Reduction in Fear of Falling Through Intense Tai Chi ExerciseTraining in Older, Transitionally Frail AdultsRichard W. Sattin, MD,* Kirk A. Easley, MS/Steven L. Wolf, PhD, FAPTA,^ Ying Chen, MS/andMichael H. Kutner, ^

OBJECTIVES: To determine whether an intense tai chiexercise program could reduce fear of falling better than awellness education (WE) program in older adults who hadfallen previously and meet criteria for transitioning to frail-ry-DESIGN: Cluster-randomized, controlled trial of 48weeks' duration.SETTING: Ten matched pairs of congregate living facili-ties in the greater Atlanta area.PARTICIPANTS: Sample of 291 women and 20 men, aged70 to 97.MEASUREMENTS: Activity-related fear of falling usingthe Activities-Specific Balance Confidence Scale (ABC) andthe Fall Efficacy Scale at baseline and every 4 months for 1year. Demographics, time to first fall and all subsequentfalls, functional measures. Centers for Hpidemiologic Stud-ies Depression Scale, medication use, level of physical ac-tivity, comorbidities, and adherence to interventions.

RESULTS: Mean ABC was similar in both cohort groups atthe time of randomization but became significantly higher(decreased fear) in the tai chi cohort at 8 months (57.9 vs49.0, P<.001) and at study end (59.2 vs 47.9, P<.001).After adjusting for covariates, the mean ABC after 12months of intervention was significantly greater in the taichi group than in the WE group, with the differences in-creasing with time (mean difference at 12 months = 9.5points, 95% confidence interval = 4.8-14.2, P<.001).

CONCLUSION: Tai chi led to a significantly greater re-duction m fear of falling than a WE program in transition-

From the *National Center for Injury Prevention and Conrrol, Centerstor Disease Ontrol ;ind Prevention, Atlanta, Georgia; and Department ofRiiistatistics, Rollin!> Schoc)! of Public Health, and 'Department ofRehabilitation Medicine. School of Medicine, Kmory University, Atlanta,Georgia.

This research was supported by National Institutes of Health Grant AG14767 from the National Institute on Aging and coupons for redeemableproducts from the Kroger Clorporatinn and CVS Pharmacies for eachparticipant upon completion of participatitin.

Address correspondence to Richard W. Sattm, MD, KAGP, Division of hijuryand Disability Outcomes and Programs, National Center for injuryIVevi-ntion and Control, Centers for Disease (]nntrnl and PreventionlF-411,4770 Buford Highway, Atlanta, GA 30341. F-mail: rsattin@cdi;.gov

DOI: 10.111 l/j. 1532-5415.2005..53375.X

ally frail older adults. The mean percentage change in ABCscores widened between tai chi and WE participantsover the trial period. Tai chi should be considered in anyprogram designed to reduce falling and fear of fallingin transitionally frail older adults. J Am Geriatr Soc53:1168-1178,2005.

Key words: fear of falling; exercise; balance; aging; tai chi

Fear of falling ranks as the top fear of community-dwelling older persons.' The wide range of prevalence

estimates for fear of falling (29-77%) may be due todifferences between samples in age, sex, activity level,history of falls, measuresof fear, and other comorbidities.""^Individuals who have fear of falling are at greater risk offalling, and those who have fallen are at greater risk ofdeveloping fear of falling.'' An escalating cycle of fear andfalling or falling and fear can lead to decreased physicalactivity, decreased functional ability, and decreased qualityof life.'"'" This escalating cycle can, thus, adversely affectefforts to socialize and engage in activities that would lessenthe risk of falling and other serious health conditions. Ef-fective interventions to lessen fear of falling could thereforehave significant consequences for individuals, their fami-lies, and healthcare delivery systems. Understanding fear offalling alone, without consideration of other chronic con-ditions and depressive symptoms, may not be sufficient toguide intervention activities."*

Few studies have evaluated interventions to reduce fearof falling. Recently, one' ' showed that seniors involved in aphysical activity program or an education program had adecrease in fear of falling from their baseline measures offear. Another study of 434 persons aged 60 and oldershowed that fear of falling could be reduced using a cog-nitive-behavioral program but only if the results were re-stricted to participants comphant with the program ratherthan ail randomized participants.'- Part of the Frailty andInjuries: Cooperative Studies of Intervention Techniques(FICSIT) trial'^ showed that a multifactorial interventioncould reduce fear of falling in those who, on average, hadminimal fear of falling at baseline. Also as part of FICSIT, it

JAGS .^3:1168-1178,2005{" 2005 by the American Geriatrics Society 0002-8614/().S/$15.00

JAGS JULY 2005-VOL. 53, NO. 7 TAI CHI AND FEAR OF FALLING IN OLDER ADULTS 1169

was found th;it, in robust older persons, a 16-week tai chiintervention resulted in a statistically significantly greaterreduction in fear of falling than in an exercise controlgroup.''*

Recently, it was shown that a 48-weck tai chi inter-vention for transitionally frail older adults can lead to areduction in falls (primary outcotne) after a 3-month la-tency period in learning the basic tai chi exercise forms.'"'Tai chi training may improve strength and flexibility,"'"'^balance^" blood pressure,''' cardiorespiratory function/"and self-esteem-' in older adults.

The prevalence of fear of falling and its associationwith several baseline demographic, functional, and behavi-oral characteristics in older transitionally frail older adultswas previously reported.-^^ Given the appareiit effects of taichi on balance, strength, and self-esteem and reduction infalls, the aim of this analysis was to determine the extent towhich tai chi affects fear of falling, a secondary outcome ofthe 48-week single-blinded cluster-randonii/.ed controlledintervention trial in transitionally frail older adults.

METHODS

FacilitiesForty-nine independent living facilities in the greater At-lanta metrt)politan area were considered, and 43 were vis-ited.^* From this list of 49, the study was conducted in 20congregate living facilities between September 1997 andAugust 2001. The two primary reasons for why an inde-pendent living facility was excluded were evidence of poorstaff support for implementing the study or too small a poolof participants, usually fewer than 75 potential candi-dates/^ The target recruitment goal was to enroll at least 15participants from each facility. Twelve facilities had morethan 15 participants (maximum of 19), and all participantswere allowed to enroll. Facilities were recruited in pairs bywhether they were Housing and Urban Development(HUD) (n = 14) or private (n ^ 6) sites and then randomi-zed to the tai chi or WF intervention group. A total capacityof 4,032 older adults could reside in the 20 congregate liv-ing facilities; the number per facility ranged from 98 to 349.Fvaluators were blinded to intervention allocation, instruc-tors were blinded to outcome measures, and participantswere instructed not to disclose the intervention they re-ceived. The Emory University human investigation com-mittee and the Centers for Disease Control and Preventioninstitutional review board approved this study. Written in-formed consent was obtained from all participants beforetheir enrollment.

Study ParticipantsDetails regarding the design and methodology of this trialhave been reported elsewhere.^* Briefly, inclusion criteriawere age 70 and older, ambulatory (with or without assist-ive device), no severe or unstable medical conditions, ab-sence of severe psychological conditions, no significantcognitive impairment (Mini-Mental State Examination(MMSF') score >24), history of one or more falls in theprevious year, and transitioning to frailty (based uponSpeechley and Tinetti's classification-'^ of older adults intovigorous, transitional, and frail categories). Staff and ad-

ministrators at the 20 facilities identified 2,288 of the 4,032residents as being aged 70 and older, ambulatory, and cog-nitively competent. Forty-eight percent (n = 1,095) of theseidentified residents attended a recruitment event held ateach facility. After each recruitment presentation, everyone(n ^ 688) who had signed interest sheets at the event wascalled. Working from a supplementary facility list, 696 ad-ditional potential participants were contacted via cold calls,yielding a total of 1,384 persons contacted after recruitmentevents. The main reasons for not scheduling persons forscreening were an absence of falls within the previous 12months (n = 359), self-report of poor health (n=152),noninterest (n = 146), reluctance to commit to the time re-quired for study participation (n — 116), reluctance to beingrandomized (n = 32), not being aged 70 and older (n = 32),too busy (n ^ 30), and other (n = 81). Of the remaining 436persons who scheduled a screening appointment, 82 with-drew or canceled before the appointment, leaving 354 whowere screened for participation.

InterventionsTai chi consists of slow, rhythmic movements that empha-size trunk rotation, weight shifting, coordination, and agradual narrowing of lower extremity stance. Six of the 24simplified tai chi forms that best exemplified these move-ments and could be combined into a final product (2 con-tinuous minutes of unassisted tai chi exercise) wereused. ""—' All tai chi exercise was standardized by havingthe two instructors practice with one another until theirexecution of the movement forms to be taught in each classwere identical. One instructor was a tai chi grand master,and the other was his student, who had studied with him formore than 5 years. Participants progressed from often beingdependent on assistive devices for upright support to per-forming 2 continuous minutes of tai chi without support.Intense tai chi was defined as two sessions per week at in-creasing durations starting at 60 minutes contact time andprogressing to 90 minutes over the course of 48 weeks. Theactual work time, exclusive of warm up and cool down,progressed from about 10 minutes to 50 minutes over thecourse of the 48 weeks.

The wellness education (WF) program was given atparticipating facilities for an hour each week and consistedof instruction about falls prevention; exercise and balance;diet and nutrition; pharmacological management; legal is-sues relevant to health; changes in body function; and men-tal health issues, such as stress, depression, and life changes.Interactive handout materials were provided, but there wasno formal instruction in exercise. The total time for indi-vidual attention from instructor to participants in eachgroup was comparable.

Measurement of Fear of FallingThe Falls Efficacy Scale (FES)-'' and the Activities-SpecificBalance Confidence Scale (ABC)-* are the two most fre-quently used surrogate measures for fear of falling in olderpersons. The reliability and validity of ABC and FES havebeen previously reported.-''"-'^ The FES and the \EQ] weremeasured at baseline and at 4-month intervals for 12months.

1170 SATTIN ETAL. JULY 2005-VOL. 53, NO. 7 JAGS

Falls Efficacy ScaleThe FES is based on the operational definition of fear as"low perceived self-confidence at avoiding falls during es-sential, relatively nonhazardous activities." A modified FESversion used in the FICSIT trials was used here.^" Briefly,participants were asked how concerned they were about thepossibility of falling while performing 10 different activitieson a four-category scale from 1 (not at all concerned) to 4(very concerned}. If participants indicated that they did notperform or were unable to perform the activity, they wereencouraged to respond hypothetically. The FES emphasizesmainly indoor, home based activities.

Activities-Specific Balance Confidence ScaleThe ABC assesses the confidence with which one can engagein a wide range of activities of daily living, including thoseperformed outside the home. For 16 specific activities, theparticipants were asked how confident they were in notlosing their balance or becoming unsteady. Answers wererated on a 0% (no confidence) to 100% (complete confi-dence) scale. The ABC represents a wider activity rangethan the FES.

Additional Measurements and DefinitionsTrained evaiuators collected demographic, medical, func-tional, and behavioral data at baseline and every 4 monthsfor each study participant. Interratcr reliability of the eval-uation team was maintained at a kappa of 0.98 or greater.For this analysis, demographic variables and selected meas-ures were included based on their proven reliability andvalidity or their known association with falling or fear offalling.'"'"'"''"" Depression was assessed using the Centerfor Epidemiological Studies Depression Scale (CES-D)."This test for depression consists of 20 items measuring fourdomains of depressive symptoms, including depressive af-fect, positive affect, somatic symptoms, and interpersonalrelationships. Scores of 16 or greater are indicative of de-pression. Gait speed was assessed using time to complete a10-m walk;'^ the median value for all study participantswas 0.97 m/s. Activity level at baseline was defined as activeif participants reported exercising for at least 1 hour perweek before study enrollment; otherwise they were con-sidered sedentary. Other measures included the distance onecould reach the arm forward at 90 shoulder flexion with-out moving the feet (functional reach test), -^ use of seda-tives (narcotics, benzodiazepines, antidepressants, orphenothiazines), and numher of fall events. Fall eventswere defined as events in which the participant uninten-tionally came to rest on an object (e.g., person, table, chestof drawers) that prevented center of mass from exceedingthe base of support or came to rest on the floor or a lowerobject becatise the center of mass exceeded the base of sup-port.'"^ Details on the methodology to ascertain falls havehcen dcscrlhed previously.' "-^^

DropoutsDrtjpouts were defined as noncompliance or ahsences byany participant who missed more than 8 consecutive weeksof the intervention. Beyond this interval, learning tai chimovement forms or WF materials already covered by therespective classes would have impeded the progress of those

classes. Every effort was made to secure quarterly data fromall randomized participants.

Statistical AnalysesSample size and power considerations for the primary out-come, number of falls, have been described previously.'^'^*The primary analyses of the data were performed accordingto patients' original treatment assignment (intent-to-treatprinciple) and the inclusion of all data from all patientsrandomized in the final analysis. Baseline measurements oftai chi and WE participants were compared using a per-mutation test in which the unit of randomization is theindependent living facility pair. ' Baseline characteristics ofcompleters and noncompleters for each intervention werecompared with a one-sample / test on the difference, inwhich the unit of analysis was the facility.

Repeated-measures analyses using mixed linear modelswere performed for ABC and FES. These analyses includedparticipant-level characteristics and congregate living facil-ities variation or clustering. Participants were clustered ornested within living facilities, with living facility pairs as arandom effect and the participant as the experimental unit.A priori, selected demographic characteristics (age, sex,ethnicity) and risk factors for fear of falling (activity level atbaseline, depression, gait speed, functional reach, use ofsedatives, number of falls) were included as covariates,based upcm previous literature.'"'"-''^ For each covariate-adjusted analysis in Table 2, a saturated model was fit thatincluded the main effects, the two-way interaction effects,and the three-way interaction. For example, for activitylevel at baseline, the two-way statistical interaction includ-ed intervention by time, intervention by activity level, andactivity level by time, and the three-way interaction includ-ed intervention by activity level hy time. Repeated-measuresanalyses of ABC and FES scores and percentage changefrom baseline were performed using a means model usingSAS Proc Mixed (version 8, SAS Institute, Inc., Cary, NC),providing separate estimates of the means hy time on studyand intervention group. A compound-symmetry form in therepeated measurements was assumed for each outcome, androbust estimates of the standard errors of parameters wereused to perform statistical tests and construct 95% confi-dence intervals (CIs). * ' ^ Variance components includedthe cluster and the between- and within-participant com-ponents. The model-based means are unbiased with unbal-anced and missing data, so long as the missing data arenoninformative (missing at random). Sensitivity analysesfor missing data, including those for participants who com-pleted follow-up only, last observation carried forward,baseline observation carried forward, and multiple impu-tation (implemented using SAS Proc MI), showed similarresults to repeated measures analyses, which suggest thatthe missing data were noninformative. Statistical tests weretwo-sided. A Bonferroni adjustment (P<.01) was used forthe comparison of tai chi versus WE made at each of thefour time points.

RESULTSFrom December 1997 to September 1999, 311 personswere enrolled in this trial. A recruitment and randomizationflow chart is shown in Figure 1. Twenty-four participants

JAGS JULY 2005-VOI.. 53, NO. 7 TAI CHI AND FEAR OF FALLING IN OLDER ADULTS 1171

Screened by telephone(n= 1.384)

Assessed for eligibility

Did nol meet inclusion criteria

Randiimi/ed

Allocated Uiiui chi (n=l58lDid mil enler sludy (n - 12]

Did nol wunl to participate |n = 3|Health status too poor |n = 4|Catastrophic event (n = I)Died (n = 01Doctor refused con.scnl In = 4)Excluded due to unreporled

Parkinson's disease |n = 11

Allocated lo wellness In = L'i3)Did mil enter .sludy (n = 12)

Did mit want to partieipate (n = 4)Heallh slalus too poor (n = I)Caiasirophic event In - 2)Died In = 2)Doctor refused consent In = 3)Rxcluded due to unreixined

Parkinson's disease (n = 0)

Discontinued intervention |n = 37)Declining heallh-In = 28)Inadequate lime for cla.ss |n = 4)Died In ^21Did not like intervention In = I)Depressed In = 2)Spouse's declining heallh |n = 0)

Discontinued intervention (n = 32)Declining heallh-In = 24)Inadequate lime for class (n = 2)Died (n = 2)Did not like intervention |n = 2)Depressed Ui= 0]Spouse's declining health (n = 2)

Included in analysis|n= I5S)

Included in analysisI n = I .'i.i 1

Figure 1. Progress through the stages of the rai chi (TC) andWellness Education (WE) study, including How of participants,withdrawals, and inclusion in analyses. Not meeting inclusioncriteria: too trail (n - 7), chose to stop screening procedure(n — 7), progressive neurological disease (n — 6), too rohust(n - 5)., unstable coronary status (n - 5], Mini-Mcntal State Ex-amination score <24 (n-4) , orthopedic problems (n = 3),emotionally unstable/early dementia (n - 2), recent cerebrovas-cular accident [n - 2), untreated abdominal aneurysm (n = 1),and severely compromised vision (n - 2).' Declining health rea-sons: injuries/fractures (n - 4, TC; n - 5, WE), deteriorating vi-sion (n - 3, TC), cardiac (n = .3, TC; n - ^, WE),musculoskeletal/neurological impairment (n = 10, TC; n — 4,WE), loss oi independence [n - 1, TC; n = 6, WE), stroke |n — I,TC; n = 2, WE), cancer (n = 1, TC; n - 2, WE), diabetes mellitus(n — 1, TC; n - I, WE), diverticulitis (n - 1, TC), major surgery(n - 1, TC), asthma (n — 1, TC), hospitalized at time of postin-tervention mterview {participant later died) {n = 1, TC), andprolonged illness (n — 1,

randomized to tai chi (n = 12} and WE (n = 12) withdrewimmediately after randomization because they were nolonger interested (n ^ 7), perceived their health to be toopoor (n — 5), were denied physician approval to participate(n = 7), experienced catastrophic health events (n = 3), ordied (n = 2). One additional participant was later excludedbecause of a previously undetected diagnosis of Parkinson'sdisease.

The mean ABC and FES score at baseline was similarbetween the tai chi and the WE groups (Table 1). Most

other baseline characteristics of the tai chi and WE groupswere also similar, except that persons in the tai chi grouphad better functional reach (?=.O5} than those in the WEgroup. The 24 persons who dropped out after randomizedassignment had similar baseline characteristics to those intheir respective treatment groups. Of the remaining 286participants, 69 (24%) did not complete the intervention:37 in the tai chi group and 32 in the WE group (P ^.58).

The baseline characteristics t)f the 217 persons whocompleted the entire 48-week trial were compared withthose of the 94 persons (50 in the tai chi group and 44 in theWE group) who did not complete the entire trial (Table 1).Overall, completers and noncompleters were similar in age,sex, race, education, presence of depression, activity level,gait speed, impairment in gait/balance, presence of lowerextremity disability, use of sedatives, body mass index, andfear of falling according to the ABC!! and EES. Noncomple-ters had less functional reach than did completers(P<.001). Tai chi noncompleters were more likely thantai chi completers to have slower gait speed (P = .02), moreimpairment in gait/balance (P=.O4), and worse functionalreach (P<.001), and WE noncompleters were more likelythan WE completers to have worse functional reach (P =.05).

ABC in the two study cohorts changed in significantlydifferent ways during the intervention trial (P<.001, testfor interaction between intervention group and time) (Table2, Figure 2A,C). Mean ABC was similar in both cohortgroups at the time of randomization but became signifi-cantly higher (better) in the tai chi cohort at 8 months (57.9vs 49.0, P < .001) and at 12 months (59.2 vs 47.9, P< .001).Mean differences at 8 months and 12 months were 8.9pomts [95% CI-3.8-13.9) and 11.3 points (95%CI — 6.3-16.4), respectively. Mean EES became significant-ly lower (better) in the tai chi cohort at 8 months (18.4 vs20.5, P = .01) and at 12 months (17.6 vs 21.2, P< .001). Allsubsequent analyses reported herein use the ABC~ only.

The pattern of change in ABC means over time bytreatment group was consistently different for one level of acovariate than for the other level for baseline activity level,baseline depression, baseline functional reach, use of sed-atives, and number of falls during the study period (Table2}. Eor each level of these covariates, the tai chi group had astatistically significantly greater mean ABC over time thanthe WF ; this difference between the tai chi and WE groupswas greatest at 12 months. Fiowever, for race and gaitspeed, an intervention effect was seen for Caucasians butnot African Americans.

All of the variables listed in Table 2 were included in amultivariable model. After adjusting for each covariate, themean ABC after 12 months was significantly greater in thetai chi group than in the WF! group, with increasing differ-ences over time (Table 3, Figure 2B). The mean differenceafter 12 months of intervention was 9.5 points higher in thetai chi group than the WE group (95% CI - 4.8-14.2). Thepercentage change in ABC from baseline increased over the12 months in the tai chi group (mean percentage increase =13.4%, 95% C! = 6.4-20.4), but, in the WE group, thepercentage change in ABC was less than baseline at 12months (mean percentage decrease = 4.2%) (Eigure ID).

Adherence to the tai chi exercise and to the WFL pro-gram was assessed by maintaining weekly attendancerecords and calculating the percentage of total sessions

1172 SATTIN ETAL.

Table 1. Baseline Characteristics of the

Characteristic

Sex, n (%)MaleFemale

Race, n (%)CaucasianNon-Caucasian*

Education, n (%)< 12 years>12+years

Depression, n (%)"AbsentPresent

Activity level, n (%)SedentaryActive

Gait speed, n (%)< 0.97 m/s>0.97 m/s

Impairment gait/balance, n (%)ImpairedNormal

Lower extremity disability, n (%)*AbsentPresent

Functional reach, n (%)< 10 inches>10 inches

Use of sedatives, n (%)"YesNo

Age, mean ± SDBody mass index, kg/m^, mean ± SDFear of falling

Study Completcrs

Compieters(n= 108)

5(5)103(95)

86 (80)22 (20)

20(19)88(81)

86 (80)22 (20)

41 (38)67 (62)

48 (44)60 (56)

96 (89)12(11)

41 (38)67 (62)

21 (19)^87(81)^

25 (23)83 (77)

80.4 ± 3.126.9 ± 2.5

Activities-specific Balance and Confidence 53.5 ± 9.2Scale, mean ± SD'

Falls Efficacy Scale, mean l- SD" 19.9 ±2.2

Versus Noncompleters

Tai Chi

Noncompleters(n = 50)

5(10)45 (90)

40 (80)10 (20)

11 (22)39 (78)

34 (68)16 (32)

23 (46)27 (54)

28 (56)22 (44)

50(100)0(0)

19(38)31 (62)

23 (46)^27 (54)^

14(29)6(71)

82.6 ± 3.728.1 ± 3.6

49.9 ± 12.2

20.4 = 2.6

P-value

.35

.80

.16

.19

.55

.02

.04

.62

.0002

.29

.12

.16

.41

.63

JULY 2005-VOL. 53, NO. 7

for the Two Intervention Group:

Wellness Education

Completers(n= 109)

7(6)102 (94)

92 (84)17(16)

24 (22)85 (78)

84 (77)25 (23)

45(41)64 (59)

63 (58)46 (42)

102(94)7(6)

42 (39)67(61)

42 (39)^67(61)^

26 (24)83 (76)

80.5 ± 3.226.5 + 2.2

52.1 ± 6.0

19.9 ± 1.5

Noncompleters(n = 44)

3(7)41 (93)

32 (73)12 (27)

11 (25)33 (75)

29 (66)15(34)

26 (59)18(41)

26 (59)18(41)

42 (95)2(5)

18(41)26 (59)

25 (57)^19(43)^

12 (27)32 (73)

80.9 n 2.827.3 = 2.5

47.1 ± 12.8

21.2 i. 4.9

JAGS

P-value

.61

.29

.63

.22

.22

.82

.92

.92

.05

.58

.78

.12

.20

.39

Fifty-five Atriuan-American and si\ other.Using Center for Kpiilemiolnjjii: .Studies Depression Scale; a score of > 16 indicates presence of depression.

'Gross lower extremity motor impairment; disahility and impairment hasftl upon reference 24 definitions.• All participants combmed m the tai chi group had better functLonal reach than those in the wellness education group [P = .05)." Narcotics, benzodiazepines, antiiiepressants, or phenothiazines.

Confidence m perccntaj e not to lose balance.10 = not at all concerned about falling; 40 = very concerned about falling.

SD — standarii deviation.

attended. The average attendance ± standard deviation inthe tai chi group was 76 ± 19% (range 6-100%), whereasthe average attendance for the WE group was 81 ± 17%(range 10-100%). No adverse events occurred during thetai chi or WE intervention. One participant sustained anankle abrasion during the medicai evaluation.

DISCUSSIONThe results from this study show that a 48-week tai chiintervention for transitionally frail older persons led to a

significant reduction in fear of falling. Older persons inthe tai chi group had a continued improvement in averagefear of falling at each subsequent time period. The differ-ence in effect with the WE group increased over eachtime interval and was greatest at 12 months. Fear offalling was lower in tai chi than in WE participants, re-gardless of activity level, presence or ahsence of depression,use of sedatives, or functional reach at baseline or thenumber of falls during the study. The effect of tai chi onreducing fear of falling remained even after adjusting forcovariates.

JAGS JULY 2005-VOL. 53, NO. 7 TAI CHI AND FEAR OF FALLING IN OLDER ADULTS 1173

cUX;

3

Cffl

oco

co(5'J

-aUJ

ac

U

o

UCQ<

t2

CD

O

o

o£

CO

CDC

"55toCD

CO

liJCO

LU

cn

UJ

UJCO

oO

oo

oo

COoo

oCM

CJ

ai

roo

CJ

oCM

•H

cn

oCM

-HCD

eno

CO

ooiH

o

#oq

Oq

en COen en

^ bCDLO

127

.90

IIocdLO

CM

in

130

.92

Ti <^+1 OJ

Ln

CO ro

£ 5

CO0)

I I_ CO

J = X)O «--

CO _

S oJ £

OJ Oq COcd CM «

OI CO PC D • ^

LIJO^CM

-HCO

cd

CDr--.

oi-HOJ

CO

S, CDCM CM

" = To

CMoo

CJoo

oo

roocd-H

T -

CDLO

OJ

oi-Hocd

*oq

ininCvi-HOJ

CD

CD

CM

-HoCM

CO LOLO CDCM Cvj

II HT- Ol

oi cdLn LO

O CM1^- CD

CD ro• ^ L O

CO L O

CO Lri

CO 1 -

co in

4 §

03J2

oo

Oo

oo

CDOCM

-H(DO)Ln

CM

cn

l O

oOJ-H

1 -

QCD

LOCD

O

oj-HOcdLO

CD

o

i ncn

-H"^CMLO

oCM

Ol

OJHCO

CJLO

OJCO

CO

CM

41t t

Ln

oCO

ooi^1

LO

COLO

COCJ)

roen

-Hoo]Ln

COT

*

oqV

#oq

roq

CDCO

CDCD

cdII

CO

t o

CD

CDCOCO

-H

CO

OCM

COCO

cd4CMCD

CM

CDh-oj4-1

LO

roCO

COCD

ojH

CD

d

CM

[—

oj-HCD

cd

CDCO

OJ

oi• 1

CO

CO

CJ

CM

-H

O

*

oqV

CM

CD

*CDOo

CO

oOJII

CM

l Ol O

unCM

+1CO

CD

1- lOcd oi

CDCM

CD

oj11

tiLO

48,

COOJ

CM

CO

48.

CO

CD

s ^^ CflCfl CD> JD

CO OJ Om "^ (n * " CO to" oi ^ 'oi ^ °^

5 I I "S I I o IQ. C V

LOO

OO

enCM-HCO

CMLn

en

cj

rooj

«,I—

oqV

oj

CO

^—CD

COCM

CM

41€0y

Ln

«

oqV

1 - COro CD

OJ

-HCOenLn

COCO

CO

o

O l

-H

LO

CO

,001

LO

101

roCD1 - ^

-H

CMLn

CO00

OJ

oi4HCO

• ^ CD7- ro

oCD

to * " <{> (ft

= CO C O : iCD > • " • = 0 )

Al

CM-HT -

LO

COCO

CO

roCM

HLn

CM-HCO

40.

oCOOJH

OJ

oqV

HroCD

HOJ

CDCO

0 o

0 1 CM

CO LnCO m

en mCO CM

CM C J cno

CO

CD

Ln a>m CD

> S CJ •= CD

CO C JCD a>

-H -HCM CO

CD CD^ CO

O CO

" ^ Op

LO •'J

LO IO

•H

CD

CM

OJ

-HoCO

oCJ

COCO

CDtnCO

JD

*-^

"D

iCO

o

cn

7m,

CD

CDV

COo

OJ

CD

cc

1174 SATTIN ETAL. JULY 200i-VOL. Si, NO. 7 JAGS

CO

O

O

LUen

UJCO

+1c

LUCO

-Mcro

9~oitnCOCQ

LUU)^ca(U

CO OCO m

oi CM *-H -H oo) q 9• ^ C D

o oCO ^

CM - ^

CD m

CD CMOJ - ^

h - CO

ci cd

• ^ C MC M ^CM CM

H ^ [ocn r-

*

CDCO

CD

CO q^ ccjun m

COCO

I- CO

CM •^Lfi Ln

iri

~HCOCM

COCD

~HCD[

mLO

CD CD• I - C O

CD

cn

15 dCO A l

CO CD

CJoj-H

CDCM

oi41oo

CJ • ^

ccj cd

CO COCO CO

CMCM

41CO

58.

CDCO

OJCM

•M

ocn

CMCM4o

56,

CD

O

ojJ J11

CMoi

CMoi-HCD

51.

cncn

CM

,

CD

o>

cd cd

Ln CDCM CM

cd4LnCD

mCDOJ

1-9

COliincdLn

oCO

COcd41CDcdLn

COOJ

COCO

cd••H

CDCMLn

CO

EDCM COLri CM

a> COCO CO

5 C

o> CD

CD CO

CO

CM-H

59.1

CM

CDOcd4r\ 1

50.;

5-

CO

q

cd

63.C

CDOJ

cd *

+* §cd

CO

COCO

cd4

56.-

CM

O

cd

45.^

CO

CDCDCM

HCMCO

COLD

COCD

CD

cd•H

CM

d

CO

CO

cnCM CM

-H -H

OJ OCD m

CD CDTj 03CM o i-H -HO CD

cd d

CO

1-

CO CO

4' MEC5 CD

ci d

<D CDCM CO

r - COCD CO

cd cd

T t CD

CO CNJ

n

OJ cnCO CO

o oCD C^

cd cdII Hf2

CO COd CO

<J = CD•= a> >

Ln ^r-. COcd cd

LO

oCD CD

in rt

o rCO CO

r O(D CM

cd cd4^ ^r-.. CD

in oi

COco

CD CM

CO cd

^g(D caCD ^in Ln

CM

Al

JAGS JULY 2l)0.5-VOL. 5.5, NO. 7 TAI CHI AND FEAR OF FALLING IN OLDER ADULTS 1175

60 -

50 -

40 -

30 -

—-—— ————'

Wellness

60 -

50 -

40 -

30 -Tai ChiWellness

12 12Months on Study Months on Study

oOJaCO

O ^Qj a)o) c/)CO CO

Q.

6CQ<

20 -

10 -

0 -

-10 .

-20 -

.-^—"

•'• :

Wellness

20 -

10 -

0 -

-10 .

-20 -

- ^

-• *

Wellness

12 12

Months on Study Months on Study

Figure 2. Longitudinal changes in activities-related fear of falling measured using the Activities Balance Confidence Scale (ABC) byintervention (tai chi exercise program or wellness education program). The vertical bars indicate the 95% confidence intervals for themean. A: Mean ABC scores from univariate analysis (Table 2). B: Mean ABC scores from multivariable analysis (Table 3). C: Meanpercentage change from baseline in ABC scores from univariate analysis. D: Mean percentage change from baseline in ABC scoresfrom multivariable analysis.

Tbe results also show tbat African Americans tended tohave a greater fear of falling than Caucasians., regardless ofbaseline gait speed. Tai chi was no better at reducing fear offalling for African Americans tban WE in those with a gaitspeedof 0.97 m/s or greater or in those with a gait speed lessthan 0.97 m/s. Tbe reason for tbis lack of effect is unclearbut may reflect differences in perceived safety in one's en-vironment or small numbers of African Americans in thesample; tberefore, these data sbouid be interpreted cau-tiously. Little information exists about differences of andreasons for fear of falling in racial/etbnic groups otber tban

Wben tbe data were reanalyzed using tbe FES., similarresults to tbose derived using tbe ABC score were obtained.Because of the known strong correlation between thesemeasures,^^'^"'^^ and the larger variety of environmentsin which to assess confidence in not falling using the ABCversus the FES, the ABC was chosen for reportingmost of tbe final results. Although frequently used assurrogates for fear of falling, the FES and the ABC scalesused in this study have some limitations. Tbe FES is re-stricted to self-reports regarding fear of falling in the homeonly, whereas tbe ABC addresses confidence that one willnot fall under a variety of circumstances but docs notcapture the bebavior of fear. In this regard, tbe develop-ment of a behavioral scale that better engages the fear of

falling in this transitional or intermediate frail group iswarranted.'"

Tai chi is a martial art, and its underlying philosophywas developed in China many centuries ago, * -"*" but its usein Western culture is relatively new. It teaches precise bodymovements while maintaining a tranquil, concentratedmind, and its performance relies more on image recollectiontban memorization of instruction."" Tai cbi has postulatedeffects on balance, mental outlook, and stress tbat mightdirectly affect fear of falling. A mnltivariable analysis of tbedata from FICSIT revealed tbat fear of falling was a pre-dictor of subsequent falls.'•* Findings from tbat study alsosuggested tbat tai cbi practice increased older participants'sense of well-being wben performing other activities'^ andfavorably affected older participants' activities of dailylivitig."^^ Another study showed tbat women practicing anexercise program like tai chi experienced significant reduc-tions in tension, depression, and anger, witb an overallimprovement in mood.'*' Also, experienced tai chipractitioners have lower levels of salivary cortisol concen-tration, suggesting reduced stress, a notion confirmed byparticipant reports of reduced tension and anxiety com-pared witb sedentary control participants."^"^ Other studiesbave shown that tai cbi can reduce resting systolic bloodpressure,'**•''*••*'' can reduce tbe rate of cardiorespiratorydecline compared with age-matched sedentary controls,'''^

1176 SATTIN ET AL. JULY 2005-VOL. 5i, NO. 7 JAGS

Table ^. Multivariable Analysis of Activities-Specific Balance and Confidence Scale (ABC): Adjusted Mean ABC at 12Months After Baseline by Treatment Intervention Program, Depression, Functional Reach, Race, Gait Speed, and Numberof Falls

Covariate Mean ABC 95% Confidence Interval P-value

Treatment interventionTai chiWellness

DepressionAbsentPresent

Functional reach<10 inches>10 inches

Race and gait speedWhite/other<0.97m/sAfrican-American<0.97 m/sWhite/other >0.97 m/sAfrican-American >0.97 m/s

Number of falls*012345+

52.342.8

49.843.8

45.248.3

45.839.654.347.4

49.450.450.141.747.441.6

(48.1-56.5)(38.7-46.8)

(46.4-53.1)(40.0-47.5)

(41.7-48.8)(44.8-51.8)

(42.4-49.2)(34.2-44.9)(50.9-57.6)(41.1-53.7)

(46.6-52.3)(47.0-53.7)(45.7-54.6)(35.5-47.9)(40.1-54.6)(35.5-47.8)

.008

<.OO1

.006

<.OO1

.01

Note: Reportfci means are least-squares means char adjust for other f.ictors in the model. The miiltivariahlt model for ABC scores included, initially, the treatmentintervention, baseline depression, haselitie fiiiiLtiona! reach, baseline race anci gait spcfd, use of sedatives, am! number of falls as a time-dependent covariate. Use ofseclatives was niit statistically significant in the multivariable analysis, and the mode! was refit without "use of sedatives" for the final analysis. The treatmentintervention means are the estimated ABC means after 12 months of follow-up, averaged over levels (catej^ories) of tbe other factors. The estimated means for levels ofthe other covariates (e.;;., depression) represent the ABt: scores, averaged over levels of the other factors. Reported P-valiies are for comparing the means acrosscatct^ories of each variable, attcr adjusting for the other variables in the mode!.'Time-dependent covariate. Participants were classified acLording to the number of falls at the time of each ABC measurement. As the number of falls increased overtime, the mean ABC] scores tended to decline.

can lead to a reduction in self-perceived stress scores andstress assessed hy skin temperature measurements,''' canproduce greater enjoyment compared with the activities of acontrol group,'*** and can produce a feeling of self-efficacy toperform movements in a continuous and qualitativelypleasing way.'*' Findings from the latter study also suggest-ed that there may he a relationship between self-efficacy inmovement performance and adherence to the tai chi exer-cise program.'^''

bear of falling is common in older persons and can leadto gait alteration, deconditioning, activity restriction, de-creased social interaction, worsening perceived health sta-tus, and increasing risk of falling.'""^-'-^ It was recentlyreported that older persons with fear of falling had greaterdeclines in ability to perform activities of daily living thandid those with no fear of falling over a 12-month period.^Persons who had not fallen hut were fearful had a greaterrisk t)f admission to an aged care institution. Persons whohave fear of falling alone may differ from persons who areafraid and restrict their activities.'" The current study foundthat tai chi significantly increased confidence in preventingfalls and reduced fear of falling in those who were sedentaryor active at baseline. Moreover, fear of falling was lower intai chi than in WE participants, independent of the numberof falls sustained by the participants.

The result showing that nearly half of all participantswere afraid of falling at baseline is consistent witb a tran-

sitionally frail population because it lies between previousprevalence estimates from vigorous and frail groups.'"^-'"That tai chi can reduce fear of falling in transitionally frailolder persons is important because the transitional stagemay he the last opportunity to bave an effect on frailty." ^ Italso builds on previous work that showed that tai chi canreduce fear of falling in robust older adults.'''

Tbis study has several limitations. First, tbe .study satn-ple was not large enough, nor was the original study de-signed to determine whether tai chi could alter changes infear of falling immediately after a fall. Knowing whether taichi had immediate effects in reducing fear of falling after afall and how quickly after a fall tai chi sbouid be begunwould be important in designing a targeted interventionprt)gram. Nevertheless, all of the study participants had ahistory of one or more falls in the previous year. Second,data on fear of falling were not collected beyond tbe 48-week study period, so it is unknown whether the changesseen in tbis study will persist. In a previous Atlanta FICSITtrial, more tban 3.")% (personal communication, SL Wolf,2004) and up to nearly half'"* of the tai chi participantscontinued tai chi exercise after the study concluded, sug-gesting that the general sense of well-being can lead olderadults to exercise more. Third, male participation was low-er tban expected. Anecdotal comments from spouses sug-gested that their male partners were too frail to participateor were too busily engaged in other activities.'"' Finally, 94

JAGS JULY 2lffl.S-V0L. ax NO. 7 TAI CHI AND FRAR OF FALLING IN OLDER ADULTS 1177

(30%) of the rransirionfllly frail participants did not com-plete the 48-week trial. It is possible that the changes ob-served over time in the tai chi group were due to differentialdropout t)f more frail individuals. The mixed model anal-ysis was used to reduce this bias."*" In addition, many tran-sitional older people declined to participate in this study.Therefore, the study results may not generalize to the entirepopulation of transitional older people or to those withlesser or greater degrees of physical frailty.

Tai chi is a safe exercise form for older adults willingand able to participate and fosters focused mental controlover body movements. Given that tai chi can improve bal-ance confidence, tai chi cotiid be considered a therapeuticexercise to build falls confidence in the face of decliningfunctional and cognitive abilities in older adults and tocounteract fears about losing balance or falling, Tai chishould be considered in any program designed to reducefalling and fear of falling in transitionally frail older adults.

ACKNOWLEDGMENTSWe are indebted to Nana Freret, RN, MS, for her efforts atcoordinating this project; Susan Murphy for her assistancein project coordination; Lois Ricci, RN, EdD, and CarolHolbert, PhD, RN, for provision of the weilness program;Michael O'Grady, MD,, Mary Jowers, RN, and ArleneGreenspan, DrPH, for assistance with evaluations and fallssurveillance; Tingsen Xu, PhD, and tai chi Grand Masterfor devising and implementing the tai chi program; andGregory Bailey for provision of tai chi exercise classes athalf the facilities.

Financial Disclosures: None.Author Contributions: Richard W. Sattin, the primary

author, developed the study concept and design, performedanalysis and interpretation of data, and oversaw prepara-tion of manuscript. Kirk A. Easley consulted on study con-cept and design, performed analysis and interpretation ofdata, and played a major role in preparation of the man-uscript. Steven L. Wolf consulted on the study concept anddesign, played a key role in acquisition of subjects and data,helped with analysis and interpretation of data, and playeda maj(jr role in preparation of the manuscript. Ying Chenconsulted on the study concept and design, performed anal-ysis and interpretation of data, and provided help in prep-aration of manuscript. Michael H. Kutner consulted on thestudy concept and design, played a key role in acquisition ofsubjects and data, performed analysis and interpretation ofdata, and played a major role in preparation of manuscript.

Sponsor's Role: None.

REFERENCES1. Hdwiand J, Peterson EW, Levin WC et al. Fear of falling amonf! the fdmniii-

nity-dwelliiig eldtrly. J Aging Htalth l993;,i:229-24,^,2. Tinettj MK, Mi-ndes de Leon C¥, Dimcctte JT et al, Kcsr of falling and f.ill-

rclated efficacy Jn relationship to functioning among comniiiiuty-living I'ldcrs.J Gerontol 1994;49;M I4i)-M 147.

^. Vellas BJ, W.iyncSJ, RomLTii I.Jct.)!, Fear of fallin!; and rtsirictiun <if inobilityill c'kiiTly faliers. Age Ageing 1997;26:IS'*-I'»3,

4, Arfken i.l, I.aih HW, Birge S| L-t al. The prevalence and correlates of fear offalling in elderly persons living in the coinniunity. Am J Public Health|994;H4:566-,^70.

5. Lawrence RH.IennsredtSL, Kasten LHct al. Intensity and correlates uffearuffalling and hurting oneself in the next year. J Aging Health I998;10:267-2H(i.

6. Friedman SM, Muim/ B, West SK et al. Falls and ft-.ir ot falling: Which comesfirst? J Am fieriatr Soc 2D02;5D:Li29-I.13,i,

7. Gumming KG, Salkeki G, Thomas M et al. Prospective study of the impact offtar of falling on activities of daiiy living, SF-36 scores, and nursing home^ldmission. j Gerontoi A Biol Sci Med Sci 2t)()0;,S,iA:M29')-M305.

S. HowlandJ, Lachnian MK, Peterson F.Wcral. Govariates (if lear ot falling andassociated activity curtailment. Gerontologist 1998;,3H;549-5,^5.

9. I achman MF,, Howland J, Tennstedt S et al. hear <if falling and activityrestriction: The survey of activities and fear of falling in the elderly (SAFE).,[ Gerontol B l'sychol Sci Soc Sci I99S;53B:P43-P5U.

lU, Murphy SL, Williams GS, Gili TM. C^haracn-ristics associated with fear offalling and activity restriction in com mil nit>'-living older persons, J Am GeriatrSoc 2(l()2;51):,^lh-520.

11, Brouwer BJ, Walker G, Rydahl SJ et a!, Reducmg tear ot tailing m seniorsthrough education and activity programs: A randomi/ed trial. J Am GeriatrSoc 2O0,5;.5 1:829-834.

12, Tennstedt S,, HowlandJ, 1,achman M et al, A randomized,, controlled trial o( agroup intervention to reduce fear of fallinj; and associated activity' restrictionin older adults, J Gerontol B I'sychol Sci Soc Sci 199R;53B:P,?S4-P392.

13, Tinetti Mh, Baker Dl, McAvay G et al. A multifactorial intervention to reducethe risk of falling among elderly people living in the community. N Eng! J Med!994;,5,il;82l-827.

14, Wolf SL, Barnhart HX, Kutner NG et al. Exercise training and subsequent fallsamimg older persons: Gomparison of tai chi and computerized balance train-ing.,I Am ticriarr Soc I996;44:489^97.

15, Wolf SL, Sattin RW, Kutner M et al. Intense tai chi exercise training, and falloccurrences in older, transitionally frail adults: A randomized, controlled trial.,1 Am Getiatr Soc2003;51:169.i-I70l.

Ifi, Laii C, Lai |S, Chen SYet al. 12-month tai chi training in the olderly: Its effectson health fitness. Med Sci Sports Fxerc 1998;,10:34,S-,VS L

17, Hong Y, LiJK, Robinson PD. Balance control, flexibility and cardiorespiratorytitness among older tai chi practitioners. Br J Sports Med 2000;34:29-34.

18, Yan JH, Tai chi practice improves senior citizens' balance and arm movementcontrol.,[ Aging I'hys Activity 1998;6:271-294.

19, Young DR, Appel LJ, JeeSHetal. The effects of aerobic exercise and t'ai chi onblood pressure m older peiiple: Results of a randomized trial. | Am Cieriatr SocI999;47;277-2S4.

21), l.an C,Ghen SY, Lai JS etal. Effect of Tai Gbi on cardiorespiratory function inpatients with coronary artery bypass surgery. Med Sci Sports Exerc1999;31:634-63«,

21, Li F, Harmer P, Gbaumeton NRet al.r.ii Ghi as a means to enhance self esteem:A randomized controlled trial, J AppI (ietontol 2002;2i:70-S9.

22, KtessigRW, Wolf SL, Sattin RWer al. Associations of demographic, functional,and behavioral characteristics with activity-related fear of falling among olderadults transitioning to frailt\', J Am tieriatr Soc 2001;49:14,56-1462.

23, Wolf SL,SartinRW, O'Grady Metal , A study design to investigate the effect ofintense tai chi in reducing falls among older adults transition ing to frailty.Control Clin Trials 20l)l;22:6«9-704,

24, Speecbley M, Tinetti ME. Falls and injuries in trail and vigorous communityL-Iderly persons, J Am Geriatr Soc l99l;39:46-52,

25, Wo!f SL, Coogler C, Xu T. Exploring the basis for tai chi qiian as a therapeuticexercise approach. Arch I'hys Med Rehabil l997;78:886-892.

26, Ghinese Sports Editorial Board, Simplified iaijiquan, 3rd Ed, Beijing, China:Eoreign Languages Printing House, 1986.

27, Tinetti ME. Richman 1), I'owell L. Ealls efticacy as a measure of fear of falling,i Gerontol I99O;45:P239-P243,

2H. Powell L, Myers AM. The acrivitius-'ipecific balance confidence (ABC) scale.,[ Gerontol B ISychol Sci Soc Sci 1995;50B:M28-M34,

29. Hotchkiss A, hisher A, Robertson R ct al, (xmvergent and predictive validityof tbiee scales related to falls in the elderly. Am \ Occup Ther 2004;58:100-103,

30. Buchncr DM, Hornbrook MG, Kutner NG et a!. Development of rbe commondata base for the FIGSIT trials. J Am Geriatr Soc ]993;41:297-308,

31. Radloff LS, I be t;FS D scale, A sett report depression scale for research in thegeneral popLilation, AppI Psychol Meas I977;I:385-4O1.

32. Gill TM, Williams C:S, Tinetti M. Assessing risk for the onset of functionaldependence among older adults: Tbe role of physical performance. J AmGeriatr Soc 1995:43:603-609,

33. Duncan PW, Studenski S, Chandler | et al. HuiKtional teach: Predictive validityin a sample of elderly male veterans. J Gerontol 1992;47:M93-M9R.

34. Fcinstein AK. C linical Biostatistics. St. Louis: Mosby, 1977,VS, Maki B. Clait changes in older adults. Predictors of falls or indicators of fear?

J Am Geriatr Soc 1997;45:3 13-320.36. Diggle I'j, Liang KY, Zeger SL. Analysis of Longitudinal Data. Oxford:

Clarendon Press, 1994.37. Donner A. Some aspects ot tbe design and analysis of cluster randomization

trials. AppI Stat I998;47:95-1 13.

1178 SATTIN ET AL. JULY 2005-VOL. .53, NO. 7 JAGS

38. Fried LP, Tangen CM, Walston J ct al. Frailty in older adults: Evidence for aphenotypt. ,1 Cerontol A Biol Sci Med Sei 2001:56A:M146-M 156.

39. TseS, Bailey DM. T'ai t^hi and postural control m the well elderly. Am J OccupTher 1992;46:29.5-.?0().

40. YanJH.DowningJH.TaiChi: An altcrnarive exercise hirm for seniors. I A^in;;Phys Activity 1998;6:3.S0-362.

4L Bottomly JM. Tai chi. Chiireography of hody and mind. In: Davis CM, ed.Complementary Therapies in Rehabilitation: Holistic Approaches tor Preven-tion and Wellness. Thorofar, NJ: Slack Inc., 1997, pp. I3.i-I S6.

42. Kutner N, Barnhart H, Wolf SLet al. Self-report benefits of Tai Chi practice byolder adults. J Gerontol B Psycho! Sti Sot Sci 1997;52B:I'242-P246.

43. Brown DR, Wang Y, Ward A et al. Chronic psychologiizal effecrs of exerciseand exercise plus cognitive strategies. Med Sci Sports F.xerc 1995;27;765-775.

44. Jin P. F-ftieacy of Tai Chi, brisk walking, meditation, and reading in reducingmental and emotional stress. J Psychosom Res t992;36:3()l-370.

45. Channer KS, Brown D, Barrow R et al. Changes in haemodynamic para-meters following Tai C hi Chuan and aerobic exercise in patientsrecoverinj; from acute myocardial infarction. Postgrad Med J 1996;72:349-351.

46. Lai |S, Lan C, Wong .V1K et al. Two-year trends in cardiorespiratory functionamong older Tai Chi Chiian practitioners and sedentary subietts. J Am GeriatrSnc I995;43:I222-1227.

47. Sun WY, Dosch M, Gilmore GD er al. Effects ot tai chi chiian prof^ram onHm(mg American older adults. Ednc Gerontol I996;22:l6l-I67.

4H. Van Deusen I, Harlowe D.T he efficacy of the ROM Dance Program for adultswith rheumatoid arthritis. Am J OcLiipThtT 19H7;41:90-95.

49. Li K, McAuley L, Harmer P et al. Tai Chi enhances self-efficacy and excrt;isehehavior in older adults. J Aging Phys Activity 2001;9:161-171.

50. l.achin JM. Statistical considerations in the in tent-to-treat principle. ControlClin Trials 2{)OO;21:I67-I89.