Teri L, Gibbons LE, McCurry SM, Logsdon RG, Buchner DM, Barlow WE, Kukull WA, LaCroix AZ, McCormick W, Larson EB. Exercise Plus Behavioral Management in Patients With Alzheimer DiseaseA Randomized Controlled Trial. JAMA. 2003;290(15):2015-2022. doi:10.1001/jama.290.15.2015
Author Affiliations: Departments of Psychosocial and Community Health (Drs Teri, Gibbons, McCurry, and Logsdon, Biostatistics (Dr Barlow), Epidemiology (Dr Kukull), and Medicine (Drs McCormick and Larson), University of Washington, Seattle; Division of Nutrition and Physical Activity, Centers for Disease Control and Prevention, Atlanta, Ga (Dr Buchner); and Center for Health Studies, Group Health Cooperative, Seattle, Wash (Drs Barlow, LaCroix, and Larson).
Context Exercise training for patients with Alzheimer disease combined with
teaching caregivers how to manage behavioral problems may help decrease the
frailty and behavioral impairment that are often prevalent in patients with
Objective To determine whether a home-based exercise program combined with caregiver
training in behavioral management techniques would reduce functional dependence
and delay institutionalization among patients with Alzheimer disease.
Design, Setting, and Patients Randomized controlled trial of 153 community-dwelling patients meeting
National Institute of Neurological and Communicative Diseases and Stroke/Alzheimer
Disease and Related Disorders Association criteria for Alzheimer disease,
conducted between June 1994 and April 1999.
Interventions Patient-caregiver dyads were randomly assigned to the combined exercise
and caregiver training progam, Reducing Disability in Alzheimer Disease (RDAD),
or to routine medical care (RMC). The RDAD program was conducted in the patients'
home over 3 months.
Main Outcome Measures Physical health and function (36-item Short-Form Health Survey's [SF-36]
physical functioning and physical role functioning subscales and Sickness
Impact Profile's Mobility subscale), and affective status (Hamilton Depression
Rating Scale and Cornell Depression Scale for Depression in Dementia).
Results At 3 months, in comparison with the routine care patients, more patients
in the RDAD group exercised at least 60 min/wk (odds ratio [OR], 2.82; 95%
confidence interval [CI], 1.25-6.39; P = .01) and
had fewer days of restricted activity (OR, 3.10; 95% CI, 1.08-8.95; P<.001). Patients in the RDAD group also had improved
scores for physical role functioning compared with worse scores for patients
in the RMC group (mean difference, 19.29; 95% CI, 8.75-29.83; P<.001). Patients in the RDAD group had improved Cornell Depression
Scale for Depression in Dementia scores while the patients in the RMC group
had worse scores (mean difference, −1.03; 95% CI, −0.17 to −1.91; P = .02). At 2 years, the RDAD patients continued to have
better physical role functioning scores than the RMC patients (mean difference,
10.89; 95% CI, 3.62-18.16; P = .003) and showed a
trend (19% vs 50%) for less institutionalization due to behavioral disturbance.
For patients with higher depression scores at baseline, those in the RDAD
group improved significantly more at 3 months on the Hamilton Depression Rating
Scale (mean difference, 2.21; 95% CI, 0.22-4.20; P =
.04) and maintained that improvement at 24 months (mean difference, 2.14;
95% CI, 0.14-4.17; P = .04).
Conclusion Exercise training combined with teaching caregivers behavioral management
techniques improved physical health and depression in patients with Alzheimer
It is well-known that Alzheimer disease adversely affects cognitive,
emotional, and behavioral functioning.1 Less
well-known are the deleterious effects of Alzheimer disease on physical conditioning.
However, there are a number of studies linking Alzheimer disease with physical
deterioration. For example, when compared with age-matched controls, Alzheimer
disease patients show more signs of undernutrition,2 higher
risk of falls and fractures,3- 6 and
more rapid decline on measures of mobility.7,8 Once
injured, Alzheimer disease patients are at greater risk of subsequent injury
than age- and sex-matched controls.3 Reduced
muscle mass has also been associated with loss of independence.9 Consequently,
improved physical conditioning for patients with Alzheimer disease may extend
their independent mobility and enhance their quality of life despite progression
of the disease.
Research is accumulating to suggest that even the oldest adults can
improve cardiovascular function and increase flexibility, balance, and strength
with systematic exercise training.10,11 In
one uncontrolled study of 11 patients with Alzheimer disease, patients benefited
from a hospital-based exercise program.12 Furthermore,
exercise programs have been shown to improve function even in frail nursing
Exercise can yield additional benefits for elderly dementia patients.
In elderly individuals without dementia, randomized controlled clinical trials
have demonstrated that exercise successfully reduces depression.13- 16 Between
17% and 86% of dementia patients are depressed.17 Exercise
may provide the added benefit of reducing their levels of depression.
Because caregivers are responsible for structuring the patient's day-to-day
activities and providing ongoing care, teaching them effective caregiving
strategies to encourage exercise and avoid behavioral problems associated
with increased activity may make exercise training most effective and most
reasonable given the circumstances. Caregivers have been successfully trained
to reduce patient depression,18 agitation,19,20 and delay institutionalization.21 The attention and activity inherent in exercise programs
can be an opportunity to improve patient-caregiver interactions. If positive
behavioral strategies are used for encouraging exercise participation, exercise
may increase opportunities for pleasant interactions between patient and caregiver
and conflicts may be reduced. Thus, a caregiver-supervised exercise program
for patients with Alzheimer disease may yield significant improvements in
physical health, affect, and behavioral distress.
This study was undertaken to determine whether a home-based exercise
program combined with caregiver training in behavioral management techniques
would reduce functional dependence and delay institutionalization among patients
with Alzheimer disease. The Reducing Disability in Alzheimer Disease (RDAD)
program was compared with routine medical care in a randomized controlled
clinical trial. It was hypothesized that patients in the RDAD program would
show significant improvement on measures of physical frailty and depression
compared with those obtained among patients in the routine care group.
A total of 153 patients were randomized from an ongoing, community-based
Alzheimer disease patient registry22 and through
referrals from physician practices and community advertisements. Enrollment
began in June 1994 and follow-up ended in April 1999. The study was approved
by institutional review boards of both the University of Washington and Group
Health Cooperative. Written consent was obtained from both patient and caregiver.
Additionally, caregivers (next of kin or legal guardians) provided consent
on behalf of patients. All patients received a comprehensive, multidisciplinary
diagnostic evaluation, and results were reviewed at consensus meetings attended
by a geriatrician, neurologist, psychologist, epidemiologist, nurse, and research
All patients met National Institute of Neurological and Communicative
Diseases and Stroke/Alzheimer Disease and Related Disorders Association criteria23 for probable or possible Alzheimer disease, were
required to be community-dwelling, ambulatory, and to have a caregiver who
was willing to participate in training sessions. Patients ranged in age from
55 to 93 years, were predominantly male (59%), white (89%), and had dementia
for an average of 4.3 years. Patients' mean (SD) Mini-Mental State Examination
(MMSE) score was 16.8 (7.1), which placed them in the moderate to severe range
of cognitive impairment.
The caregivers were spouses or adult relatives who lived with or spent
a minimum of 4 hours every day with the patient. Caregivers' ages ranged from
24 to 91 years; 70% were female, 87% were white, and 80% were spouses.
Patient-caregiver dyads were randomly assigned to exercise plus behavior
management techniques (RDAD program) or routine medical care. The random allocation
sequence was obtained from a computer program that blocked groups of 8 patients.
Dyads were randomized after the baseline assessment by research coordinators.
Assessments were conducted at screening, baseline, after 3 months (posttreatment)
and at 6, 12, 18, and 24-month follow-up by interviewers blind to treatment
assignment (Figure 1).
The active treatment program (RDAD) was adapted from 2 previously established
treatments—one to reduce behavioral problems in Alzheimer disease18 and one to increase exercise among older adults.24 Patient-caregiver dyads assigned to this program
were seen in their own homes for 12 hour-long sessions on a schedule of 2
sessions per week for the first 3 weeks, followed by weekly sessions for 4
weeks, and then biweekly sessions over the next 4 weeks. Three follow-up sessions
were conducted over the next 3 months to answer questions and consolidate
treatment gains. Home health professionals experienced in dementia care conducted
The exercise component of the RDAD program included aerobic/endurance
activities, strength training, balance, and flexibility training. The goal
was for patients to engage in a minimum of 30 min/d of moderate-intensity
exercise. In the behavioral management component of the RDAD program, caregivers
were taught to identify and modify patient behavioral problems that impaired
day-to-day function and adversely affected patient-caregiver interactions.
Caregivers were given specific instructions about how to reduce the occurrence
of these problems while also teaching them skills to identify and modify precipitants
of patient distress. Caregivers were also educated about dementia, its impact
on patient behavior and function, and how to modulate their own responses
to problems. The caregivers were encouraged to identify pleasant activities
for their patients to encourage positive interactions and to increase physical
and social activity.
In each session, exercises were demonstrated and practiced, caregivers
were taught how to encourage and help patients with their exercises, and behavioral
plans were developed and implemented. A new topic was introduced in each of
the first 10 sessions; subsequent sessions focused on solidifying gains and
helping caregivers and patients maintain exercise and behavioral management
after the study ended. A complete RDAD treatment manual is available from
the corresponding author and a description of the intervention has been published
Patients in the control group received routine medical care, including
acute medical or crisis intervention provided at community health care centers.
This could include nonspecific advice and support routinely provided by nurses
and primary care physicians or community support services. Specific exercise
and behavioral management training was not provided to control patients.
Treatment adherence was maintained and monitored by weekly supervision
of each trainer by clinical geropsychologists (L.T., S.M.M.) and a physical
therapist. Treatment sessions were videotaped and reviewed by independent
raters to ensure that trainers followed the treatment protocol.
The primary patient outcomes were physical health and function and affective
status. We hypothesized that patients receiving RDAD would improve in each
area. However, because not all patients entered the study with affective or
behavioral disturbance, we believed that these areas would show improvement
only in patients for whom such problems were evident at baseline. We also
hypothesized that physical health and function would show the most gain because
this domain was relevant across all patients.
Physical Health and Function. Two subscales
of the Medical Outcome Study 36-item Short-Form Health Survey (SF-36; physical
functioning and physical role functioning) and 3 from the Sickness Impact
Profile (SIP; body care and movement, mobility, and home management subscales)
were obtained. The SF-36 and SIP assess general health status, are psychometrically
sound, and have been used extensively with older adults.26,27 Higher
SF-36 scores indicate better health functioning; higher SIP scores indicate
worse function. Caregivers completed these measures based on their personal
experience with the patients.
Affective Status. Professional interviewers
assessed depression based on direct observations of the patient and caregiver
interviews. We used the Hamilton Depression Rating Scale28,29 and
the Cornell Scale for Depression in Dementia,30 which
are psychometrically sound measures that have been used to assess depression
in older adults with dementia.31,32 Higher
scores indicate greater impairment. Interviewers were trained and periodically
monitored by a clinical geropsychologist (S.M.M.) to ensure interrater reliability.
Patient performance-based and caregiver-report assessments were obtained.
Patient walking speed, functional reach, and standing balance were measured.
These measures of physical health and function have been used extensively
in trials of healthy older adults.33- 35 Caregiver
reports included the number of minutes spent walking or doing another aerobic
activity for exercise in the past week; the number of restricted activity
days and days spent in bed during the past 2 weeks; and falls and near falls
during the past month. These measures have demonstrated validity in healthy
older adults36,37 and are responsive
to change.38 To our knowledge, this is the
first trial using these measures with dementia patients. To assess the level
of patient behavioral disturbance and caregiver distress, the Revised Memory
and Behavior Problem Checklist was used. Good psychometric reliability and
validity have been reported with use of this checklist.39
Baseline demographic information was reported for patient and caregiver
age, sex, ethnic group, education, and relationship. Patient age at onset
and duration of dementia were obtained at screening. In addition, patient
cognitive status was assessed using the MMSE.40 No
change was hypothesized in the MMSE as a function of treatment; it was obtained
for descriptive purposes. Caregivers completed an adverse symptom checklist
at each visit. No unexpected or serious adverse events were attributed to
the RDAD program, and there was no difference between active and control groups
in adverse symptoms. Exercise compliance was assessed using daily exercise
logs completed by caregivers, and ratings of exercise homework completed by
trainers.25 Completion of assigned behavioral
management homework (eg, recommended readings, viewing a training videotape,
or implementing a behavioral change plan) was also rated after each session
by study trainers.
The study was designed to have 80% power (α = .05) to detect at
least half SD difference on the primary outcome measure. Between-group comparisons
of baseline covariates were conducted using Fisher exact tests, t tests, or nonparametric Kruskal-Wallis tests. Cox proportional hazards
survival analyses were used to determine which baseline characteristics significantly
predicted patient attrition. Hazard ratios and 95% confidence intervals (CIs)
Outcome analyses compared the RDAD group with the routine care group
using generalized estimating equations for linear, logistic, and Poisson regression.41 Mean (SD) differences, odds ratios, and relative
risks were calculated with 95% CIs.
For pretrial and posttrial analyses, the outcome at the 3-month visit
(corresponding to the end of the intervention) was regressed on treatment
group, controlling for the baseline value of that factor. Pretrial and postrial
analyses were based on the intention-to-treat (ITT) principle, using all randomized
patients. Baseline values were carried forward for patients missing the posttest.
In secondary analyses, these analyses were repeated without imputation for
missing posttests, and nonparametric Kruskal-Wallis tests were conducted on
the change scores. SAS statistical software was used to perform analyses (Version
6.12; SAS Institute Inc, Cary, NC).
Longitudinal analyses used all 5 posttreatment visits (3, 6, 12, 18,
and 24 months) and time, controlling for the baseline value of the outcome.
This is a repeated-measures design, with up to 5 observations per person.
To account for within-patient correlation of scores at different time points,
an autoregressive correlation structure was modeled, which assumed that consecutive
visits were more highly correlated than nonconsecutive visits. Time by group
interactions were assessed with the same model structure and were reported
if significant. Potential confounders (age, sex, MMSE score at baseline, and
duration of dementia) were evaluated by entering their baseline values as
covariates and noting the change in the estimated treatment coefficient. Changes
of more than 15% were considered to be evidence of confounding. Two sets of
longitudinal analyses were conducted. The primary longitudinal analyses used
all available data for each patient. In the complete follow-up analyses, only
patients with 24 months of follow-up were included. In addition, a sensitivity
analysis of missing data was conducted. Outcome data were ranked by visit
and missing data were assigned the best or the worst rank in the following
combinations: (1) all worst, (2) all best, (3) control worst and RDAD best,
and (4) control best and RDAD worst.
There were no significant differences at baseline in any patient or
caregiver characteristics (such as age, sex, duration of dementia, and medication
use) or assessment measures (Table 1).
At 3 months (posttest), significant differences were obtained between
groups for the primary measures of physical role function and affective status.
In regression analyses, the mean estimated difference for the SF-36 was 19.29
(95% CI, 8.75-29.83; P<.001), and it was −1.03
(95% CI, −0.17 to 1.19; P = .02) for the Cornell
Depression Scale. For both outcomes, patients in the RDAD group improved while
routine care patients declined. Using ITT, an improvement of 5.9 points on
the SF-36 physical role functioning subscale was obtained for RDAD patients
compared with a decline of 16.6 points for control patients; on the Cornell
scale, an improvement of 0.5 points for RDAD patients compared with a decline
of 0.5 for control patients. Significant differences were also obtained on
secondary physical and health function measures. At baseline, 56% of those
in both the control and RDAD groups reported exercising at least 60 min/wk.
At posttest, 79% of those in the RDAD group and 62% of those in the routine
care group reported exercising 60 min/wk, which is an improvement of 23% for
the RDAD group vs 6% for the control group (odds ratio, 2.82; 95% CI, 1.25-6.39; P = .01). Restricted activity days decreased by an average
of 0.5 days in the RDAD group, but increased by 0.2 days in the control group
(odds ratio, 3.10; 95% CI, 1.08-8.95; P<.001).
The percentage of RDAD patients experiencing restricted days was also lower.
No other significant differences were obtained. Nonparametric Kruskal-Wallis
tests were conducted on the pretrial and posttrial change scores to verify
distributional assumptions. Results were similar.
At 24 months of follow-up, significant differences remained between
patients in the RDAD group and the control group on the SF-36 physical role
functioning subscale (mean difference, 10.89; 95% CI, 3.62-18.16; P = .003). Significant differences between RDAD and routine care patients
emerged on the SIP Mobility scale (relative risk, 1.27; 95% CI, 1.03-1.56; P = .02). Table 2 presents
the observed data before imputation for missing posttests. The longitudinal
analyses were repeated with the 89 patients who completed all 24 months of
follow-up. Significant differences were again obtained between the RDAD group
and the routine care group on the SF-36 physical role functioning subscale
(mean difference, 10.59; 95% CI, 2.22-18.96; P =
.01) and SIP Mobility scale (relative risk, 1.37; 95% CI, 1.07-1.75; P = .01).
In the sensitivity analyses, the SF-36 physical role functioning scores
were robust for most of the possible assumptions about the 22% of patients
who were missing data. Under most scenarios, the RDAD intervention improved
patients' physical functioning (mean difference, 9.18; 95% CI, 1.57-16.78
[all worst]; mean difference, 8.81; 95% CI, −2.62 to 20.24 [all best];
mean difference, 48.26; 95% CI, 38.67-57.85 [RDAD group best and control worst];
mean difference, 10.45; 95% CI, 4.06-16.85 [complete data]). Only in the most
extreme combination, in which all the missing control patients would have
had the best scores and all the missing RDAD patients would have had the worst
scores, would the routine care have been better than RDAD intervention. For
the SIP Mobility scale, assigning RDAD the best ranks and control the worst
retained the significance of the findings.
Of 153 patients who began the study, 140 (92%) completed posttest assessment.
Of the 13 who discontinued at 3 months, 8 had been assigned to the RDAD group
and 5 to the routine care group (Fisher exact test, P =
.40). Eighty-nine (58%) completed 24-month assessments. Patients who completed
this last assessment were less cognitively impaired (baseline MMSE score hazard
ratio, 1.36; 95% CI, 1.15-1.61) and had an average MMSE score that was 5 points
lower (P<.001). There were no other significant
demographic differences between those who completed this assessment and those
who did not.
Patient institutionalization was the major reason patients did not complete
all assessments, with no significant differences between treatment arms (67%
for the control group and 68% for the RDAD group; χ21 = 0.04 and P = .84). However, the reasons
for patient institutionalization did differ (Table 3). Eleven control patients (50%) were institutionalized because
of patient behavioral problems compared with 4 RDAD patients (19%). One control
patient (5%) was institutionalized because of caregiver health or caregiver
availability compared with 8 RDAD patients (38%).
We further investigated patients with preexisting mood disturbance to
determine whether those with problems were more or less likely to benefit
from intervention. Patients with a score of 6 or greater on the Cornell scale
were selected because this score was above our sample mean score at baseline.
For these patients, Hamilton Depression Rating scale scores were examined.
In ITT analysis, RDAD patients improved at posttest (mean [SD], 2.0 [4.9])
while control patients worsened (mean [SD], 0.6 [5.1]). The mean adjusted
difference was 2.21 (95% CI, 0.22-4.20; P = .04).
Over 24 months of follow-up, this difference of 2.14 (95% CI, 0.14-4.17) remained
statistically significant (P = .04).
Baseline cognitive status (MMSE score), patient or caregiver sex, and
duration of dementia did not affect results. Analyses of treatment compliance
data completed by study trainers showed that 91% of RDAD patients attempted
their exercise homework (79% completed >75% of assigned homework). Only 9%
of RDAD patients did not complete any homework.
This study demonstrated that an integrated treatment program designed
to train dementia patients and their caregivers in exercise and behavioral
management techniques was successfully implemented in a community setting.
Caregivers were able to learn how to encourage and supervise exercise participation,
and patients participating in this program achieved increased levels of physical
activity, decreased rates of depression, and improved physical health and
function. Patients in the RDAD group fared significantly better than those
in the control group. Scores on the SF-36 Physical Role Functioning subscale,
Cornell Depression scale, and the number of restricted activity days all significantly
improved. Posttest physical function improvements were maintained at 24-month
follow-up and, for those patients entering with higher levels of initial depression,
improvements in depression were maintained after 24 months.
The reasons for patient institutionalization throughout the 24-month
follow-up period differed between patients in the RDAD group and in the routine
care group. Eleven routine care patients (50%) who discontinued the study
due to institutionalization did so because he/she experienced an increase
of behavioral problems compared with only 4 (19%) of 21 in the RDAD group
(Table 3). These numbers are small
but suggest that the RDAD program may have influenced patients and caregivers
to a significant enough degree to delay institutionalization caused by an
inability to manage patient behavioral disturbances.
Adherence to program recommendations was quite high demonstrating that
community-dwelling caregivers can be successfully trained to supervise a home
exercise program for persons with dementia. Although other studies have found
that exercise professionals in structured institutional settings can implement
exercise programs,10,42- 44 this
is the first to show that a simplified exercise program can be taught to caregivers
of dementia patients residing in the community. Furthermore, this study involved
a heterogeneous array of patients and caregivers, lending support to the generalizability
of these findings.
We included both caregivers and patients in this intervention. We did
not, however, assess the degree to which caregivers felt satisfied with what
they were learning, nor did we assess the potential outcomes of treatment
on caregivers. It became clear as we conducted the study that caregivers were
benefiting from participation, but the nature of that benefit and its potential
impact on their caregiving was not determined. We also did not investigate
the relative efficacy of exercise or behavioral management in producing the
results reported herein. We were interested in the combined effect of both
treatment components because we were interested in improving both physical
and affective health. Now that the combination has been shown efficacious
in producing change, we would be interested in exploring the relative effectiveness
of each component alone as well as the impact of treatment on caregivers themselves.
While we hypothesize that the combination is superior and that caregivers
benefit from training, these are empirical questions worthy of inquiry.
Future research is needed to determine whether the effects obtained
herein can be replicated or improved. Given that our depressed patients improved,
a more targeted approach may show stronger results. That would be consistent
with our own earlier work in which we found that a targeted behavioral approach
was successful in reducing the levels of depression in dementia patients and
their caregivers.18 Because exercise is also
associated with reduced depression in adults without dementia, targeting patients
with coexisting depression and dementia might enhance treatment effects. Given
these results and the consistently strong association between physical exercise
and health in older adults without dementia,13- 15 the
potential health benefits of a simple exercise program for older adults with
dementia should not be overlooked.