Reliability and Validity of Static Knee Strength Measurements Obtained With a Chair-Fixed...

O

ROHDM

BkmM

aoc(

so

i

h

iwl

tlmmcmspka

kkrv

ltt

0

so

Re

1998

A

RIGINAL ARTICLE

eliability and Validity of Static Knee Strength Measurementsbtained With a Chair-Fixed Dynamometer in Subjects Withip or Knee Arthroplasty

any Gagnon, MSc, Sylvie Nadeau, PhD, Denis Gravel, PhD, Josée Robert, DEC, Dominique Bélanger, BSc,
olly Hilsenrath, BSc
v

cR

Okoccpph3wd

otirbetbsds

iffecetuof(ctsste

mge

ABSTRACT. Gagnon D, Nadeau S, Gravel D, Robert J,élanger D, Hilsenrath M. Reliability and validity of staticnee strength measurements obtained with a chair-fixed dyna-ometer in subjects with hip or knee arthroplasty. Arch Physed Rehabil 2005;86:1998-2008.

Objectives: To evaluate the reliability (intertrial, interevalu-tor) and the concurrent validity of strength measurementsbtained with a chair-fixed dynamometer and to recommend alinical protocol that minimizes standard error of measurementSEM).

Design: Within-session repeated measures of maximal statictrength of knee flexors and extensors at 30° and 60° of flexionn the chair-fixed and Cybex dynamometers.Setting: Ambulatory physiotherapy department of a rehabil-

tation hospital.Participants: Convenience sample of 50 subjects with total

ip (n�25) or knee (n�25) arthroplasty.Interventions: Not applicable.Main Outcome Measures: Reliability was quantified by

ndices of dependability and corresponding SEMs estimatedith the generalizability theory, whereas coefficients of corre-

ation were used to assess the validity.Results: Indices of dependability confirmed excellent inter-

rial (0.98–1.00) and a very good interevaluator (.92–.99) re-iability for the strength measures obtained for different move-

ents and positions. An average of 3 trials minimized theagnitude of the SEMs (�2Nm for all measurements). When

omparing the strength values obtained with the Cybex dyna-ometer to those measured with the chair-fixed dynamometer,

trongest relations were attained when the tested knee wasositioned at 60° compared with 30° for subjects with a totalnee (.78–.92 vs .87–.93) or hip (.73–.85 vs .86–.91)rthroplasty.

Conclusions: A clinical protocol averaging 3 trials with thenee positioned at 60° after a familiarization period, both fornee flexors and extensors, performed by a trained therapist isecommended to minimize measurement errors on strengthalues measured with the chair-fixed dynamometer.

From the Hôpital de Réadaptation Lindsay, Montréal, QC (Gagnon, Robert, Bé-anger, Hilsenrath); Centre de Recherche Interdisciplinaire en Réadaptation du Mon-réal Métropolitain, Montréal, QC (Gagnon, Nadeau, Gravel); and École de Réadap-ation, Université de Montréal, Montréal, QC (Gagnon, Nadeau, Gravel), Canada.

Supported by the Fondation de l’Hôpital de Réadaptation Lindsay (grant no.4-2004).No commercial party having a direct financial interest in the results of the research

upporting this article has or will confer a benefit upon the authors(s) or upon anyrganization with which the author(s) is/are associated.Reprint request to Dany Gagnon, MSc, Services de Réadaptation, Hôpital de

éadaptation Lindsay, 6363 chemin Hudson, Montréal, QC H3S 1M9, Canada,-mail: [email protected].

a0003-9993/05/8610-9781$30.00/0doi:10.1016/j.apmr.2005.04.013

rch Phys Med Rehabil Vol 86, October 2005

Key Words: Arthroplasty; Rehabilitation; Reliability andalidity.© 2005 by the American Congress of Rehabilitation Medi-

ine and the American Academy of Physical Medicine andehabilitation

STEOARTHRITIS IS THE LEADING cause of long-termdisability in Canada.1 The annual rates of elective hip and

nee arthroplasty have increased by 17.9% and 50.6% in Canadaver the past few years, respectively.1 This rate is expected toontinue to grow because the population over 65 years of ageonstitutes one of the fastest growing segments of the Canadianopulation. Of all these orthopedic procedures, 66% are currentlyerformed on individuals over 65 years of age.1 The rate of totalip (THA) and knee arthroplasty (TKA) increased surprisingly by0% and 90%, respectively, among people under 55 years of ageithin the past 6 years, whereas the revision rate rose by 10%uring that same period.1

It is well recognized that people who undergo elective THAr TKA benefit from rehabilitation services particularly duringhe immediate postoperative period.2 It is believed that earlynitiation of intensive functional rehabilitation may result inapid attainment of an optimal level of function.3-5 The reha-ilitation treatment regimen generally includes therapeutic ex-rcises to increase range of motion, to strengthen muscles, ando enhance balance and functional training related to bed mo-ility, transfer abilities, gait, and stair activities.2,6 As a result,everity of impairment and disability decreases progressivelyuring this period, ensuring a high rate of (re)integration intoociety for these clients.7

Precise quantification of motor performance is of majormportance during the rehabilitation process of these patientsor clinical evaluation, decision making, and prediction ofunctional capacity. Because strength appears to be a cruciallement in the realization of many functional activities, espe-ially among patients over 60 years of age, it is relevant tovaluate this motor component after TKA and THA.8,9 Al-hough manual muscle testing (MMT) remains the most widelysed clinical method for categorizing muscle strength using anrdinal scale,10,11 numerous authors have confirmed that staticorce measurements obtained with a handheld dynamometerHHD) are more reliable as well as more responsive to detecthange over time than traditional MMT.12-15 Furthermore, in-ra- and interevaluator reliability of the HHD has been widelytudied in both disabled and able populations.16 However,tandard error of measurement (SEM) was rarely reported inhese studies despite the value of this indicator of measurementrror in the appreciation of changes over time.17,18

Improvement of the reliability of handheld force measure-ents is still a concern especially when assessing muscle

roups producing high-force levels.19 Because knee flexors andxtensors are among the strongest muscles of the human body,
n alternative procedure should be explored to ensure reliable

aumtsatmndtH(cs

ckdascvamao

(mcwtflasnT

P

ioeads

atajssapbtp

E

tebttwmdtpi

E

sssttTfaTfugmiatwrwMtomcptue

aCdlma

A

1999STRENGTH MEASUREMENTS WITH A CHAIR-FIXED DYNAMOMETER, Gagnon

nd valid strength evaluations. The strength of the dominantpper extremity of the evaluator (maximal vertical push force)ay limit the extent of weakness that can be detected because

he evaluator is minimally required to accommodate for thetrength generated by the patient assessed (make test).17,20-22 Inddition, the evaluator may also encounter difficulty in at-empting to stabilize both the HHD perpendicular to the seg-ent being tested and the segment of the patient to simulta-

eously avoid muscle substitutions and standardize joint angleuring the test.23 To this effect, it has been suggested recentlyhat the use of external mechanical support to stabilize theHD (belt method) or to fix the HHD to a rigid segment

platform method) may represent a modification that can over-ome some of these issues related to the upper-extremity testertrength.24-27

Testing angle is critical in strength measurements becausehanges in mechanical advantages and muscle lengths arenown to influence strength values.8 Moreover, knee angleuring strength testing is relevant for other reasons such asrthrogenous muscle inhibition,28 variable contribution of pas-ive resistance of the joint, and anterior tibial shear forces thathange with knee angles.8 All these factors can increase theariance across subjects in interaction with angles that mayffect the reliability of the measurements. In this context, forceeasurements done on joint directly affected by the pathology

nd the surgery may be less reliable than force measurementsbtained on adjacent joints.The first objective of our study was to evaluate the reliability

intertrial, interevaluator) of knee flexors and extensors strengtheasurements obtained at 2 angles with a newly developed

hair-fixed dynamometer among people who recently under-ent a TKA or a THA. The second objective of our study was

o assess the validity and the associated agreement of the kneeexors and extensors strength measurements obtained at 2ngles with the chair-fixed dynamometer by comparing thesetrength values with those obtained with an instrumented dy-amometer among people who recently underwent TKA orHA.

METHODS

articipantsA group of 25 volunteers with elective TKA (cemented

ntramedullary femoral and tibial components; no patellectomyr patellar reconstruction) and a group of 25 volunteers withlective THA (cemented intramedullary femoral stem and fixedcetabular cup) were recruited at the ambulatory physiotherapyepartment of the Lindsay Rehabilitation Hospital (table 1). All

Table 1: Characteristics of Participants With THA or TKA

Characteristics

TKA (n�25) THA (n�25)

Mean � SD 95% CI Mean � SD 95% CI

Age (y) 68.2�9.1 64.6–71.8 67.0�10.4 62.9–71.1Sex (n)

Male 12 9Female 13 16

Height (cm) 166.6�9.3 162.9–170.2 165.4�6.4 162.9–167.9Mass (kg) 87.6�17.8 80.6–94.6 80.2�14.1 74.7–85.8Time since

surgery (d)37.8�20.0 29.9–45.6 38.3�23.5 29.1–47.5

bbreviations: CI, confidence interval; SD, standard deviation.

ubjects were weight bearing as tolerated, attained a minimum p

ctive range of motion of 20° to 70° of knee flexion, were ableo move their knee against resistance in flexion and extension,nd were ambulatory with or without assistive devices. Sub-ects were excluded from this study if they could not followimple commands or had known medical conditions for whichtatic strength measurements would be contraindicated. Ethicspproval was obtained from the research ethics committee. Allarticipants reviewed and signed an informed consent formefore entering the study. Nominative information was omit-ed, and research-related data were stored anonymously torotect confidentiality.

valuatorsThe 3 evaluators were 2 physical therapists and 1 physical

herapist assistant with an average of 20.3�5.7 years of clinicalxperience in privately and publicly funded orthopedic reha-ilitation settings. Because of limited clinical experience withhe evaluation techniques proposed, all clinical evaluators par-icipated in 3 training sessions given by a physical therapistith clinical and research experience with manual and instru-ented dynamometers mostly to standardize evaluation proce-

ures before initiating the study. Every evaluator was blindo the strength values obtained during each trial for a givenarticipant and by the other evaluators to correctly test thentertrial and interevaluator reliabilities, respectively.

quipmentA handheld MicroFET2 dynamometer,a using sophisticated

train gauge element (load-cell technology), was mounted on ateel support (figs 1A, B). The rigid support can move along aquare shaft fixed to a therapeutic chairb (fig 1C). The angle ofhe shaft can be adjusted to test the maximal static strength ofhe knee flexors and extensors at 30° or 60° of knee flexion.he MicroFET2 dynamometer was factory calibrated, and the

orce gauge was properly zeroed before each measurementccording to the recommended procedure by the manufacturer.he MicroFET2 dynamometer was designed to record peak

orces in 4.4-N (1-lb) increments up to 445N (100lb) whensing the high threshold option according to the manufactureruidelines, although forces exceeding 445N may clinically beeasured. It has been reported to be a reliable and valid

nstrument for strength assessment.16,29,30 To facilitate momentrm calculation and ensure that the support was perpendicularo the leg, the axis of rotation of the shaft was properly alignedith the center of the external condyle of the knee (axis of

otation). The support of the dynamometer was positionedithin the distal third of the tibia on the tested lower extremity.oments of force generated with the MicroFET2 dynamome-

er were corrected for gravitational effects created by the weightf the accessory and tested leg segment. The weight of the seg-ent was measured by testing the flexor force under resting

ondition. This force was added to the one generated by thearticipant during knee extension. For the knee flexor muscleesting, the MicroFet2 dynamometer was simply reset to zeronder passive condition just before initiating the knee flexionffort.

For the validation study, assessment of static knee flexionnd extension strength at 30° and 60° was repeated by using aybex isokinetic system.c Before each testing day, the Cybexynamometer was self-calibrated and moment of force calcu-ation was verified with a known weight according to theanufacturer’s instruction. The axis of the dynamometer was

ligned with the axis of rotation at the knee. This system
rovides a valid and reliable tool to assess knee extensors and
Arch Phys Med Rehabil Vol 86, October 2005

flct

(epemf

T

aeephewplcVcvsscp5duesuo6i

S

fsw

vmomtsi((brpfoateesifwaaatrrnrda

2000 STRENGTH MEASUREMENTS WITH A CHAIR-FIXED DYNAMOMETER, Gagnon

A

exors strength.31-34 The generated static strength values wereorrected for gravitational effects associated to the weight ofhe tested leg segment and the Cybex accessories.

For both testing procedures, optimal external stabilizationstraps) was provided to avoid compensation at the lowerxtremities, pelvis, and trunk while the force pad was helderpendicular to the limb being tested. Standardized verbalncouragement was given during all tests. All static strengtheasures were converted to moments of force (in Nm) to

acilitate interpretation.

esting ProtocolA physical therapist (evaluator 1) and a physical therapy

ssistant (evaluator 2) each completed a knee flexors andxtensors strength assessment with the chair-fixed dynamom-ter. All equipment required to perform the evaluation waslaced on a rehabilitation treatment table, and each therapistad to install properly each participant in conformity with thexperimental protocol. After a familiarization period, subjectsere reminded to avoid explosive contraction and instructed torogressively increase their effort to their maximal strengthevel for every maximal static contraction. The maximal staticontraction was maintained for a minimum of 5 seconds.erbal encouragement was provided during maximum static

ontractions. Evaluators could only read the highest strengthalues on the MicroFET2 dynamometer after the 5-secondtatic contraction because measurements were automaticallyaved to adequately test intertrial reliability. The protocol in-luded 2 trials for every position and a minimal 30-second resteriod between each contraction. Each subject was given a-minute rest period before repeating the entire procedure at aifferent angle, for a different movement or with a new eval-ator. For the evaluation performed with the Cybex dynamom-ter, a physical therapist (evaluator 3) completed in the sameession the knee flexors and extensors strength assessment bysing a similar testing protocol. A random testing sequencerder of direction (knee flexion, knee extension), position (30°,0°), and evaluators (evaluators 1, 2, 3) was established before
nitiating the study. e

tatistical AnalysisDescriptive statistics. Descriptive statistics were calculated

or demographic (age, weight, height, time since surgery) andtrength measures obtained for the TKA and THA groups. Dataere analyzed by using the SPSS software, version 10.5.d

Reliability. The generalizability theory (G theory) pro-ided a framework for investigating the reliability of strengtheasurements.18,35 In the G theory, the subject (S) is the object

f measurement and the factors, called facets, are the measure-ent conditions. In this study, the facets are the trials (T) and

he evaluator (E). The G theory is divided into 2 steps. The firsttep is the generalizability study (G study), which aims todentify the magnitude of the variance attributed to the subjectsS), to the systematic errors related to trial (T) and evaluatorE), and to the random errors associated to the interactionsetween different sources of variance (ST, SE, ET, SET). In theesults section, variance magnitude will be reported as theercentage of the total variance because interpretation may beacilitated. The second step, the decision study (D study), reliesn evidences generated by the G study to determine the reli-bility of a particular protocol. In our study, mixed design withhe evaluator as a fixed facet (appendix 1) was considered tostimate the intertrial reliability for hypothetical protocols of 1valuator using 1 trial, or the average of 2 or 3 trials. In thisituation, the variance associated with the evaluator is not takennto consideration. The interevaluator reliability with the trialacet as a fixed element (see appendix 1) was also estimatedith a D study that included 1 evaluator using 1 trial or the

verage of 2 or 3 trials as the criterion of strength. With the trials a fixed facet, a better estimation of the interevaluator reli-bility is possible because measurement error associated withhe trial facet is removed. Moreover, this approach is alsoelevant if comparisons are to be made with interevaluatoreliabilities reported in other studies in which the trial facet wasot included as such in the design. For these 2 D studies, theeliability coefficient retained in our study is the index ofependability (�),36 which takes into account the systematicnd random errors associated with the trial and evaluator fac-

Fig 1. Chair-fixed MicroFet2dynamometer system used tomeasure maximum staticknee flexor and extensorstrength values. (A) Anteriorview of the steel support, (B)lateral view of the steel sup-port, (C) lateral view of thesteel support fixed to the ther-apeutic chair.

ts. Like the intraclass correlation coefficient, it ranges between

0Wsemtaamcw

tdtottataioTsawAsmfefitewa

a(vvmiaewaval(a

2a1msdiSiSme

adimaa

N* dynd


(null reliability) and 1 (perfect reliability). Portney andatkins37 have suggested that values above .75 can be con-

idered good reliability and those below .75 as poor to mod-rate. To appreciate the errors in terms of the unit of measure-ent, the SEM was computed. The SEM is the root square of

he error variances associated with each D-study protocol (seennex for computational details). The SEMs were described inbsolute strength unit (in Nm) and relative (in percent) to theean strength values obtained by the 2 evaluators with the

hair-fixed dynamometer. Generalizability analysis was doneith the GENOVA statistical software.e

Validity. Pearson product-moment coefficients of correla-ion and absolute agreement between the chair-fixed MicroFet2ynamometer and the Cybex dynamometers results were usedo document the correspondence between the strength valuesbtained with the 2 different measurement techniques. Becausehe unit of measurement is the same for these 2 techniques, thisype of validity can also be considered as parallel-form reli-bility. For each participant group, the coefficients of correla-ion were calculated for knee flexors and extensors at bothngles for each evaluator. For each participant, the scores usedn the correlation analysis were the average value of the 2 trialsbtained by each method of testing and by each evaluator.wo-tailed tests combined to a level of significance of .05 wereelected for these statistical analyses. Thereafter, the measuringgreement between the 2 different measurement techniquesas calculated according to the method proposed by Bland andltman.38 To do so, strength values obtained with the 2 mea-

urement techniques were averaged to represent the best esti-ate of the strength value for each movement, at each angle,

or each evaluator (1, 2). Then, the difference between the beststimate and the mean strength value measured with the chair-xed dynamometer was calculated for each evaluator using this

echnique along with 95% limits of agreement (mean differ-nce �1.96 standard deviation [SD] of the difference). Dataere analyzed by using the SPSS, version 10.5, statistical

Table 2: Mean Strength Values and 95% CIs of the Mean ObtaineKnee Angles and Measured by 3 Different Evaluators With the

Differences Between the

Static Strength Tests

Chair-Fixe

Evaluator 1

Mean � SD 95% CI Differences

TKAKnee positioned at 30°

Flexion 47.5�25.4 37.0–58.0 3.4�13.Extension 33.2�23.1 23.6–42.7 �5.6�16.

Knee positioned at 60°Flexion 42.2�21.3 33.4–51.0 �4.9�9.4Extension 43.6�27.0 32.4–54.7 12.0�16.

THAKnee positioned at 30°

Flexion 45.0�18.7 37.2–52.7 7.1�10.Extension 46.6�21.5 37.7–55.5 2.9�14.

Knee positioned at 60°Flexion 45.0�17.9 37.6–52.5 �3.0�7.6Extension 54.3�26.9 43.1–65.4 18.7�15.

OTE. Values are in newton meters.Differences for evaluator 1 equal mean of the 2 trials on the Cybexynamometer for evaluator 1.

nalysis software.d d

RESULTSDemographic characteristics of subjects with TKA or THA

re presented in table 1, whereas clinical characteristicsstrength values) are shown in table 2. The magnitude ofariance components (expressed as percentage of the totalariance) calculated in the G study is presented in table 3. Theain source of variance was generated by the subject for

ndividuals with THA or TKA and fluctuated between 84.95%nd 96.70% for all movements and positions. The systematicrror variance associated with the evaluator (E) and trial (T)as under .91% of the total variance. The highest error vari-

nce is related to the subject-evaluator interaction (SE), withalues ranging between 1.46% and 13.40% of the total vari-nce. The other variance components (ST, ET) contributed toess than .37% of the total variance. Finally, the residual errorSET interaction) represents only .79% to 2.02% of total vari-nce.

The intertrial reliabilities for the D studies (1 trial, mean ofand 3 trials) are reported in table 4. The indices of depend-

bility (�) highlight the excellent intertrial reliability (0.98–.00) for the strength measures obtained for different move-ents and positions with the chair-fixed dynamometer among

ubjects with THA and TKA. As evidenced by the results, theifferences between 1 and the mean of 2 or 3 trials are notmportant for the indices of dependability, but the SEM and theEM/mean values are reduced by using the average of 3 trials

nstead of just 1 or 2 trials. Of clinical relevance, all relativeEMs were below 10% of the mean for the different move-ents and positions measured with the chair-fixed dynamom-

ter among subjects with THA and TKA.The interevaluator reliabilities calculated with the trial facet

s a fixed element are reported in table 5. The indices ofependability (�) confirm the very good interevaluator reliabil-ty (.92–.99) for the strength measures obtained for differentovements and positions. Interestingly, the index of depend-

bility (�) revealed that adding trials did not, in most cases,ffect the level of interevaluator reliability. Yet, the SEM

ring Static Knee Flexion and Extension Initiated From 2 Differentx Dynamometer and the Chair-Fixed Dynamometer and Meaneasurement Techniques

amometer Cybex Dynamometer

Evaluator 2 Evaluator 3

Mean � SD 95% CI Differences Mean � SD 95% CI

47.4�24.8 37.1–57.6 3.5�14.8 50.9�23.8 41.0–60.731.1�21.9 22.0–40.1 �3.5�14.8 27.5�19.5 19.5–35.5

41.3�23.0 31.8–50.7 �4.0�14.8 37.3�17.8 29.9–44.739.8�23.7 30.0–49.6 15.8�14.1 55.5�31.6 42.5–68.6

45.0�20.5 36.6–53.5 7.0�9.0 52.1�22.7 42.7–61.447.8�26.1 37.0–58.5 1.8�13.3 49.5�21.3 40.8–58.3

45.5�20.4 37.1–53.9 �3.4�8.9 42.0�16.5 35.2–48.856.4�26.3 45.6–67.3 16.5�12.3 72.9�31.7 59.8–86.0

amometer minus the mean of the 2 trials with the chair-fixed HHD

d DuCybe2 M

d Dyn

*

81

1

62

7

ecreases with 3 trials as compared with 1 or 2 trials. When a


t8w

tdtStvst6.

cbmhCMb

TnokobC

troaTbvmcm

Ae* ortedi 1�38


A

otal of 3 trials were performed, the SEM/mean varied between.34% and 9.92% among subjects with TKA, whereas thoseith THA ranged between 7.46% and 12.69%.The Pearson product-moment coefficients of correlation and

he scatterplot diagrams of the strength values collected toetermine the association between the chair-fixed dynamome-er and the Cybex dynamometer are displayed in figure 2.ignificant positive coefficients of correlation fluctuating be-

ween .73 and .93 were generated when comparing the strengthalues obtained with the Cybex dynamometer to those mea-ured with the chair-fixed dynamometer. Stronger linear rela-ionships were attained when the tested knee was positioned at0° compared with 30° both for subjects with TKA (.78–.92 vs87–.93) or THA (.73–.85 vs .86–.91).

Despite these high Pearson product-moment coefficients oforrelation reported, the scatterplot diagram of the differencesetween the 2 measurement techniques against the best esti-ate of the strength values calculated for evaluators 1 and 2

ighlights a clinically relevant lack of agreement between theybex dynamometer and the chair-fixed dynamometer (fig 3).ean differences ranging between �5.6 and 15.8Nm have

een found for participants with TKA, whereas those with

Table 3: Magnitude of Variance Components in Percentage of thInteractions for Strength Values Obtained During Static K

Variance Components

Knee Positioned at 30°

Flexion Ex

THA TKA THA

Subject (S) 90.91 94.46 92.62Evaluator (E) 0.00 0.00 0.00Occasion (O) 0.05 0.05 0.00SE 7.18 4.67 6.00SO 0.29 0.02 0.34EO 0.00 0.00 0.32SEO error 1.57 0.79 0.71Total variance (%) 100.00 100.00 100.00Total variance* (Nm) 388.57 634.15 712.42

bbreviations: EO, evaluator-occasion interaction; SE, subject-evaluarror); SO, subject-occasion interaction.These variances are used to calculate the variance in percentage repn newton meters associated with the subject (S) component is 90.9

Table 4: Intertrial Reliability (Evaluator Facet Fixed) Obtained forTrials as the

Static Strength Tests

Coefficient of Dependability (�)

1 Trial 2 Trials 3 Trials


Flexion 0.99 1.00 1.00Extension 0.99 0.99 1.00

Knee positioned at 60°Flexion 0.99 0.99 1.00Extension 0.99 0.99 1.00



Knee positioned at 60°Flexion 0.98 0.99 0.99
Extension 0.98 0.99 0.99 3.

HA varied between –3.4 and 18.7Nm (see table 2). Thearrowest 95% limit of agreement (mean difference �1.96 SDf the difference) was found when evaluator 1 was assessingnee flexion at 60° among participants with THA, where a limitf agreement confirmed that the chair-fixed dynamometer maye 17.9 below or 11.9Nm above the value obtained with theybex dynamometer.

DISCUSSIONThe main objectives of our study were initially to evaluate

he reliability (intertrial, interevaluator) as well as the concur-ent validity of knee flexor and extensor strength measurementsbtained with a newly developed chair-fixed dynamometermong people who recently underwent an elective TKA orHA. Excellent intertrial and interevaluator reliabilities haveeen found in this study. In addition, very high concurrentalidity was shown with the Cybex dynamometer measure-ents. For these reasons, it is possible to recommend a clini-

ally relevant standardized measurement protocol that mini-izes error of measurement.

tal Variance Calculated in the G Study for Each Facet and Theirlexion and Extension Initiated From 2 Different Positions

Knee Positioned at 60°

n Flexion Extension

TKA THA TKA THA TKA

96.70 92.47 92.59 84.95 93.830.35 0.00 0.00 0.00 0.910.00 0.00 0.00 0.05 0.001.46 5.46 6.05 13.40 3.840.12 0.06 0.37 0.27 0.000.00 0.00 0.00 0.00 0.001.38 2.02 0.99 1.32 1.42

100.00 100.00 100.00 100.00 100.00675.93 372.32 493.46 714.00 656.87

teraction; SEO, subject-evaluator-occasion interaction (includes the

in the corresponding column. Thus, in the first column, the variance8.57�353.24Nm.

d D-Study Designs With 1 Evaluator Using 1, Average of 2, or 3e Criterion

SEM (Nm) SEM/Mean Strength (%)

rial 2 Trials 3 Trials 1 Trial 2 Trials 3 Trials

35 1.66 1.36 4.95 3.50 2.8618 2.25 1.84 9.88 7.00 5.71

59 1.83 1.50 6.20 4.38 3.5906 2.16 1.77 7.34 5.18 4.24

73 1.93 1.57 6.07 4.29 3.4913 2.21 1.58 6.63 4.68 3.35

78 1.97 1.61 6.14 4.35 3.55

e Tonee F

tensio

tor in

MixeForc

1 T

2.3.

2.3.

2.3.

2.
43 2.43 1.98 6.19 4.39 3.57

R

26pvsttSstshet

lwtis(rrkwaac

owwstvSce

sr

itdogtp

taerTfitoicipWb

w1(tHgrpcviierstasa

7.


eliabilityAs pointed out by the results of the G studies applied to the

subject groups (THA, TKA) at 2 different testing angles (30°,0°) for 2 muscle groups (knee flexors, knee extensors), theercentage of variance components are generally similar acrossarious G studies. There are no large differences related toubject groups, testing angles, and muscle groups. One excep-ion is the G study result obtained for the THA sample whenesting knee extension at 60°. The percentage of variance of theE component (13.4%) is more important than for other Gtudies (1.46%–7.18%). There is no specific explanation forhis finding, but from the results reported in table 2, the averagetrength recorded in this test for the 2 evaluators was theighest among all testing and it can be speculated that variancerror across evaluator (SE interaction) is possibly affected byhe magnitude of strength values.

In general, the reliabilities found in the D studies are excel-ent because the main source of variance in the measurementas related to differences among participants. This indicates

hat the chair-fixed dynamometer can successfully discriminatendividuals with a TKA or THA based on knee strength mea-ures. Narrowing the inclusion criteria of this research projectage, sex, time since surgery) would have possibly led to aeduced variance among subjects and consequently led to lowereliabilities. Despite high indexes of dependability, it should beept in mind that the SEM parameters are also of importancehen evaluating an individual patient. The SEM parameters

re related to the magnitude of error variances associated withspecific facet and will be discussed later under each variance

omponents analysis.Among the error variance components, the systematic effect

f the evaluator (E) was negligible. Effectively, mean valuesere similar for the 2 evaluators (see table 2). The exceptionas the knee extension strength measured at 60° on TKA

ubjects with evaluator 2 having lower value on the averagehan evaluator 1. Thus, it is for this test that maximal errorariance was obtained (5.95Nm or .91% of the total variance).imilarly, from the low systematic trial variance (T), it can beoncluded that neither significant muscular fatigue nor learningffects were present across trials.

The error variances associated with the interactions betweenubject and trial (ST), between evaluator and trial (ET), and the

Table 5: Interevaluator Reliability (Trial Facet Fixed) Obtained forTrials as the

Static Strength Test

Coefficient of Dependability (�)

1 Trial 2 Trials 3 Trials







esidual error (SET) have all low values in comparison with the p

nteraction between subject and evaluator (SE). The interpre-ation of this last interaction implies that some patients pro-uced a larger moment with a particular evaluator, whereasthers did the reverse. This specific effect cannot be disentan-led especially if scores obtained from different observers areo be compiled in clinical or experimental settings for the sameatient.Having generated the variance estimate in the G study, it is

hen possible to explore the effects of certain decisional situ-tions (D studies) representative of clinical situations experi-nced by rehabilitation specialists involved with people withecent elective TKA or THA undergoing strength assessment.he intertrial reliability (mixed design with the evaluator as axed facet) suggest that static strength values obtained after 1

rial determines an excellent estimate of reliability with indexesf dependability ranging from .98 to .99. These near-perfectntertrial reliability indices are marginally affected by an in-rease of up to 3 trials (0.99–1.00). Alternatively, the effect ofncreasing the number of trials is more evident for the SEMarameters with a reduction of about 40% of the parameters.hen testing a single patient, using the mean of 3 trials should

e more appropriate than using 1 or the mean of 2 trials.The interevaluator reliability by using a mixed D design in

hich the trial facet was fixed covers a range of .92 to .97 withtrial and is marginally improved by using the mean of 2

.92–.99) or 3 trials (.93–.99). These results provide evidencehat the parsimony of a single trial should not be dismissed.owever, the promotion of multiple trials remains of interestiven its positive impact on the error of measurement, which iselevant for decision making on an individual subject in clinicalractice. Generally, weaker interevaluator reliability coeffi-ients have been reported in studies using nonfixed HHD, withalues ranging from .14 to .96 when testing the lower extrem-ty.16 The SEM parameter is higher than those found forntertrial reliability, and this is a consequence of the higherrror variance of the evaluator facet relative to the trial facet aseported in the G study. The SEM/mean strength parameters isometimes too large with values higher than 10%. For thoseests, it should be important to identify the source of errorssociated with the evaluator. Because intertrial reliability pre-ented better than interevaluator reliability, the error was prob-bly caused by the positioning of subjects or the dynamometer

d D-Study Designs With 1 Evaluator Using 1, Average of 2, or 3e Criterion

SEM (Nm) SEM/Mean Strength (%)

rial 2 Trials 3 Trials 1 Trial 2 Trials 3 Trials

16 4.01 3.96 8.75 8.45 8.3428 2.90 2.77 10.18 9.02 8.60

17 4.02 3.97 9.97 9.63 9.5050 4.24 4.14 10.8 10.17 9.92

12 3.93 3.87 9.16 8.73 8.6000 4.82 4.76 10.59 10.21 10.10

73 3.47 3.38 8.23 7.67 7.4630 7.09 7.03 13.17 12.81 12.69

MixeForc

1 T

4.3.

4.4.

4.5.

3.

lacement, which were not exactly the same for the 2 evalua-


Fao


A

ig 2. Scatterplot diagrams and Pearson product-moment coefficients of correlation of the strength values of the chair-fixed dynamometer
gainst those obtained with the Cybex dynamometer. The number of points, representing each participant, may be inferior to 25 becausef similar strength values (superposition) of few participants.

Fai


ig 3. Scatterplot diagrams of the differences in static strength values between the Cybex dynamometer and the chair-fixed dynamometer
gainst the best estimate of the static strength values for evaluators 1 and 2. The number of points, representing each participant, may benferior to 25 because of similar strength differences (superposition) of few participants.

ttm

bDs3mvpisironuitoetbswd

V

mTvsbwmccpc6(wttptwasd(tsbcdwmumpmt

Ts

tscbvhciceatecnt

nplhTsw(ce1timsodw(wtwaa

dsttBuaamcotodpvTa


A

ors. Better evaluation protocol with additional details abouthese elements would increase the reliability of the measure-ent across evaluators.At this point of the discussion, further explanations may

e welcome to facilitate understanding of the results of thestudies in which reliability indices were calculated for the

ame design (1 evaluator using 1 trial, mean of 2 or mean oftrials) applied either to an intertrial or to an interevaluatoreasurement protocol. On one hand, the intertrial reliability

alues reported in our study are limited to measurementrotocol in which the same evaluator is involved. In the casen which the mean of many trials is used as the criterion, theame evaluator should do all trials. Thus, the reliabilityndices (coefficient of dependability, SEM) indicate the er-or of measurement relative to a hypothetical true scorebtained by the same evaluator (fixed facet) on a largeumber of trials (random facet). On the other hand, in these of the interevaluator reliability indices, the constraintmposed by the design is related to the trial facet. Thus, 1rial (the first 1 or any 1 in a series of trials) or the averagef the first 2 or 3 trials can be used as criterion as long as therror of measurement is relative to the true score obtained onhe same trial or similar combination of trials (fixed facet)y a large number of evaluators (random facet). In conclu-ion, the interpretation of the reliability is always associatedith a hypothetical true score obtained under specific con-itions and not with the design of the D studies.

alidityBecause therapists use the chair-fixed dynamometer toeasure muscle performance among patients with THA orKA, they must consider whether the measurements providealid information. Consequently, a comparison of thetrength values obtained with the chair-fixed dynamometery the physical therapist and physical therapist assistantere compared with the ones obtained on the Cybex dyna-ometer (criterion device). Overall, the significant positive

oefficient of correlations (r range, .73–.93) supports theonvergent construct validity of these 2 measures. Morerecisely, the mutual reflection of a common underlyingonstruct is stronger when the tested knee is positioned at0° compared with 30° both for participants with TKA.78 –.92 vs .87–.93) or THA (.73–.85 vs .86 –.91). Theeakest association was reported when assessing knee ex-

ensors with the knee positioned at 30° of flexion. Correla-ions obtained by the physiotherapy technician were com-arable, even better in many cases than the ones reached byhe physiotherapist. This confirms that clinical evaluatorsith extensive experience in orthopedic rehabilitation can

dequately use the chair-fixed dynamometer and follow atandardized measurement protocol. The capacity of theynamometer for automatic correction for gravity effectsweight of the tested segment) is only unidirectional becausehe MicroFet2 dynamometer is solely sensitive to compres-ion. For this reason, the MicroFet2 dynamometer needs toe moved when testing knee extension to allow gravityorrection for the weight of the segment. The use of aynamometer sensitive to compression as well as tractionould greatly facilitate clinical use of the chair-fixed dyna-ometer. The fact that only 1 type of HHD (MicroFet2) was

sed potentially limits the generalizability of the results toeasurements obtained with other dynamometers. At this

oint, the fact that the validity of the measurement instru-ent was found to be better when the tested knee is posi-

ioned at 60° compared with 30° both for individuals with o


KA or THA should also be considered in the final deci-ional process.

Despite the good correlation between the 2 measurementechniques, the Bland and Altman approach showed thattrength values measured with the chair-fixed dynamometeran differ from those obtained with the Cybex dynamometery an appreciable amount. The range of mean differencesaries between –5.6 to 18.7Nm with SD reaching values asigh as 16Nm. In an individual subject, these differencesan represent more than 10% of the strength values, whichs often referred to as a threshold for strength evaluation inlinical practice and research.8 Even if part of these differ-nces can be explained by the lack of a perfect reliabilitycross trials (one can see the Cybex measure as 1 trial andhe chair-fixed HHD measure as another trial), the consid-rable discrepancies between the 2 measurement techniqueslearly support the contention that the 2 measurements tech-iques should not be used interchangeably in clinical prac-ice at this time.

To conclude, it should be mentioned that neither incidentor accident were reported during the study. According torevious articles, prosthetic components were fully stabi-ized in bone39 and repair of muscle sutures have reachedigh resistance to failure (tensile strength) at this time.28

hus, strength assessment was considered by orthopedicurgeons as a secure and applicable procedure for subjectsith TKA or THA. However, almost half of the subjects

THA�48%, TKA�44%) reported mild-to-moderate dis-omfort in the area of the interface between the dynamom-ter and the skin overlying the distal third of the tibia duringor more tests despite the addition of high density foam to

he original padding. Increasing the surface area of thenterface might help to eliminate the problem. The dynamo-etric testing was done on average about 38 days after the

urgery (see table 1). Finally, no patient generated forcever the suggested maximal capacity of the MicroFET2ynamometer (445N [100lb]). In fact, the highest scoresere 381N (85.6lb) for the knee extension at 60° and 437N

98.3lb) for the knee extension at 60° among participantsith THA and TKA, respectively. Obviously, the capability

o manually resist these high levels of force being generatedith or without a HHD may appreciably vary among ther-

pists based on their personal characteristics and clinicalbilities.17,22

CONCLUSIONSThe results of this study support the use of a chair-fixed

ynamometer for clinical assessment of static knee muscletrength (flexors, extensors) among subjects with recent elec-ive THA or TKA (population specific) because excellent in-ertrial and interrater indices of reliability were documented.ased on these findings, a clinical protocol that suggests these of the mean of 3 trials with the knee positioned at 60° afterfamiliarization period, both for knee flexors and extensors, bytrained therapist is strongly encouraged to minimize errors ofeasurement on static knee strength values measured with the

hair-fixed dynamometer. Factors like safety, portability, easef use, limited expense, and space requirement further confirmhat the chair-fixed dynamometer is highly valuable. It is rec-mmended that the chair-fixed dynamometer and the Cybexynamometer should not be used interchangeably in clinicalractice at this time because of a lack of agreement in strengthalues measured at the knee among patients with TKA andHA. These results are of clinical importance as knee flexorsnd extensors strengthening exercises constitute the foundation
f the majority of the rehabilitation programs offered to indi-

vifiaa

ast

A

B

1

1

1

1

1

1

1

1

1

1

2

2

2


iduals after THA or TKA. Future research should focus on thentersession reliability and responsiveness of this specific chair-xed dynamometer over the course of rehabilitation whiledditional chair adaptations to include hip flexors, extensors,nd abductors will be explored.

Acknowledgments: We thank Marc Guérard in the developmentnd adaptations of the chair-fixed dynamometer, France Piotte for thetatistical analysis, and Luci F. Teixeira-Salmela and Nathalie Fish forhe final review of this manuscript.

APPENDIX 1: GENERALIZABILITY THEORY

: Components of variance

�S2 � subject variance

�E2 � evaluator variance �systematic�

�T2 � trial variance �systematic�

�ET2 � evaluator-trial interaction (random)

�SE2 � subject-evaluator interaction (random)

�ST2 � subject-trial interaction (random)

�SET2 � subject-evaluator-trial interaction (residual variance)

: Calculation of the reliability

Mixed design (intertrial coefficient)

● Index of dependability (�T)

Absolute error variance

�T �

�S2 �

�SE2

nE

�S2 �

�SE2

nE

��T

2

nT

��ET

2

nEnT

��ST

2

nT

��SET

2

nEnT

● Standard error of measurement (SEMT)

SEMT ��T2

nT

��ET

2

nEnT

��ST

2

nT

��SET

2

nEnT

Mixed design (interevaluator coefficient)

● Index of dependability (�E)

Absolute error variance

�E �

�S2 �

�ST2

nT

�S2 �

�ST2

nT

��E

2

nE

��ET

2

nEnT

��SE

2

nE

��SET

2

nEnT

● Standard error of measurement (SEME)

SEME ��E2

nE

��ET

2

nEnT

��SE

2

nE

��SET

2

nEnT

nE�number of evaluators in the D studynT�number of trials in the D study

In this study, all coefficients for the D studies were
computed with nE�1 and nT�1, 2, or 3.
References1. Canadian Joint Registry Report. Total hip and total knee replace-

ments in Canada. Ottawa: Canadian Institute of Health Informa-tion; 2003. p 1-78.

2. Enloe L, Shields R, Smith K, Leo K, Miller B. Total hip and kneereplacement treatment programs: a report using consensus. J Or-thop Sports Phys Ther 1996;23:3-11.

3. Freburger J. An analysis of the relationship between the utilizationof physical therapy services and outcomes of care for patients aftertotal hip arthroplasty. Phys Ther 2000;80:448-58.

4. Kirk-Sanchez N, Roach K. Relationship between duration oftherapy services in a comprehensive rehabilitation program andmobility at discharge in patients with orthopedic problems. PhysTher 2001;81:888-95.

5. Munin M, Rudy T, Glynn N, Crosset L, Rubash H. Early inpatientrehabilitation after elective hip and knee arthroplasty. JAMA1998;279:847-52.

6. Kuster M. Exercise recommendations after total joint replace-ment: A review of the current literature and proposal of scientif-ically based guidelines. Sports Med 2002;32:433-45.

7. Fougeyrollas P, Noreau L, Boschen K. Interaction of environmentwith individual characteristics and social participation: theoreticalperspectives and applications in persons with spinal cord injury.Top Spinal Cord Inj Rehabil 2002;7:1-16.

8. Sapega AA. Muscle performance evaluation in orthopaedic prac-tice. J Bone Joint Surg Am 1990;72:1562-74.

9. Clarkson HM. Musculoskeletal assessment: joint range of motionand manual muscle strength. Philadelphia: Lippincott, Williams &Wilkins; 2000.

0. Brown M, Sinacore DR, Host HH. The relationship of strength tofunction in older adult. J Gerontol A Biol Sci Med Sci 1995;50:55-9.

1. Bohannon RW. Manual muscle testing of the limbs: consider-ations, limitations and alternatives. Phys Ther Pract 1992;2:11-21.

2. Bohannon RW, Corrigan D. A broad range of forces is encom-passed by the maximum manual muscle test grade of five. PerceptMot Skills 2000;90(3 Pt 1):747-50.

3. Bohannon RW. Measuring knee extensor muscle strength. Am JPhys Med Rehabil 2001;80:13-8.

4. Knepler C, Bohannon RW. Subjectivity of forces associated withmanual-muscle test grades of 3�, 4�, and 4. Percept Mot Skills1998;87:1123-8.

5. Bohannon RW. Adoption of hand-held dynamometry. PerceptMot Skills 2001;87:1123-8.

6. Bohannon RW. Intertester reliability of hand-held dynamometry:a concise summary of published research. Percept Mot Skills1999;88(3 Pt 1):899-902.

7. Roebroeck ME, Harlaar J, Lankhorst GJ. Reliability assessment ofisometric knee extension measurements with a computer-assistedhand-held dynamometer. Arch Phys Med Rehabil 1998;79:442-8.

8. Roebroeck ME, Harlaar J, Lankhorst GJ. The application ofgeneralizability theory to reliability assessment: an illustrationusing isometric force measurements. Phys Ther 1993;73:386-401.

9. Bohannon RW. Hand-held dynamometer: factors influencing re-liability and validity. Clin Rehabil 1997;11:263-4.

0. Wikholm JB, Bohannon RW. Hand-held dynamometer measure-ments: tester strength makes a difference. J Orthop Sports PhysTher 1991;13:191-8.

1. Smidt GL, Rogers MW. Factors contributing to the regulating andclinical assessment of muscular strength. Phys Ther 1982;62:1283-90.

2. Mulroy SJ, Lassen KD, Chambers SH, Perry J. The ability of maleand female clinicians to effectively test knee extension strengthusing manual muscle testing. J Orthop Sports Phys Ther 1997;26:
192-9.

2

2

2

2

2

2

2

3

3

3

3

3

3

3

3

3

3

a

b

c

de


A

3. Brinkmann JR. Comparison of a hand-held and fixed dynamom-eter in measuring strength of patients with neuromuscular disease.J Orthop Sports Phys Ther 1994;19:100-4.

4. Desrosiers J, Prince F, Rochette A, Raîche M. Reliability of lowerextremity strength measurements using the belt-resisted method. JAging Phys Act 1998;6:317-26.

5. Karner PM, Thompson AL, Connelly DM, Vandervoort AA.Strength testing in elderly women using a portable dynamometer.Physiother Can 1998;50:35-9

6. Scott DA, Bond EQ, Sisto SA, Nadler SF. The intra- and interraterreliability of hip muscle strength assessments using a handheldversus a portable dynamometer anchoring station. Arch Phys MedRehabil 2004;85:598-603.

7. Kramer JF, Vaz MD, Vandervoort AA. Reliability of isometrichip abductor torques during examiner- and belt-resisted tests. JGerontol 1991;46:M47-51.

8. Stevens JE, Mizner RL, Snyder-Mackler L. Quadriceps strengthand volitional activation before and after total knee arthroplastyfor osteoarthritis. J Orthop Res 2003;21:775-9.

9. Kwoh CK, Petrick MA, Munin MC. Inter-rater reliability forfunction and strength measurements in the acute care hospital afterelective hip and knee arthroplasty. Arthritis Care Res 1997;10:128-34.

0. Click Fenter P, Bellew JW, Pitts T, Kay R. A comparison of 3hand-held dynamometers used to measure hip abduction strength.J Strength Cond Res 2003;17:531-5.

1. Emery CA, Maitland ME, Meeuwisse WH. Test-retest reliabilityof isokinetic hip abductor and flexor muscle strength. Clin J SportMed 1999;30:79-85.

2. Li RC, Wu Y, Maffulli N, Chan KM, Chan JL. Eccentric andconcentric isokinetic knee flexion and extension: a reliabilitystudy using the Cybex 6000 dynamometer. Br J Sports Med
1996;30:156-60.

3. Li RC, Wu Y, Maffulli N, Chan KM, Chan JL. Eccentric andconcentric isokinetic knee flexion and extension: a reliabilitystudy using the Cybex 6000 dynamometer. Br J Sports Med1996;30:156-60.

4. Madsen OR, Lauridsen UB. Knee extensor and flexor strength inelderly women after recent hip fracture: assessment by the Cybex6000 dynamometer of intra-rater inter-test reliability. Scand JRehabil Med 1995;27:219-26.

5. Shavelson RJ, Webb NM. Generalizability theory: a primer. New-bury Park: Sage; 1991.

6. Brennan RL, Kane MT. An index of dependability for masterytests. J Educ Meas 1977;14:277-89.

7. Portney LG, Watkins MP. Foundations of clinical research: ap-plications to practice. 2nd ed. Upper Saddle River: Prentice-Hall;2000.

8. Bland JM, Altman DG. Statistical methods for assessing agree-ment between two methods of clinical measurement. Lancet 1986;1:307-10.

9. Shepard MF, Kabo JM, Lieberman Jr. The Frank StinchfieldAward. Influence of cement technique on the interface strengthof femoral components. Clin Orthop Relat Res 2000;Dec(381):26-35.

Suppliers. Hoggan Health Technologies Inc, Medical Products Div, PO 488,

West Jordan, UT 84084.. BTE Technologies Inc, 7455-L New Ridge Rd, Hanover, MD

21076.. Computer Sport Medicine Inc, 101 Tosca Dr, Stoughton, MA

02072-1505.. SPSS Inc, 233 S Wacker Dr, 11th Fl, Chicago, IL 60606.. JE Crick, National Board of Medical Examiners, 3930 Chestnut St,

Philadelphia, PA 19104.

Reliability and Validity of Static Knee Strength Measurements Obtained With a Chair-Fixed...

Documents

Transcript of Reliability and Validity of Static Knee Strength Measurements Obtained With a Chair-Fixed...