Home
Search
Collections
Journals
About
Contact us
My IOPscience
Inertia based functional scoring of the shoulder in clinical practice
This content has been downloaded from IOPscience. Please scroll down to see the full text. 2014 Physiol. Meas. 35 167 (http://iopscience.iop.org/0967-3334/35/2/167) View the table of contents for this issue, or go to the journal homepage for more
Download details: IP Address: 141.117.125.1 This content was downloaded on 25/05/2014 at 19:11
Please note that terms and conditions apply.
Institute of Physics and Engineering in Medicine Physiol. Meas. 35 (2014) 167–176
Physiological Measurement
doi:10.1088/0967-3334/35/2/167
Inertia based functional scoring of the shoulder in clinical practice 1 ¨ R J P Korver , I C Heyligers, S K Samijo and B Grimm
AHORSE Research Foundation, Department of Orthopaedic Surgery and Traumatology, Atrium Medical Center Parkstad Heerlen, Henri Dunantstraat 5, 6419 PC Heerlen, The Netherlands E-mail: [email protected], [email protected], [email protected] and [email protected] Received 2 October 2013, revised 24 November 2013 Accepted for publication 3 December 2013 Published 7 January 2014 Abstract
Shoulder-related dysfunction is the second most common musculoskeletal disorder and is responsible for an increasing burden on health-care systems. Commonly used clinical outcome scores suffer from subjectivity, pain dominance and a ceiling effect. Objective functional measurement has been identified as a relevant issue in clinical rehabilitation. In recognition of this goal simple techniques for routine clinical application have been investigated with some success. Inertia based motion analysis (IMA) is a new generation of objective outcome assessment tool; it can produce objective movement parameters while being fast, cheap and easy to operate. This study investigates if a simple IMA shoulder test is suitable as a functional outcome measure for routine clinical follow-up. We measured 100 healthy subjects and 50 patients with confirmed unilateral shoulder pathology. Two motion tasks were performed on both shoulders and two simple motion parameters based on angular rate and acceleration were calculated. Patients were also assessed by the disability of arm, shoulder and hand (DASH) and the simple shoulder test. IMA produced high intra- (ICC = 0.94) and inter-assessor reliability (ICC = 0.90). Asymmetry was >3 times higher in patients than in healthy controls (p < 0.01). Healthy and pathological subjects could be distinguished with high diagnostic sensitivity (>84.0%) and specificity (>81.0%). There was a weak correlation between the IMA shoulder score and the clinical questionnaires (Pearson R < 0.25), as it may add an objective functional dimension to outcome assessment. The fast assessment (t < 5 min) of a simple motion task makes it workable for routine clinical follow-up. The IMA shoulder test adds objective information on functional capacity to the clinical scores and may help the physician in his 1
Author to whom any correspondence should be addressed.
0967-3334/14/020167+10$33.00
© 2014 Institute of Physics and Engineering in Medicine Printed in the UK
167
¨ R J P Korver et al
Physiol. Meas. 35 (2014) 167
decision-making, follow-up of treatment, effect of training and possibly lead to the development of new therapeutic interventions. Keywords: shoulder, outcome assessment, inertia sensors, healthy reference data, clinical validation, diagnostic power, clinical scores (Some figures may appear in colour only in the online journal)
1. Introduction Shoulder-related dysfunction with limited range of motion (ROM) and restricted activities of daily living (ADL) is a common health problem with a 34% prevalence in subjects >65 years of age (Chakravarty and Webley 1990). In clinical practice, outcome can be assessed by a large variety of questionnaire based scores of which the disability of arm, shoulder and hand (DASH) score and the simple shoulder test (SST) are the most commonly used (Jester et al 2004, Lippitt et al 1992). However, poor reliability has been reported (Kirkley et al 2003). Even for the Constant–Murley score (CMS), which adds an objective component (e.g. ROM) to its questionnaire poor reliability is documented (Conboy et al 1996, Constant et al 2008, Rocourt et al 2008). In general, questionnaires suffer from subjectivity, a ceiling effect and are highly painrelated, often masking function-related results (Coley et al 2007, Lin et al 2005). In addition, since the conception of the scores, patients characteristics and their demands have shifted toward younger patients being treated earlier in the disease development and claiming for higher functional improvement after treatment (Nilsdotter et al 2009). Rising health-economic demands require objective evidence of new treatment effects in clinical rehabilitation, which the questionnaires can hardly provide (Lansky and Milstein 2009). In research settings, advanced motion analysis systems are used for quantitative movement analysis using opto-electronic systems, force plates or electromyography (EMG) (Bey et al 2006, Bolink et al 2012, Garcia-Alsina et al 2005, Klopcar and Lenarcic 2006, Leggin et al 1996, Lin and Kuli´c 2012, Pitts et al 2013, Van Andel et al 2008, Veeger et al 2006). Although these methods can provide objective evaluation of shoulder movements, they are less practical for routine clinical use because they require technically skilled personnel, are time-consuming, relatively expensive and need a laboratory that may cause unnatural movements. In addition, kinematic analysis of the upper extremity has not received as much scientific attention as that of the lower limb. Last decade, a major contribution as potential alternative for lab-based motion analysis has been made in the development of inertia based motion analysis (IMA) (Coley et al 2007, Cutti et al 2008, Parel et al 2012, de Vries et al 2010). Previous studies used accelerometers and gyroscopes to measure the overall shoulder movement at the arm during ADL tasks, developed motion parameters and validated these parameters against opto-electronic systems (Coley et al 2007). IMA is a feasible method for describing shoulder kinematics during ADL tasks and can serve as an activity monitor (Coley et al 2007, 2008b, Parel et al 2012). However, these studies were performed in small groups, including heterogeneous pathologies (e.g. rotator cuff disease, osteoarthritis and postoperative shoulder arthroplasty). Furthermore, these studies did not investigate the diagnostics power of IMA and its correlation to clinical questionnaires, which makes the level of clinical validation, is still weak. The aim of this study is, (A) to clinically validate this score by testing its reliability and diagnostic power in distinguishing healthy from pathologic shoulders and (B) compare this IMA shoulder score against standard clinical questionnaires. 168
¨ R J P Korver et al
Physiol. Meas. 35 (2014) 167
2. Methods 2.1. Subjects
A total of 175 subjects were recruited and bilateral shoulder measurement was performed according to a protocol approved by our hospital’s Institutional Ethics Board Committee. Group A consisted of 100 healthy subjects without shoulder complaints, who did not undergo any treatment of the shoulder joint currently or in the past (40.6 ± 15.7 years, 16–81 year, male/female = 37/63). Group B consisted of 50 subjects (55.6 ± 11.5 years, 22–76 year, male/female = 17/33) with unilateral shoulder pathology (39 subacromial impingement, 9 rotator cuff tears, 2 other) confirmed by physical examination, x-rays, ultrasound and/or MR imaging. Exclusion criteria applied to the healthy and patient group were: anatomical abnormality (e.g. congenital, osteoarthritis), previous trauma and/or treatment (surgical or non-surgical) for shoulder joint pathology and a neuromuscular, musculoskeletal or systemic disorder (e.g. Parkinson’s disease, rheumatoid arthritis). Group C consisted of 25 subjects (50 shoulders, 45.2 ± 16.2 years, 21–78 year, male/female = 15/10) in whom 13 subjects (52%) did not have any shoulder complaints and 12 subjects with unilateral (n = 8, 32%) or bilateral (n = 4, 16%) affected shoulders of any pathology (e.g. subacromial impingement, rotator cuff tears, osteoarthritis). This group was included to investigate test–retest reliability of the IMA scores in subjects with different degree of impairment and different pathology. Subjects in groups A and B were recruited and measured during their visit at the orthopaedic outpatient department. Group C consists of subjects recruited at the inpatient orthopaedic ward during hospitalization, this allowed us to measure them on multiple occasions (by independent assessors) without the need for multiple visits to our outpatient department. Demographic data (i.e. length, weight, gender, date of birth, arm length, dominant side and shoulder pathology) was collected for each subject. In addition two of the most common questionnaire based shoulder questionnaires, the DASH (range 30–150 points, 30 points indicates unimpaired shoulder function) and SST (range 0–12, 12 points indicates unimpaired shoulder function) scores were used (Jester et al 2004, Lippitt et al 1992). 2.2. Equipment
Humeral accelerations and angular rates were measured using a wireless miniature inertial sensor (Inertia-Link-2400-SK1, MicroStrain, Inc., Williston, VT, USA, 41 × 63 × 24 mm, 39 g) containing a three-dimensional (3D) accelerometer ( ± 5 g) and gyroscope ( ± 300 ◦ s−1). The sensor was fixed by an adhesive patch onto the humerus in a standardized position on the distal and posterior part of the humerus, at the centre of the equilateral triangle formed by both condyles and the apex of the olecranon when the elbow is flexed 90◦ . This position minimizes movement of the skin and is also used in earlier studies (Coley et al 2007). The data is transmitted to a nearby portable computer (3DM-GX2 Software Development Kit). Power supply was performed by a small battery that was fixed at the anterior side of the forearm, not interfering with the arm movements (figure 1(A)). The distance between the lateral acromion and the centre of the sensor was measured in order to analyse the effect of arm length. To reproducibly indicate the point of task achievement (PTA) during specific functional motion tasks, an elastic belt with a central indicator was positioned cranial to the subjects’ sacrum (figure 1(B)). 169
¨ R J P Korver et al
Physiol. Meas. 35 (2014) 167
(A)
(B)
Figure 1. (A) Orientation of the 3D kinematics inertia sensor module and its fixation
on the subject’s humerus. (B) The two motion tasks: ‘arm to the back’ (left) and ‘arm behind the head’ (right).
2.3. Measurements
After inclusion, first all subjects completed the DASH and SST questionnaires. Subsequently, two functional motion tasks while sitting on a stable non-rotating stool, 90◦ of hip, knee and ankle flexion were performed. An inertial sensor was fixed by an adhesive patch onto the humerus in a standardized position on the distal and posterior part of the humerus as described above (figure 1(A)) and bilateral shoulder measurements were performed. Two functional (ADL based) motion tasks that are part of standard clinical questionnaires (i.e. CMS, DASH and SST) and have previously been used for IMA shoulder assessment were performed three times at self-selected speed (Coley et al 2007) (figure 1(B)). (1) Hand to the back (‘toilet hygiene’). Subjects started with the arm in the neutral anatomical position with their hand hanging beside their body and their thumb pointing to anterior (starting position). The PTA was reached when the hand palm touched the belt indicator. The position was held for 2 s before moving back to the starting position. (2) Hand behind the head (‘combing hair’). This tasks starts with the hand in the starting position, as mentioned above. The point of task achieved was reached when their hand palm was placed on the occiput with the elbow straight out to the side. After holding this position for 2 s the subjects were instructed to move their arm back to the starting position. The complete exercise was repeated with the same individual on the same day by the same assessor to investigate the test–retest reliability and on the same day with another assessor to investigate the inter-assessor reliability. 2.4. Data analysis
Specific algorithms, based on previous reports, were developed and used to analyse the raw inertia signal for estimating the relative difference in shoulder kinematics between both sides in healthy subjects and the affected and non-affected side in patients (Coley et al 2007). Two parameters, derived from the inertia signal, were used to calculate the asymmetry as the relative difference (%) between both sides. (1) COMP score: calculating and averaging the surface areas described by combining the angular rate signal and acceleration signal of each independent axis (figure 2(A)). This 170
¨ R J P Korver et al
Physiol. Meas. 35 (2014) 167
(A)
(B)
Figure 2. (A) COMP score, area described by combining the angular rate signal and
acceleration signal for healthy (left) and pathological (right) side. (B) ARS score, the peak-to-peak difference in the angular rate signal for the three axes.
score is the product of angular rate and acceleration and can be considered as the control of the humerus velocity by its acceleration (Coley et al 2007). Every graph was visually inspected for spurious peaks, which (especially in the acceleration signal) may influence the area described by the two signals. (2) ARS score: measuring and averaging the peak-to-peak difference in the angular rate signal for the three axes (figure 2(B)). The ARS score is based on angular rate only and represents the velocity of the humerus during ADL movements. For both asymmetry scores, the average of three repetitions was documented. Higher IMA asymmetry values indicate an increasing difference in shoulder function between both sides. 2.5. Statistical analysis
Intra- and inter-assessor reliability of IMA was calculated using two way random intra-class correlation coefficient (ICC) testing. Group comparisons (i.e. healthy versus pathological) were performed using the Student t-test after verification of normality (Kolmogorov-Smirnoff test). 171
¨ R J P Korver et al
Physiol. Meas. 35 (2014) 167
Table 1. Demographics of test subjects.
Group A N = 100 (♂ = 37) Group B N = 50 (♂ = 17) Group C N = 25 (♂ = 15)
Age (years)
Weight (kg)
Height (m)
BMI (kg m−2)
Arm length (cm)
40.6 ± 15.7 55.6 ± 11.5 45.2 ± 16.1
74.6 ± 15.2 73.4 ± 14.7 75.7 ± 13.3
1.72 ± 0.08 1.68 ± 0.09 1.75 ± 0.09
25.1 ± 4.4 26.0 ± 4.4 24.6 ± 3.1
30.0 ± 2.1 29.6 ± 1.6 29.9 ± 1.8
An analysis of variance (ANOVA) with Tukey’s post-hoc tests (HSD 0.05) were done to determine differences in demographics between the three test groups (groups A, B, and C). Correlations between the asymmetry scores (COMP, ARS) and subjects’ demographics (e.g. age, gender or body mass index) or standard clinical questionnaires (DASH, SST) were tested using Pearson’s R correlation. Diagnostic sensitivity and specificity of the IMA shoulder test to distinguish pathological from healthy shoulders were calculated based on asymmetry threshold values which were set after threshold optimizing to obtain high diagnostic power (figure 4). A possible effect of training or fatigue during the three repetitions was investigated by comparing scores of the subsequent individual repetitions using repeated measures ANOVA. The healthy shoulders of group B were compared to group A. Inter-group asymmetry was calculated by comparing mean COMP and ARS scores of the healthy shoulders (group A) to the non-affected shoulder of patients (group B) to study the effect on diagnostic power when measuring only the affected side and compare this to the healthy reference database. 3. Results No differences in demographics were found between the three groups (p > 0.05; table 1). The majority of the subjects (92%) were dominant with the right hand. Test duration was 0.05) in the IMA asymmetry scores between the measurements performed during the first and second observation by the same assessor and a third observation performed by a second assessor, indicating that shoulder movements were similar over time. The COMP and ARS scores showed a high repeatability (ICC = 0.94 and 0.95 respectively) and inter-assessor reliability (ICC = 0.90 and 0.91 respectively). The DASH and SST scores were also the same between different assessors (p > 0.38), with inter-assessor reliability significant lower compared to the IMA scores (ICC = 0.63 and 0.70 respectively; p < 0.05). 3.2. Reference data and diagnostic power
In general, healthy subjects (group A) had an average asymmetry of 14.6 ± 10.6%; 0.15– 44.7% (COMP score) and 9.6 ± 7.1%; 0.19–28.4% (ARS score). No influence of demographic data on the asymmetry scores could be found (p > 0.05). There were no differences in asymmetry between subsequent repetitions (p > 0.05) and hand dominance could not be distinguished (p > 0.48). 172
¨ R J P Korver et al
Physiol. Meas. 35 (2014) 167
Figure 3. Boxplot of asymmetry between healthy subjects and patients with a
pathological shoulder (∗∗ p < 0.05).
Figure 4. Diagnostic sensitivity and specificity for the ARS and COMP asymmetry
scores based on thresholds.
Patients had three–four times higher average asymmetry values (COMP: 44.2%, 20.6– 79.5%; ARS: 36.3%, 15.4–61.5%) compared to the healthy subjects (figure 3). The results of the clinical questionnaires showed a wide variety (SST: 6 ± 3; 1–12 and DASH: 51 ± 20; 0–79) with some values associated with a perfect healthy status. The healthy shoulders in group B and group A produced the same IMA values (p > 0.18). After threshold optimizing asymmetry thresholds were set for a pathological shoulder at COMP: >27% and ARS: >16%, healthy and pathological subjects could be distinguished with high diagnostic power (figure 4). The ARS score showed a higher sensitivity (98.0%; 95%-CI: 88.9–99.9%) and specificity (81.0%; 95%-CI: 71.7–87.9%) compared to the COMP score (84.0%; 95%-CI: 70.3–92.4% and 85.0; 95%-CI: 76.1–91.1% resp.). Inter-group comparison between a patient’s affected shoulder and the healthy reference database only slightly lowered the sensitivity (COMP: 78.0% versus ARS: 82.0%) and specificity (86.0% versus 85.0%). Only a weak correlation (Pearson R < 0.25) was found between the asymmetry scores and the clinical questionnaires (i.e. DASH and SST). Correlation between both clinical questionnaires is fair (Pearson R = 0.62) and between the two asymmetry scores good (Pearson R = 0.79). 173
¨ R J P Korver et al
Physiol. Meas. 35 (2014) 167
4. Discussion The aim of this study was first to establish the reliability of the IMA shoulder test in routine clinical practice. Secondly, steps toward clinical validity were performed by calculating the diagnostic power to distinguish healthy from pathological shoulder movement and by correlating the new asymmetry score to standard clinical scores. Previous studies on shoulder outcome assessment used questionnaires or lab-based techniques largely involving the entire upper extremity instead of the shoulder exclusively. These are less practical for routine clinical use and only little is known about the diagnostic power and reliability (Bey et al 2006, Bolink et al 2012, Garcia-Alsina et al 2005, Kirkley et al 2003, Klopcar and Lenarcic, 2006, Leggin et al 1996, Lin and Kuli´c 2012, Pitts et al 2013, Van Andel et al 2008, Veeger et al 2006). The IMA shoulder test and asymmetry score assesses relevant ADL movements and objectively quantifies them using the angular rate and acceleration signals. Due to its wireless design, ease of fixation, simplicity of both patient instructions and test operation (even by non-medical personal) and the short test time (84%, specificity >81%). exceeding the diagnostic power (sensitivity
Search
Collections
Journals
About
Contact us
My IOPscience
Inertia based functional scoring of the shoulder in clinical practice
This content has been downloaded from IOPscience. Please scroll down to see the full text. 2014 Physiol. Meas. 35 167 (http://iopscience.iop.org/0967-3334/35/2/167) View the table of contents for this issue, or go to the journal homepage for more
Download details: IP Address: 141.117.125.1 This content was downloaded on 25/05/2014 at 19:11
Please note that terms and conditions apply.
Institute of Physics and Engineering in Medicine Physiol. Meas. 35 (2014) 167–176
Physiological Measurement
doi:10.1088/0967-3334/35/2/167
Inertia based functional scoring of the shoulder in clinical practice 1 ¨ R J P Korver , I C Heyligers, S K Samijo and B Grimm
AHORSE Research Foundation, Department of Orthopaedic Surgery and Traumatology, Atrium Medical Center Parkstad Heerlen, Henri Dunantstraat 5, 6419 PC Heerlen, The Netherlands E-mail: [email protected], [email protected], [email protected] and [email protected] Received 2 October 2013, revised 24 November 2013 Accepted for publication 3 December 2013 Published 7 January 2014 Abstract
Shoulder-related dysfunction is the second most common musculoskeletal disorder and is responsible for an increasing burden on health-care systems. Commonly used clinical outcome scores suffer from subjectivity, pain dominance and a ceiling effect. Objective functional measurement has been identified as a relevant issue in clinical rehabilitation. In recognition of this goal simple techniques for routine clinical application have been investigated with some success. Inertia based motion analysis (IMA) is a new generation of objective outcome assessment tool; it can produce objective movement parameters while being fast, cheap and easy to operate. This study investigates if a simple IMA shoulder test is suitable as a functional outcome measure for routine clinical follow-up. We measured 100 healthy subjects and 50 patients with confirmed unilateral shoulder pathology. Two motion tasks were performed on both shoulders and two simple motion parameters based on angular rate and acceleration were calculated. Patients were also assessed by the disability of arm, shoulder and hand (DASH) and the simple shoulder test. IMA produced high intra- (ICC = 0.94) and inter-assessor reliability (ICC = 0.90). Asymmetry was >3 times higher in patients than in healthy controls (p < 0.01). Healthy and pathological subjects could be distinguished with high diagnostic sensitivity (>84.0%) and specificity (>81.0%). There was a weak correlation between the IMA shoulder score and the clinical questionnaires (Pearson R < 0.25), as it may add an objective functional dimension to outcome assessment. The fast assessment (t < 5 min) of a simple motion task makes it workable for routine clinical follow-up. The IMA shoulder test adds objective information on functional capacity to the clinical scores and may help the physician in his 1
Author to whom any correspondence should be addressed.
0967-3334/14/020167+10$33.00
© 2014 Institute of Physics and Engineering in Medicine Printed in the UK
167
¨ R J P Korver et al
Physiol. Meas. 35 (2014) 167
decision-making, follow-up of treatment, effect of training and possibly lead to the development of new therapeutic interventions. Keywords: shoulder, outcome assessment, inertia sensors, healthy reference data, clinical validation, diagnostic power, clinical scores (Some figures may appear in colour only in the online journal)
1. Introduction Shoulder-related dysfunction with limited range of motion (ROM) and restricted activities of daily living (ADL) is a common health problem with a 34% prevalence in subjects >65 years of age (Chakravarty and Webley 1990). In clinical practice, outcome can be assessed by a large variety of questionnaire based scores of which the disability of arm, shoulder and hand (DASH) score and the simple shoulder test (SST) are the most commonly used (Jester et al 2004, Lippitt et al 1992). However, poor reliability has been reported (Kirkley et al 2003). Even for the Constant–Murley score (CMS), which adds an objective component (e.g. ROM) to its questionnaire poor reliability is documented (Conboy et al 1996, Constant et al 2008, Rocourt et al 2008). In general, questionnaires suffer from subjectivity, a ceiling effect and are highly painrelated, often masking function-related results (Coley et al 2007, Lin et al 2005). In addition, since the conception of the scores, patients characteristics and their demands have shifted toward younger patients being treated earlier in the disease development and claiming for higher functional improvement after treatment (Nilsdotter et al 2009). Rising health-economic demands require objective evidence of new treatment effects in clinical rehabilitation, which the questionnaires can hardly provide (Lansky and Milstein 2009). In research settings, advanced motion analysis systems are used for quantitative movement analysis using opto-electronic systems, force plates or electromyography (EMG) (Bey et al 2006, Bolink et al 2012, Garcia-Alsina et al 2005, Klopcar and Lenarcic 2006, Leggin et al 1996, Lin and Kuli´c 2012, Pitts et al 2013, Van Andel et al 2008, Veeger et al 2006). Although these methods can provide objective evaluation of shoulder movements, they are less practical for routine clinical use because they require technically skilled personnel, are time-consuming, relatively expensive and need a laboratory that may cause unnatural movements. In addition, kinematic analysis of the upper extremity has not received as much scientific attention as that of the lower limb. Last decade, a major contribution as potential alternative for lab-based motion analysis has been made in the development of inertia based motion analysis (IMA) (Coley et al 2007, Cutti et al 2008, Parel et al 2012, de Vries et al 2010). Previous studies used accelerometers and gyroscopes to measure the overall shoulder movement at the arm during ADL tasks, developed motion parameters and validated these parameters against opto-electronic systems (Coley et al 2007). IMA is a feasible method for describing shoulder kinematics during ADL tasks and can serve as an activity monitor (Coley et al 2007, 2008b, Parel et al 2012). However, these studies were performed in small groups, including heterogeneous pathologies (e.g. rotator cuff disease, osteoarthritis and postoperative shoulder arthroplasty). Furthermore, these studies did not investigate the diagnostics power of IMA and its correlation to clinical questionnaires, which makes the level of clinical validation, is still weak. The aim of this study is, (A) to clinically validate this score by testing its reliability and diagnostic power in distinguishing healthy from pathologic shoulders and (B) compare this IMA shoulder score against standard clinical questionnaires. 168
¨ R J P Korver et al
Physiol. Meas. 35 (2014) 167
2. Methods 2.1. Subjects
A total of 175 subjects were recruited and bilateral shoulder measurement was performed according to a protocol approved by our hospital’s Institutional Ethics Board Committee. Group A consisted of 100 healthy subjects without shoulder complaints, who did not undergo any treatment of the shoulder joint currently or in the past (40.6 ± 15.7 years, 16–81 year, male/female = 37/63). Group B consisted of 50 subjects (55.6 ± 11.5 years, 22–76 year, male/female = 17/33) with unilateral shoulder pathology (39 subacromial impingement, 9 rotator cuff tears, 2 other) confirmed by physical examination, x-rays, ultrasound and/or MR imaging. Exclusion criteria applied to the healthy and patient group were: anatomical abnormality (e.g. congenital, osteoarthritis), previous trauma and/or treatment (surgical or non-surgical) for shoulder joint pathology and a neuromuscular, musculoskeletal or systemic disorder (e.g. Parkinson’s disease, rheumatoid arthritis). Group C consisted of 25 subjects (50 shoulders, 45.2 ± 16.2 years, 21–78 year, male/female = 15/10) in whom 13 subjects (52%) did not have any shoulder complaints and 12 subjects with unilateral (n = 8, 32%) or bilateral (n = 4, 16%) affected shoulders of any pathology (e.g. subacromial impingement, rotator cuff tears, osteoarthritis). This group was included to investigate test–retest reliability of the IMA scores in subjects with different degree of impairment and different pathology. Subjects in groups A and B were recruited and measured during their visit at the orthopaedic outpatient department. Group C consists of subjects recruited at the inpatient orthopaedic ward during hospitalization, this allowed us to measure them on multiple occasions (by independent assessors) without the need for multiple visits to our outpatient department. Demographic data (i.e. length, weight, gender, date of birth, arm length, dominant side and shoulder pathology) was collected for each subject. In addition two of the most common questionnaire based shoulder questionnaires, the DASH (range 30–150 points, 30 points indicates unimpaired shoulder function) and SST (range 0–12, 12 points indicates unimpaired shoulder function) scores were used (Jester et al 2004, Lippitt et al 1992). 2.2. Equipment
Humeral accelerations and angular rates were measured using a wireless miniature inertial sensor (Inertia-Link-2400-SK1, MicroStrain, Inc., Williston, VT, USA, 41 × 63 × 24 mm, 39 g) containing a three-dimensional (3D) accelerometer ( ± 5 g) and gyroscope ( ± 300 ◦ s−1). The sensor was fixed by an adhesive patch onto the humerus in a standardized position on the distal and posterior part of the humerus, at the centre of the equilateral triangle formed by both condyles and the apex of the olecranon when the elbow is flexed 90◦ . This position minimizes movement of the skin and is also used in earlier studies (Coley et al 2007). The data is transmitted to a nearby portable computer (3DM-GX2 Software Development Kit). Power supply was performed by a small battery that was fixed at the anterior side of the forearm, not interfering with the arm movements (figure 1(A)). The distance between the lateral acromion and the centre of the sensor was measured in order to analyse the effect of arm length. To reproducibly indicate the point of task achievement (PTA) during specific functional motion tasks, an elastic belt with a central indicator was positioned cranial to the subjects’ sacrum (figure 1(B)). 169
¨ R J P Korver et al
Physiol. Meas. 35 (2014) 167
(A)
(B)
Figure 1. (A) Orientation of the 3D kinematics inertia sensor module and its fixation
on the subject’s humerus. (B) The two motion tasks: ‘arm to the back’ (left) and ‘arm behind the head’ (right).
2.3. Measurements
After inclusion, first all subjects completed the DASH and SST questionnaires. Subsequently, two functional motion tasks while sitting on a stable non-rotating stool, 90◦ of hip, knee and ankle flexion were performed. An inertial sensor was fixed by an adhesive patch onto the humerus in a standardized position on the distal and posterior part of the humerus as described above (figure 1(A)) and bilateral shoulder measurements were performed. Two functional (ADL based) motion tasks that are part of standard clinical questionnaires (i.e. CMS, DASH and SST) and have previously been used for IMA shoulder assessment were performed three times at self-selected speed (Coley et al 2007) (figure 1(B)). (1) Hand to the back (‘toilet hygiene’). Subjects started with the arm in the neutral anatomical position with their hand hanging beside their body and their thumb pointing to anterior (starting position). The PTA was reached when the hand palm touched the belt indicator. The position was held for 2 s before moving back to the starting position. (2) Hand behind the head (‘combing hair’). This tasks starts with the hand in the starting position, as mentioned above. The point of task achieved was reached when their hand palm was placed on the occiput with the elbow straight out to the side. After holding this position for 2 s the subjects were instructed to move their arm back to the starting position. The complete exercise was repeated with the same individual on the same day by the same assessor to investigate the test–retest reliability and on the same day with another assessor to investigate the inter-assessor reliability. 2.4. Data analysis
Specific algorithms, based on previous reports, were developed and used to analyse the raw inertia signal for estimating the relative difference in shoulder kinematics between both sides in healthy subjects and the affected and non-affected side in patients (Coley et al 2007). Two parameters, derived from the inertia signal, were used to calculate the asymmetry as the relative difference (%) between both sides. (1) COMP score: calculating and averaging the surface areas described by combining the angular rate signal and acceleration signal of each independent axis (figure 2(A)). This 170
¨ R J P Korver et al
Physiol. Meas. 35 (2014) 167
(A)
(B)
Figure 2. (A) COMP score, area described by combining the angular rate signal and
acceleration signal for healthy (left) and pathological (right) side. (B) ARS score, the peak-to-peak difference in the angular rate signal for the three axes.
score is the product of angular rate and acceleration and can be considered as the control of the humerus velocity by its acceleration (Coley et al 2007). Every graph was visually inspected for spurious peaks, which (especially in the acceleration signal) may influence the area described by the two signals. (2) ARS score: measuring and averaging the peak-to-peak difference in the angular rate signal for the three axes (figure 2(B)). The ARS score is based on angular rate only and represents the velocity of the humerus during ADL movements. For both asymmetry scores, the average of three repetitions was documented. Higher IMA asymmetry values indicate an increasing difference in shoulder function between both sides. 2.5. Statistical analysis
Intra- and inter-assessor reliability of IMA was calculated using two way random intra-class correlation coefficient (ICC) testing. Group comparisons (i.e. healthy versus pathological) were performed using the Student t-test after verification of normality (Kolmogorov-Smirnoff test). 171
¨ R J P Korver et al
Physiol. Meas. 35 (2014) 167
Table 1. Demographics of test subjects.
Group A N = 100 (♂ = 37) Group B N = 50 (♂ = 17) Group C N = 25 (♂ = 15)
Age (years)
Weight (kg)
Height (m)
BMI (kg m−2)
Arm length (cm)
40.6 ± 15.7 55.6 ± 11.5 45.2 ± 16.1
74.6 ± 15.2 73.4 ± 14.7 75.7 ± 13.3
1.72 ± 0.08 1.68 ± 0.09 1.75 ± 0.09
25.1 ± 4.4 26.0 ± 4.4 24.6 ± 3.1
30.0 ± 2.1 29.6 ± 1.6 29.9 ± 1.8
An analysis of variance (ANOVA) with Tukey’s post-hoc tests (HSD 0.05) were done to determine differences in demographics between the three test groups (groups A, B, and C). Correlations between the asymmetry scores (COMP, ARS) and subjects’ demographics (e.g. age, gender or body mass index) or standard clinical questionnaires (DASH, SST) were tested using Pearson’s R correlation. Diagnostic sensitivity and specificity of the IMA shoulder test to distinguish pathological from healthy shoulders were calculated based on asymmetry threshold values which were set after threshold optimizing to obtain high diagnostic power (figure 4). A possible effect of training or fatigue during the three repetitions was investigated by comparing scores of the subsequent individual repetitions using repeated measures ANOVA. The healthy shoulders of group B were compared to group A. Inter-group asymmetry was calculated by comparing mean COMP and ARS scores of the healthy shoulders (group A) to the non-affected shoulder of patients (group B) to study the effect on diagnostic power when measuring only the affected side and compare this to the healthy reference database. 3. Results No differences in demographics were found between the three groups (p > 0.05; table 1). The majority of the subjects (92%) were dominant with the right hand. Test duration was 0.05) in the IMA asymmetry scores between the measurements performed during the first and second observation by the same assessor and a third observation performed by a second assessor, indicating that shoulder movements were similar over time. The COMP and ARS scores showed a high repeatability (ICC = 0.94 and 0.95 respectively) and inter-assessor reliability (ICC = 0.90 and 0.91 respectively). The DASH and SST scores were also the same between different assessors (p > 0.38), with inter-assessor reliability significant lower compared to the IMA scores (ICC = 0.63 and 0.70 respectively; p < 0.05). 3.2. Reference data and diagnostic power
In general, healthy subjects (group A) had an average asymmetry of 14.6 ± 10.6%; 0.15– 44.7% (COMP score) and 9.6 ± 7.1%; 0.19–28.4% (ARS score). No influence of demographic data on the asymmetry scores could be found (p > 0.05). There were no differences in asymmetry between subsequent repetitions (p > 0.05) and hand dominance could not be distinguished (p > 0.48). 172
¨ R J P Korver et al
Physiol. Meas. 35 (2014) 167
Figure 3. Boxplot of asymmetry between healthy subjects and patients with a
pathological shoulder (∗∗ p < 0.05).
Figure 4. Diagnostic sensitivity and specificity for the ARS and COMP asymmetry
scores based on thresholds.
Patients had three–four times higher average asymmetry values (COMP: 44.2%, 20.6– 79.5%; ARS: 36.3%, 15.4–61.5%) compared to the healthy subjects (figure 3). The results of the clinical questionnaires showed a wide variety (SST: 6 ± 3; 1–12 and DASH: 51 ± 20; 0–79) with some values associated with a perfect healthy status. The healthy shoulders in group B and group A produced the same IMA values (p > 0.18). After threshold optimizing asymmetry thresholds were set for a pathological shoulder at COMP: >27% and ARS: >16%, healthy and pathological subjects could be distinguished with high diagnostic power (figure 4). The ARS score showed a higher sensitivity (98.0%; 95%-CI: 88.9–99.9%) and specificity (81.0%; 95%-CI: 71.7–87.9%) compared to the COMP score (84.0%; 95%-CI: 70.3–92.4% and 85.0; 95%-CI: 76.1–91.1% resp.). Inter-group comparison between a patient’s affected shoulder and the healthy reference database only slightly lowered the sensitivity (COMP: 78.0% versus ARS: 82.0%) and specificity (86.0% versus 85.0%). Only a weak correlation (Pearson R < 0.25) was found between the asymmetry scores and the clinical questionnaires (i.e. DASH and SST). Correlation between both clinical questionnaires is fair (Pearson R = 0.62) and between the two asymmetry scores good (Pearson R = 0.79). 173
¨ R J P Korver et al
Physiol. Meas. 35 (2014) 167
4. Discussion The aim of this study was first to establish the reliability of the IMA shoulder test in routine clinical practice. Secondly, steps toward clinical validity were performed by calculating the diagnostic power to distinguish healthy from pathological shoulder movement and by correlating the new asymmetry score to standard clinical scores. Previous studies on shoulder outcome assessment used questionnaires or lab-based techniques largely involving the entire upper extremity instead of the shoulder exclusively. These are less practical for routine clinical use and only little is known about the diagnostic power and reliability (Bey et al 2006, Bolink et al 2012, Garcia-Alsina et al 2005, Kirkley et al 2003, Klopcar and Lenarcic, 2006, Leggin et al 1996, Lin and Kuli´c 2012, Pitts et al 2013, Van Andel et al 2008, Veeger et al 2006). The IMA shoulder test and asymmetry score assesses relevant ADL movements and objectively quantifies them using the angular rate and acceleration signals. Due to its wireless design, ease of fixation, simplicity of both patient instructions and test operation (even by non-medical personal) and the short test time (84%, specificity >81%). exceeding the diagnostic power (sensitivity
