Full text of DOI:10.1207/S15324796ABM2303_6

The Effects of Exercise on the Quality of Life of Frail Older Adults:
A Preplanned Meta-Analysis of the FICSIT Trials
Kenneth B. Schechtman, Ph.D.
Washington University School of Medicine
Marcia G. Ory, Ph.D., M.P.H.
National Institute on Aging, Behavioral and Social Research
The FICSIT Group
ABSTRACT
search on exercise interventions has been conducted in healthy
younger populations or based on small sample sizes, inadequate
controls, or little standardization of outcome assessments. An
exception is a set of linked randomized clinical trials, the Frailty
and Injuries: Cooperative Studies of Intervention Techniques
(FICSIT) (14). These trials involved eight clinical sites and were
designed to test the effectiveness of several exercise-oriented in-
terventions in the frail elderly. The eight sites recruited more
than 2,500 older persons at risk for fall-related injury.
Although FICSIT focused primarily on physiologic and fit-
ness benefits, these trials also examined quality of life (QOL), an
outcome category that has been shown to be potentially modifiable
by exercise (15) and that has several domains which can be differ-
entially affected by exercise interventions (16). The inclusion of a
well-tested and widely used standardized measure makes FICSIT
one of the largest controlled studies to assess the benefits of exer-
cise on different domains of QOL in frail older adults.
The purpose of this article is to discuss the effects of exer-
cise in older adults on four QOL scales of the Medical Outcomes
Study (MOS) Short Form–36 (SF–36). We were particularly in-
terested in whether the potential QOL benefits of exercise are
partially offset by increased bodily pain resulting from the phys-
ical burden of the exercise. To address these issues, we em-
ployed meta-analysis in a unique context: when raw data were
available to compute prespecified site specific covariate-ad-
justed summary statistics, and when there was no possibility of
publication bias (17,18) and selection bias (19,20) because in-
cluded studies were chosen in advance. Specifically, we per-
formed covariate-adjusted meta-analyses to (a) assess the effect
of exercise on the four QOL outcomes, (b) investigate the effects
of exercise on QOL measures according to the predominant in-
tervention type (resistance, endurance, flexibility, balance), and
(c) examine the association between changes in QOL and physi-
cal function.
The Frailty and Injuries: Cooperative Studies of Interven-
tion Techniques (FICSIT) was a linked series of randomized
clinical trials focused on the benefits of exercise in the frail el-
derly. This article uses covariate-adjusted preplanned
meta-analyses of FICSIT data to evaluate the effect of exercise
on quality of life (QOL) outcomes (N = 1,733; age = 73.4 ± 6.1
years). Results indicate that (a) exercise produced a small but
significant improvement in the emotional health component of
QOL, trended toward an improved social component, and did
not effect perceptions of general health; (b) exercise-related
joint and muscle stresses did not increase bodily pain; and (c)
QOL improvements were independent of changes in physical
functioning. We conclude that exercise can improve QOL in the
frail elderly but that the magnitude of the improvement is modest
in size. The benefits of a meta-analytical approach for docu-
menting efficacy outcomes across different types of interven-
tions are discussed.
(Ann Behav Med
2001, 23(3):186–197)
INTRODUCTION
The health benefits of physical activity are well docu-
mented in epidemiological research (1). Long-term observa-
tional data from the Alameda County Study (2,3) show that lei-
sure time physical activity is associated with reduced morbidity
and mortality and better functioning in older adults. Regular ex-
ercise has been hypothesized to improve health and well-being
through mechanisms that include improved self-efficacy, ex-
panded social networks, attention effects, and direct physiologi-
cal responses (4,5). Yet, until recently, prevention oriented exer-
cise interventions have not been conducted in older populations
that have been assumed to be uninterested in or unable to adopt
changed lifestyles (6). These preconceptions are being chal-
lenged by recent studies that demonstrate that older people can
be successfully recruited and retained in exercise programs, al-
though long-term maintenance of health behaviors still remains
an issue (7–9).
METHOD
Data discussed herein were collected as part of the National
Institute on Aging and National Center of Nursing Research
sponsored study, known as FICSIT. FICSIT was a multicenter
investigation of exercise-related interventions, with outcomes
including functional status, falls, fall-related injuries, and QOL.
The selection of FICSIT QOL measures, and issues related to
QOL assessment in the elderly, are discussed elsewhere by
Kutner et al. (16).
Although some intervention studies have shown that exer-
cise may benefit frail older people (10–13), most previous re-
Reprint Address: K. Schechtman, Ph.D., Washington University
School of Medicine, Division of Biostatistics, Box 8067, 660 South Eu-
clid Avenue, St. Louis, MO 63110. E-mail: ken@wubios.Wustl.EDU
186

Volume 23, Number 3, 2001
Exercise and QOL
187
The eight FICSIT clinics defined their own interventions
using site-specific endpoints and evaluations and, aside from the
requirement that all participants be elderly, their own eligibility
requirements. At the same time, an extensive database, devel-
oped jointly by the clinical sites and the Coordinating Center at
Washington University in St. Louis, contains data common to all
or to several FICSIT sites. Thus, the unique FICSIT design had
the flavor of a multicenter clinical trial in that eight studies were
linked with a coordinating center and a large common database.
Results from the four FICSIT sites that collected QOL data are
included in this report. More general findings on physiological
mediators and fall outcomes are found elsewhere (21). Details
of the four designs, each of which is a randomized exercise trial
involving community-dwelling participants, are discussed in
other reports (22–25). Table 1 provides summary information
about each trial, with Panel A providing general information and
Panel B providing the details of the intervention.
was clearly the least intense because it was both less vigorous
and for fewer minutes per week than in the other sites.
Statistical Analysis of Data
Our primary data analytic tool was a meta-analysis per-
formed separately for each outcome measure. The purpose was
to assess the covariate-adjusted combined effect of interventions
on poststudy values. The meta-analytic approach is based on the
two-stage procedure developed by DerSimonian and Laird (31)
and discussed by Fleiss and Gross (32). The first stage generated
summary statistics for each treatment arm in each FICSIT site.
Within site regression analyses were performed with the
poststudy value of the outcome measure as the dependent vari-
able and with covariates and treatment group as predictors. For
each treatment arm in each FICSIT site, these analyses yielded
adjusted means plus or minus standard errors of the poststudy
value of each dependent variable. Control group means were
then subtracted from treatment group means, and the difference
and associated standard error served as a covariate adjusted
measure of the effect of each intervention in each site.
The second stage in this process used the aforementioned
adjusted mean differences and standard errors as raw data that
generated pooled estimates of the overall effect of interventions
on each outcome measure. The pooled estimates were computed
using the DerSimonian and Laird meta-analysis procedure and
are applied to all nine interventions evaluated in the four FICSIT
sites included in this article. We also applied these methods to
four overlapping subsets of the nine interventions: those that
contained a resistance component (n = 4 interventions), a bal-
ance component (n = 4), an endurance component (n = 3), and a
flexibility component (n = 4).
Variables and Measures
The four dependent variables were the general health per-
ceptions, emotional health, social functioning, and bodily pain
scales of the MOS SF–36 (26–27), in which the pain scale is
viewed primarily as an adverse symptom in our analysis. All
scores range from 0 to 100. FICSIT began prior to the publica-
tion of the final version of the SF–36, so the prepublication ver-
sion was used. However, because differences between prepubli-
cation and final versions are small, they should not affect the
interpretation of results. The physical functioning subscale of
the MOS was not included because this construct was treated as
a covariate in these analyses.
Prospectively defined covariates were age, gender, and gait
speed. Gait speed was measured in meters per second using a
previously described common protocol (28). Gait speed was se-
lected as an indicator of physical functioning because it corre-
lates well with other physical function measures and with the
risk of falling (29), and because it correlates better with self-per-
ceived physical function than any other available physical per-
formance measure (30). To investigate the effect of specific
characteristics of the exercise programs on the outcome mea-
sures, we present each arm of each intervention according to
which of four training components it included (resistance, en-
durance, flexibility, balance) and according to a subjective as-
sessment of the intensity of the intervention (low, medium, or
high intensity).
All data were analyzed using SAS (33–34). Chi-square
tests and a one-way analysis of variance tested the equality of
means and proportions across sites. Because the intensity mea-
sure was subjective and could not include the component of in-
tensity that reflects participant compliance, we had no prede-
fined analytic strategy for dealing with this variable. Instead, a
graphical presentation of the impact of each intervention is pre-
sented according to the level of intensity. Subjective evaluation
of these graphs provided no obvious trends, so no formal statisti-
cal analysis of the intensity variable was performed.
RESULTS
A total of 1,733 participants from four FICSIT sites are dis-
cussed in this article. Mean age (± SD) was 73.4 ± 6.1 years, and
55.6% of participants were women. Table 2 contains a detailed
description, by site and treatment arm, of baseline characteris-
tics of the sample. As a reflection of the different entry criteria
across sites, Table 2 reports significant between site differences
in six of the seven tabulated variables.
Table 3 contains site and intervention-specific information
about the covariate-adjusted effect of the individual interven-
tions. The table describes each intervention according to
whether it contains a resistance, endurance, flexibility, or bal-
ance component. Reported mean values are the adjusted
postintervention values of each outcome measure. They were
Intensity of the Intervention
Our meta-analytic results are presented according to a sub-
jective categorization of the intensity of the relevant interven-
tion as high, medium, or low. The primary determinant was the
vigor of the prescribed exertion, but the ordering would corre-
late precisely with a measure of kilocalories burned per week
that combined information about the number and duration of the
sessions as well as the level of exertion. There was little ambigu-
ity in defining the categories because the Seattle target of 75%
of maximal heart rate after 2 months was by far the most vigor-
ous, and the program in Portland that involved brisk walking

TABLE 1
Characteristics of the Four Included FICSIT Trials
Panel A: General Characteristics of Trial
Site No. and
Location
Primary
Outcomes
Intervention (Details in Panel B)
N
Eligibility Criteria
Site 1:
Portland, OR
(7)
Control group versus endurance
exercise group. Modify
environmental risks at home to
reduce falls.
1,323 in two groups At least age 65; community
dwelling; ambulatory; at least
one fall in last year, a near fall
in last month, or at least age
75
Falls and
fall-related
injuries
Site 3:
Seattle, WA
(23)
Modified 2 × 2 factorial design
using resistance and endurance
training. Reduced endurance
exercise time in group receiving
both interventions.
100 in four groups
Age 68–85, community
dwelling, ambulatory, no
severe cognitive impairment,
no participation in vigorous
exercise, moderate gait and
balance impairments
Strength,
aerobic capacity,
gait,and balance
Site 5:
Atlanta, GA
(24)
Three groups: Control group
meets once a week for education
session; interventions groups are
a static balance exercise group
using a balance platform and a
dynamic balance group using Tai
Chi
200 in three groups
At least age 70, community
dwelling, ambulatory, no
major debilitating illness
Balance, range
of motion,
activities of
daily living,
instrumental
activities of
daily living
Site 8:
Farmington,
CT (25)
2 × 2 factorial design with
balance and resistance training
110 in four groups
At least age 75, community
dwelling, ambulatory, no
cognitive impairment or
terminal illness
Balance on
balance
platform, gait,
functional
mobility
Panel B: General Characteristics of Interventions
Duration of
Trial
Site No.
Type of Intervention
Method
Large group
(one/week),
unsupervised
Frequency of Sessions
Session Length
1
Endurance componenta
4–6/week total
30 min in group
and 15 min at
home
4 months
vigorous walking
(3–5/week) at home
1
1
3
3
3
Flexibility componenta
Education componenta
Endurance + flexibility
Resistance + flexibility
Small group
Small group
Small group
Small group
Small group
11 sessions
11 sessions
3/week
30 min
30 min
1 hr
4 months
4 months
6 months
6 months
6 months
3/week
1 hr
Endurance + resistance +
flexibility
3/week
1 hr
5
5
Balance
Tai Chi
Individual in lab
1/week
2/week
50 min
1 hr
15 weeks
15 weeks
Small group
(1/week),
unsupervised
(1/week)
8
8
8
Balance
Individual
3/week
3/week
3/week
45 min
45 min
90 min
13 weeks
13 weeks
13 weeks
Resistance
Small group
Balance + resistance
Individual + small
group
Note. All trials are randomized. FICSIT = Frailty and Injuries: Cooperative Studies of Intervention Techniques.
aThe three components at Site 1 are each part of the intervention for the only intervention group at that site. For all other sites, the number of tabulated groups
corresponds to the number of intervention groups at the site.
188

Volume 23, Number 3, 2001
Exercise and QOL
189
TABLE 2
Baseline Characteristics by FICSIT Site and Treatment Group
Covariates
Outcome Measures
Baseline MOS
General
Baseline MOS
Emotional
Health
Baseline
MOS Pain
Score
Site and
Treatment Arm
%
Female
Baseline Gait
Speed
Baseline MOS
Social Scale
Age
Health
All participantsa
Portlandb
73.4 ± 6.1
73.4 ± 6.1
72.3 ± 6.3
72.5 ± 5.9
75.2 ± 4.8
76.4 ± 4.9
73.8 ± 4.1
75.4 ± 5.1
75.2 ± 4.7
55.6
53.3
52.6
54.0
51.0
52.0
52.0
52.0
48.0
0.99 ± 0.2
0.92 ± 0.2
0.93 ± 0.2
0.91 ± 0.2
1.32 ± 0.2
1.23 ± 0.2
1.35 ± 0.2
1.34 ± 0.2
1.35 ± 0.2
69.4 ± 18.8
67.6 ± 19.2
67.1 ± 19.4
68.1 ± 19.1
75.6 ± 14.4
77.7 ± 12.9
75.8 ± 11.1
78.5 ± 17.7
70.6 ± 14.7
77.0 ± 33.8
76.2 ± 34.8
76.3 ± 34.3
76.0 ± 35.3
79.3 ± 30.3
73.3 ± 34.7
86.7 ± 28.9
78.7 ± 27.0
78.7 ± 30.2
69.9 ± 23.4
67.0 ± 23.8
66.7 ± 23.7
67.3 ± 23.9
76.8 ± 20.4
76.4 ± 17.1
74.7 ± 20.7
79.6 ± 23.3
76.4 ± 21.1
86.4 ± 19.7
85.5 ± 20.2
84.7 ± 20.6
86.4 ± 19.7
87.9 ± 18.1
89.3 ± 15.9
81.3 ± 24.6
89.8 ± 15.7
91.1 ± 13.6
Controlc
Individualizedd
Seattlee
Controlf
Resistancef
Endurancef
Resistance and
endurancef
Atlantag
75.7 ± 4.7
74.9 ± 4.1
75.8 ± 5.1
76.4 ± 4.9
79.3 ± 3.9
80.1 ± 4.5
79.8 ± 4.0
78.4 ± 2.8
79.0 ± 4.1
80.5
84.4
76.6
80.6
41.8
40.7
39.3
42.9
44.4
1.16 ± 0.2
1.19 ± 0.2
1.15 ± 0.2
1.15 ± 0.2
1.09 ± 0.2
1.07 ± 0.2
1.05 ± 0.2
1.12 ± 0.2
1.10 ± 0.2
74.7 ± 17.2
72.6 ± 17.8
76.9 ± 16.2
74.7 ± 17.5
75.4 ± 15.8
75.5 ± 14.0
73.7 ± 15.3
76.4 ± 16.0
76.0 ± 18.3
78.3 ± 31.3
74.5 ± 36.5
76.6 ± 31.2
83.3 ± 25.6
83.2 ± 28.0
82.1 ± 28.6
85.3 ± 25.6
84.0 ± 29.8
81.3 ± 29.0
79.6 ± 18.9
80.2 ± 17.0
79.0 ± 19.5
79.6 ± 20.1
82.1 ± 17.4
80.8 ± 15.6
81.3 ± 21.9
85.6 ± 15.4
80.4 ± 16.8
90.1 ± 16.6
87.7 ± 19.5
91.1 ± 16.5
91.4 ± 13.5
89.6 ± 19.9
90.2 ± 17.9
88.9 ± 24.0
86.0 ± 23.6
93.8 ± 12.0
Controlh
Balanceh
Tai Chii
Farmingtonj
Controlk
Resistancel
Balancel
Resistance and
balancek
p value comparing
sitesm
< .0001
< .0001
< .0001
< .0001
.170
< .0001
.005
Note. FICSIT = Frailty and Injuries: Cooperative Studies of Intervention Techniques; MOS = Medical Outcomes Study.
aN = 1,733. bn= 1,323. cn = 662. dn = 661. en = 100. fn = 25. gn = 200. hn = 64. in = 72. jn = 110. kn = 27. ln = 28. mp values generated using analysis of variance
and, for gender, a chi-square test.
generated by a linear regression in which the dependent variable
was the poststudy value of the outcome measure and the inde-
pendent variables were age, gender, baseline gait speed, the
baseline value of the relevant outcome measure, and the treat-
ment group to which the patient was assigned. Reported p values
are adjusted for covariates, with each representing a test of the
null hypothesis that the postintervention adjusted mean value of
the outcome measure in the control group equals the corre-
sponding mean in the relevant intervention group. Because
postintervention outcome data were missing for some
participants, sample sizes in Table 3 are somewhat reduced.
Figures 1 through 4 report the results of the meta-analyses,
one figure per outcome. Each graph contains information about
the difference between the covariate-adjusted final value of the
outcome in the control group and the corresponding adjusted
value in the relevant intervention group. Graphed values are
means and 95% confidence bounds, with positive values indi-
cating that the intervention was associated with improved QOL.
The top nine bars in each figure present means and confi-
dence bounds that compare control groups with individual inter-
ventions in each site. They are divided into three groups accord-
ing to the intensity of the intervention. The final five bars in each
figure report the meta-analytic results. Because all interventions
involved exercise, the pooled estimate labeled Exercise in each
figure reflects a meta-analysis of all nine interventions. This is
the primary, predefined hypothesis test. The specific interven-
tions that generate the pooled estimates of resistance, balance,
endurance, and flexibility can be determined from the Type of
Exercise column of Table 3. Table 4 provides covariate-adjusted
measures of the between group differences associated with each
meta analysis summarized in Figures 1 through 4. Confidence
bounds and p values associated with these effect sizes are also
contained in Table 4.
The Effect of Interventions on the
General Health Perceptions Scale
Figure 1 and the first results column of Tables 3 and 4 sum-
marize the effect of FICSIT on the general health scale of the
MOS. Table 3 indicates that with one exception, none of the ad-
justed within site evaluations of the effect of specific interven-
tions on general health was significant. The exception was that
in Site 8 (Farmington), the combined resistance and balance in-
tervention was associated with improved general health (p =
.021). However, the meta-analyses in Figure 1 indicate that none
of the five intervention categories showed even a trend toward
significance, with the smallest p value at .300.

190
Schechtman et al.
Annals of Behavioral Medicine
TABLE 3
Effect of FICSIT Interventions on the General Health, Emotional Health, Pain, and Social Subscales of the MOS
General Health Subscale
Adjusted
Emotional Subscale
Adjusted
Pain Subscale
Social Subscale
Adjusted
Treatment
Group
Adjusted
M ± SE
FICSIT Site
M ± SE
p
M ± SE
p
p
M ± SE
p
Site 1:
Portland,ORa
1. Control
69.0 ± 0.5
77.2 ± 1.1
69.7 ± 0.8
85.7 ± 0.7
2. Individual-
ized (endur +
flex)
68.7 ± 0.5
.667
80.9 ± 1.2
.029
70.2 ± 0.8
.673
87.4 ± 0.7
.060
R2 = .551
75.9 ± 2.1
R2 = .217
79.3 ± 5.9
R2 = .339
78.4 ± 3.6
R2 = .310
90.2 ± 3.9
Site 3:
Seattle, WAb
1. Control
2. Resist + flex
3. Endur + flex
4. Resist +
76.9 ± 2.3
77.4 ± 2.2
75.0 ± 2.1
.757
.624
.756
8 8.7 ± 6.4
83.8 ± 6.2
79.5 ± 5.8
.289
.599
.982
79.3 ± 3.8
77.7 ± 3.8
76.7 ± 3.5
.869
.893
.731
90.5 ± 4.2
88.9 ± 4.0
85.2 ± 3.8
.958
.818
.353
endur + flex
R2 = .565
74.9 ± 1.4
R2 = .124
71.3 ± 3.7
R2 = .334
76.6 ± 2.1
R2 = .073
85.3 ± 2.6
Site 5:
Atlanta, GAc
1. Control
2. Static
balance with
platform
3. Tai Chi
(dynamic
balance)
73.1 ± 1.4
74.8 ± 1.3
.360
.966
82.3 ± 3.8
80.6 ± 3.5
.039
.073
78.9 ± 2.1
77.8 ± 1.9
.442
.680
90.5 ± 2.7
84.1 ± 2.4
.164
.738
R2 = .683
74.6 ± 2.3
R2 = .257
84.2 ± 5.7
R2 = .424
74.2 ± 3.9
R2 = .189
91.3 ± 4.0
Site 8:
1. Control
Farmington,
CTd
2. Resistance
3. Balance
4. Resistance +
balance
75.1 ± 2.3
74.4 ± 2.3
82.3 ± 2.3
.872
.952
.021
83.2 ± 5.7
68.9 ± 5.8
85.7 ± 6.0
.899
.066
.854
72.0 ± 3.9
73.6 ± 4.0
82.4 ± 4.0
.687
.913
.150
86.0 ± 4.0
85.5 ± 4.2
86.5 ± 4.2
.355
.320
.408
R2 = .519
R2 = .223
R2 = .395
R2 = .109
Note. Least square means are adjusted for age, gender, baseline gait speed, and baseline score on the relevant subscale in within site analysis of variance
models that use the posttreatment score on the relevant subscale as the dependent variable. For each subscale, p values compare each treatment arm with the
control arm in the same site. FICSIT = Frailty and Injuries: Cooperative Studies of Intervention Techniques; MOS = Medical Outcomes Study.
an = 1,218. bn = 90. cn = 175. dn = 91.
The Effect of Interventions on the Emotional
Health Scale
the balance group (68.9 ± 5.8) was almost significantly less than
the score in the control group (84.2 ± 5.7, p = .066).
The meta-analyses presented in Figure 2 indicate that for
three of the five intervention categories, the pooled adjusted fi-
nal emotional health score was significantly greater in the inter-
vention than in the control group. Specifically, Table 4 indicates
that the adjusted effect of all interventions combined was to in-
crease the emotional health score by 3.97 ± 2.0 (p = .043). In ad-
dition, endurance exercise programs were associated with a sig-
nificant increase (after subtracting off the control group change)
in the emotional health score of 3.59 ± 1.6 (p = .027), and flexi-
bility programs were associated with a greater final emotional
health score than control programs (3.78 ± 1.6, p = .018). Due to
the anomalous negative result in Farmington (Site 8), there was
significant heterogeneity (p = .041) with respect to the balance
Figure 2 and the second results column of Tables 3 and 4
summarize the effect of FICSIT interventions on the emotional
health scale. Table 3 indicates that after adjusting for covariates,
the final emotional health score at Site 1 (Portland) was signifi-
cantly greater (i.e., an improvement) in the intervention group
(80.9 ± 1.2, p = .029) than in the control group (77.2 ± 1.1). In
Site 5 (Atlanta), the adjusted final score was greater in the bal-
ance group (82.3 ± 3.8, p = .039) than in the control group (71.3
± 3.7). Also in Atlanta, the emotional health score in the Tai Chi
group (80.6 ± 3.5) was almost significantly greater (p = .073)
than in the control group. Finally, Table 3 demonstrates that in
Site 8 (Farmington), the adjusted final emotional health score in

Volume 23, Number 3, 2001
Exercise and QOL
191
FIGURE 1 Effect of exercise interventions on the general health
scale of the Medical Outcomes Study Short Form–36. The top nine bars
contain point estimates and 95% confidence bounds for individual esti-
mates ordered by intensity. The final five bars contain meta-analytic re-
sults for all nine interventions and separately for each of four exercise
categories (which evaluate, respectively, four, four, three, and four in-
terventions). The horizontal axis measures pre- to postintervention
changes in raw scores. The numbers preceding the labels of the type of
exercise refer to FICSIT sites: 1 = Portland, 3 = Seattle, 5 = Atlanta, and
8 = Farmington.
FIGURE 2 Effect of exercise interventions on the emotional health
scale of the Medical Outcomes Study Short Form–36. The top nine bars
contain point estimates and 95% confidence bounds for individual esti-
mates ordered by their intensity. The final five bars contain meta-ana-
lytic results for all nine interventions and separately for each of four ex-
ercise categories (which evaluate, respectively, four, four, three, and
four interventions). The horizontal axis measures pre- to
postintervention changes in raw scores. The numbers preceding the la-
bels of the type of exercise refer to FICSIT sites: 1 = Portland, 3 = Seat-
tle, 5 = Atlanta, and 8 = Farmington.
intervention. There was no evidence of heterogeneity with re-
spect to any of the other four meta-analyses.
combined and for the endurance and flexibility exercises only
was repeated. To further assess changes in the emotional health
scale, we note that there was no association between changes in
this scale and age (–.006, p = .810) or gender (p = .914 by t test).
Because there were significant changes in the emotional
health scale, we evaluated factors associated with changes in the
scale, with a particular focus on whether they were associated
with changes in gait speed. There was almost no correlation
(.021) between changes in gait speed and changes in the emo-
tional health scale (p = .433). This lack of correlation was inde-
pendent of treatment group assignment. When the meta-analy-
ses presented in Figure 2 were repeated with change in gait
speed replacing baseline gait speed as a covariate, the results
were almost identical to those in the figure. In particular, the
same pattern of significant pooled results for all interventions
The Effect of Interventions on the Pain Scale
Figure 3 and the third results column of Tables 3 and 4 sum-
marize the effect of interventions on the MOS pain scale. Table
3 indicates that none of the adjusted within site evaluations ap-
proached significance. Moreover, the meta-analyses at the bot-
tom of Figure 3 indicate that none of the five intervention cate-
gories showed a significant difference between treatment and

192
Schechtman et al.
Annals of Behavioral Medicine
FIGURE 3 Effect of exercise interventions on the pain scale of the
Medical Outcomes Study Short Form–36. The top nine bars contain
point estimates and 95% confidence bounds for individual estimates or-
dered by their intensity. The final five bars contain meta-analytic results
for all interventions and separately for each of four exercise categories
(which evaluate, respectively, four, four, three, and four interventions).
The horizontal axis measures pre- to postintervention changes in the
raw scores. The numbers preceding the labels of the type of exercise re-
fer to FICSIT sites: 1 = Portland, 3 = Seattle, 5 = Atlanta, and 8 =
Farmington.
FIGURE 4 Effect of exercise interventions on the social scale of the
Medical Outcomes Study Short Form–36. The top nine bars contain
point estimates and 95% confidence bounds for individual estimates or-
dered by their intensity. The final five bars contain meta-analytic results
for all nine interventions and separately for each of four exercise cate-
gories (which evaluate, respectively, four, four, three, and four inter-
ventions). The horizontal axis measures pre- to postintervention
changes in raw scores. The numbers preceding the labels of the type of
exercise refer to FICSIT sites: 1 = Portland, 3 = Seattle, 5 = Atlanta, and
8 = Farmington.
control. The smallest of the five p values is shown in Table 4 to
be .239.
significant. The endurance (p = .103) and flexibility (p = .106)
exercise programs came closest (see Table 4). There was no sug-
gestion of heterogeneity in any of these meta-analyses.
Because there were suggested changes in the social scale,
we evaluated factors associated with these changes more
closely. As a consequence of the large sample size, the small
correlation between changes in gait speed and changes in the so-
cial scale (.052) was nearly significant (p = .055). However,
when the meta-analyses presented in Figure 4 were repeated
with the change in gait speed replacing baseline gait speed as a
covariate, results were almost identical to those in the figure.
There was no correlation between changes in the social scale
and either age or gender.
The Effect of Interventions on the Social Scale
Figure 4 and the final results column of Tables 3 and 4 sum-
marize the effect of the FICSIT interventions on the social scale.
Table 3 indicates that none of these adjusted within site evalua-
tions was significant. However, there was a clear trend in Port-
land (Site 1) where the intervention was almost significantly as-
sociated with a greater final adjusted social scale score (87.4 ±
0.68, p = .060) than the control group (85.7 ± 0.68). Although
there was a suggested effect in Portland, the other sites did not
add to the trend and none of the meta-analyses in Figure 4 was

Volume 23, Number 3, 2001
Exercise and QOL
193
TABLE 4
Effect Sizes and Significance Levels Associated With Results of Meta-Analyses
Type of
No. of
Subscale
Intervention in
Meta-Analysis
Interventions in
Analysis
General Health
Emotional
Pain
Social
Exercise (all
9
–0.05 ± 0.6 (–1.2, 1.1)
3.97 ± 2.0 (0.1, 7.8)
0.71 ± 0.9 (–1.0, 2.5)
1.10 ± 0.8 (–0.6, 2.8)
interventions)
p
.937
1.96 ± 1.9 (–1.7, 5.7)
.300
0.08 ± 1.8 (–2.7, 4.4)
.648
–0.25 ± 0.7 (–1.6, 1.1)
.724
–0.19 ± 0.7 (–1.5, 1.2)
.781
.043
2.28 ± 4.2 (–5.9, 10.5)
.586
3.07 ± 5.4 (–7.5, 13.7)
.570
3.59 ± 1.6(0.4, 6.8)
.027
3.78 ± 1.6 (0.6, 6.9)
.018
.427
1.07 ± 2.7 (–4.2, 6.3)
.690
2.15 ± 1.8 (–1.4, 5.7)
.239
0.32 ± 1.1 (–1.8, 2.4)
.762
0.35 ± 1.0 (–1.7, 2.4)
.741
.192
–3.79 ± 2.8 (–9.4, 1.8)
.182
–0.43 ± 2.5 (–5.4, 4.5)
.865
1.53 ± 0.9 (–0.3, 3.4)
.103
1.50 ± 0.9 (–0.3, 3.3)
.106
Resistance
p
Balance
p
Endurance
p
Flexibility
p
4
4
3
4
Note. Entries are M ± SE of the covariate adjusted change in the outcome measure generated by the specific meta-analysis. Numbers in parentheses are
95% confidence bounds on the covariate adjusted differences. p values test the null hypothesis that mean differences are zero.
DISCUSSION
level of physical activity suggest more positive outcomes (42).
Other recent studies are indicating that particular intervention
strategies may be more effective in different populations and
that future intervention efforts should be better tailored to partic-
ipant’s functional levels and preferences (43).
Although the literature on effects of exercise on QOL do-
mains measured by the SF–36 is extensive, intervention results
have been inconsistent. McAuley (35) reviewed 23 studies that
evaluated the association between exercise and psychological
well-being in people of all ages and concluded that 69% were
suggestive of positive associations. The results are equally in-
consistent in older populations. In a randomized study of 357
older adults, for example, King, Taylor, and Haskell (36) found
that exercise reduced anxiety, depressive symptoms, and per-
ceived stress. Although other intervention studies (37–38) have
provided similar evidence of a positive association between ex-
ercise and measures of psychological well-being in older adults,
studies finding no positive differences have been equally com-
mon (12,39–41).
One reason for the inconsistent results noted in previous
studies may be diversity in the intensity, duration, and other as-
pects of the exercise programs themselves. Although the epide-
miological literature suggests a dose response effect with addi-
tional health benefits being achieved through greater amounts of
physical activity (1), exercise intensity did not seem to be a ma-
jor factor in our analysis.
Alternatively, it is possible that differences in outcome
measures are responsible, with the MOS not being as sensitive
as more detailed measures of anxiety or depression. It is also
possible that participants recruited into studies may be at a high
level of psychological well-being at the beginning of the study,
thus making it more difficult to detect improvements on generic
QOL scales. Based on the FICSIT experience, we would recom-
mend that future studies incorporate more detailed psychologi-
cal assessment measures if QOL is a major outcome variable of
interest.
Finally, the problem may lie in the fact that intention to treat
analysis may minimize overall intervention effects if compli-
ance to prescribed interventions is low or if study participants in
control groups increase activities on their own. Studies examin-
ing QOL effects in those who adopted and maintained a new
A major concern regarding vigorous exercise in older adults
is the risk of injury and pain (44). Figure 3 demonstrates that
there was no trend toward exercise-induced increases in pain, ei-
ther in the individual FICSIT sites or in the meta-analyses. This
result suggests that none of the exercise programs created
chronic muscle or joint pain, or other injuries, that were suffi-
cient to be noticed above the general background of pain and
discomfort experienced by these elderly participants. Because
FICSIT participants were elderly individuals with many poten-
tial sources of pain (e.g., 64% of our sample reported having ar-
thritis), and because none of the interventions was focused spe-
cifically on any of those sources, we are not surprised that pain
was not reduced. On the contrary, one may reasonably expect an
increase in pain in a frail elderly population subjected to the re-
peated stress of vigorous and frequent exercising. We view the
absence of such an increase as an important positive finding, al-
though recent findings demonstrating that flexibility and
stretching types of exercises can reduce pain in sedentary older
adults are quite intriguing (45).
There was a significant improvement in the emotional
health score that essentially translates into perceptions of less
interference in work and other daily activities due to emotional
problems. This finding is consistent with the general conclusion
from the 1996 Surgeon General’s Report on Physical Activity
and Health (1). That report documented the potential of physical
activity for enhancing psychological well-being by improving
mood, relieving depression, and providing a personal sense of
accomplishment. Although not directly tested in this meta-anal-
ysis, FICSIT interventions may have increased self-efficacy and
the sense of mastery that help to provide focus and meaning to
one’s life. It is noteworthy also that the changes we observed
were not correlated with changes in gait speed and that they

194
Schechtman et al.
Annals of Behavioral Medicine
were maintained after adjusting for such changes. Thus, our data
suggest that the impact of exercise on self-perceived emotional
health cannot be explained by improved physical functioning in
these participants, at least as assessed by gait speed.
teractions between dummy variables and covariates (12 interac-
tions in all). Interpreting these results would be far more com-
plex that the meta-analytic results summarized in the figures and
would not yield the simple intervention-specific summaries that
are provided by these graphs and by Table 3.
The general health scale scores in the combined resistance
and balance group in the Farmington site improved significantly
following the intervention. However, no other intervention
showed a trend toward an improvement, and all meta-analyses
of this measure demonstrated changes that were clustered
around zero. The lack of a response in this domain suggests, as
one may expect, that the FICSIT interventions had limited abil-
ity to broadly affect perceived health. Indeed, respondents who
scored poorly on this parameter generally had the burden of one
or more chronic conditions that the interventions were not de-
signed to affect.
There was a nonsignificant trend toward endurance and flex-
ibility exercises having a beneficial effect on factors that are asso-
ciated with social functioning. Figure 4 indicates that the sug-
gested benefit was small in magnitude and that the effect results
entirely from Portland where the site-specific analysis demon-
strated that the intervention was associated with a nearly signifi-
cant improvement in the social scale (p = .060). As distinct from
other FICSIT sites, Portland conducted a group intervention em-
phasizing social reinforcement, the implication is that any impact
of exercise on social functioning may be mediated more by the
social nature of the intervention than by enhanced mobility and
self-confidence resulting from the exercises themselves.
The decision to use meta-analysis in this article was pro-
spective and based on several considerations. First, publication
bias (17,18) and selection bias (19,20) are common problems in
many meta-analyses, with substantial evidence showing that re-
search with positive results is more likely to be published than
research with negative findings (17,18). Thus, literature-based
meta-analyses tend to be biased in favor of treatment. Addi-
tional problems related to the selection of articles can occur
when researchers are aware of the results of articles being con-
sidered for inclusion, when time has changed the standard of
care, and when large variability in the quality of selected articles
means that meta-analytic conclusions reflect research quality as
much as treatment efficacy. Because FICSIT preselected the
data to be analyzed and provided central oversight of research,
these problems are either nonexistent or limited.
Limitations
Because FICSIT was not designed to compare exercise pro-
grams across sites, our analyses provide limited information
about the type of exercise program that is most likely to improve
QOL. Although we categorize interventions according to their
predominant characteristics, those distinctions do not reflect the
intensity of the exercise, compliance with the program, or other
potentially predictive characteristics of specific interventions.
Although precise compliance figures are not available for the
four sites, study investigators reported relatively high group at-
tendance levels (e.g., in the 70% and higher range during the
course of the active intervention). Nevertheless, because we do
not have precise compliance data and because our subjective in-
tensity measure did not prove useful, we are unable to make the
basic assertion that there is a dose–response relation in the ef-
fects we have observed. We would expect such a relation, but we
do not have the data to prove it.
A further limitation in our results is that the overlapping
components of the intervention characteristics mean that the
characteristic-specific meta-analyses are not independent of one
another. Thus, although the final four bars in Figure 4 suggests
that the modest trend toward exercise having a positive impact
on the social subscale can be explained by the endurance and
flexibility components, there is overlap in the component-spe-
cific analyses so that these conclusions should be interpreted
cautiously. Also, because FICSIT participants came from dispa-
rate sources and because participant selection was not popula-
tion based, we must be cautious in generalizing to other popula-
tions. Finally, we were careful in this study to prospectively
define only four primary analyses: one covariate-adjusted
meta-analysis for each outcome measure. For this reason, multi-
ple comparisons are a minor issue as they relate to the primary
hypotheses. However, there were numerous secondary analyses,
which implies the need for caution in interpreting the associated
p values unless they are substantially below .05.
SUMMARY AND CONCLUSIONS
A second benefit of meta-analysis in this context is that
most such analyses combine findings from studies that differ in
sample and intervention characteristics and in outcome mea-
sures. Although FICSIT participants and interventions varied
across sites, outcome measures were prospectively defined us-
ing a common protocol. Moreover, we were not restricted to
comparing aggregate values only. Thus, unlike the standard
meta-analysis, we adjusted for between-site differences using
covariates with common definitions.
A final benefit of the meta-analytic approach is that it was
simpler than the alternative option of pooling across sites and
evaluating covariate by site interactions. Indeed, an appropriate
pooled analysis would include three dummy variables to repre-
sent the four sites along with factors reflecting the statistical in-
The marked contrast between FICSIT and other investiga-
tions is highlighted by the fact that most previous studies in this
area employed younger and healthier populations and used ei-
ther nonrandomized designs or lacked control groups altogether.
Moreover, most have had small sample sizes and limited statisti-
cal power—facts that may largely explain the inconsistent pub-
lished results. The results of this study suggest that exercise can
improve QOL in at least one important domain without causing
an increase in pain. At the same time, because the magnitude of
the effects are limited and because the confidence bounds in the
figures are narrow, this large study of 1,733 randomized partici-
pants suggests strongly that the QOL benefits of these exercise
interventions in frail elderly adults are modest in size. Finally,

Volume 23, Number 3, 2001
Exercise and QOL
195
(13) Morris JN, Fiatarone M, Kiely DK, et al.: Nursing rehabilita-
tion and exercise strategies in nursing homes. Journals of Ger-
ontology: Medical Sciences. 1999, 54:M494–M500.
(14) Ory MG, Schechtman KB, Miller JP, et al.: The FICSIT Group:
Frailty and injuries in later life: The FICSIT trials. Journal of
the American Geriatric Society. 1993, 41:283–296.
(15) Abeles RP, Gift HC, Ory MG (eds): Aging and Quality of Life:
Charting New Territories in Behavioral Sciences Research.
New York: Springer, 1994.
(16) Kutner NG, Ory MG, Baker DI, et al.: Quality of life assess-
ment issues in a health-promotion intervention trial with older
people. Public Health Reports. 1992, 107:530–539.
(17) Sacks SH, Berrier J, Reitman D, Ancova-Berk VA, Chalmers
TC: Meta-analyses of randomized controlled trials. New Eng-
land Journal of Medicine. 1987, 316:450–455.
the observed effects appear to be independent of improvements
in physical functioning at least as assessed by gait speed.
As a next research step, we recommend additional studies
with designs focused on identifying the type of older adults who
can be expected to achieve the greatest QOL benefits from vari-
ous exercise interventions. In addition, we also recommend that
future studies incorporate more sensitive measures of QOL in
their assessment batteries to ascertain intervention impacts
more precisely and that future analyses examine the effects of
adherence to the prescribed exercise regimens.
The FICSIT model of linking distinct clinical studies with a
coordinating center and a common database has facilitated the
successful implementation of preplanned meta-analyses that
have several methodological advantages. Thus, the FICSIT
model has provided the sample size benefits of a common effi-
cacy analysis while yielding site specific information about lo-
cally defined interventions. Both the FICSIT study design and
the data-analytic approach deserve careful consideration as a
potential model for future multicenter investigations.
(18) Einarson TR, Leeder SJ, Koren G: A method for meta-analysis
of epidemiological studies. Pharmacoepidemiology. 1988,
22:813–824.
(19) Dickersin K, Chan S, Chalmers TC, Sacks HS, Smith Jr. H:
Publication bias and clinical trials. Controlled Clinical Trials.
1987, 8:343–353.
(20) Easterbrook PJ, Berlin JA, Gopalan R, Matthews DR: Publi-
cation bias in clinical research. The Lancet. 1991,
337:867–872.
REFERENCES
(1) U.S. Department of Health and Human Services. US Surgeon
General’s Report of Physical Activity and Health. Atlanta, GA:
Centers for Disease Control and Prevention, 1996.
(2) Kaplan GA, Strawbridge WJ, Cohen WJ, et al.: Natural history
of leisure-time physical activity and its correlates. American
Journal of Epidemiology. 1996, 144:793–797.
(3) Kaplan GA, Seeman TE, Cohen RD, et al.: Mortality among el-
derly in the Alameda County Study: Behavioral, and demo-
graphic risk factor. American Journal of Public Health. 1987,
77:307–312.
(4) Stewart AL, King AC: Evaluating the efficacy of physical activ-
ity for influencing quality of life outcomes in older adults. An-
nals of Behavioral Medicine. 1991, 13:108–116.
(5) Lachman ME, Jette A, Tennstedt J, et al.: A cognitive-behav-
ioral model for promoting regular physical activity in older
adults. Psychology, Health and Medicine. 1997, 2:251–261.
(6) Ory MG, Defriese, GH: Self-Care in Later Life: Research, Pro-
gram, and Policy Issues. New York: Springer, 1998.
(7) Hornbrook MC, Stevens VJ, Wingfield DJ, et al.: Preventing
falls among community dwelling older persons: Results from a
randomized trial. The Gerontologist. 1994, 34:16–23.
(8) Jette AM, Rooks D, Lachman ME, et al.: Home-based resis-
tance training: Predictors of participation and adherence. The
Gerontologist. 1998, 38:412–421.
(9) King AC, Haskell WL, Young D, et al.: Long-term effects of
varying intensities and formats of physical activity on participa-
tion rates, fitness, and lipoproteins in men and women aged
50–65 years. Circulation. 1995, 91:2596–2604.
(10) Fiatrone M, O’Neill E, Ryan ND, et al.: Exercise training and nu-
tritional supplementation for physical frailty in very elderly peo-
ple. New England Journal of Medicine. 1994, 330:1769–1775.
(11) Tinetti ME, Baker DI, McAvay G, et al.: A multifactorial inter-
vention to reduce risk of falling among elderly people living in
the community. New England Journal of Medicine. 1994,
331:821–827.
(21) Province MA, Hadley EC, Hornbrook MC, et al.: Assessing the
effects of exercise on falls and injuries in the frail elderly from
the FICSIT Trials using a pre-planned meta-analysis of individ-
ual patient data. Journal of the American Medicine Association.
1995, 273:1341–1347.
(22) Hornbrook MC, Stevens VS, Wingfield DJ: Seniors’ program
for injury control and education. Journal of the American Geri-
atric Society. 1993, 41:309–314.
(23) Buchner DM, Cress ME, Wagner EH, et al.: The Seattle
FICSIT/Moveit study: The effect of exercise on gait and bal-
ance in older adults. Journal of the American Geriatric Society.
1993, 41:321–325.
(24) Wolf SL, Kutner NG, Green RC, McNeely E: The Atlanta
FICSIT study: Two exercise interventions to reduce frailty in
elders. Journal of the American Geriatric Society. 1993,
41:329–332.
(25) Wolfson L, Whipple R, Judge J, et al.: Training balance and
strength in the elderly to improve function. Journal of the Amer-
ican Geriatric Society. 1993, 41:341–343.
(26) Ware Jr. JE, Sherbourne CD: The MOS 36-item Short-Form
Health Survey (SF-36): I. Conceptual framework and item se-
lection. Medical Care. 1992, 30:473–483.
(27) Stewart AL, Ware JE (eds): Measuring Functioning and
Well-Being: The Medical Outcomes Study Approach. Durham,
NC: Duke University Press, 1992.
(28) Buchner DM, Hornbrook MC, Kutner NG, et al.: The FICSIT
Group: Development of the common database for the FICSIT
trials. Journal of the American Geriatric Society. 1993,
41:297–308.
(29) Tinetti ME, Speechley M, Ginter SF: Risk factors for falls
among elderly persons living in the community. New England
Journal of Medicine. 1989, 319:1701–1707.
(30) Cress ME, Schechtman KB, Mulrow CD, et al.: Relationship
between physical performance and self-perceived physical
function. Journal of the American Geriatric Society. 1995,
43:93–101.
(12) Jette AM, Lachman M, Giorgetti MM, et al.: Exercise—It’s
never too late: The Strong for Life Program American Journal
of Public Health. 1999, 89:66–72.
(31) DerSimonian R, Laird N: Meta-analysis in clinical trials. Con-
trolled Clinical Trials. 1986, 7:177–188.

196
Schechtman et al.
Annals of Behavioral Medicine
Mark C. Hornbrook, Ph.D.
Victor J. Stevens, Ph.D.
Darlene J. Wingfield, M.P.A., P.T.
Merwyn R. Greenlick, Ph.D.
(32) Fleiss JL, Gross AJ: Meta-analysis in epidemiology, with spe-
cial reference to studies of the association between exposure to
environmental tobacco smoke and lung cancer: A critique.
Journal of Clinical Epidemiology. 1991, 44:127–139.
(33) SAS Institute Inc.: SAS/STAT User’s Guide, Version 6 (4th Ed.,
Vol. 1). Cary, NC: Author, 1989.
Program on Aging, Yale University School of
Medicine (U01 AG09087)
(34) SAS Institute Inc.: SAS/STAT User’s Guide, Version 6 (4th Ed.,
Vol. 2). Cary, NC: Author, 1989.
(35) McAuley E: Physical activity and psychosocial outcomes. In
Bouchard C, Shepard R, Stephens T (eds), Physical Activity, Fit-
ness and Health. International Proceedings and Consensus State-
ment. Champaign, IL: Human Kinetics, Inc., 1994, 551–568.
(36) King AC, Taylor CB, Haskell WL: Effects of differing intensi-
ties and formats of 12 months of exercise training on psycho-
logical outcomes in older adults. Health Psychology. 1993,
12:292–300.
(37) Riddick CC, Freitag R: The impact of an aerobic fitness pro-
gram on the body image of older women. Activities, Adapta-
tion, and Aging. 1984, 6:59–70.
(38) Parent CJ, Whall AL: Are physical activity, self-esteem, and depres-
sion related? Journal of Gerontological Nursing. 1984, 10:8–11.
(39) Gitlin LN, Lawton MP, Windsor-Landsberg LA, et al.: In search
of psychological benefits: Exercise in healthy elderly adults.
Journal of Aging and Health. 1982, 4:174–192.
Mary E. Tinetti, M.D.
Dorothy I. Baker, Ph.D., R.N.C.
Elizabeth B. Claus, Ph.D.
Ralph I. Horwitz, M.D.
Department of Health Sciences, University of
Washington (U01 AG09095)
David M. Buchner, M.D., M.P.H.
Edward H. Wagner, M.D., M.P.H.
Barbara J. de Lateur, M.D.
M. Elaine Cress, Ph.D.
Robert Price, M.M.S.E.
Itmar B. Abrass, M.D.
(40) Emery CF, Gatz M: Psychological and cognitive effects of an
exercise program for community-residing older adults. The
Gerontologist. 1990, 30:184–192.
Peter Esselman, M.D.
Tony Marguerita, M.D.
(41) Blumenthal AJ, Emery CF, Madden PJ, et al.: Cardiovascular
and behavioral effects of aerobic exercise training in healthy
older men and women. Journal of Gerontology. 1989,
44:M147–M157.
(42) Stewart AL, Mills KM, Sepsis PG, et al.: Evaluation of
CHAMPS, a physical activity program for older adults. Annals
of Behavioral Medicine. 1997, 19:353–361.
(43) King AC, Rejeski WJ, Buchner D: Physical activity interven-
tions targeting older adults: A critical review and recommenda-
tions. American Journal of Preventive Medicine. 1998,
15:316–330.
(44) Clark DO: Perceived barriers and facilitators to exercise among
older low income adults. Journal of Clinical Geropsychology.
1999, 1:51–62.
Geriatric Research, Education, and Clinical Center
Audie L. Murphy Veterans Hospital (U01 AG09117)
Cynthia D. Mulrow, M.D.
Meghan B. Gerety, M.D.
John E. Cornell, Ph.D.
Louis A. DeNino, Ph.D.
Deanna Kanten, R.N.
Department of Rehabilitation Medicine
Emory University School of Medicine
(U01 AG09124)
(45) King A, Pruit LA, Phillips W, et al.: Comparative effects of two
physical activity programs on measured and perceived physical
functioning and other health-related quality of life outcomes in
older adults. Journals of Gerontology: Medical Sciences (in
press).
Steven L. Wolf, Ph.D. FAPTA
Nancy G. Kutner, Ph.D.
Robert C. Green, M.D.
Elizabeth McNeely, Ph.D.
Carol Coogler, P.T., DSc.
APPENDIX
Hebrew Rehabilitation Center for Aged
Harvard Medical School
The FICSIT Group consists of:
Division of Biostatistics, Washington University
School of Medicine (U01 AG09098)
USDA Human Nutrition Research Center on Aging
(U01 AG09078)
J. Philip Miller, A.B.
Maria A. Fiatarone, M.D.
Evelyn F O’Neill, C.T.R.S
Nancy Doyle Ryan, Diet T.
Karen M. Clements, M.P.H.
Lewis A. Lipsitz, M.D.
Michael A. Province, Ph.D.
Kenneth B. Schechtman, Ph.D.
Cynthia L. Arfken, Ph.D.
Jane E. Rossiter-Fornoff, D. Phil
Joseph J. Kehayias, Ph.D.
Susan B. Roberts, Ph.D.
William J. Evans, Ph.D.
Center for Health Research, Kaiser Permanente North-
west Region (U01 AG09089)

Volume 23, Number 3, 2001
Exercise and QOL
Paula M. Amerman, R.N., M.S.N.
197
University of Iowa
Iowa State University (U01 AG09123)
James O. Judge, M.D.
Carol A. Derby, Ph.D.
Mary B. King, M.D.
Robert Wallace, M.D.
Jo Ellen Ross, M.A.
Jeffrey C. Huston, Ph.D.
Carolyn J. Kundel, Ph.D.
Michael S. Selberg, B.S.
NIA/NCNR
Evan Hadley, M.D.
Ashlesha Tamboli, M.D., Ph.D.
Marcia G. Ory, Ph.D., M.P.H.
Sharlene Weiss, Ph.D., R.N.
Department of Neurology,
University of Connecticut Health Center
(U01 AG09096)
Centers for Disease Control and Prevention
Richard W. Satin, M.D.
Leslie I. Wolfson, M.D.
Robert H. Whipple, M.A.