. All the measurement data were collected by a research assistant who did not participate in the examinations. The validity of the Duncan-Ely test The validity of the Duncan-Ely test was assessed for its sensitivity and specificity. The test results were compared with the kinematic data derived from the 3DGA, and
(a)
(b)
(c)
Figure 1: The estimated knee flexion angle during the Duncan-Ely test. (a) With the patient in the prone position; (b) the lower leg is brought into flexion; (c) the knee flexion angle (i.e. the point at which the hip rises or resistance is felt by the examiner) is measured.
receiver operator characteristics (ROC) curves were evaluated to determine sensitivity and specificity. Sensitivity was the true positive rate and specificity was 1 minus false-positive rate of validity when the peak knee flexion in the swing phase, the knee range of motion total, or timing of peak knee flexion in the swing phase of gait analysis was abnormal. The cut-off value was determined at the point of trade-off between the highest sensitivity and specificity. The ROC curve provides a graphical illustration of these trade-offs at each cut-off for any diagnostic test that uses a continuous variable.18 Physical examination findings were compared at three key kinematic data points, which were derived from the 3DGA and were commonly used for determining the RFT outcomes: (1) peak knee flexion in the swing phase; (2) knee range of motion total (i.e. the range of motion from the maximum extension in the stance phase to peak flexion in swing);19 and (3) timing of the peak knee flexion in the swing phase (as a % of the swing phase).20,21 A few days before surgery, 3DGA was performed using a Vicon 370 system (Oxford Metrix, Oxford, UK), which was equipped with seven cameras and two force plates. Markers were placed by a single operator according to the Helen Hayes marker set.22 Kinematic data were corrected as patients walked barefoot on a 9m walkway during an interval of approximately 30 seconds. Three
trials were averaged to determine the values of the index variables. The measurement of each patient’s 3DGA was compared with the data in normal23 gait to assess the cut-off values for the peak knee flexion in the swing phase, knee range of motion total, and timing of peak knee flexion in the swing phase. Gait parameters were considered abnormal at values of one standard deviation (SD) above or below the mean of the normal value.
Statistical methods Descriptive analysis of the entire data set was performed, which yielded data, including the average and SDs or proportions. The data were assessed for normality using a Kolmogorov–Smirnov test. Prior precision power analysis, which is used in studies estimating a parameter at a fixed confidence level, was performed to identify the minimal sample size required for the analysis. This study was designed to enable the intraclass correlation coefficients (ICCs) of reliability to be calculated at a target value of 0.8. In addition, we used the approximation by Bonett.24 Accordingly, when we set the width of the 95% confidence interval (CI) to 0.2 for the three examiners, the minimal sample size was calculated to be 36. The ICCs and 95% CI were used to summarize the inReliability and Validity of the Duncan-Ely Test Seung Yeol Lee et al.
965
terobserver reliability and were calculated in the setting of a two-way random effects model, assuming a single measurement and absolute agreement. An ICC value of 1 indicated perfect reliability. ROC curves were plotted as the false-positive rate (i.e. 1 minus specificity) versus sensitivity for all cut-off values in the range of velocity values observed. The area under the ROC curve was used to measure the performance accuracy of velocity as a predictor of gait parameters. An area under the ROC curve of 1.0 represented an error-free prediction of abnormal gait status in all samples, whereas an area of 0.50 represented a 50% likelihood of a correct prediction of abnormal gait status (similar to a coin toss). One limb of the patients with bilateral involvement and the affected limb of the patients with unilateral involvement were selected and included in the data analysis to ensure statistical independence.25 The limb in patients with bilateral involvement was randomly selected for data collection. Statistical analyses used R software, version 2.15.2 (R Foundation for Statistical Computing, Vienna, Austria; http://www.r-project.org/) with the PPCOR, pROC, and NLME packages. All the statistical tests were two-tailed, and p values less than 0.05 were considered significant.
RESULTS The means and SDs of the knee flexion angle (i.e. the point at which the hip rise was felt by the examiner) were 87.5° (SD 22.1°) in V1, 75.7° (SD 19.9°) in V2, and 65.7° (SD 22.3°) in V3 (Table II).
The Duncan-Ely test had favourable ICC values for interobserver reliability. The interobserver reliability of the test was the highest during V3 (ICC 0.819), while the lowest interobserver reliability of the test was during V2 (ICC 0.626; Table II). The performance accuracy (area under the ROC curve) of V1 as a predictor for gait parameters was not statistically significant (Table III). The cut-off values of the Duncan-Ely test between the normal and pathological values of gait analysis were as follows: 78.3° during V2 and 65° during V3 for knee range of motion in the swing phase (Table III). The percentages of abnormal knee range of motion in the swing phase were correctly identified by measuring the knee flexion angles, which were 63% during V2 and 66.7% during V3 (Fig. 2a–c).
DISCUSSION There is a lack of research into the reliability and validity of the Duncan-Ely test. Therefore, the present study aimed to evaluate the interobserver reliability and validity of this test. The ICC, which measured the interobserver reliability of the Duncan-Ely test, showed the highest ICC value during fast velocity (V3). The sensitivity and specificity of the test during fast velocity (V3) were highest with a cut-off value of 65°. Before discussing the study results, a limitation of the current study should be considered. Sensitivity and specificity were used to assess the strength of the Duncan-Ely test’s validity. Previous literature shows that the peak knee flexion, knee range of motion, and timing of peak knee
Table II: Interobserver reliability of the Duncan-Ely test Interobserver reliability Measurements
Knee flexion angle° (SD; range)
Measurement
ICC
95% CI
V1 (slow)
87.5 (22.1; 40–150)
V2 (gravity)
75.7 (19.9; 35–153)
V3 (fast)
65.7 (22.3; 28–152)
Single Average Single Average Single Average
0.721 0.865 0.626 0.848 0.819 0.914
0.537–0.854 0.795–0.915 0.455–0.768 0.766–0.904 0.636–0.952 0.869–0.946
ICC, intraclass correlation coefficient; Single, the single measure is an index for the reliability of the ratings of one, typical, single rater; Average, the average measure is an index for the reliability of the ratings of three different raters averaged together.
Table III: Cut-off values of the Duncan-Ely test Velocity
Gait parameter
Reference value of 3DGA23 (SD)
Area under the ROC curve
95% CI
p
V1 (slow)
Pk flexion in swing phase Knee ROM total Timing of pk flexion in swing Pk flexion in swing phase Knee ROM total Timing of pk flexion in swing Pk flexion in swing phase Knee ROM total Timing of pk flexion in swing
55.6° 51.6° 72% 55.6° 51.6° 72% 55.6° 51.6° 72%
0.663 0.753 0.562 0.644 0.790 0.598 0.562 0.846 0.589
0.469–0.825 0.562–0.891 0.370–0.743 0.449–0.809 0.603–0.916 0.404–0.772 0.370–0.742 0.667–0.951 0.395–0.764
0.104 0.066 0.582 0.350 0.017 0.376 0.718
ORIGINAL ARTICLE
Reliability and validity of the Duncan-Ely test for assessing rectus femoris spasticity in patients with cerebral palsy SEUNG YEOL LEE 1 *
| KI HYUK SUNG 2 * | CHIN YOUB CHUNG 3 | KYOUNG MIN LEE 3 | SOON-SUN KWON4 | TAE GYUN KIM 5 | SANG HYEONG LEE 6 | IN HYEOK LEE 7 | MOON SEOK PARK 3 1 Department of Orthopaedic Surgery, Ewha Womans University Mokdong Hospital, Seoul; 2 Department of Orthopaedic Surgery, Myongji Hospital, Kyungki; 3 Department of Orthopaedic Surgery, Seoul National University Bundang Hospital, Kyungki; 4 Biomedical Research Institute, Seoul National University Bundang Hospital, Kyungki; 5 Department of Orthopaedic Surgery, Konyang University Hospital, Daejon; 6 Department of Orthopaedic Surgery, Dongguk University Ilsan Hospital, Kyungki; 7 Department of Orthopaedic Surgery, Sungkyunkwan University Samsung Changwon Hospital, Gyeongnam, Korea. Correspondence to Moon Seok Park at Department of Orthopedic Surgery, Seoul National University Bundang Hospital, 300 Gumi-Dong, Bundang-Gu, Sungnam, Kyungki 463-707, Korea. E-mail: [email protected] *These authors contributed equally to this work. This article is commented on by Stott on pages 895–896 of this issue.
PUBLICATION DATA
Accepted for publication 27th February 2015. Published online 6th April 2015. ABBREVIATIONS
3DGA ICC RFT ROC
Three-dimensional gait analysis Intraclass correlation coefficient Rectus femoris transfer Receiver operator characteristics
AIM The aim of this study was to clarify the method of the Duncan-Ely test and to estimate its interobserver reliability and validity by comparing it with three-dimensional gait analysis (3DGA). METHOD This study included 36 consecutive ambulatory patients with cerebral palsy (CP) who underwent preoperative 3DGA. The Duncan-Ely test was performed during three different velocities (slow, gravity, and fast). The interobserver reliability was assessed by three examiners. The results of the test were compared with kinematic variables derived from the gait analysis to assess the sensitivity and specificity of the test. The cut-off value was determined at the point of trade-off between the highest sensitivity and specificity. RESULTS The intraclass correlation coefficient measuring interobserver reliability of the Duncan-Ely test was greatest during fast velocity (0.819). The sensitivity and specificity of the test during gravity velocity for knee range of motion total were 63.0% and 100% respectively, with a cut-off value of 78.3°. The sensitivity and specificity of the test during fast velocity for knee range of motion total were 66.7% and 100% respectively, with a cut-off value of 65°. INTERPRETATION The Duncan-Ely test shows excellent reliability in fast knee-flexion velocity, and good sensitivity and specificity compared with 3DGA during physical examination as a preoperative assessment of rectus femoris spasticity in patients with CP.
Stiff-knee gait, which is defined as a restricted arc in knee motion during swing,1 is a common gait disorder in ambulatory patients with cerebral palsy (CP), regardless of the anatomical type of CP.2 Stiff-knee gait results in foot clearance problems and reduces gait velocity and step length. Some patients complain of their shoes wearing out rapidly. Although there is still controversy,3 it is now widely accepted that one of the causes of stiff-knee gait in patients with CP is spasticity of the rectus femoris.4–7 Since rectus femoris transfer (RFT) has been recommended for restoring knee motion during the swing phase, it is now the standard surgical treatment for stiff-knee gait in patients with CP and is frequently performed as part of a single-event multilevel surgery.1,5,8,9 Before single-event multilevel surgery, a physical examination is essential to assess the patient’s physical status. The Duncan-Ely test, among others, has been recommended as a clinical tool for assessing rectus femoris spasticity.10 In the classic Duncan-Ely test, the examiner passively flexes the knee rapidly while the patient lies prone © 2015 Mac Keith Press
in a relaxed state. The test is considered positive if the patient simultaneously flexes their ipsilateral hip or if resistance is felt by the examiner during passive knee flexion.10,11 A positive test result reflects the velocitydependent spasticity of the rectus femoris muscle during the rapid movement test as well as muscle contracture during the slow movement test.12 Although a previous study reported that the hip rise was not necessarily specific for rectus femoris spasticity and that it might be due to the electromyographic response in both the rectus femoris and the iliopsoas,11 the Duncan-Ely test is a helpful predictor of rectus femoris dysfunction during gait and of outcomes in children being considered for RFT surgery.12,13 The test is easy to perform; however, the test method is vague in terms of the knee flexion velocity, and the results are based on the examiners’ judgements and are therefore subjective. In addition, the test cannot reflect the severity of spasticity of the rectus femoris. Therefore, physicians might be unsure of whether surgery is required when the patient’s hip rise is felt by the examiner at the end of knee flexion. DOI: 10.1111/dmcn.12761
963
Currently, three-dimensional gait analysis (3DGA) is commonly used as a tool in guiding the treatment of gait disorders.14 In this analysis, rectus femoris spasticity or contracture may result in a decrease in knee range of motion, a decrease in peak knee flexion, and a decrease in its flexion velocity during the swing phase. This analysis produces objective values for the kinematics and kinetics of the joint using optical tracking. Although 3DGA is relatively reliable and accurate,14 it might not always be available before surgery. The Duncan-Ely test showed only moderate levels of reliability in one study,15 but the method of the test was not clarified. Therefore, we performed this study to clarify the method of the Duncan-Ely test and to estimate the interobserver reliability and validity by comparing it with 3DGA.
METHOD The present prospective study was approved by the institutional review board at Seoul National University Bundang Hospital, a tertiary referral centre for CP. Informed consent was obtained from all the patients or their legal guardians. Between March 2011 and October 2013, we screened consecutive ambulatory patients with CP (Gross Motor Function Classification System [GMFCS] levels I–III) who showed positive results in the classic Duncan-Ely tests and underwent preoperative 3DGA for surgical planning purposes. Preoperative 3DGA was performed within 1 month before surgery, and the Duncan-Ely test (i.e. the reliability test) was performed 1 day before surgery. Patients with a history of orthopaedic surgical procedures or with neuromuscular diseases other than CP were excluded. Thirty-six consecutive patients with CP were enrolled. The median age of the patients at the time of examination was 10 years (interquartile range 8–19y; Table I). Consensus building and the test protocol Six orthopaedic surgeons with a median of 12 years’ orthopaedic experience (interquartile range 8y 8mo–17y 4mo)
Table I: Patient demographics Parameter
Value
Number of patients (male/female) Median age (y)
36 (25/11) 10 (interquartile range 8–19)
Involvement of CP
Number of patients (male/female)
Unilateral Bilateral
8 (5/3) 28 (20/8)
GMFCS level
Number of patients (male/female)
I II III
8 (4/4) 19 (16/3) 9 (5/4)
CP, cerebral palsy; GMFCS, Gross Motor Function Classification System.
964 Developmental Medicine & Child Neurology 2015, 57: 963–968
• •
What this paper adds The interobserver reliability of the Duncan-Ely test was greatest during fast velocity. The sensitivity and specificity of the Duncan-Ely test were greatest during fast velocity.
held a consensus-building session before the patients’ physical examination measurements were taken. We established the details of the test method, including the velocity. Three orthopaedic surgeons (KML, KHS, and SYL) with 11, 9, and 8 years’ orthopaedic experience respectively, were selected as examiners for the session. All decisions required unanimity.16 Then the test was performed on the patients. First, with the patient in the prone position, one hip was stabilized in extension while the lower leg was brought into flexion. Next, the lower leg was placed in flexion to test velocity, which was divided into three levels according to the Tardieu scale (V1, V2, and V3).17 V1 is as slow as possible for the examiner. It is slower than the rate of the natural drop of the limb segment under gravity. V2 is the velocity of the limb segment naturally falling under gravity. We tried to simulate the similar velocity with gravity, because it was not possible to evaluate the natural fall into knee flexion in the prone position. V3 is as fast as possible, and it is faster than the rate of the natural drop of the limb segment under gravity. The slow velocity test assesses the shortening of the rectus femoris, and the fast velocity test assesses the velocity-dependent spasticity of the muscle. V1 served as a comparison for V2 and V3. V2 was used as a reference velocity. Because the measurement of the angular velocity of the knee joint was difficult to obtain during the test in a general clinical setting, the gravity velocity was selected as a reference velocity for reliability, according to the Tardieu scale. Finally, the knee flexion angle, at which the hip rise was felt by the examiner, was measured and recorded (Fig. 1).
The reliability of the Duncan-Ely test The reliability of the Duncan-Ely test was assessed by calculating the interobserver reliability for the three examiners who had 13, 10, and 9 years’ experience in orthopaedics. The tests were conducted during a single day before single-event multilevel surgery for all patients. Examiners were blinded to the measurements of the other examiners as well as to the patients’ clinical information. Examinations were performed by three teams; each team was headed by one of the three orthopaedic surgeons and included physician assistants with more than 4 years’ experience. An orthopaedic surgeon positioned the patients’ limbs and made measurements using a goniometer, while a physician assistant held the limb in place. All the measurement data were collected by a research assistant who did not participate in the examinations. The validity of the Duncan-Ely test The validity of the Duncan-Ely test was assessed for its sensitivity and specificity. The test results were compared with the kinematic data derived from the 3DGA, and
(a)
(b)
(c)
Figure 1: The estimated knee flexion angle during the Duncan-Ely test. (a) With the patient in the prone position; (b) the lower leg is brought into flexion; (c) the knee flexion angle (i.e. the point at which the hip rises or resistance is felt by the examiner) is measured.
receiver operator characteristics (ROC) curves were evaluated to determine sensitivity and specificity. Sensitivity was the true positive rate and specificity was 1 minus false-positive rate of validity when the peak knee flexion in the swing phase, the knee range of motion total, or timing of peak knee flexion in the swing phase of gait analysis was abnormal. The cut-off value was determined at the point of trade-off between the highest sensitivity and specificity. The ROC curve provides a graphical illustration of these trade-offs at each cut-off for any diagnostic test that uses a continuous variable.18 Physical examination findings were compared at three key kinematic data points, which were derived from the 3DGA and were commonly used for determining the RFT outcomes: (1) peak knee flexion in the swing phase; (2) knee range of motion total (i.e. the range of motion from the maximum extension in the stance phase to peak flexion in swing);19 and (3) timing of the peak knee flexion in the swing phase (as a % of the swing phase).20,21 A few days before surgery, 3DGA was performed using a Vicon 370 system (Oxford Metrix, Oxford, UK), which was equipped with seven cameras and two force plates. Markers were placed by a single operator according to the Helen Hayes marker set.22 Kinematic data were corrected as patients walked barefoot on a 9m walkway during an interval of approximately 30 seconds. Three
trials were averaged to determine the values of the index variables. The measurement of each patient’s 3DGA was compared with the data in normal23 gait to assess the cut-off values for the peak knee flexion in the swing phase, knee range of motion total, and timing of peak knee flexion in the swing phase. Gait parameters were considered abnormal at values of one standard deviation (SD) above or below the mean of the normal value.
Statistical methods Descriptive analysis of the entire data set was performed, which yielded data, including the average and SDs or proportions. The data were assessed for normality using a Kolmogorov–Smirnov test. Prior precision power analysis, which is used in studies estimating a parameter at a fixed confidence level, was performed to identify the minimal sample size required for the analysis. This study was designed to enable the intraclass correlation coefficients (ICCs) of reliability to be calculated at a target value of 0.8. In addition, we used the approximation by Bonett.24 Accordingly, when we set the width of the 95% confidence interval (CI) to 0.2 for the three examiners, the minimal sample size was calculated to be 36. The ICCs and 95% CI were used to summarize the inReliability and Validity of the Duncan-Ely Test Seung Yeol Lee et al.
965
terobserver reliability and were calculated in the setting of a two-way random effects model, assuming a single measurement and absolute agreement. An ICC value of 1 indicated perfect reliability. ROC curves were plotted as the false-positive rate (i.e. 1 minus specificity) versus sensitivity for all cut-off values in the range of velocity values observed. The area under the ROC curve was used to measure the performance accuracy of velocity as a predictor of gait parameters. An area under the ROC curve of 1.0 represented an error-free prediction of abnormal gait status in all samples, whereas an area of 0.50 represented a 50% likelihood of a correct prediction of abnormal gait status (similar to a coin toss). One limb of the patients with bilateral involvement and the affected limb of the patients with unilateral involvement were selected and included in the data analysis to ensure statistical independence.25 The limb in patients with bilateral involvement was randomly selected for data collection. Statistical analyses used R software, version 2.15.2 (R Foundation for Statistical Computing, Vienna, Austria; http://www.r-project.org/) with the PPCOR, pROC, and NLME packages. All the statistical tests were two-tailed, and p values less than 0.05 were considered significant.
RESULTS The means and SDs of the knee flexion angle (i.e. the point at which the hip rise was felt by the examiner) were 87.5° (SD 22.1°) in V1, 75.7° (SD 19.9°) in V2, and 65.7° (SD 22.3°) in V3 (Table II).
The Duncan-Ely test had favourable ICC values for interobserver reliability. The interobserver reliability of the test was the highest during V3 (ICC 0.819), while the lowest interobserver reliability of the test was during V2 (ICC 0.626; Table II). The performance accuracy (area under the ROC curve) of V1 as a predictor for gait parameters was not statistically significant (Table III). The cut-off values of the Duncan-Ely test between the normal and pathological values of gait analysis were as follows: 78.3° during V2 and 65° during V3 for knee range of motion in the swing phase (Table III). The percentages of abnormal knee range of motion in the swing phase were correctly identified by measuring the knee flexion angles, which were 63% during V2 and 66.7% during V3 (Fig. 2a–c).
DISCUSSION There is a lack of research into the reliability and validity of the Duncan-Ely test. Therefore, the present study aimed to evaluate the interobserver reliability and validity of this test. The ICC, which measured the interobserver reliability of the Duncan-Ely test, showed the highest ICC value during fast velocity (V3). The sensitivity and specificity of the test during fast velocity (V3) were highest with a cut-off value of 65°. Before discussing the study results, a limitation of the current study should be considered. Sensitivity and specificity were used to assess the strength of the Duncan-Ely test’s validity. Previous literature shows that the peak knee flexion, knee range of motion, and timing of peak knee
Table II: Interobserver reliability of the Duncan-Ely test Interobserver reliability Measurements
Knee flexion angle° (SD; range)
Measurement
ICC
95% CI
V1 (slow)
87.5 (22.1; 40–150)
V2 (gravity)
75.7 (19.9; 35–153)
V3 (fast)
65.7 (22.3; 28–152)
Single Average Single Average Single Average
0.721 0.865 0.626 0.848 0.819 0.914
0.537–0.854 0.795–0.915 0.455–0.768 0.766–0.904 0.636–0.952 0.869–0.946
ICC, intraclass correlation coefficient; Single, the single measure is an index for the reliability of the ratings of one, typical, single rater; Average, the average measure is an index for the reliability of the ratings of three different raters averaged together.
Table III: Cut-off values of the Duncan-Ely test Velocity
Gait parameter
Reference value of 3DGA23 (SD)
Area under the ROC curve
95% CI
p
V1 (slow)
Pk flexion in swing phase Knee ROM total Timing of pk flexion in swing Pk flexion in swing phase Knee ROM total Timing of pk flexion in swing Pk flexion in swing phase Knee ROM total Timing of pk flexion in swing
55.6° 51.6° 72% 55.6° 51.6° 72% 55.6° 51.6° 72%
0.663 0.753 0.562 0.644 0.790 0.598 0.562 0.846 0.589
0.469–0.825 0.562–0.891 0.370–0.743 0.449–0.809 0.603–0.916 0.404–0.772 0.370–0.742 0.667–0.951 0.395–0.764
0.104 0.066 0.582 0.350 0.017 0.376 0.718
