Kumar, P., Mardon, M., Bradley, M., Gray, S. and Swinkels, A.(2014) Assessment of glenohumeral subluxation in post-stroke hemi-plegia: Comparison between ultrasound and fingerbreadth palpationmethods. Physical Therapy, 94 (11). pp. 1622-1631. ISSN 0031-9023Available from: http://eprints.uwe.ac.uk/24970

We recommend you cite the published version.The publisher’s URL is:http://dx.doi.org/10.2522/ptj.20130303

Refereed: No

This article is Copyright ?2014? American Physical Therapy Association. Arti-cle on publisher’s website available at http://dx.doi.org/10.2522/ptj.20130303

Disclaimer

UWE has obtained warranties from all depositors as to their title in the materialdeposited and as to their right to deposit such material.

UWE makes no representation or warranties of commercial utility, title, or fit-ness for a particular purpose or any other warranty, express or implied in respectof any material deposited.

UWE makes no representation that the use of the materials will not infringeany patent, copyright, trademark or other property or proprietary rights.

UWE accepts no liability for any infringement of intellectual property rightsin any material deposited but will remove such material from public view pend-ing investigation in the event of an allegation of any such infringement.

PLEASE SCROLL DOWN FOR TEXT.

doi: 10.2522/ptj.20130303Originally published online July 24, 2014

2014; 94:1622-1631.PHYS THER. Selena Gray and Annette SwinkelsPraveen Kumar, Marianne Mardon, Michael Bradley,Ultrasound and Fingerbreadth Palpation MethodsPoststroke Hemiplegia: Comparison Between Assessment of Glenohumeral Subluxation in

http://ptjournal.apta.org/content/94/11/1622found online at: The online version of this article, along with updated information and services, can be

Online-Only Material 30303.DC1.html

http://ptjournal.apta.org/content/suppl/2014/10/24/ptj.201

Collections

Tests and Measurements     Stroke (Neurology)    

Stroke (Geriatrics)     Injuries and Conditions: Shoulder    

Hemiplegia/Paraplegia/Quadriplegia     in the following collection(s): This article, along with others on similar topics, appears

e-Letters

"Responses" in the online version of this article. "Submit a response" in the right-hand menu under

or click onhere To submit an e-Letter on this article, click

E-mail alerts to receive free e-mail alerts hereSign up

by Journals librarian on February 11, 2015http://ptjournal.apta.org/Downloaded from by Journals librarian on February 11, 2015http://ptjournal.apta.org/Downloaded from

Assessment of GlenohumeralSubluxation in Poststroke Hemiplegia:Comparison Between Ultrasound andFingerbreadth Palpation MethodsPraveen Kumar, Marianne Mardon, Michael Bradley, Selena Gray,Annette Swinkels

Background. Glenohumeral subluxation (GHS) is a common poststroke compli-cation. Treatment of GHS is hampered by the lack of objective, real-time clinicalmeasurements.

Objective. The aims of this study were: (1) to compare an ultrasound method ofGHS measurement with the fingerbreadth palpation method using a receiver oper-ating characteristic curve (ROC) and (2) to report the sensitivity and specificity of thismethod.

Design. A prospective study was conducted.

Setting. The study was conducted in local hospitals and day centers in thesouthwest of England.

Patients. One hundred five patients who had one-sided weakness following afirst-time stroke (51 men, 54 women; mean age�71 years, SD�11) and who gaveinformed consent were enrolled in the study.

Measurements. Ultrasound measurements of acromion–greater tuberosity(AGT) distance were used for the assessment of GHS. Measurements were under-taken on both shoulders by a research physical therapist trained in shoulder ultra-sound with the patient seated in a standardized position. Fingerbreadth palpationassessment of GHS was undertaken by a clinical physical therapist based at thehospital, who also visited the day centers.

Results. The area under the ROC curve was 0.73 (95% confidence interval [95%CI]�0.63, 0.83), suggesting that the ultrasound method has good agreement com-pared with the fingerbreadth palpation method. A cutoff point of �0.2 cm AGTmeasurement difference between affected and unaffected shoulders generated asensitivity of 68% (95% CI�51%, 75%), a specificity of 62% (95% CI�47%, 80%), apositive likelihood ratio of 1.79 (95% CI�1.1, 2.9), and a negative likelihood ratio of0.55 (95% CI�0.4, 0.8).

Limitations. Clinical therapists involved in the routine care of patients conductedthe fingerbreadth palpation method. It is likely that they were aware of the patients’subluxation status.

Conclusion. The ultrasound method can detect minor asymmetry (�0.5 cm) andhas the potential advantage over the fingerbreadth palpation method of identifyingpatients with minor subluxation.

P. Kumar, PhD, MCSP, MIAP,MSPA, Department of AlliedHealth Professions, Faculty ofHealth and Life Sciences, Univer-sity of the West of England, Room1K05, Glenside Campus, Black-berry Hill, Stapleton, Bristol, BS161DD, United Kingdom. Address allcorrespondence to Dr Kumar at:Praveen.Kumar@uwe.ac.uk.

M. Mardon, BSc, Department ofPhysiotherapy, University HospitalBristol, Bristol, United Kingdom.

M. Bradley, MD, Department ofRadiology, Southmead Hospital,North Bristol NHS Trust, Bristol,United Kingdom.

S. Gray, MBChB, MD, Centre forClinical and Health ServicesResearch, University of the West ofEngland.

A. Swinkels, PhD, Department ofPhysiotherapy, University of theWest of England.

[Kumar P, Mardon M, Bradley M,et al. Assessment of glenohumeralsubluxation in poststroke hemi-plegia: comparison between ultra-sound and fingerbreadth palpa-tion methods. Phys Ther.2014;94:1622–1631.]

© 2014 American Physical TherapyAssociation

Published Ahead of Print:July 24, 2014

Accepted: July 4, 2014Submitted: July 15, 2013

Research Report

Post a Rapid Response tothis article at:ptjournal.apta.org

1622 f Physical Therapy Volume 94 Number 11 November 2014 by Journals librarian on February 11, 2015http://ptjournal.apta.org/Downloaded from

Glenohumeral subluxation(GHS) is a recognized compli-cation in people with post-

stroke hemiplegia. The reported inci-dence of GHS ranges from 17% to81% of patients, depending on themeasurement methods used and thetime frames over which it isassessed.1,2 Severe loss of motorfunction and apparent absence ofsupraspinatus muscle contractionare potential risk factors for GHS, butscapular orientation does not con-tribute to GHS, as was originallythought.3 The association betweenGHS and other poststroke complica-tions such as pain and poor motorrecovery is uncertain. When presentin combination, however, thesecomplications could have a signifi-cant impact on upper limb func-tion.4 The management of GHS,therefore, is an important therapeu-tic goal, and various approacheshave been used in its prevention andtreatment.5,6 Current approacheshave significant problems and limita-tions to their use, and the effective-ness of any one of these approachesfor the treatment of GHS is inconclu-sive.7 A potential reason for thesefindings is the lack of reliable, objec-tive, real-time clinical measure-ments.8 Current clinical measure-ments include the fingerbreadthpalpation method9 and plainradiographs.6,10

The fingerbreadth palpation methodlacks the sensitivity to detect earlysigns of GHS or minor subluxations.8

There is a concern that without treat-ment, subluxation can progress to anuncorrectable level over time.6 EarlyGHS can contribute to irreversiblepartial or complete tears of the non-elastic shoulder capsule.5,6,11 Radio-graphs are considered to be objec-tive and have high reliability andvalidity,12 but problems relating tocost, time involved, and risks inher-ent to exposure to radiation13 limittheir utility in the clinical setting. Inaddition, radiographic diagnosis is

not generally recommended for clin-ical evaluation of GHS.14

Diagnostic ultrasound is now rou-tinely used for clinical imaging of theshoulder region in patients withmusculoskeletal conditions.15–18

Recently, several studies used diag-nostic ultrasound to evaluate theincidence and prevalence of soft tis-sue injuries (rotator cuff tears,biceps tendinitis) in the shoulders ofpeople with poststroke hemiple-gia.19–26 The ultrasound method iscurrently being investigated anddeveloped for the assessment of GHSin these patients27,28; however, it isnot routinely used in clinical set-tings. Using a large, static ultrasoundmachine, Park et al27 reported highintrarater reliability (intraclass corre-lation coefficient [ICC]�.979) ofultrasound measurements of GHS.More recently, Kumar et al28

recruited 26 patients with strokeand, using a new standardized posi-tion with the forearm supported,found that bedside assessment ofacromion–greater tuberosity (AGT)distance, undertaken by a physicaltherapist trained in shoulder ultra-sound, demonstrated good intraraterreliability (ICC�.980) and discrimi-nant validity.

The purposes of this study were: (1)to compare ultrasound and finger-breadth palpation methods of GHSmeasurements using a receiver oper-ating characteristic (ROC) curve and(2) to report the sensitivity and spec-ificity of these methods. The finger-breadth palpation method is rou-tinely used in clinical practice andhas been tested for both reliabilityand validity.29–31 Hall et al29 investi-gated the concurrent validity of thismethod by comparing it with plainradiographs. They reported a Spear-man correlation coefficient of .760between the fingerbreadth palpationmethod and plain radiographs. Thisstudy continued this research by

comparing ultrasound and finger-breadth palpation methods.

MethodParticipantsThe study used a prospective designand received approval from theNational Health Service (NHS)Research Ethics Committee, NorthBristol Trust, United Kingdom.Patients over 50 years of age whohad stroke resulting in one-sidedweakness and who were able to situpright were eligible to participate.Patients with aphasia also were eligi-ble to participate in the study. Apha-sia was confirmed if a patient haddifficulty following simple com-mands, understanding questions(receptive aphasia), or speaking(expressive aphasia). Diagnosis orpresence of GHS was not a require-ment to be able to participate in thestudy. Patients with other neurologicconditions, traumatic brain injury,brain tumors or other serious comor-bidities, shoulder pathology, orrecent surgery to the neck, arm, orshoulder; those who were unavail-able for testing; and those who wereunable to volunteer due to any rea-son were excluded.

An a priori power calculation wasperformed for assessing the clinicalutility of the ultrasound method asquantified by the area under thereceiver operating characteristic(AUROC) curve. To our knowledge,this is the first study of this topicusing AUROC curve statistics. There-fore, power calculations were con-ducted for 2 AUROC curve values.For standard level of significance

Available WithThis Article atptjournal.apta.org

• eTable: AUROC Curve StatisticsWith 95% CI for UltrasoundMethod Using Optimal CutoffPoint

Comparison of Ultrasound and Fingerbreadth Palpation Methods

November 2014 Volume 94 Number 11 Physical Therapy f 1623 by Journals librarian on February 11, 2015http://ptjournal.apta.org/Downloaded from

(��.05, ��.20), a minimum samplesize of n�72 and n�114 would haveat least 80% power to determine sta-tistical significance if the trueAUROC was equal to 0.70 and 0.65,respectively, assuming a 1:1 ratiobetween negative and positive casesin the sample (calculations were per-formed using MedCalc Software, ver-sion 11.1, Mariakerke, Belgium).Therefore, the aim of this study wasto recruit up to 114 patients withstroke.

Patients were recruited from 4 localhospital trusts in the southwest ofEngland and from the community byaccessing the Bristol Area StrokeFoundation (BASF), a voluntary orga-nization that operates social clubs ina number of day centers for patientswith stroke in Bristol. Of the severalBASF social clubs, 6 centers locatedin and around the Bristol area wereapproached for the recruitment ofpatients. Each patient gave informedwritten consent to take part, and, forthose who lacked mental capacity,appropriate procedures were fol-lowed and involved a family membersigning the personal consultee agree-ment form in the presence of thepatient.

Apparatus and RatersPrior to commencement of the datacollection process, a portable diag-nostic ultrasound machine (TITANmodel, L38/10-5 MHz broadband,Sonosite Ltd, Hitchin, United King-dom) was tested and calibratedaccording to the manufacturer’sguidelines.

Ultrasound measurements of AGTdistance were undertaken by a phys-ical therapist (P.K.) at all of theresearch sites (hospital and day cen-ters). The training protocol con-sisted of a 1-day manufacturer’scourse, supervised training from aconsultant radiologist, pilot work on6 healthy volunteers, and reliabilitystudies on healthy volunteers32 andpatients with stroke (n�64).28

Clinical assessment of GHS (usingthe fingerbreadth palpation method)was performed by one of the seniorclinical physical therapists (NHSbands 6–8) at each local hospitaltrust and at the day centers. Sevenphysical therapists with 4 to 15 yearsof experience in stroke rehabilita-tion were involved with clinicalassessments of GHS. To ensure stan-dardization and familiarization withthe testing procedure, each physical

therapist practiced the standardizedprotocol on 2 patients with stroke inthe presence of the chief researcher(P.K.). Any issues arising were dis-cussed and clarified at this stage.During actual data collection, physi-cal therapists undertook measure-ments independently.

ProcedureBaseline demographic data, includ-ing age, sex, date of onset, type ofstroke, site of stroke, and sideaffected, were collected frompatients’ medical records by thechief researcher. For patients at daycenters, only age, sex, and date ofstroke were gathered directly fromthe patients, as no medical recordswere available. Assessments wereconducted at the hospital bedside orin the day centers. The therapistundertaking clinical assessment ofGHS was blinded to ultrasound mea-surements of AGT distance, and thetherapist undertaking ultrasoundmeasurements was blinded to clini-cal assessment. The order of data col-lection was as follows.

Clinical assessment of GHS by aclinical physical therapist usingthe fingerbreadth palpationmethod. A standardized protocolwas used.29 Patients were seated in achair or wheelchair with both feetflat on the ground or on a footrest.The physical therapist first assessedthe unaffected side to palpate thegap between the acromion and thehead of the humerus, and this assess-ment was repeated on the affectedshoulder. Shoulders were positionedin neutral rotation, with the armhanging by the side (thumb pointingforward) close to the body with noabduction (Fig. 1). Some patientswho demonstrated high tone wereunable to hang their affected armfreely by the side. For these patients,the shoulder was maintained in inter-nal rotation with slight elbow flexionand the forearm resting on their lap.Glenohumeral subluxation was

Figure 1.Participants’ standardized position for data collection: (A) fingerbreadth palpationmethod, (B) ultrasound method.

Comparison of Ultrasound and Fingerbreadth Palpation Methods

1624 f Physical Therapy Volume 94 Number 11 November 2014 by Journals librarian on February 11, 2015http://ptjournal.apta.org/Downloaded from

defined as a palpable gap betweenthe inferior aspect of the acromionand the superior aspect of thehumeral head that is 1⁄2 fingerbreadthor more. A 0–5 grading scheme wasused: 0�no subluxation, 1�1⁄2 fin-gerbreadth gap, 2�1 fingerbreadthgap, 3�11⁄2 fingerbreadth gap, 4�2fingerbreadth gap, and 5�21⁄2 finger-breadth gap.29

Ultrasound measurements ofAGT distance by the chiefresearcher. For ultrasound mea-surements of AGT distance, eachpatient was placed in the standard-ized position to allow measurementof AGT distance (Fig. 1).32 The shoul-der was in neutral rotation, with theelbow at 90 degrees of flexion andthe forearm in pronation. The fore-arms rested on a pillow placed onthe patient’s lap with the elbow jointitself remaining unsupported. Assis-tance was provided by theresearcher if the patient was unableto move the arm. The ultrasoundtransducer then was placed over thelateral border of acromion alongthe vertical/longitudinal axis of thehumerus to scan the shoulder. TheAGT distance was recorded on thefrozen image using an on-screen cal-iper that automatically calculates dis-tances (Fig. 2). The AGT distancewas defined as the relative lateral dis-tance between the lateral edge of theacromial process of the scapula andthe nearest margin of the superiorpart of the greater tuberosity of thehumerus.32 A dark linear acousticshadow beneath the acromionhelped to identify the lateral edge ofthe acromion. The supraspinatustendon was clearly visible as a thickband (acoustic hyperechoic appear-ance) at its point of insertion, whichfacilitated identification of thegreater tuberosity (Fig 2). Threeultrasound images of the right shoul-der were obtained, and AGT distancewas measured on each image. Thisprocess was repeated on the leftshoulder. In order to ensure the rater

was blind to measurements, the val-ues displayed were obscured byplacing a sticker on the ultrasoundscreen.

A general neurological clinicalexamination of the upper limb bythe chief researcher. The generalneurological examination includedassessment of muscle strength in theshoulder muscles (Medical ResearchCouncil Scale)33 and muscle tone34,35

on both affected and unaffectedsides. Muscle tone was classified aslow (grade 0), normal (grade 1), andhigh (grades 2–5), as described byCulham et al.35 For both musclestrength and tone, the shoulder flex-ors, abductors, and internal andexternal rotators were assessed.

Data AnalysisData were analyzed using SPSS (ver-sion 19.0, IBM UK, Business Analyt-ics, Middlesex, United Kingdom).Descriptive statistics were used tocalculate the mean and standarddeviation of AGT distance measure-ments for both affected and unaf-fected shoulders. The differencebetween affected and unaffected

shoulders was considered a measureof GHS based on the ultrasoundmethod and was analyzed usingrepeated-measures analysis of vari-ance (ANOVA), and both sides(affected and unaffected) and timewere considered as within-subjectfactors. The standard error of mea-surement (SEM) was calculated fromthe ANOVA output. The minimumdetectable change with 90% confi-dence interval (MDC90) was calcu-lated using the formula: MDC90�SEM � 1.65 � �2.36,37

The association between the finger-breadth palpation method (differ-ence between affected and unaf-fected shoulders) and the ultrasoundmethod was tested using Spearmanrank correlation coefficients. Thisstatistical test is used when one ofthe methods (in this case, the finger-breadth palpation method) gener-ates ordinal data rather than intervalor ratio data. Agreement betweenthe ultrasound and fingerbreadth pal-pation methods was tested using theROC curve, the AUROC curve, sen-sitivity, specificity, negative and pos-itive predictive values, and likeli-

Figure 2.Measurement of acromion–greater tuberosity (AGT) distance between the lateral bor-der of the acromion (AC) and the nearest superior margin of the greater tuberosity (GT).Dotted caliper represents AGT distance. Sup�supraspinatus muscle.

Comparison of Ultrasound and Fingerbreadth Palpation Methods

November 2014 Volume 94 Number 11 Physical Therapy f 1625 by Journals librarian on February 11, 2015http://ptjournal.apta.org/Downloaded from

hood ratios for different values ofultrasound measurements of AGTdistance.

Role of the Funding SourceThis research was undertaken as partof Dr Kumar’s doctoral thesis, whichwas funded by the University of theWest of England, Bristol, UnitedKingdom.

ResultsOver a 16-month period, 115patients with stroke wereapproached to participate in thestudy. Ten patients were excludedbecause they did not meet the inclu-sion criteria. Of these patients, 3 hadserious comorbidities (intestinal can-cer, heart problems), 1 was enrolled

but withdrew before finishing datacollection, 1 was discharged fromthe hospital prior to data collection,and 5 could not visit the day centeron the day of data collection becauseof personal reasons. Therefore, 105patients agreed to participate andwere enrolled in the study: 70patients were from hospital settings,and 35 patients were from stroke daycenters. Of the recruited patients, 22(21%) had aphasia. Seven patientsrequired alternative positioning (anonstandard modified sitting posi-tion) due to the presence of hightone.

A summary of the demographic char-acteristics of the participants isshown in Table 1. The mean AGT

distance for the total sample(n�105) was 2.2 cm (SD�0.6) and1.8 cm (SD�0.4) for the stroke-affected and stroke-unaffected shoul-ders, respectively. On the stroke-affected side, the minimum andmaximum AGT values recordedacross patients were 1.0 and 3.7 cm,respectively, and the 95% confi-dence intervals (95% CIs) rangedfrom 2.0 to 2.3 cm. Correspondingvalues for the unaffected shoulderwere 0.7, 3.2, and 1.7 to 1.9 cm. Therepeated-measures ANOVA showeda significant mean AGT differencebetween affected and unaffectedshoulder measurements (X�0.4 cm,SD�0.5) (F5,520�53.101, P�.001).The SEM for the between-shoulderdifference in AGT was 0.08 cm, andthe MDC90 was �0.2 cm.

Shoulder subluxation was present in71 patients (67%) and absent in 34patients (33%) using the finger-breadth palpation method of assess-ment. Of those with GHS, 31/71(44%) had grade 1 (1⁄2-finger gap),28/71 (39%) had grade 2 (1-fingergap), 8/71 (11%) had grade 3 (11⁄2-finger gap), and 4/71 (6%) had grade4 (2-finger gap) subluxation.

The Spearman rank correlation coef-ficients showed a moderate correla-tion (rs�.52) between the 2 meth-ods, and this correlation wasstatistically significant (P�.001). TheROC curve allows seeing, in a simplevisual display, how sensitivity andspecificity vary around different cut-off points (curved line) (Fig. 3). TheAUROC curve can have any valuebetween 0 and 1, and a test could beregarded as excellent or not usefulbased on the following categories:0.9–1.0 (excellent), 0.8–0.9 (verygood), 0.7–0.8 (good), 0.6–0.7 (suf-ficient), 0.5–0.6 (bad), and �0.5(test has no diagnostic value).38,39 Ifthe AUROC curve value is 0.9 to 1.0(ie, closer to the upper left-hand cor-ner of the ROC curve), it demon-strates excellent agreement between

Table 1.Demographic Characteristics of Patients With Stroke (n�105)

Variable Measurements

Age (y)

X 71

SD 11

Range 50–90

Sex, n (%)

Male 51 (48)

Female 54 (52)

Type of stroke, n (%)

Cerebral infarction 66 (62)

Intracerebral hemorrhage 10 (9)

Unspecified 29 (29)

Side affected, n (%)

Right 51 (48)

Left 54 (52)

Aphasia, n (%) 22 (21)

Muscle strength42 (shoulder flexors, abductors, rotators), n (%)

�3 79 (75)

�4 26 (25)

Muscle tone44 (shoulder flexors, abductors, rotators), n (%)

Low 42 (40)

Normal 40 (38)

High 23 (22)

Time since onset of stroke (wk)

Median 5.6

Range 0.4–728

Comparison of Ultrasound and Fingerbreadth Palpation Methods

1626 f Physical Therapy Volume 94 Number 11 November 2014 by Journals librarian on February 11, 2015http://ptjournal.apta.org/Downloaded from

the tests. In contrast, if the value is�0.5 (ie, on or below the straightline), it suggests that there is pooragreement between the tests39

(Fig. 3). The AUROC curve was 0.73(95% CI�0.63, 0.83). Based on theAUROC curve, there was a goodlevel of agreement between theultrasound and fingerbreadth palpa-tion methods.

Conventionally, on an ROC graph, apair of diagnostic sensitivity andspecificity values for every individualcutoff is plotted, with the sensitivityon the y-axis and 1 minus specificityon the x-axis. The sensitivity andspecificity for various cutoff pointsare presented in Table 2. A cutoffpoint of �0.2 cm AGT measurementdifference between affected andunaffected shoulders could be con-sidered optimal because it providesthe best trade-off between sensitivityand specificity, with a sensitivity of68% (95% CI�51%, 75%) and a spec-ificity of 62% (95% CI�47%, 80%).Using this cutoff point, the true valuefor the sensitivity of the ultrasoundmethod is likely to be between 0.51(the lower boundary of the CI), 0.68(the point estimate), and 0.75 (theupper boundary of the CI).

Using the optimal cutoff point of�0.2 cm, the usefulness of the ultra-sound method is illustrated in theeTable (available at ptjournal.apta.org). Likelihood ratios summarizehow many times more or less likelypatients with subluxation are to havea particular test result than patientswithout subluxation. The positivelikelihood ratio of 1.79 suggeststhat a patient with subluxation(defined as grade of 1 or higher inthe 5-point fingerbreadth palpationmethod) is 1.79 more likely to havean AGT difference greater than0.2 cm on ultrasound than a patientwithout palpable subluxation. Theflow diagram presented in Figure 4illustrates the comparison of ultra-sound method versus fingerbreadth

palpation method. Using the ultra-sound method, 61/105 patients(58%) had a mean AGT differenceof �0.2 cm between the affectedand unaffected shoulders. Of thosepatients with �0.2 cm AGT distance,33/61 (54%) demonstrated a meanAGT difference of between 0.2 and0.5 cm, indicating minor asymmetrybetween the unaffected and affectedshoulders.

DiscussionThe primary aims of this study were:(1) to compare an ultrasoundmethod of GHS measurement withthe fingerbreadth palpation methodusing an ROC curve and (2) to reportthe sensitivity and specificity ofthese methods. The AUROC curvefrom this study was 0.73. Presentedwith pairs of randomly selectedpatients, one with GHS and one

Table 2.Sensitivity and Specificity With 95% CI for Ultrasound Method (AGT DistanceMeasurement Difference Between Affected and Unaffected Shoulders)a

CutoffPoint Sensitivity 95% CI Specificity 95% CI

�0.5 40% 28%, 52% 89% 73%, 96%

�0.4 47% 33%, 57% 83% 69%, 95%

�0.3 55% 39%, 63% 74% 62%, 91%

�0.2 68% 51%, 75% 62% 47%, 80%

�0.1 76% 57%, 80% 50% 38%, 73%

a 95% CI�95% confidence interval, AGT�acromion–greater tuberosity.

Figure 3.Receiver operating characteristic (ROC) curve with area under the receiver operatingcharacteristic (AUROC) curve value of 0.73 (95% confidence interval�0.63–0.83).Curved line shows sensitivity and specificity varied around different cutoff points.Straight line indicates AUROC curve �0.5 (test not useful). Cutoff points �0.1 to �0.5cm correspond to sensitivity and specificity. Optimal cutoff point of �0.2 cm indicates68% sensitivity and 62% specificity. Diagonal segments produced by ties.

Comparison of Ultrasound and Fingerbreadth Palpation Methods

November 2014 Volume 94 Number 11 Physical Therapy f 1627 by Journals librarian on February 11, 2015http://ptjournal.apta.org/Downloaded from

without GHS as determined by thefingerbreadth palpation method, anexaminer would classify 73% of thepairs correctly by choosing the onewhose AGT distance on ultrasoundwas the larger of the two.

For the diagnostic ultrasoundmethod to be useful, it is importantto select a trade-off between sensitiv-ity and specificity.40 The cutoff pointof �0.5 cm generated a sensitivity of40% (eTable). Examiners who applya cutoff point of 0.5 cm while usingthe ultrasound method would fail toidentify 6/10 patients judged by thefingerbreadth palpation method tohave GHS. In contrast, a cutoff pointof �0.1 cm generates a sensitivity of76%, indicating that the ultrasoundmethod would identify 8/10 patientswith subluxation. However, thisvalue is associated with a low speci-ficity of 50%, indicating that 5/10patients whose ultrasound measuresare asymmetrical by �0.1 cm ormore would have no evidence ofGHS on the fingerbreadth palpationtest. Unlike these cutoff points,which generate high sensitivity andlow specificity or vice versa, a cutoffpoint of �0.2 cm generates a sensi-tivity of 68% and a specificity of 62%.Based on the sensitivity statistic,when the fingerbreadth palpationtest indicates GHS, in 68% of thosecases, the ultrasound measure alsowould indicate GHS.

The cutoff point of �0.2 cm (wherethe sensitivity is 0.68) may be con-sidered optimal because it helps to“rule out” GHS by indicating that,among patients with subluxation(defined as a grade of 1 or higher onthe 5-point palpation scale), 68% willhave an AGT difference of at least 0.2cm. This finding suggests that ultra-sound potentially could be used as ascreening tool, which is criticalbecause early diagnosis of GHSwould facilitate the application ofappropriate treatment and therebypotentially prevent the long-term

complications associated with GHS.Furthermore, a cutoff point of �0.2cm (where the test’s specificity is0.62) indicates that, with a predictedfalse-positive result of 0.38, amongthose who demonstrate no sublux-ation (a grade of 0 on the 5-pointfingerbreadth palpation scale), 62%will have an AGT difference of lessthan 0.2 cm. Specificity is equallyimportant because applying treat-ment such as positioning (armtroughs, lap boards), shoulder slings,or strapping to a patient withoutGHS could reduce the normal gapbetween the acromion and the headof the humerus. This position couldalter the normal scapulohumeralrhythm required for smooth move-ment at the shoulder joint resultingin compression of the rotator cufftendons under the acromion pro-cess, which can cause tearing ofthese structures and result in sub-acromial impingement.19

The cutoff point of �0.2 cm alsocoincides with the MDC90 value of�0.2 cm, which is in agreementwith the findings of a previousstudy.28 Kumar et al,28 in a study of26 patients with stroke, reported amean AGT difference of 0.4 cm andan MDC90 value of �0.2 cm betweenaffected and unaffected shoulders. Astudy of healthy individuals (n�32;mean age 64 years, SD�11) showeda mean AGT difference of 0.1 cm(SD�0.18) (95% CI�0.03, 0.17) andMDC90 value of �0.07 cm betweenright and left shoulders.32 Based onthe MDC90 values from these studies,it could be predicted that a changeof �0.2 cm in AGT distance mea-surements between affected andunaffected shoulders would be nec-essary to indicate an asymmetry thatis not due to measurement error.

In this study, a mean AGT differenceof �0.5 cm between affected andunaffected shoulders was observedin 33 patients, suggesting minor sub-luxation. It is critical to identify

minor subluxation in its early stage,as application of appropriate treat-ment can improve upper limb motorfunction.5,6 Several studies havereported on the benefits of func-tional electrical stimulation in theprevention and treatment of GHS inearly stages of rehabilitation5,41–43

but not in patients with chronicstroke (�6 months).6 Findings fromthese studies suggest that GHS canbe prevented by the application ofappropriate treatment but that with-drawal of treatment can lead to sub-sequent subluxation, especially inpatients with loss of voluntary con-trol. In the United Kingdom, there-fore, the latest national guidelines forstroke44 recommend application offunctional electrical stimulation tothe supraspinatus and deltoid mus-cles for any patient with stroke whohas developed, or is at risk of devel-oping, GHS. Ultrasound has thepotential advantage of identifyingpatients with even minor sublux-ation (�0.5 cm) and can provideobjective measurements in the earlystages of rehabilitation.

In contrast, the fingerbreadth palpa-tion method has the potential advan-tage of being a quick, equipment-free method of identifying significantsubluxation. However, it lacks theability to detect early signs of sublux-ation,8 is subjective,8,12,31 and isinsensitive, as it cannot detect differ-ences of less than 0.5 cm.12 Further-more, the reported correlations forthe concurrent validity of the finger-breadth palpation method in com-parison with radiographic measure-ments range from .69 to .76,29–31

which are described as relativelylow.45 Limitations of the finger-breadth palpation method couldresult in an underestimation of thetrue prevalence of GHS, which couldcontribute to the moderate correla-tion and relatively low sensitivity andspecificity values for the ultrasoundmethod found in this study. Due toresource, cost, and ethical con-

Comparison of Ultrasound and Fingerbreadth Palpation Methods

1628 f Physical Therapy Volume 94 Number 11 November 2014 by Journals librarian on February 11, 2015http://ptjournal.apta.org/Downloaded from

straints, it was not possible to under-take radiographs of 210 shoulders.

Our study suggests that ultrasoundmeasurements of AGT have potentialvalue in the prevention and manage-ment of GHS in people with stroke.The technique is safe, noninva-sive,15,16,46,47 allows real-time mea-surements,48 and requires limitedtraining to produce reliable measure-ments of AGT distance.28,32,49,50 Sev-eral other benefits of diagnosticultrasound have been reported byrecent studies of people withstroke.25,26 A recent study demon-strated that subluxation occurredmore frequently in patients (n�182)with a known presence of fluid inthe subhumeral and subdeltoid bur-sae and in patients with reduced

functional capacity.25 Ultrasoundwas used both as a diagnostic tooland to monitor the effectiveness ofthe exercise program targetingreduction of subluxation and bursalfluid. Similarly, another study inves-tigated the association between GHSand soft tissue injuries in 39 peoplewith stroke.26 Similar to our study,the diagnosis of GHS was done bymeasuring the lateral AGT distance.The study showed that ultrasoundcomplements the assessment of softtissue injuries in shoulders of peoplewith stroke.26

Given these findings, ultrasound haspotential usefulness in both researchand clinical practice. Clinically, ultra-sound may be used to assess andmonitor the effectiveness of treat-

ment interventions for GHS in peo-ple with severe paralysis, especiallyduring the early stage of rehabilita-tion. It also has potential to diagnosesoft tissue injuries in people withstroke, both with and without GHS,and thus can facilitate managementof shoulder pain. In particular, it hasutility as an outcome measure inintervention studies. The ultrasoundmethod is objective, quantitative,and has the potential to detect evensmall changes in AGT distancemeasurements.

The current study had some limita-tions. First, there was a difference inthe patients’ starting position for the2 methods. For the fingerbreadth pal-pation method, patients were in anupright sitting position with their

Recruited patients(N=115)

Excluded patients (n=10)

3–serious comorbidities1–withdrew prior to data collection1–discharged from hospital5–could not visit the day center on the day of data collection

Enrolled(n=105)

Fingerbreadth Palpation Method

GHS present(n=71)

Fingerbreadth Palpation Method

GHS absent (n=34)

Inconclusive (n=0)

Inconclusive (n=0)

Ultrasound MethodPresent(n=48)

Ultrasound MethodAbsent(n=23)

Ultrasound MethodPresent(n=13)

Ultrasound MethodAbsent (n=21)

Figure 4.Flow diagram illustrating comparison between the ultrasound method and the fingerbreadth palpation method based on cutoff pointof �0.2 cm acromion-greater tuberosity difference. GHS�glenohumeral subluxation.

Comparison of Ultrasound and Fingerbreadth Palpation Methods

November 2014 Volume 94 Number 11 Physical Therapy f 1629 by Journals librarian on February 11, 2015http://ptjournal.apta.org/Downloaded from

arms hanging freely by their sidesand without arm support. In con-trast, for the ultrasound method, thepatients’ forearm was placed in theirlap. Patients with loss of motor con-trol are potentially at risk of develop-ing GHS, and gravity-dependent posi-tions of the shoulder, therefore,should be avoided.3 Kumar et al28

developed a new standardized posi-tion with the forearm supported on apillow, but the elbow itself remainedunsupported to allow the effect ofgravity. The study demonstratedexcellent intrarater reliability anddiscriminant validity for the ultra-sound method, suggesting it has theability to diagnose GHS even whenthe forearm is supported. It was notpossible to use this newly developedposition for the fingerbreadth palpa-tion method because there are noreports on reliability and validity forthis method in this position. Second,patients were recruited from multi-ple sites, which meant 7 differentphysical therapists were involved inthe assessment of GHS using the fin-gerbreadth palpation method. Thereported interrater reliability (ICC)for the fingerbreadth palpationmethod is between .770 and .792.12

This reliability might have contrib-uted to greater variability and mighthave had some effect on the cor-relation between fingerbreadth pal-pation and ultrasound methods.Finally, the assessors conducting thefingerbreadth palpation method alsowere the clinical therapists involvedin the routine care of study patientsat the hospital. It is likely, therefore,that they were aware of the patients’subluxation status, which mighthave influenced their judgment onthe day of data collection. However,patients (n�35) at the day centerwere not known to the clinicaltherapists.

In conclusion, on the basis of theAUROC curve value of 0.73, thisstudy demonstrated a good level ofagreement between the ultrasound

and fingerbreadth palpation meth-ods. The ROC curve findings fromthis study indicate that a cutoff pointof �0.2 cm could be used to deter-mine the sensitivity and specificity ofthe ultrasound method to identifyasymmetry between affected andunaffected shoulders and facilitatediagnosis of GHS. The ultrasoundmethod has the potential advantageover the fingerbreadth palpationmethod of identifying patients witheven minor subluxation (�0.5 cm).Ultrasound measurements provideratio-level data and have clinical util-ity as an outcome measure in inter-vention studies. Future studiesshould investigate the diagnosticaccuracy of the ultrasound methodin the assessment of GHS in compar-ison with radiographic measure-ments. From an ethical perspective,investigation of diagnostic accuracyof the ultrasound method could beincorporated into intervention stud-ies that routinely use radiographs toevaluate outcomes.

All authors provided concept/idea/researchdesign and consultation (including review ofmanuscript before submission). Dr Kumarprovided writing and data analysis. DrKumar and Ms Mardon provided data col-lection, facilities/equipment, and institu-tional liaisons. Dr Kumar and Dr Swinkelsprovided project management. Ms Mardonprovided study participants.

The authors thank all participants, SonositeLimited, Hitchin, United Kingdom (loan ofultrasound equipment), the manager andsupport staff at the Bristol Area Stroke Foun-dation (help with recruitment), and Dr PaulWhite and Dr Jon Pollock (statistical help).Special thanks to the following physical ther-apists (clinical investigators) for help withrecruitment and data collection: Carol Jen-kins, Karen-Martyn Jones, Paul Cunningham(University Hospital Bristol), Fiona Henchie(Weston General Hospital), and Lucy Holland Rebecca Wedeman (Chippenham Com-munity Hospital).

This research was undertaken as part of DrKumar’s doctoral thesis, which was fundedby the University of the West of England,Bristol, United Kingdom.

DOI: 10.2522/ptj.20130303

References1 Vuagnat H, Chantraine A. Shoulder pain in

hemiplegia revisited: contribution of func-tional electrical stimulation and other ther-apies. J Rehabil Med. 2003;35:49–54.

2 Ada L, Foongchomcheay A, Canning C.Supportive devices for preventing andtreating subluxation of the shoulder afterstroke. Cochrane Database Syst Rev.2009;1:CD003863.

3 Kumar P, Kassam J, Denton C, et al. Sys-tematic review: risk factors for inferiorshoulder subluxation in patients withstroke. Phys Ther Rev. 2010;15:3–11.

4 Kumar P, Swinkels A. A critical review ofshoulder subluxation and its associationwith other post-stroke complications.Phys Ther Rev. 2009;14:13–25.

5 Linn SL, Granat MH, Lees KR. Preventionof shoulder subluxation after stroke withelectrical stimulation. Stroke. 1999;30:963–968.

6 Wang RY, Yang YR, Tsai MW, et al. Effectsof functional electric stimulation on upperlimb motor function and shoulder range ofmotion in hemiplegic patients. Am J PhysMed Rehabil. 2002;81:283–290.

7 Stolzenberg D, Siu G, Cruz E. Current andfuture interventions for glenohumeral sub-luxation in hemiplegia secondary tostroke. Top Stroke Rehabil. 2012;19:444–456.

8 Yu D, Chae J. Neuromuscular stimulationfor treating shoulder dysfunction in hemi-plegia. Crit Rev Phys Rehabil Med. 2002;14:1–23.

9 Paci M, Nannetti L, Taiti P, et al. Shouldersubluxation after stroke: relationshipswith pain and motor recovery. PhysiotherRes Int. 2007;12:37–50.

10 van Langenberghe HV, Hogan BM. Degreeof pain and grade of subluxation in thepainful hemiplegic shoulder. Scand JRehabil Med. 1988;20:161–166.

11 Miglietta O, Lewitan A, Rogoff JB. Sublux-ation of the shoulder in hemiplegicpatients. N Y State J Med. 1959;59:457–460.

12 Boyd EA, Torrance GM. Clinical measuresof shoulder subluxation: their reliability.Can J Public Health. 1992;83(suppl)2:S24–S28.

13 Berrington de Gonzalez A, Darby S. Risk ofcancer from diagnostic X-rays: estimatesfor the countries. Lancet. 2004;363:345–351.

14 Gamble GE, Barberan E, Laasch HU, et al.Poststroke shoulder pain: a prospectivestudy of the association and risk factors in152 patients from a consecutive cohort of205 patients presenting with stroke. Eur JPain. 2002;6:467–474.

15 Hides J, Richardson C, Jull G, Davies S.Ultrasound imaging in rehabilitation. AustJ Physiother. 1995;41:187–193.

16 Chang CY, Wang SF, Chiou HJ, et al. Com-parison of shoulder ultrasound and MRimaging in diagnosing full-thickness rota-tor cuff tears. Clin Imaging. 2002;26:50–54.

Comparison of Ultrasound and Fingerbreadth Palpation Methods

1630 f Physical Therapy Volume 94 Number 11 November 2014 by Journals librarian on February 11, 2015http://ptjournal.apta.org/Downloaded from

17 Schibany N, Zehetgruber H, Kainberger F,et al. Rotator cuff tears in asymptomaticindividuals: a clinical and ultrasonographicscreening study. Eur J Radiol. 2004;51:263–268.

18 Lento PH, Primack S. Advances and utilityof diagnostic ultrasound in musculoskele-tal medicine. Curr Rev MusculoskeletMed. 2008;1:24–31.

19 Ikai T, Tei K, Yoshida K, et al. Evaluationand treatment of shoulder subluxation inhemiplegia: relationship between sublux-ation and pain. Am J Phys Med Rehabil.1998;77:421–426.

20 Aras MD, Gokkaya NK, Comert D, et al.Shoulder pain in hemiplegia: results from anational rehabilitation hospital in Turkey.Am J Phys Med Rehabil. 2004;83:713–719.

21 Lee IS, Shin YB, Moon TY, et al. Sonogra-phy of patients with hemiplegic shoulderpain after stroke: correlation with motorrecovery stage. AJR Am J Roentgenol.2009;192:W40–W44.

22 Pong YP, Wang LY, Wang L, et al. Sonog-raphy of the shoulder in hemiplegicpatients undergoing rehabilitation after arecent stroke. J Clin Ultrasound. 2009;37:199–205.

23 Huang YC, Liang PJ, Pong YP, et al. Phys-ical findings and sonography of hemiple-gic shoulder in patients after acute strokeduring rehabilitation. J Rehabil Med.2010;42:21–26.

24 Pompa A, Clemenzi A, Troisi E, et al.Enhanced-MRI and ultrasound evaluationof painful shoulder in patients after stroke:a pilot study. Eur Neurol. 2011;66:175–181.

25 Pop T. Subluxation of the shoulder joint instroke patients and the influence ofselected factors on the incidence of insta-bility. Ortop Traumatol Rehabil. 2013;15:259–267.

26 Huang SW, Liu SY, Tang HW, et al. Rela-tionship between severity of shoulder sub-luxation and soft-tissue injury in hemiple-gic stroke patients. J Rehabil Med. 2012;44:733–739.

27 Park GY, Kim JM, Sohn SI, et al. Ultra-sound measurement of shoulder sublux-ation in patients with post-stroke hemiple-gia. J Rehabil Med. 2007;39:526–530.

28 Kumar P, Bradley M, Gray S, Swinkels A.Reliability and validity of ultrasound mea-surement of acromion-greater tuberositydistance in post-stroke hemiplegia. ArchPhys Med Rehabil. 2011;92:731–736.

29 Hall J, Dudgeon B, Guthrie M. Validity ofclinical measures of shoulder subluxationin adults with post stroke hemiplegia.Am J Occup Ther. 1995;49:526–533.

30 Boyd EA, Goudreau L, O’Riain MD, et al. Aradiological measure of shoulder sublux-ation in hemiplegia: Its reliability andvalidity. Arch Phys Med Rehabil. 1993;74:188–193.

31 Prevost R, Arsenault AB, Dutil E, Drouin G.Shoulder subluxation in hemiplegia: aradiologic correlational study. Arch PhysMed Rehabil. 1987;68:782–785.

32 Kumar P, Bradley M, Swinkels A. Within-day and between day intra-rater reliabilityof ultrasound measurements of acromion-greater tuberosity distance in healthy peo-ple. Physiother Theory Pract. 2010;26:347–351.

33 Medical Research Council. Aids to Exam-ination of the Peripheral Nervous Sys-tem. London, United Kingdom: Her Majes-ty’s Stationary Office; 1976. MemorandumNo. 45.

34 Bohannon RW, Smith MB. Interrater reli-ability of a modified Ashworth scale ofmuscle spasticity. Phys Ther. 1987;67:206–207.

35 Culham EG, Noce RR, Bagg SD. Shouldercomplex position and glenohumeral sub-luxation in hemiplegia. Arch Phys MedRehabil. 1995;76:857–864.

36 Haley SM, Fragala-Pinkham MA. Interpret-ing change scores of tests and measuresused in physical therapy. Phys Ther. 2006;86:735–743.

37 Kolber MJ, Saltzman SB, Beekhuizen KS,Cheng MS. Reliability and minimal detect-able change of inclinometric shouldermobility measurements. Physiother The-ory Pract. 2009;25:572–581.

38 Selvin S. Statistical Tools for Epidemio-logic Research. Oxford, United Kingdom:Oxford University Press; 2001.

39 Ana-Maria S. Measures of diagnostic accu-racy: basic definitions. Med Biological Sci.2008;22:61–65.

40 de Jesus JO, Parker L, Frangos AJ, NazarianLN. Accuracy of MRI, MR arthrography,and ultrasound in the diagnosis of rotatorcuff tears: a meta-analysis. AJR Am J Roent-genol. 2009;192:1701–1707.

41 Baker LL, Parker K. Neuromuscular elec-trical stimulation of the muscles surround-ing the shoulder. Phys Ther. 1986;66:1930–1937.

42 Faghri PD, Rodgers MM, Glaser RM, et al.The effects of functional electrical stimu-lation on shoulder subluxation, arm func-tion recovery, and shoulder pain in hemi-plegic stroke patients. Arch Phys MedRehabil. 1994;75:73–79.

43 Wang RY, Chan RC, Tsai MW. Functionalelectrical stimulation on chronic and acutehemiplegic shoulder subluxation. Am JPhys Med Rehabil. 2000;79:385–390.

44 Royal College of Physicians. National clin-ical guideline for stroke. Available at:http://www.rcplondon.ac.uk/sites/default/files/national-clinical-guidelines-for-stroke-fourth-edition.pdf. PublishedSeptember 2012. Accessed January 15,2014.

45 Dunn WW. Validity. In: Miller LJ, ed.Developing Norm-Referenced Standard-ized Tests. New York, NY: The HaworthPress; 1989:149–168.

46 Ardic F, Kahraman Y, Kacar M, et al. Shoul-der impingement syndrome: relationshipsbetween clinical, functional, and radio-logic findings. Am J Phys Med Rehabil.2006;85:53–60.

47 Awerbuch M. The clinical utility of ultra-sonography for rotator cuff disease, shoul-der impingement syndrome and subacro-mial bursitis. Med J Aust. 2008;188:50–53.

48 Stokes M, Hides J, Nassiri D. Musculoskel-etal ultrasound imaging: diagnostic andtreatment aid in rehabilitation. Phys TherRev. 1997;2:73–92.

49 Kumar P, Bourke C, Flanders J, et al. Theeffect of arm position on the ultrasono-graphic measurements of the acromion-greater tuberosity distance. PhysiotherTheory Pract. 2014;30:171–177.

50 Kumar P, Chetwynd J, Evans A, et al. Inter-rater and intrarater reliability of ultrasono-graphic measurements of acromion-greater tuberosity distance in healthypeople. Physiother Theory Pract. 2011;27:172–175.

Comparison of Ultrasound and Fingerbreadth Palpation Methods

November 2014 Volume 94 Number 11 Physical Therapy f 1631 by Journals librarian on February 11, 2015http://ptjournal.apta.org/Downloaded from

doi: 10.2522/ptj.20130303Originally published online July 24, 2014

2014; 94:1622-1631.PHYS THER. Selena Gray and Annette SwinkelsPraveen Kumar, Marianne Mardon, Michael Bradley,Ultrasound and Fingerbreadth Palpation MethodsPoststroke Hemiplegia: Comparison Between Assessment of Glenohumeral Subluxation in

References

http://ptjournal.apta.org/content/94/11/1622#BIBLfor free at: This article cites 46 articles, 4 of which you can access

Information Subscription http://ptjournal.apta.org/subscriptions/

Permissions and Reprints http://ptjournal.apta.org/site/misc/terms.xhtml

Information for Authors http://ptjournal.apta.org/site/misc/ifora.xhtml

by Journals librarian on February 11, 2015http://ptjournal.apta.org/Downloaded from