Effects of Exercise Training on Frailty in Community-Dwelling Older Adults: Results of a Randomized, Controlled Trial


Address correspondence to Ellen F. Binder, MD, WU Older Adult Health Center, 4488 Forest Park Blvd., Suite 201, St. Louis, MO 63108. E-mail: ebinder@im.wustl.edu


OBJECTIVES: Although deficits in skeletal muscle strength, gait, balance, and oxygen uptake are potentially reversible causes of frailty, the efficacy of exercise in reversing frailty in community-dwelling older adults has not been proven. The aim of this study was to determine the effects of intensive exercise training (ET) on measures of physical frailty in older community-dwelling men and women.

DESIGN: Randomized controlled trial.

SETTING: Medical school research center.

PARTICIPANTS: One hundred fifteen sedentary men and women (mean age ± standard deviation = 83 ± 4) with mild to moderate physical frailty, as defined by two of the following three criteria: Modified Physical Performance Test (modified PPT) score between 18 and 32, peak oxygen uptake (.VO2 peak) between 10 and 18 mL/kg/min, and self-report of difficulty or assistance with one basic activity of daily living (ADL), or two instrumental ADLs.

INTERVENTION: Participants were randomly assigned to a control group that performed a 9-month low-intensity home exercise program (control) or an exercise-training program (ET). The control intervention primarily consisted of flexibility exercises. ET began with 3 months of flexibility, light-resistance, and balance training. During the next 3 months, resistance training was added, and, during the next 3 months, endurance training was added.

MEASUREMENTS: Modified PPT score, .VO2 peak, performance of ADLs as measured by the Older Americans Resources and Services instrument, and the Functional Status Questionnaire (FSQ).

RESULTS: ET resulted in significantly greater improvements than home exercise in three of the four primary outcome measures. Adjusted 95% confidence bounds on the magnitude of improvement in the ET group compared with the control group were 1.0 to 5.2 points for the modified PPT score, 0.9 to 3.6 mL/kg/min for .VO2 peak, and 1.6 to 4.9 points for the FSQ score.

CONCLUSIONS: Our results show that intensive ET can improve measures of physical function and preclinical disability in older adults who have impairments in physical performance and oxygen uptake and are not taking hormone replacement therapy better than a low-intensity home exercise program.

Physical frailty is a syndrome characterized by declines in multiple physiological domains, including muscle mass and strength, flexibility, balance and neuromuscular coordination, and cardiovascular function. In addition to aging per se, factors that contribute to the development of frailty include chronic illnesses, poor nutrition, and physical inactivity.1 Frailty greatly increases the risk for functional decline, institutionalization, morbidity, and mortality;2,3 There is evidence supporting the beneficial effects of exercise on functional capacity in older people.4 However, studies of the effects of exercise on frailty have been limited to patients with specific diseases such as arthritis,5 nursing home residents,6,7 and younger individuals8 or have focused on selected intermediate outcome measures such as strength, balance, and gait,9–11 with mixed results. As a result, there is insufficient information as to whether community-dwelling frail older people can adapt to exercise training (ET) with a physiological response sufficient to delay or reduce frailty. We performed a randomized controlled study to evaluate whether a multidimensional exercise-training program can significantly reduce frailty in community-dwelling older men and women.



Men and women aged 78 years and older were recruited from the community using mass media, direct mailings, and community efforts to participate in studies of exercise. Participants provided written informed consent to participate in this study, which the Institutional Review Board of the Washington University School of Medicine approved.

Preliminary screening evaluations included a medical history, physical examination, chest x-ray, blood and urine chemistries, standardized interviews about performance of activities of daily living (ADLs), the Short Blessed Test of Orientation, Memory and Concentration,12 a modified Physical Performance Test (PPT), and measurement of peak oxygen uptake (VO2 peak).

ADL Measures

The Older American Resources and Services (OARS) instrument13,14 was used to collect information about the use of human assistance or assistive technology for basic activities of daily living (ADLs) and instrumental activities of daily living (IADLs). The Physical Function subscale of the Functional Status Questionnaire (FSQ)15 was used to collect information about difficulty with task performance over the previous month. The OARS ADL and IADL scales each have a possible maximum score of 14; the FSQ subscale has a possible maximum score of 36.

Modified Physical Performance Test

We used an objective evaluation of physical function that is a modification of the PPT developed by Reuben et al.16 that correlates with degree of disability, loss of independence, and mortality.16,17 Our modification, which focuses on gross motor function, substitutes a chair-rise task and a balance task18 for the writing and eating tasks in the original PPT. This modified PPT includes seven standardized tasks that are timed (50-foot floor walk, putting on and removing a laboratory coat, picking up a penny from the floor, standing up five times from a 16-inch chair, lifting a 7-pound book to a shelf, climbing one flight of stairs, and standing with feet in side-by-side, semitandem and full-tandem positions) and two additional tasks (climbing up and down four flights of stairs and performing a 360° turn). The score for each item ranges between 0 and 4, with 36 representing a perfect total score for the test. Test-retest reliability for the total modified PPT score in our laboratory for this population is 0.96.

Peak Oxygen Uptake

.VO2peak was assessed during graded treadmill walking. During a 3- to 5-minute warm-up on the treadmill at 0% grade, the speed was adjusted to identify the fastest comfortable walking speed for the individual. Speed was then held constant during the test, and elevation was progressively increased 2% to 3% every 2 minutes. As a safety measure, participants were allowed to hold on to a handrail to maintain their balance during the test. Cardiorespiratory data were collected at 30-second intervals using a computerized system.19 The test was terminated when the participant became too fatigued to continue the test.

We used objective criteria that were developed a priori to define mild to moderate physical frailty. Because there are no established criteria or criterion standard for frailty, we included measures with established predictive validity for disability and mortality in older populations.2,16–18,20 To be eligible for this study, individuals had to meet at least two of the following three criteria: (1) score between 18 and 32 on the modified PPT, (2) report of difficulty or need for assistance with up to two IADLs or one ADL, or (3) achievement of a .VO2peak between 10 and 18 mL · kg−1· min−1 (range of age-predicted .VO2peak for healthy sedentary adults aged 75–80 is 18–30 mL · kg−1· min−1).21

Individuals were excluded for the following indications: (1) did not meet two of our three frailty criteria; (2) medical conditions that would contraindicate vigorous exercise; (3) neuromuscular disorders unlikely to improve with exercise; (4) chronic use of corticosteroids, immunosuppressive drugs, or androgen-, estrogen-, or progestin-containing compounds (because we did not want to confound analyses of secondary outcome measures of body composition and bone density); (5) cigarette use within the previous year; (6) diagnosis of cancer within the previous 5 years; (7) sensory impairments that would interfere with following instructions for testing or ET; or (8) significant cognitive impairment.

Baseline and Follow-up Assessments

Individuals enrolled in the study underwent a series of assessments at baseline and upon completion of each intervention phase (every 3 months: baseline, Test 2, Test 3, Test 4).

A physical therapy evaluation included the following tests: modified PPT, maximal voluntary muscle strength for knee extension and flexion using Cybex® isokinetic dynamometry, and range of motion of the hip, knee, ankle, and shoulder using standard goniometry. Balance was assessed using the Functional Reach Test,22 a balance beam, single-limb stance time, and the Berg Balance instrument.23 The physical therapy assessment procedures have been previously described.24 Personnel who performed the physical therapy assessments were not involved in training the participants.

In addition to the ADL questionnaires, the Medical Outcomes Short Form (SF-36)25 and the Geriatric Depression Scale26 were administered. For ET participants, .VO2 peak was measured at baseline, and before and after endurance training. For control participants, .VO2peak was measured at baseline and end of study.

Diet Evaluation

After receiving detailed instruction, participants completed a 3-day food record at baseline and every 3 months until the end of the study, under the supervision of a registered dietitian. The food-intake records were analyzed using Nutritionist IV (First Databank, San Bruno, CA). Participants were not prescribed a specific diet.

Randomization Procedures

After the baseline assessments, study participants were randomly assigned to the ET or control group in a 3:2 ratio, using a computer-generated random permutation procedure27 and a block design. More participants were assigned to the ET group because we anticipated that attrition would be higher in that group. The study personnel who maintained the randomization log were not involved in screening, testing, or training procedures.

ET Program

The ET program was conducted in our indoor exercise facility and supervised by exercise physiology technicians. It consisted of three approximately 3-month-long phases of ET. The primary rationale for an exercise program with successive phases was that each type of exercise was designed to help prepare the participants for the subsequent phase, to enable them to perform the exercises more effectively and minimize injury. The training components were selected specifically to target physical impairments that are associated with frailty.

Participants were required to attend exercise sessions three times per week and complete 36 sessions of each exercise phase before follow-up assessments and progression to the next phase of ET.

The first exercise phase used a group format and included 22 exercises that focused on flexibility, balance, coordination, speed of reaction, and, to a modest extent, strength. It has been described previously.28 The exercises were made progressively more difficult over time by increasing the number of repetitions and by performing them in more challenging ways.

During the second phase, progressive resistance training was added. One-repetition maximum (1RM) voluntary strength measurements were performed at baseline for each of six different exercises (knee extension, knee flexion, seated bench press, seated row, leg press, and biceps curl), which were performed on a Hoist weightlifting machine (Hoist Fitness Systems, San Diego, CA). Initially, the participants performed one to two sets of six to eight repetitions of each exercise at 65% of the 1RM. By the end of the first month of weight training, they progressed to three sets of eight to 12 repetitions performed at 85% to 100% of the initial 1RM. The 1RM measurements were repeated at monthly intervals to provide information for adjusting each individual's exercise prescription. Abdominal muscle exercises and some free-weight exercises were added after the first month. The participants also continued to perform a shortened version of the Phase 1 exercises.

In the third phase, endurance training was performed using treadmills, stationary bicycles, Aerodyne bicycles, or rowing machines. Initially, the intensity of the exercise was set at a level that elicited approximately 65% to 70% of .VO2peak, and the participants exercised at this intensity for 15 minutes. The duration of exercise was increased progressively to 20 minutes. The training regimen was then supplemented with interval training, consisting of several 3- to 5-minute exercise bouts requiring 85% to 90% of the subject's .VO2peak, interspersed with 2 to 3 minutes of rest. The exercise intensity was adjusted to the prescribed level using heart rate, measured with radiotelemetry, relative to the subjects' heart rate measured during .VO2peak testing. The duration of endurance training was increased to a maximum of 30 minutes. Shortened programs of Phase 1 and Phase 2 exercises were continued during Phase 3.

Home Exercise Program (Control Group)

The rationale for the home exercise program was to provide the participants with a low-intensity exercise program that would not be expected to improve our primary outcome measures. The home exercise program included nine of the 22 core exercises included in Phase 1 of the supervised exercise program and focused primarily on flexibility.28 Participants in the control group attended a 1-hour training session in our exercise facility. They were asked to perform the exercises at home two to three times per week. To enhance adherence, control participants attended a monthly exercise class at our exercise facility. Participants in the control group performed the exercises for three consecutive 3-month intervals. Follow-up testing was performed at the end of each 3-month interval.

Participants in both groups were provided transportation to our research facility as needed for assessment and training sessions.

Statistical Analysis

Between-group comparisons of continuous variables were performed using t tests, except for a few cases where required conditions were not satisfied, and Wilcoxon test was used as a nonparametric alternative. Chi-square tests were used for between-group comparisons of categorical variables at baseline, except when cell sample sizes in the contingency table were small, in which case Fisher exact test was used. All subjects who provided follow-up data at any time point were included in the analysis. Because of missing data, all longitudinal analyses for variables that were measured at more than two time points were performed using mixed-model repeated-measures analysis of variance. These analyses were performed using the MIXED procedure in SAS (SAS Institute, Inc., Cary, NC). The primary focus of these analyses was on the significance of the interaction between group and test and interactions that tested hypotheses regarding the equality of changes over time in the two groups. Within the framework of the mixed-model analysis, when interactions were significant, the appropriate statistical contrasts were used in testing the null hypothesis that changes between two specific time points in one group were equal to corresponding changes in the other group. For primary outcome measures that yielded interactions with a P-value that was less than .1, adjusted means and standard errors were calculated and used to compute 95% confidence bounds on the magnitude of the between-group difference at 9 months. Because these confidence bounds were adjusted for any between-group differences at earlier time points, they provide a best estimate of the true effect of the intervention. For the outcome measurement that was done only at baseline and at Test 4 in the control group (.VO2peak), longitudinal analyses were performed using analysis of covariance with the final test value as the dependent variable and the baseline value as a covariate. All data are presented as mean ± standard deviation (SD) unless otherwise stated. All investigators involved in data analysis were blinded to treatment assignment.



Four hundred forty-four individuals (152 men; 292 women) underwent the preenrollment evaluations (Figure 1). One hundred sixty-five were excluded from participation; 95 did not meet the selection criteria (79 too fit, 16 too frail), and 70 had medical exclusions, of whom 18 were women undergoing hormone replacement therapy. In addition, 67 women elected to enroll in a concurrent study of exercise combined with hormone replacement therapy, and 93 individuals declined participation.

Figure 1.

Results of recruitment and randomization.

Participant Characteristics

One hundred nineteen individuals were randomized: 69 to ET and 50 to the control group. Four individuals (3 ET; 1 control) were excluded from this analysis because data for the primary outcome measurements were not available for at least one follow-up visit. All of these individuals dropped out within the first 3 months of the study. Therefore, 115 individuals were included in this analysis. The baseline characteristics of the participants are presented in Table 1. The mean age ± SD of participants in both groups was 83 ± 4. Similar to the general community population in this age group, 82% were Caucasian and 12% were African Americans. There were no significant differences in the baseline characteristics of the two intervention groups. Twenty-eight participants (23%; 20 ET, 8 control) dropped out before the final assessment visit. All of these individuals provided follow-up data and were included in the analyses. Reasons for drop-out were (1) medical problems unrelated to study participation (n = 18, 64% of dropouts), (2) personal reasons unrelated to study participation (n = 4, 14%), (3) noncompliance with study procedures (n = 2, 7%), (4) personal reasons related to the study (time commitment, inconvenience, etc.) (n = 2, 7%), (5) medical reasons related to study participation (n = 1, 4%), and (6) death unrelated to study participation (n = 1, 4%). The 28 individuals who dropped out of the study had a lower baseline PPT score (26 ± 6 vs 29 ± 5, P = .01) and a lower baseline .VO2peak (13 ± 2 vs 16 ± 3, P = .0001) than individuals who completed the study.

Table 1.  Baseline Characteristics of the Sample
VariableHome Exercise
(n = 49)
Exercise Training
(n = 66)
  • Note: P-values compare baseline means in the two groups and, unless otherwise indicated, are based on t tests or chi-square tests.

  • *

    P-value based on Wilcoxon test or Fisher exact test.

  • SD = standard deviation.

Age, mean ± SD83 ± 483 ± 4.80
Female, %5352.87
Education, %   
 Not high school graduate2016.87
High school graduate1615 
Some college2725 
College graduate3744 
Ethnic origin, % Caucasian8877.15
Short Blessed test score, mean ± SD3.0 ± 2.92.3 ± 2.4.28*
Geriatric Depression Score, mean ± SD2.0 ± 2.02.4 ± 2.4.45*
Body mass index, kg/m2, mean ± SD26.5 ± 3.826.9 ± 4.5.61
Number of routine medications, mean ± SD4 ± 24 ± 3.96*
History of, %   
Diabetes mellitus129.58
Coronary artery disease2523.78
Heart failure82.16*
Atrial fibrillation1312.99*
Cigarette smoking5342.26
 Joint replacement2312.14
Taking lipid-lowering agents812.76*

Compliance with the Protocol and Adverse Events

Because participants in the ET group were required to complete 36 sessions of each exercise phase, their compliance was 100%. Control participants were instructed to record exercise sessions on a calendar, but their compliance was not monitored rigorously. ET participants completed the intervention in 422 ± 80 days; home participants completed it in 350 ± 65 days (P = .001). One individual in the ET group sustained a rotator cuff injury during resistance training, and a second individual experienced worsening of an existing shoulder problem during resistance training; both dropped out of the study. There were no other adverse events.

Primary Outcomes

ET participants had greater improvements than those in the control group in the modified PPT score over the course of the study, with significant group-by-test comparisons by the end of Phase 1 (Table 2, Figure 2). Although the score on all PPT items improved in the ET group from baseline to the end of the study, the chair-rise score was the only PPT item with a significant group-by-time effect (data not shown). ET participants also had a significant improvement in .VO2peak and FSQ score in comparison with control participants, the latter indicating less difficulty with ADL performance and mobility. These differences persisted after controlling for age, sex, and number of days between baseline and follow-up. Adjusted 95% confidence bounds on the magnitude of the improvement in the ET as compared with the control group between baseline and the final assessment were 1.0 to 5.2 for the modified PPT score, 0.9 to 3.6 mL/kg/min for .VO2peak, and 1.6 to 4.9 for the FSQ score. There were no significant changes in the OARS ADL scales (data not shown), indicating no change in the amount of human assistance or assistive technology.

Table 2.  Primary Outcome Measures
VariableIntervalHome ExerciseExercise TrainingEffect
P-value*Significant Contrasts
(n = 49)
(n = 23)
(n = 26)
(n = 66)
(n = 32)
(n = 34)
Mean ± Standard Deviation
  • Note: Means and standard deviations are reported for each primary outcome at baseline and at each follow-up visit until completion of the intervention. Means were generated using participants with data for at least two time points for the outcome of interest. P-values are based on mixed-model repeated measures analysis of variance.

  • *

    P-values are interpreted from the results of comparisons between specific time points. When the P-value interpreted is from the group-by-test interaction effect, the change between two time points for the two groups is compared.

  • VO2peak = peak oxygen uptake.

 Test score
Test 2
Test 3
Test 4
28.3 ± 5.9
29.0 ± 5.8
28.7 ± 5.5
29.2 ± 5.1
27.5 ± 6.2
27.6 ± 6.4
28.5 ± 5.3
28.6 ± 4.8
29.0 ± 5.8
30.2 ± 5.1
28.9 ± 5.8
29.7 ± 5.3
28.4 ± 4.7
30.3 ± 4.3
31.0 ± 4.4
31.8 ± 3.9
28.9 ± 4.0
30.8 ± 3.8
31.2 ± 3.5
32.0 ± 4.1
27.9 ± 5.3
29.7 ± 4.7
30.8 ± 5.3
31.7 ± 3.8
Group* Test.02Baseline to Test 2; P = .04
Baseline to Test 3; P = .02
Baseline to Test 4; P = .002
Test 3 to Test 4; P =.08
Test 3
Test 4
15.6 ± 2.6

15.2 ± 3.3
15.8 ± 2.1

15.1 ± 4.0
15.4 ± 2.9

15.2 ± 2.9
15.4 ± 2.9
16.2 ± 3.4
17.4 ± 3.4
16.2 ± 2.9
17.0 ± 3.5
18.5 ± 3.2
14.4 ± 2.6
15.2 ± 3.1
16.3 ± 3.3
Group* Test.0001
Test 2
Test 3
Test 4
26.6 ± 4.5
28.3 ± 3.7
27.8 ± 4.2
27.0 ± 4.2
27.4 ± 4.9
28.0 ± 4.5
27.5 ± 5.1
25.9 ± 4.2
25.9 ± 4.0
28.5 ± 2.9
28.0 ± 3.5
27.8 ± 4.0
26.6 ± 4.2
28.6 ± 3.5
30.0 ± 3.3
30.4 ± 4.0
28.0 ± 3.1
29.3 ± 3.1
31.0 ± 2.5
31.3 ± 2.7
25.2 ± 4.7
27.9 ± 3.6
28.9 ± 3.8
29.3 ± 5.0
Group* Test.01Baseline to Test 3; P = .01
Baseline to Test 4; P = .002
Test 2 to Test 3; P = .01
Figure 2.

Changes in total modified Physical Performance Score (PPT) from baseline to end of study. Values are means ± standard deviation. Significantly different from baseline, *P < .05; P < .01.

Secondary Outcome Measures

Baseline values for knee extension torque and knee flexion torque were significantly higher for men than women. The effects of the exercise programs on selected secondary outcome measurements are shown in Table 3. ET participants had significant improvements in maximum voluntary knee extensor and knee flexor torque and one-leg stance time, and these changes were significantly greater than the changes for control participants. Performance on the Berg Balance Test also improved to more in the ET group than in the control group. Responses on the Change in Health subscale of the SF-36 improved more in the ET group (baseline score = 39 ± 16, Test 4 score = 53 ± 17) than in the control group (baseline score = 40 ± 17, Test 4 score = 39 ± 19; group-by-test effect = 0.01). These group-by-test differences persisted after controlling for age, sex, and number of days between baseline and follow-up. The observed group-by-test differences also persisted after adjustment for the change in score for the SF-36 Social Functioning and Mental Health subscales. There was no significant group-by-test effect for changes in body weight (data not shown).

Table 3.  Measurements of Strength, Balance, Body Composition
VariableIntervalHome ExerciseExercise TrainingEffect
(n = 49)
(n = 23)
(n = 26)
(n = 66)
(n = 32)
(n = 34)
Mean ± Standard DeviationSignificant Contrasts
  • Note: Means and standard deviations are reported for each primary outcome at baseline and at each follow-up visit until completion of the intervention. Means were generated using participants with data for at least two time points for the outcome of interest. P-values are based on mixed-model repeated measures analysis of variance.

  • *

    P-values are interpreted from the results of comparisons between specific time points. When the P-value interpreted is from the group-by-test interaction effect, the change between two time points for the two groups is compared.

Cybex knee
 extension 60°,
Test 2
Test 3
Test 4
58.9 ± 18.1
56.8 ± 19.1
59.0 ± 16.7
55.7 ± 19.1
66.4 ± 17.3
65.9 ± 17.5
64.9 ± 14.7
67.0 ± 15.0
52.6 ± 16.6
49.3 ± 17.2
54.1 ± 17.1
48.3 ± 18.0
59.8 ± 20.7
63.2 ± 18.7
71.2 ± 21.0
70.9 ± 20.5
71.3 ± 19.4
74.5 ± 17.1
83.4 ± 20.5
85.1 ± 13.8
49.1 ± 15.7
52.7 ± 13.3
58.6 ± 12.3
56.1 ± 15.1
Group* Test.004Baseline to Test 2; P = .01
Baseline to Test 3; P = .0005
Baseline to Test 4; P = .002
Test 2 to Test 3; P = .03
Cybex knee
 flexion 60°,
Test 2
Test 3
Test 4
41.2 ± 15.3
40.3 ± 14.1
43.3 ± 13.4
43.1 ± 14.0
46.1 ± 16.6
45.1 ± 15.6
49.6 ± 14.1
52.3 ± 12.6
37.1 ± 13.0
36.4 ± 11.5
38.1 ± 10.5
37.1 ± 11.6
40.5 ± 14.3
44.2 ± 13.0
48.6 ± 14.7
49.9 ± 13.9
47.2 ± 14.6
51.6 ± 12.9
57.9 ± 12.1
58.4 ± 12.5
34.3 ± 10.9
37.4 ± 8.7
39.0 ± 10.5
40.9 ± 8.8
Group* Test.02Baseline to Test 2; P = .004
Baseline to Test 3; P = .005
Baseline to Test 4; P = .01
Single limb
 stance time, sec
Test 2
Test 3
Test 4
4.6 ± 5.0
 5.8 ± 6.2
 4.4 ± 4.7
 5.4 ± 4.6
4.0 ± 4.6
 5.9 ± 7.2
 4.5 ± 5.6
 4.2 ± 3.7
5.1 ± 5.4
 5.6 ± 5.4
 4.2 ± 4.0
 6.2 ± 5.0
3.6 ± 3.2
 6.5 ± 6.2
 6.8 ± 5.9
 6.8 ± 6.5
3.3 ± 2.4
 7.2 ± 6.5
 7.2 ± 6.0
 5.9 ± 4.5
4.0 ± 3.9
 5.9 ± 5.9
 6.5 ± 6.0
 7.7 ± 8.0
Group* Test.05Baseline to Test 2; P = .05
Baseline to Test 3; P = .005
Baseline to Test 4; P = .07
Berg Balance
Test 2
Test 3
Test 4
49.8 ± 4.5
50.0 ± 5.9
50.5 ± 5.5
50.9 ± 4.1
49.2 ± 5.1
48.3 ± 7.4
50.4 ± 6.4
50.9 ± 3.0
50.2 ± 4.1
51.4 ± 3.7
50.6 ± 4.8
50.8 ± 4.7
49.8 ± 3.8
51.2 ± 4.4
51.9 ± 4.4
52.5 ± 3.8
49.7 ± 3.6
51.4 ± 4.8
51.6 ± 4.5
52.5 ± 3.7
49.8 ± 4.1
51.0 ± 4.1
52.2 ± 4.4
52.6 ± 4.0
Group* Test.06Baseline to Test 2; P = .02
Baseline to Test 3; P = .08
Baseline to Test 4; P = .02


The results of this study extend and broaden the evidence supporting the efficacy of ET for enhancing functional capacity in frail older men and women. To the best of our knowledge, this is the first study to provide evidence of the efficacy of a multidimensional high-intensity training regimen for reducing frailty and improving oxygen uptake in frail community-dwelling older people. Improvements in skeletal muscle strength, balance, and perceived health status accompanied the observed improvements in physical function, as reflected in the FSQ score, PPT score, and oxygen uptake.

The ET program was designed with a 3-month period of physical therapy–type of stretching, flexibility, balance, coordination, and mild strengthening exercises in an effort to prepare the participants for, and protect against injury from, the second, more-strenuous, weightlifting component. The sequence of the ET components was also designed so that the increases in muscle strength induced by weightlifting would help prepare the participants for, and enable them to perform, the endurance training more effectively. The design of the exercise-training program may be responsible for the remarkably small number of exercise-induced injuries, which is particularly impressive in light of the age and comorbid conditions in the participants. Further study is necessary to determine whether a shorter period of physical therapy–type of exercises can be used before weight training, with a similar safety profile and effects on function.

The relative increases in .VO2peak of 14% for the men and 13% for the women are of similar magnitude to increases in .VO2peak that have been reported for younger, healthier individuals in response to endurance training of a similar duration.29 This is the first study to demonstrate improved aerobic capacity in this population and indicates that mildly to moderately frail older people retain the ability to adapt to endurance training with functionally significant increases in .VO2peak. This adaptive response could be expected to translate into improved function as reflected in greater endurance (development of less fatigue during activities such as shopping and housework). The increases in strength induced by the weightlifting program confirm the results of previous studies showing that frail older people retain the capacity to adapt to weightlifting with significant increases in strength.7 Our results extend these previous observations by showing that frail older people are able to perform and adapt to whole-body programs of strength training involving a variety of exercises and all major muscle groups. However, the magnitude of the increases in strength induced by this program were considerably smaller than have been reported in studies in which frail older individuals exercised only one or two major muscle groups.7 The reason for this may be the low baseline level of maximum voluntary muscle strength observed in institutionalized frail older people.7

This study has several limitations. Participants in the ET group had greater social contact than individuals in the control group. It is possible that the increased level of socialization enhanced the motivation of ET participants to a greater extent than that of control participants and may account for some of the improvements observed, particularly the SF-36 measure. It is unlikely that differences in socialization account for the changes observed in the modified PPT, .VO2peak, strength, and balance, and this is supported by analyses of covariance that included the changes in SF-36 subscale scores for Mental Health and Social Functioning. Individuals who dropped out of the study had lower PPT scores and .VO2peak, and therefore our results may not generalize to individuals with more-severe impairments. However, the overall dropout rate (27%) is comparable with that of previous randomized trials of ET for older people.4 Men and women on hormone replacement therapy were excluded from the study, and therefore our results may not generalize to that population. Lastly, because transportation services were provided for individuals unable to get to the exercise facility, the sample may not reflect the percentage of eligible individuals who would be able to participate in this type of ET program without such support.

In summary, our findings indicate that supervised high-intensity ET is a safe and effective intervention in older physically frail community-dwelling adults that reduces physical impairments and improves functional limitations more than unsupervised low-intensity home exercise. ET appears to improve measures of “preclinical ADL disability,”30 and, thus, a major benefit of an intensive multidimensional exercise program appears to be the prevention or postponement of frailty that is severe enough to cause loss of independence. Future studies will need to evaluate the long-term effects of this type of ET and its utility in individuals with more-severe levels of frailty.


The authors wish to acknowledge the excellent technical assistance of staff for the WU Claude Pepper Older Americans Independence Center and the WU General Clinical Research Center.

This work was supported by NIH Claude Pepper Older Americans Independence Center Award Grant P01-AG13629 and NIH General Clinical Research Center Grant 5-M01 RR00036. It was presented in part at the 48th annual scientific meeting of the American College of Sports Medicine, Baltimore, Maryland, May 2001.