Assessing recovery and establishing prognosis following total knee arthroplasty

Assessing Recovery and EstablishingPrognosis Following Total KneeArthroplastyDeborah M Kennedy, Paul W Stratford, Daniel L Riddle, Steven E Hanna,Jeffrey D Gollish

Background and PurposeInformation about expected rate of change after arthroplasty is critical for makingprognostic decisions related to rehabilitation. The goals of this study were: (1) todescribe the pattern of change in lower-extremity functional status of patients overa 1-year period after total knee arthroplasty (TKA) and (2) to describe the effect ofpreoperative functional status on change over time.

SubjectsEighty-four patients (44 female, 40 male) with osteoarthritis, mean age of 66 years(SD�9), participated.

MethodsRepeated measurements for the Lower Extremity Functional Scale (LEFS) and theSix-Minute Walk Test (6MWT) were taken over a 1-year period. Data were plotted toexamine the pattern of change over time. Different models of recovery were ex-plored using nonlinear mixed-effects modeling that accounted for preoperative statusand gender.

ResultsGrowth curves were generated that depict the rate and amount of change in LEFSscores and 6MWT distances up to 1 year following TKA. The curves account forpreoperative status and gender differences across participants.

Discussion and ConclusionThe greatest improvement occurred in the first 12 weeks after TKA. Slower improve-ment continued to occur from 12 weeks to 26 weeks after TKA, and little improve-ment occurred beyond 26 weeks after TKA. The findings can be used by physicaltherapists to make prognostic judgments related to the expected rate of improvementfollowing TKA and the total amount of improvement that may be expected.

DM Kennedy, BScPT, MSc, is Man-ager of Program Development, Hol-land Orthopaedic & Arthritic Cen-tre, Sunnybrook Health SciencesCentre, 43 Wellesley St E, Toronto,Ontario, Canada M4Y 1H1. She alsois Instructor, Department of Physi-cal Therapy, University of Toronto,and Part-time Assistant Clinical Pro-fessor, School of Rehabilitation Sci-ence, McMaster University, Hamil-ton, Ontario, Canada. Address allcorrespondence to Ms Kennedy at:[email protected].

PW Stratford, PT, MSc, is Profes-sor, School of Rehabilitation Sci-ence, and Associate Member, De-partment of Clinical Epidemiologyand Biostatistics, McMaster Uni-versity, and Scientific Affiliate, De-partment of Surgery, SunnybrookHealth Sciences Centre.

DL Riddle, PT, PhD, FAPTA, is OttoD Payton Professor, Departmentof Physical Therapy, Virginia Com-monwealth University, MedicalCollege of Virginia Campus, Rich-mond, Va.

SE Hanna, PhD, is Associate Pro-fessor, Department of Clinical Ep-idemiology and Biostatistics andSchool of Rehabilitation Science,McMaster University.

JD Gollish, BASc, MD, FRCSC, isMedical Director, Holland Ortho-paedic & Arthritic Centre, Sunny-brook Health Sciences Centre, andAssistant Professor, Department ofSurgery, Faculty of Medicine, Uni-versity of Toronto.

[Kennedy DM, Stratford PW, Rid-dle DL, et al. Assessing recoveryand establishing prognosis follow-ing total knee arthroplasty. PhysTher. 2008;88:22–32.]

© 2008 American Physical TherapyAssociation

Research Report

Post a Rapid Response orfind The Bottom Line:www.ptjournal.org

22 f Physical Therapy Volume 88 Number 1 January 2008

Knee osteoarthritis (OA) is one ofthe most frequent causes of dis-ability.1 For patients with end-

stage OA, which is characterized bysevere pain and poor functional sta-tus, total knee arthroplasty (TKA) isrecognized as a highly beneficial andcost-effective treatment.2–4 Despitethe benefits and the rise in utilizationof this procedure,5 questions remainunanswered, particularly in the areaof rehabilitation services. The Na-tional Institutes of Health consensusstatement on total knee replacementindicates that the use of rehabilita-tion services is one of the most un-derstudied aspects of the periopera-tive management of this population.6

One issue that clinicians face whentreating patients with TKA is the de-cision as to which outcome mea-sures to use for assessment of func-tional recovery. A growing body ofliterature indicates that self-reportmeasures of function provide differ-ent information than physical perfor-mance measures in people with OAor arthroplasty.7–11 Physical perfor-mance and self-report measures mayassess different aspects of physicalfunction.12 In the arthroplasty litera-ture, many studies have used onlyself-report measures, with the Medi-cal Outcomes Study 36-Item Short-Form Health Survey questionnaire(SF-36) and the Western Ontario andMcMaster Universities OsteoarthritisIndex (WOMAC) cited most fre-quently.13 Although performance-based measures appear to providemore information about actual phys-ical ability, consensus is still neededon what activities should be in-cluded for patients with hip or kneeOA.14 Previously, Kennedy et al15 in-vestigated the measurement proper-ties of the Six-Minute Walk Test(6MWT), the Timed “Up & Go” Test(TUG), a fast self-paced walk test,and a stair performance measure insubjects with arthroplasty.

The expected rate of change in func-tional status following surgery is ofsignificant interest to both research-ers and clinicians. Researchers canapply this information to scheduleoptimal outcome assessment pointsin a randomized trial, and clinicianscan use this knowledge to bench-mark progress and to make prognos-tic decisions related to rehabilitationneeds. Studies investigating exercise-based interventions have often as-sessed outcome up to 1 year afterarthroplasty.16–18 Long-term follow-upis essential for some interventions spe-cific to arthroplasty to prevent prob-lems such as prosthetic failure. How-ever, extended follow-up times arelikely not necessary for interventionsthat lead to rapid changes in a patient’sstatus over a relatively short period oftime. In a study examining the first 4months of recovery in patients fol-lowing hip and knee replacement,Kennedy et al12 found that the greatestperiod of postoperative change oc-curred in the first 9 weeks.

Numerous studies7,12,19–26 have ex-amined recovery patterns after TKAwith differing periods of follow-up.Several authors19,21,23 provided graph-ical representations of recovery forthe WOMAC and SF-36 but did notinclude performance measures. Thestudy by Mizner et al24 provided recov-ery curves for quadriceps femoris mus-cle strength (force-generating capac-ity), knee range of motion, the TUG, atimed stair-climbing test, SF-36 sum-mary scores, and the Knee OutcomeSurvey–Activities of Daily Living Scaleat 1, 2, 3, and 6 month postoperativetime points. Two studies12,22 exam-ined recovery in the first 4 monthsafter total hip and knee arthroplastyusing hierarchical linear modeling toillustrate trajectories of change. Signif-icant differences in the patterns andpredictors of recovery were foundwhen comparing the WOMAC and theLower Extremity Functional Scale(LEFS) with the TUG, a timed stair test,and the 6MWT. In the case of the

WOMAC and LEFS, site of arthroplastywas not a predictor and preoperativelevels of function were met and ex-ceeded much earlier (1–3 weeks) thanwhat was observed for the perfor-mance measures (6–9 weeks post-operatively). A ceiling effect around9 to 10 weeks was observed with re-spect to the TUG, indicating that thismeasure is not useful for detectingimprovement beyond 3 months. A lim-itation of these studies was the inabil-ity to predict when patients hadreached their maximal functional lev-els as measured via self-report or gaitperformance. We found no other stud-ies that determined the specific timepoint of maximal functional return fol-lowing knee arthroplasty.

The purpose of this study, therefore,was to build on the existing workby profiling the change in lower-extremity functional status of partic-ipants during the first year followingprimary TKA using the 6MWT andthe LEFS. Although the WOMAC isone of the leading outcome mea-sures for people with arthroplasty,the LEFS has demonstrated cross-sectional and longitudinal validityequal to or better than that of theWOMAC physical function sub-scale.27 Clinicians find the LEFS easyto administer in busy clinic settings,and data are published on its scoreinterpretation to a greater extentthan for the WOMAC.28–30 Ourchoice to report LEFS scores alsowas influenced by the growing bodyof evidence indicating that theWOMAC lacks factorial validity.31–34

We chose the 6MWT because it isrecognized as a useful measure offunctional status and exercise capac-ity in elderly adults.35–38 Speed anddistance abilities are both importantconsiderations for community mobil-ity in older adults. Older adults needto be able to walk, on average, 300 mduring the performance of instru-mental activities of daily living.39

Recovery and Prognosis Following Total Knee Arthroplasty

January 2008 Volume 88 Number 1 Physical Therapy f 23

The specific study goals of this studywere: (1) to describe the pattern ofchange in lower-extremity func-tional status as measured by the LEFSand 6MWT of participants over a1-year period after TKA and (2) toexplore the effect of preoperativefunctional status on the pattern ofchange. Clinicians need prognosticevidence to educate their patientsabout expected time to reach theirmaximal recovery. Having this knowl-edge allows patients and their familymembers to judge progress over timeand have realistic expectations.40 Weprovide a brief illustration using a hy-pothetical clinical vignette to illustratehow the study results can be appliedto assist clinicians in making prognos-tic decisions when treating patientsfollowing TKA.

Clinical Practice VignetteMr Smith, a 67-year-old with a long-standing history of OA of the rightknee, is referred for rehabilitation 2weeks after a right TKA. As part ofthe initial assessment, you adminis-ter the LEFS and the 6MWT and ob-tain values of 28 LEFS points and261 m, respectively. These values aresubstantially lower than Mr Smith’spreoperative values of 40 pointsfor the LEFS and 507 m of the 6MWT.Mr Smith mentions that he has a va-cation cruise scheduled in 8 weeksand asks what his function is likely tobe at that time. He also wonderswhat his maximum functional statusis likely to be and when he will reachthis level of functioning. Questionsarising from the assessment includethe following: (1) How much changeis required in these measures to bereasonably certain that a true changehas occurred? (2) What factorsshould be considered in schedulingthe next assessment, and whenshould it occur? (3) What is MrSmith’s lower-extremity functionalstatus likely to be in 8 weeks? (4)What is Mr Smith’s maximum func-tional status likely to be? (5) When is

Mr Smith likely to reach his maxi-mum functional level?

MethodAll data were collected as part of alarger observational study conductedat a tertiary care orthopedic facilityin Toronto, Canada, from Novem-ber 2001 to February 2004. Desig-nated a Centre of Excellence forhip and knee replacement, the facil-ity is one of the largest-volume ar-throplasty sites in the country. Pa-tients were recruited prospectivelyeither at point of consultation withthe orthopedic surgeon or at thepreadmission visit prior to surgery.Only those patients with follow-upfor the first year postoperativelywere eligible for this study. Duringthe larger study, there were periodsof interruption of recruitment andtracking, such as with the outbreakof severe acute respiratory syndromein Toronto from April to June 2003.At the height of the outbreak, thou-sands of people were quarantined, andthere were significant restrictions onpatient-related activities in hospitalsfor several months. None of the pa-tients took part in other interventionalstudies. However, the current sampleoverlaps samples described in earlierpublications, which used data fromthe same observational study.10,12,15,22

Participant eligibility criteria in-cluded the following: diagnosis ofOA, scheduled for primary TKA; suf-ficient language skills to communi-cate in written and spoken English;and absence of neurological, cardiac,or psychiatric disorders or othermedical conditions that would signif-icantly compromise physical func-tion. Ethics approval for the studywas received from the institution’sresearch ethics review board, and allparticipating patients provided writ-ten informed consent. Patients re-ceived standardized inpatient treat-ment following a primary total kneecare pathway. All patients were per-mitted to be full weight bearing and

participated in a progressive pro-gram of range of motion, strengthen-ing exercises, proprioceptive exer-cises, and functional training. At thetime of this study, the majority ofthe patients were transferred fromthe acute care floor on the fourth orfifth postoperative day to the on-siteshort-term rehabilitation unit to con-tinue the aforementioned programfor a maximum length of stay of 7days. All patients were dischargedwith a home exercise program, andsome patients received additionalphysical therapy treatment in thecommunity.

SubjectsPreoperatively, 88 patients con-sented to participate in the study;however, only 84 patients contrib-uted LEFS and 6MWT data follow-ing arthroplasty. Table 1 provides asummary of the participants’ char-acteristics. Female participants hada greater body mass index(t82��2.05, P2�.042); male partici-pants had higher LEFS scores(t82�3.02, P2�.003) and walkedgreater distances in 6 minutes(t82�5.28, P2�.001).

DesignWe applied a prospective study de-sign with repeated measurementsover a period of approximately 1year following arthroplasty. To pro-vide an accurate model of changeover time, participants’ follow-upmeasurements were not standard-ized to be at the same time pointsduring the first 4 postoperativemonths, the period of greatestchange.7,12 When measurementstake place at the same spaced timepoints, the shape of the curve is dic-tated by the choice of time points.Three assessments were plannedduring this time frame, and subse-quently participants were assessedat points corresponding to the nextsurgeon follow-up appointments,which typically might fall at 6 or 9months and then 12 months postop-



eratively. As noted earlier, this sched-uling of assessments facilitated ob-taining good estimates of the rate ofchange as well as the limit values orpoint of maximal return.

MeasuresPrevious work7,8,10 has suggestedthat self-report and performance-based measures capture different,but related, aspects of lower-extremity functional status. Accord-ingly, we chose 2 measures of lower-extremity functional status—one aself-report measure and the other aperformance-based measure.

LEFS. Conceived by Binkley and col-leagues,28 the LEFS is a 20-item self-report measure of lower-extremityfunctional status. It includes itemsthat assess the disablement conceptsof functional limitation (activity lim-itation) and disability (participationrestriction).41,42 Each item is scoredon a 5-point scale (0–4). Accord-ingly, total LEFS scores can vary from0 to 80 points, with higher scoresbeing associated with greater levelsof functional status. Considerablesupport for this measure’s reliability,validity, and ability to detect changeexists both for general lower-extremity conditions43–45 and spe-cific to patients with OA progressingto knee or hip arthroplasty.27,29,46

The test-retest reliability estimate(intraclass correlation coefficient,type 2,1) for the LEFS derived from asample of patients following arthro-plasty was .85, the standard error ofmeasurement (SEM) was 3.7 LEFSpoints, and the minimal detectablechange at the 90% confidence level(MDC90) was estimated to be 9 LEFSpoints.28 In patients undergoingknee or hip arthroplasty, the LEFShas been shown to detect change aswell as or better than the WOMACphysical function subscale.27,46

6MWT. Originally conceived as anoutcome measure for people withrespiratory problems, the standard-

ized 6MWT has become a popularmeasure of lower-extremity func-tional limitation for patients with OAof the lower extremity and thoseprogressing to arthroplasty.7,8,15,47,48

Participants were instructed to coveras much distance as possible duringthe 6-minute time frame. The testwas conducted on a measured 46-muncarpeted rectangular indoor cir-cuit. The course was marked off inmeters, and the distance traveled byeach participant was measured tothe nearest meter. Standardized en-couragement—“You are doing well,keep up the good work”—was pro-vided at 60-second intervals.49 Theoutcome was the distance walked in6 minutes. A previous investigationwith a similar group of subjects dem-onstrated the reliability and validityof data for this measure (intraclasscorrelation coefficient�.94 for test-retest reliability, SEM�26.3 m, andMDC90�61.34 m).15

Data AnalysisBefore beginning the modeling, weplotted the data to gain an impres-sion of the pattern of change overtime. Although one of the benefits ofusing mixed-effects modeling is thatit does not require the number andtiming of observations to be thesame across all participants, missingdata are still important. Bias will re-sult if the cause of the missing datapoints is related to the outcomethat would have been observed. Forexample, it would not be a problem

if a patient following arthroplastymissed an appointment due to achange in his or her schedule; how-ever, it would be a problem if thepatient missed the appointment be-cause of poor functioning due to anincrease in pain. Therefore, we alsoexamined the pattern of missing dataacross the time points.

Based on the plotted data, we devel-oped and tested several nonlinearmodels of change that related thedependent variable of functional sta-tus—either LEFS scores or 6MWTdistances—to the independent vari-able of number of weeks after arthro-plasty.50 The equation for our non-linear change model (model 1) was:

Functional status (LEFS or 6MWT)

� limit � (y0 � limit)

� e(�e(lnchange rate) � weeks),

where the functional status variableis the LEFS or 6MWT value, e is thebase of natural logarithms (approxi-mately 2.71828), weeks is the num-ber of weeks after arthroplasty; y0 isthe parameter that represents they-intercept value; limit is the param-eter that represents the asymptote ormaximum LEFS or 6MWT value, andlnchange rate is the natural log ofthe change rate (“change rate” re-fers to the rate of improvement atwhich patients approach their max-imum functional status). We esti-

Table 1.Preoperative Descriptive Statistics Expressed as Quartile Values (25th, 50th, 75th)

Measure FemaleParticipants(n�44)

MaleParticipants(n�40)

Age (y) 60, 64, 71 61, 67, 74

Body mass index (kg/m2) 29, 32, 34 27, 29, 32

Prearthroplasty Lower ExtremityFunctional Scale score

21, 27, 38 26, 38, 45

Prearthroplasty Six-Minute WalkTest distance (m)

312, 353, 416 397, 501, 552



mated the parameters using thenonlinear mixed-effects modelingpackage in S-Plus.50 A mixed-effectsapproach indicates that some param-eters have both fixed and randomeffects. The fixed effects describethe average change in the popula-tion (in this case, the sample ofparticipants who underwent TKA),and the random effects describethe individual differences amongparticipants. We explored dif-ferent models, and our final modelspecified the limit, y-intercept, andchange rate parameters as fixed ef-fects and the limit and y-intercept asrandom effects. Next, we examinedthe effect of including preoperativeLEFS scores and 6MWT distanceson the limit, y-intercept, and change

rate coefficients. Finally, becauseprevious work has shown that func-tional status levels differ by gen-der,22,51 we created separate modelsfor female and male participants.

ResultsAll participants completed baselinepreoperative assessments and had aminimum of 2 visits postoperatively.To summarize, 31 participants wereassessed 3 times, 18 were assessed 4times, 9 were assessed 5 times, and 2had 6 visits, with the rest of the sam-ple having 2 visits. SupplementalFigures 1 and 2 (available onlineonly at www.ptjournal.org) providespaghetti plots showing the datapoints and change profiles for eachparticipant.

All knee prostheses were posteriorstabilized, with the majority ce-mented. At our institution, the peri-operative management and rehabili-tation protocols are not influencedby prosthesis selection or method offixation. Postoperatively, one partic-ipant developed a documented deepvein thrombosis. None of the partic-ipants required revision surgerywithin the 1-year follow-up period.

Sixty-seven of the 84 participantscomposing the study sample wereassessed within 17 days of arthro-plasty. Of these 67 participants, 66completed the LEFS and 44 per-formed the 6MWT during this 17-dayperiod. The mean (�SD) preopera-tive LEFS score for those participantswho contributed LEFS data within 17days of arthroplasty was 32.3 points(SD�12.1) compared with 33.1points (SD�14.4) for those partici-pants who were not assessed withinthis period (t82�0.22, P2�.83). Sim-ilarly, the mean preoperative dis-tance for those participants whocontributed 6MWT data within 17days of arthroplasty was 428.4 m(SD�114.8) compared with 393.4 m(SD�105.8) for participants who didnot contribute 6MWT data withinthis period (t82�1.46, P2�.15). Fi-nally, the mean LEFS score assessedwithin 17 days of arthroplasty was26.0 points (SD�10.7) for the partic-ipants who contributed 6MWT dataduring this period compared with16.7 points (SD�10.2) for the partic-ipants who were assessed duringthis interval but not able to contrib-ute 6MWT data (t64�3.45, P2�.001).Post-arthroplasty assessments follow-ing the 3-week mark yielded approx-imately equal representation of LEFSand 6MWT data points.

Table 2 reports the fixed-effects pa-rameter values and the variation inrandom-effects parameter values forthe LEFS and 6MWT obtained frommodel 1. Also reported in Table 2 arethe standard deviations of individual

Table 2.Summary of Nonlinear Analysis Without Covariates

FemaleParticipants(n�44)


Lower Extremity Functional Scale analysis

Parameters of average changea

Limit (SE) 54.0 (2.3) 60.4 (2.3)

Y-intercept (SE) 10.4 (2.6) 19.0 (2.7)

Change rate (SE) �1.7 (0.1) �1.8 (0.1)

Standard deviation of individual differencesfrom average

Limit 12.1 11.6

Y-intercept 7.9 10.2

Within-patient variation 7.6 6.7

6-Minute Walk Test analysis

Parameters of average change

Limit (SE) 467.3 (15.4) 577.7 (18.2)

Y-intercept (SE) 154.7 (22.2) 185.7 (22.9)

Change rate (SE) �2.0 (0.1) �1.7 (0.1)


Limit 84.7 94.6



a Parameters of average change�fixed effects. Some parameters (ie, change rate) have only fixedeffects, indicating that there are no significant individual differences. SE�standard error.



differences from the estimated aver-age parameter values. For example,the standard deviation of individualdifferences from the average LEFSlimit value for the female partici-pants was 12.1. Accordingly, 68%of the female participants displayedlimit values from 42 to 66 LEFSpoints. The curves shown in Figure 1were generated by substituting theparameter estimates reported inTable 2 into model 1. This figureshows that most of the change oc-curred in the first 16 weeks afterarthroplasty.

Our exploration of the effect preop-erative functional status scores hadon limit values, y-intercept, andchange rate coefficients revealedthat limit values only were signifi-cantly affected. This finding indi-cates that preoperative levels offunction help to predict the maximalfunctional status that patients attainpostoperatively. Better preoperativescores will be associated with pa-tients attaining higher maximumpostoperative levels of function. Ac-cordingly, our revised model (model2) was as follows:

Functional status (LEFS or 6MWT)

� (limit � � � preoperative function)

� (y0 � limit � �

� preoperative function)

� e(�e(lnchange rate) � weeks),

where � is the regression coefficientassociated with preoperative func-tional status level. Gender- andmeasure-specific coefficients are re-ported in Table 3. Figures 2 and 3display the change curves for theLEFS and 6MWT adjusted for preop-erative scores. The 3 curves pre-sented in each figure depict thegender- and measure-specific changecurves based on the preoperativequartile values reported in Table 1.For example, the top curve in Figure

2A represents LEFS scores for maleparticipants with a preoperativeLEFS score of 45 points (ie, thirdquartile value reported in Tab. 1).The 16-week value of 61 points onthis curve was obtained by substitut-ing the coefficient values reported inTable 3 into model 2 and applying apreoperative value of 45 points.Again, these figures show that mostof the change occurred within thefirst 16 weeks after arthroplasty.

DiscussionOur goal was to describe the changein lower-extremity functional status

over a 1-year period for patients whounderwent TKA and received stan-dardized inpatient physical therapycare for 1 to 2 weeks (acute andsubacute short-term rehabilitation).The subsequent discussion will firstprovide a synthesis of our findingsand then illustrate applications ofthis information by referring to thevignette introduced early in thisarticle.

To our knowledge, this is the firststudy to sample patients at differenttime points over a 1-year period afterTKA and to apply a nonlinear mixed-

0

10

20

30

40

50

60

70

80

0 4 8 12 16 20 24 28 32 36 40 44 48 52

Weeks After Arthroplasty

Pre

dic

ted

LEF

S Sc

ore

Male Participants

Female Participants

0

100

200

300

400

500

600

700

0 4 8 12 16 20 24 28 32 36 40 44 48 52


Pre

dic

ted

6M

WT

Dis

tan

ce (

m)

Male ParticipantsFemale Participants

A

B

Figure 1.(A) Change in Lower Extremity Functional Scale (LEFS) scores over time. (B) Change in6-Minute Walk Test (6MWT) distances over time.



effects analysis to model change.Previously, members of our teamhave applied hierarchical linearmodeling and a second-degree poly-nomial to model LEFS scores and6MWT distances over the first 16weeks after arthroplasty.12,22,30 Par-ticipants’ LEFS scores and 6MWTdistances increased rapidly overthis period, and a second-degreepolynomial fit the data well withinthis interval. However, a second-degree polynomial specifies a parab-ola that clearly does not representthe change pattern of LEFS scoresor 6MWT distances over a 1-yearperiod, and it is for this reason that

we applied a nonlinear mixed-effectsanalysis. The current study’s resultsover the initial 16 weeks after TKAcompare favorably with those mod-eled using a second-degree poly-nomial in patients who similarly re-ceived a mixed-effects model ofphysical therapy intervention withunknown parameters including out-patient treatment and naturalrecovery.22,30

Our study also explored the effect ofpreoperative LEFS scores and 6-MWTdistances as potential predictors ofy-intercept, change rate, and limitvalues. Maximal LEFS scores and

6MWT distances were influencedonly by their respective preoperativevalues. Accordingly, it is importantthat clinicians take the preoperativevalue into account when making aprognosis concerning a patient’s fi-nal level of lower-extremity func-tional status.

As illustrated in the section on re-sponses to the clinical practice vi-gnette, we believe that graphical rep-resentations of recovery can be veryuseful in assisting clinicians tobenchmark recovery. The graphscan be used to compare measuredscores obtained on patients with thepredicted scores to monitor progressand guide treatment decisions. Nor-mative scores for the measures insimilar populations also are availableto enable further benchmarking.52,53

The recovery curves in our studyalso facilitate determination of thecritical time points for measuringchange. For example, if researcherswere interested in determining theeffect of interventions on improvingthe rate of recovery and the maxi-mum level of function attained, theycould apply these graphs to assist intheir decision making. More studiesare needed to determine the effectof various postoperative physicaltherapy interventions on recovery.Frequently cited assessment pointsare 3, 6, and 12 months after arthro-plasty13; however, based on the in-formation from the current study, toassess the effect of interventions, itwould be important to assess pa-tients more frequently in the first 3months. In addition, because most ofthe recovery has occurred by 6months, researchers might decide tonot assess individuals beyond thispoint to avoid unnecessary costs.

Table 3.Summary of Nonlinear Analysis With Preoperative Score as a Covariate

FemaleParticipants(n�44)


Lower Extremity Functional Scale analysis


Limit (SE)a 38.1 (5.3) 42.2 (5.6)

Preoperative (�) (SE) 0.50 (0.1) 0.50 (0.1)

Y-intercept (SE) 10.3 (2.6) 19.0 (2.7)

Change rate (SE) �1.7 (0.1) �1.8 (0.1)


Limit 10.4 9.8



6-Minute Walk Test analysis


Limit (SE) 277.2 (55.0) 326.2 (56.9)

Preoperative (�) (SE) 0.6 (0.1) 0.5 (0.1)

Y-intercept (SE) 154.7 (22.2) 188.6 (22.8)

Change rate (SE) �2.0 (0.1) �1.7 (0.1)


Limit 64.0 71.2



a SE�standard error.



Responses to the Clinical PracticeVignetteHow confident can the clinicianbe in the measured values of 28points on the LEFS and 261 mon the 6MWT, and how muchchange is required in these mea-sures to be reasonably certainthat a true change has occurred?To answer these questions, the re-sults from 2 other studies thatexamined the reliability of data forthe LEFS29 and 6MWT,15 whoseestimates are reported in theMethod section, are used. For exam-ple, the 90% confidence level (ie, 1SEM � 1.65) for an estimate of the“true score” is �6.1 points for theLEFS and 43.4 m for the 6MWT. Wecan say with 90% confidence thatMr Smith’s true LEFS score is likely tofall between 21.9 and 34.1 pointsand that his true 6MWT distance islikely to lie between 222.9 and 304.4m. To identify the minimal detect-able change (MDC), the confidencevalues reported are multiplied bythe square root of 2. For example,90% of patients who are truly stablewill display random fluctuations,when assessed on multiple occa-sions, of less than 9 points on theLEFS and 61.3 m on the 6MWT. Ac-cordingly, a change of 9 points ormore on the LEFS and of 61.3 m ormore on the 6MWT is interpretedas evidence of a true change. Esti-mates obtained from this approachoften are referred to as MDC withthe confidence value subscripted(eg, MDC90).54

What factors should the clinicianconsider in scheduling the nextassessment, and when should itoccur? Clearly, many factors, in-cluding feasibility, influence thechoice of reassessment interval. Twofactors specific to the context of thisarticle are MDC and the interval overwhich a typical patient is likely toachieve a change equal to the MDC.Mr Smith’s 2-week postoperativeLEFS value was 28, and the MDC90 is

a change of 9 points. This informa-tion is coupled with the curve forpatients with a preoperative LEFSvalue of 38 (ie, the curve closest toMr Smith’s value). Referring to themiddle curve presented in Figure 2A,it appears that a change of 9 LEFSpoints occurs between 2 and 4weeks postoperatively. Accordingly,the interval between assessments forthis specific instance is approxi-mately 2 weeks. Applying the sameapproach to 6MWT distance, it ap-pears that a minimum interval of 1

week is required for an expectedchange of 61.3 m.

What is Mr Smith’s lower-extremity functional status likelyto be in 8 weeks? To answer thisquestion, the clinician can inspectthe predicted functional status val-ues for 10 weeks after TKA (the firstassessment occurred at the 2-weekmark). For a person with Mr Smith’spreoperative values, the expectedLEFS and 6MWT scores are approxi-mately 52 points and 520 m, respec-

0

10

20

30

40

50

60

70

80


Pre

dic

ted

LEF

S Sc

ore

Preop LEFS 45Preop LEFS 38Preop LEFS 26

0

10

20

30

40

50

60

70


Pre

dic

ted

LEF

S Sc

ore

Preop LEFS 38Preop LEFS 27Preop LEFS 21

0 4 8 12 16 20 24 28 32 36 40 44 48 52

0 4 8 12 16 20 24 28 32 36 40 44 48 52

A

B

Figure 2.(A) Change in Lower Extremity Functional Scale (LEFS) scores for male participants,adjusted for preoperative (preop) LEFS scores. (B) Change in LEFS scores for femaleparticipants, adjusted for preoperative LEFS scores.



tively. Although the meaning of520 m is straightforward, the inter-pretation of an LEFS score of 52points is not intuitively obvious, andthe clinician will need to translatethis number into a narrative. Basedon information provided in a previ-ous article,30 a person with an LEFSscore of approximately 52 pointswill have: (1) “moderate difficulty”with heavy activities around thehouse, recreational activities, walk-ing a mile, or standing for 1 hour;(2) “a little bit of difficulty” with go-

ing up or down 10 stairs or lifting anobject such as a bag of groceriesfrom the floor; and (3) “no difficulty”sitting for 1 hour, putting on shoesor socks, or walking short distances.These data will likely assist the clini-cian in advising Mr Smith on what hecan expect regarding his mobility,which will likely assist Mr Smith indeciding whether he should con-sider rescheduling his trip.

What is Mr Smith’s maximumfunctional status level likely tobe? Recalling that a patient’s pre-operative level of function is a deter-minant of his or her postoperativemaximal function level, Figures 2Aand 3A are referenced to answer thisquestion. Because Mr Smith had apreoperative score of 40 points onthe LEFS, the middle curve is se-lected, and this would lead to a pre-diction that Mr Smith would have aterminal LEFS score of just over 60points. In terms of the 6MWT, usinga similar approach, the maximal dis-tance that Mr Smith would be able tocover would be around 600 m.

When is Mr Smith likely to reachhis maximum functional level?In both the case of the LEFS and6MWT, Mr Smith would reach hismaximum functional level sometimebetween 6 and 7 months.

Study LimitationsOne limitation is that all participantsin the change study were able tocomplete the LEFS and 6MWT pre-operatively. Accordingly, the gener-alizability of our findings are re-stricted to patients who are able tocomplete these tests preoperativelyand who have preoperative charac-teristics similar to those reported inTable 1. A second limitation is thatfewer participants provided 6MWTdata than LEFS data within a fewweeks of arthroplasty. Our analysisshowed that participants who wereassessed within 17 days of arthro-plasty and who did not contribute6MWT data during this period hadsignificantly lower LEFS scores at thistime point. A consequence of thismissing value pattern is that the pre-dicted 6MWT distances over the firstseveral weeks after arthroplasty arenot applicable to the entire sample,but rather are restricted to those par-ticipants who were capable of per-forming the 6MWT within this timeframe. This could have resulted inoverestimation of the predicted

0

100

200

300

400

500

600

700

0 4 8 12 16 20 24 28 32 36 40 44 48 52


Pre

dic

ted

6M

WT

Dis

tan

ce (

m)

Preop 552 m

Preop 501 mPreop 397 m

0

100

200

300

400

500

600


Pre

dic

ted

6M

WT

Dis

tan

ce (

m)

Preop 416 mPreop 353 m

Preop 312 m

A

B

0 4 8 12 16 20 24 28 32 36 40 44 48 52

Figure 3.(A) Change in 6-Minute Walk Test (6MWT) distances for male participants, adjusted forpreoperative (preop) 6MWT distances. (B) Change in 6MWT distances for femaleparticipants, adjusted for preoperative (preop) 6MWT distances.



scores for the 6MWT during thistime frame.

Although mixed-effects modelingwill stabilize the estimates of pa-tients who have limited data by an-choring them to the group average,it should be noted that 66% of theparticipants in this study were as-sessed only 2 or 3 times. More than50% of the participants were not as-sessed for both LEFS and 6MWT nearto or at the end of the study termi-nation. Finally, it was not possible todescribe with accuracy the patientswho were potentially eligible for thestudy due to study interruptionssuch as the outbreak of severe acuterespiratory syndrome.

ConclusionOur findings demonstrated that thegreatest improvement for the LEFSand 6MWT occurred in the first 12weeks after TKA. Improvement con-tinued to occur from 12 to 26 weeksafter TKA, although at a slower rater,and little improvement occurred be-yond 26 weeks after TKA. The maxi-mum scores obtained on the LEFS and6MWT were influenced by their re-spective preoperative scores. Clini-cians can use the recovery curves tomake prognoses concerning the rateof improvement in functional statusafter TKA and the expected time-specific and maximal functional statusscores. This information is critical toidentifying rehabilitation needs and as-sisting patients to set realistic goalsand plan their lives accordingly.

Ms Kennedy and Dr Gollish provided con-cept/idea/research design and fund procure-ment. Ms Kennedy, Mr Stratford, Dr Riddle,and Dr Hanna provided writing. Ms Kennedyand Mr Stratford provided data collec-tion and project management. Mr Stratfordand Dr Hanna provided data analysis. DrGollish provided subjects and institutionalliaisons. Ms Kennedy provided facilities/equipment. Mr Stratford, Dr Riddle,Dr Hanna, and Dr Gollish provided consul-tation (including review of manuscript be-fore submission). The authors acknowledge

Neil Reid for data collection, project man-agement, and clerical support.

This article was received February 11, 2007,and was accepted August 20, 2007.

DOI: 10.2522/ptj.20070051

References1 Peat G. Knee pain and osteoarthritis in

older adults: a review of community bur-den and current use of primary healthcare. Ann Rheum Dis. 2001;60:91–97.

2 Caracciolo B, Giaquinto S. Determinants ofthe subjective functional outcome of totaljoint arthroplasty. Arch Gerontol Geriatr.2005;41:169–176.

3 Hartley RC, Barton-Hanson NG, Finley R,Parkinson RW. Early patient outcomes af-ter primary and revision total knee arthro-plasty. J Bone Joint Surg Br. 2002;84:994–999.

4 Lavernia CJ, Guzman JF, Gachupin-GarciaA. Cost effectiveness and quality of life inknee arthroplasty. Clin Orthop. 1997;(345):134–139.

5 Kurtz S, Mowat F, Ong K, et al. Prevalenceof primary and revision total hip and kneearthroplasty in the United States from1990 through 2002. J Bone Joint Surg Am.2005;87:1487–1497.

6 NIH consensus statement on total knee re-placement, December 8–10, 2003. J BoneJoint Surg Am. 2004;86:1328–1335.

7 Parent E, Moffet H. Comparative respon-siveness of locomotor tests and question-naires used to follow early recovery aftertotal knee arthroplasty. Arch Phys Med Re-habil. 2002;83:70–80.

8 Maly MR, Costigan PA, Olney SJ. Determi-nants of self-report outcome measures inpeople with knee osteoarthritis. ArchPhys Med Rehabil. 2006;87:96–104.

9 Stratford PW, Kennedy DM. Performancemeasures were necessary to obtain a com-plete picture of osteoarthritic patients.J Clin Epidemiol. 2006;59:160–167.

10 Stratford PW, Kennedy DM, WoodhouseLJ. Performance measures provide assess-ments of pain and function in patientswith advanced osteoarthritis of the hip orknee. Phys Ther. 2006;86:1489–1496.

11 Terwee CB, van der Slikke RMA, van Lum-mel RC, et al. Self-reported physical func-tioning was more influenced by pain thanperformance-based physical functioningin knee-osteoarthritis patients. J Clin Epi-demiol. 2006;59:724–731.

12 Kennedy DM, Stratford PW, Hanna SE, et al.Modeling early recovery of physical functionfollowing hip and knee arthroplasty. BMCMusculoskelet Disord. 2006;7:100.

13 Ethgen O, Bruyere O, Richy F, et al.Health-related quality of life in total hipand total knee arthroplasty: a qualitativeand systematic review of the literature.J Bone Joint Surg Am. 2004;86:963–974.

14 Terwee CB, Mokkink LB, Steultjens MP,Dekker J. Performance-based methods formeasuring the physical function of pa-tients with osteoarthritis of the hip orknee: a systematic review of measurementproperties. Rheumatology (Oxford). 2006;45:890–902.

15 Kennedy DM, Stratford PW, Wessel J, et al.Assessing stability and change of four per-formance measures: a longitudinal studyevaluating outcome following total hipand knee arthroplasty. BMC Musculoske-let Disord. 2005;6:3.

16 Avramidis K, Strike PW, Taylor PN, SwainID. Effectiveness of electric stimulation ofthe vastus medialis muscle in the rehabil-itation of patients after total knee arthro-plasty. Arch Phys Med Rehabil. 2003;84:1850–1853.

17 Beaupre LA, Davies DM, Jones CA, CinatsJG. Exercise combined with continuouspassive motion or slider board therapycompared with exercise only: a random-ized controlled trial of patients followingtotal knee arthroplasty. Phys Ther.2001;81:1029–1037.

18 Kramer JF, Speechley M, Bourne R, Rora-beck C. Comparison of clinic-and home-based rehabilitation programs after totalknee arthroplasty. Clin Orthop. 2003;410:225–234.

19 Fitzgerald JD, Orav EJ, Lee TH, et al. Pa-tient quality of life during the 12 monthsfollowing joint replacement surgery.Arthritis Rheum. 2004;51:100–109.

20 Fortin PR, Clarke AE, Joseph L, et al. Out-comes of total hip and knee replace-ment: preoperative functional status pre-dicts outcomes at six months after surgery.Arthritis Rheum. 1999;42:1722–1728.

21 Fortin PR, Penrod JR, Clarke AE, et al. Tim-ing of total joint replacement affects clin-ical outcomes among patients with osteo-arthritis of the hip or knee. ArthritisRheum. 2002;46:3327–3330.

22 Kennedy DM, Hanna SE, Stratford PW,et al. Preoperative function and genderpredict pattern of functional recovery af-ter hip and knee arthroplasty. J Arthro-plasty. 2006;21:559–566.

23 Lingard EA, Katz JN, Wright EA, Sledge CB.Predicting the outcome of total knee ar-throplasty. J Bone Joint Surg Am. 2004;86:2179–2186.

24 Mizner RL, Petterson SC, Snyder-MacklerL. Quadriceps strength and the timecourse of functional recovery after totalknee arthroplasty. J Orthop Sports PhysTher. 2005;35:424–436.

25 Mizner RL, Petterson SC, Stevens JE, et al.Preoperative quadriceps strength pre-dicts functional ability one year after totalknee arthroplasty. J Rheumatol. 2005;32:1533–1539.

26 Aarons H, Hall G, Hughes S, Salmon P.Short-term recovery from hip and knee ar-throplasty. J Bone Joint Surg Br. 1996;78:555–558.

27 Stratford PW, Kennedy DM, Hanna SE.Condition-specific Western Ontario Mc-Master Osteoarthritis Index was not supe-rior to region-specific Lower ExtremityFunctional Scale at detecting change.J Clin Epidemiol. 2004;57:1025–1032.



28 Binkley JM, Stratford PW, Lott SA, RiddleDL; for the North American OrthopaedicRehabilitation Research Network. TheLower Extremity Functional Scale (LEFS):scale development, measurement proper-ties, and clinical application. Phys Ther.1999;79:371–383.

29 Stratford PW, Binkley JM, Watson J, Heath-Jones T. Validation of the LEFS on patientswith total joint arthroplasty. PhysiotherCan. 2000;52:97–105.

30 Stratford PW, Hart DL, Binkley JM, et al. In-terpreting lower extremity functional statusscores. Physiother Can. 2005;57:154–162.

31 Faucher M, Poiraudeau S, Lefevre-ColauMM, et al. Algo-functional assessment ofknee osteoarthritis: comparison of thetest-retest reliability and construct validityof the WOMAC and Lequesne indexes. Os-teoarthritis Cartilage. 2002;10:602–610.

32 Guermazi M, Poiraudeau S, Yahia M, et al.Translation, adaptation and validation ofthe Western Ontario and McMaster Uni-versities Osteoarthritis Index (WOMAC)for an Arab population: the Sfax modifiedWOMAC. Osteoarthritis Cartilage. 2004;12:459–468.

33 Kennedy DM, Stratford PW, Pagura SMC,et al. Exploring the factorial validity andclinical interpretability of the Western On-tario and McMaster Universities Osteoar-thritis Index (WOMAC). Physiother Can.2003;55:160–168.

34 Thumboo J, Chew LH, Soh CH. Validationof the Western Ontario and McMaster Uni-versity osteoarthritis index in Asians withosteoarthritis in Singapore. OsteoarthritisCartilage. 2001;9:440–446.

35 Bautmans I, Lambert M, Mets T. The six-minute walk test in community dwellingelderly: influence of health status. BMCGeriatr. 2004;4:6.

36 Bean JF, Kiely DK, Leveille SG, et al. The6-minute walk test in mobility-limited el-ders: what is being measured? J GerontolA Biol Sci Med Sci. 2002;57:M751–M756.

37 Enright PL, McBurnie MA, Bittner V, et al.The 6-min walk test: a quick measure offunctional status in elderly adults. Chest.2003;123:387–398.

38 Harada ND, Chiu V, Stewart AL. Mobility-related function in older adults: assess-ment with a 6-minute walk test. Arch PhysMed Rehabil. 1999;80:837–841.

39 Shumway-Cook A, Patla AE, Stewart A,et al. Environmental demands associatedwith community mobility in older adultswith and without mobility. Phys Ther.2002;82:670–681.

40 Rosenbaum PL, Walter SD, Hanna SE, et al.Prognosis for gross motor function in ce-rebral palsy: creation of motor develop-ment curves. JAMA. 2002;288:1357–1363.

41 Jette AM. Toward a common language forfunction, disability, and health. Phys Ther.2006;86:726–734.

42 Nagi SZ. A study in the evaluation of disabil-ity and rehabilitation potential: concepts,methods, and procedures. Am J PublicHealth Nations Health. 1964:1568–1579.

43 Alcock GK, Stratford PW. Validation of theLower Extremity Functional Scale on ath-letic subjects with ankle sprains. Phys-iother Can. 2002;54:233–240.

44 Riddle DL, Pulisic M, Sparrow K. Impact ofdemographic and impairment-related vari-ables on disability associated with plantarfasciitis. Foot Ankle Int. 2004;25:311–317.

45 Watson CJ, Propps M, Ratner J, et al. Reli-ability and responsiveness of the lower ex-tremity functional scale and the anteriorknee pain scale in patients with anteriorknee pain. J Orthop Sports Phys Ther.2005;35:136–146.

46 Pankaj J, Kramer JF, Birmingham T. Com-parison of the Western Ontario and Mc-Master Universities Osteoarthritis Index(WOMAC) and the Lower Extremity Func-tional Scale (LEFS) questionnaires in pa-tients awaiting or having undergone totalknee arthroplasty. Physiother Can. 2005;57:208–216.

47 Kreibich DN, Vaz M, Bourne RB, et al.What is the best way of assessing outcomeafter total knee replacement? Clin OrthopRelat Res. 1996;(331):221–225.

48 Ouellet D, Moffet H. Locomotor deficits be-fore and two months after knee arthroplas-ty. Arthritis Rheum. 2002;47:484–493.

49 Guyatt GH, Pugsley SO, Sullivan MJ, et al.Effect of encouragement on walking testperformance. Thorax. 1984;39:818–822.

50 Pinheiro JC, Bates DM. Mixed Effects Mod-els in S and S-Plus. New York, NY:Springer Verlag New York; 2000.

51 Kennedy DM, Stratford PW, Pagura SMC,et al. Comparison of gender and group dif-ferences in self-report and physical perfor-mance measures in total hip and knee ar-throplasty candidates. J Arthroplasty.2002;17:70–77.

52 Lieberman JR, Hawker G, Wright JG. Hipfunction in patients �55 years old: popu-lation reference values. J Arthroplasty.2001;16:901–904.

53 Steffen TM, Hacker TA, Mollinger L. Age-and gender-related test performance incommunity-dwelling elderly people: Six-Minute Walk Test, Berg Balance Scale,Timed “Up & Go” Test, and gait speeds.Phys Ther. 2002;82:128–137.

54 Beaton DE, Bombardier C, Katz JN, WrightJG. A taxonomy for responsiveness. J ClinEpidemiol. 2001;54:1204–1217.



Assessing recovery and establishing prognosis following total knee arthroplasty

Health & Medicine

Transcript of Assessing recovery and establishing prognosis following total knee arthroplasty