G Model

JPHYS-391; No. of Pages 12

Journal of Physiotherapy xxx (2018) xxx–xxx

Journal of
PHYSIOTHERAPY
journal homepage: www.elsevier.com/locate/jphys

Research

Physical exercise improves strength, balance, mobility, and endurance in people

with cognitive impairment and dementia: a systematic review
Freddy MH Lam a, Mei-Zhen Huang a, Lin-Rong Liao b, Raymond CK Chung a, Timothy CY Kwok c,
Marco YC Pang a
a
Department of Rehabilitation Sciences, The Hong Kong Polytechnic University, Hong Kong; b Department of Rehabilitation, Jiangsu Provincial Yixing Jiuru Rehabilitation Hospital,
Yixing, China; c Department of Medicine and Therapeutics, Faculty of Medicine, The Chinese University of Hong Kong, Hong Kong

K E Y W O R D S A B S T R A C T

Dementia Question: Does physical exercise training improve physical function and quality of life in people with
Exercise cognitive impairment and dementia? Which training protocols improve physical function and quality of
Mild cognitive impairment life? How do cognitive impairment and other patient characteristics influence the outcomes of exercise
Physical fitness
training? Design: Systematic review with meta-analysis of randomised trials. Participants: People with
Quality of life
mild cognitive impairment or dementia as the primary diagnosis. Intervention: Physical exercise.
Outcome measures: Strength, flexibility, gait, balance, mobility, walking endurance, dual-task ability,
activities of daily living, quality of life, and falls. Results: Forty-three clinical trials (n = 3988) were
included. According to the Grades of Recommendation, Assessment, Development and Evaluation
(GRADE) system, the meta-analyses revealed strong evidence in support of using supervised exercise
training to improve the results of 30-second sit-to-stand test (MD 2.1 repetitions, 95% CI 0.3 to 3.9), step
length (MD 5 cm, 95% CI 2 to 8), Berg Balance Scale (MD 3.6 points, 95% CI 0.3 to 7.0), functional reach
(3.9 cm, 95% CI 2.2 to 5.5), Timed Up and Go test (–1 second, 95% CI –2 to 0), walking speed (0.13 m/s, 95%
CI 0.03 to 0.24), and 6-minute walk test (50 m, 95% CI 18 to 81) in individuals with mild cognitive
impairment or dementia. Weak evidence supported the use of exercise in improving flexibility and
Barthel Index performance. Weak evidence suggested that non-specific exercise did not improve
dual-tasking ability or activity level. Strong evidence indicated that exercise did not improve quality of
life in this population. The effect of exercise on falls remained inconclusive. Poorer physical function was
a determinant of better response to exercise training, but cognitive performance did not have an impact.
Conclusion: People with various levels of cognitive impairment can benefit from supervised
multi-modal exercise for about 60 minutes a day, 2 to 3 days a week to improve physical function.
[Lam FMH , Huang MZ, Liao LR, Chung RCK, Kwok TCY, Pang MYC (2018) Physical exercise improves
strength, balance, mobility, and endurance in people with cognitive impairment and dementia: a
systematic review. Journal of Physiotherapy XX: XX–XX]
© 2017 Australian Physiotherapy Association. Published by Elsevier B.V. This is an open access article
under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Introduction with dementia, but the heterogeneous participant groups and

different outcome measures that were used made conducting and
Dementia is an increasingly important public health concern.1 It interpreting meta-analyses difficult.13–21 Meta-analyses were
is estimated that by 2050 the number of people with dementia will conducted in seven reviews to quantify the amount of improve-
reach 131.5 million worldwide.1 Apart from deficits in cognition ment gained after exercise training.14–16,20,22–24 However,
and behaviour,2 deficits in balance, gait, and movement coordina- non-randomised trials were included in some reviews, which
tion are also found in people with mild dementia and mild compromised the quality of evidence.16,23 Other reviews focused
cognitive impairment.3,4 People with dementia are less likely to on one type of exercise training, one patient subgroup,14,24 or few
participate in regular physical exercise when compared with their domains of physical function.15,20,22,23 None of the existing
counterparts with normal cognition.5 Physical inactivity may give systematic reviews conducted sensitivity analysis to specifically
rise to further decline in physical functioning.6 These factors may examine the effect of subject characteristics (eg, cognitive
partially explain the higher risk of falls and hip fractures in people impairment level) on training efficacy – probably due to the small
with dementia compared with their peers without dementia.7,8 number of trials included in the reviews. Thus, the existing reviews
Exercise training improves cognitive9 and physical10–12 func- have not provided a comprehensive understanding of the effect of
tions in healthy older adults and is feasible for people with physical exercise on physical function in people with cognitive
cognitive impairment.13,14 Previous reviews have attempted to impairment. Moreover, a good number of new exercise trials on
examine the effects of exercise on physical function in individuals people with mild cognitive impairment or dementia have been

https://doi.org/10.1016/j.jphys.2017.12.001
1836-9553/© 2017 Australian Physiotherapy Association. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.
org/licenses/by-nc-nd/4.0/).

Please cite this article in press as: Lam FMH, et al. Physical exercise improves strength, balance, mobility, and endurance in people with
cognitive impairment and dementia: a systematic review. J Physiother. (2018), https://doi.org/10.1016/j.jphys.2017.12.001
G Model
JPHYS-391; No. of Pages 12

2 Lam et al: Exercise in cognitive impairment and dementia

published in the last few years, and it is thus timely to conduct a review were based on the information obtained from the PEDro
systematic review on this topic to address the knowledge gaps website, where studies are rated in duplicate by trained raters.
identified above.
Therefore, the research questions for this systematic review Participants
were: To describe the participants in each trial, the following
information was extracted from the published report: sample
1. Does physical exercise training improve physical function and size, mean age, gender ratio, location of participants (community,
quality of life in people with cognitive impairment and institution), diagnosis, and cognitive impairment test scores
dementia? (eg, Mini Mental State Examination).
2. Which training protocols improve physical function and quality
of life? Intervention
3. How do cognitive impairment and other patient characteristics The details extracted from each included study about the
influence the outcomes of exercise training? exercise intervention were: frequency, intensity, duration and type
of physical exercise.

Method Outcome measures

Outcome data were extracted from each included study if they
Identification and selection of trials pertained to any domain of physical function or quality of life.
Corresponding authors were contacted via email in case
MEDLINE, CINAHL, PubMed, PsycINFO, and The Cochrane information needed for the meta-analysis could not be acquired
Library Databases of Systematic Reviews were searched electroni- from the original articles. When there were discrepancies
cally with search terms related to cognitive impairment, dementia, between the two researchers responsible for data extraction,
exercise, rehabilitation, and randomised trial. An example of the the information extracted was confirmed by the principal
search strategy for one database is provided in Appendix 1 on the investigator.
eAddenda. Two independent researchers screened the search
results for publications about the effect of physical exercise in Data analysis
people with mild cognitive impairment or dementia. Potentially
eligible trials were selected for further assessment of eligibility. Meta-analysis was performed for a given outcome only if at
Relevant reviews and the reference lists of all selected articles were least three similar trials used the same outcome measure. Meta-
then examined to look for potentially eligible articles. Finally, a analyses were conducted using RevMan software.a Random-effect
forward search was performed on all articles selected in the above models were used in all meta-analyses, given the large variation in
process using the Science Citation Index. The last search was study design across trials (eg, participants’ characteristics, exercise
performed in May 2016. The inclusion criteria for trials to be protocols).25 The existence of publication bias was examined
included in the review are presented in Box 1. However, trials were using Egger’s regression asymmetry test using Comprehensive
excluded if they were published only in conference proceedings or Meta-analysis software.b A p-value of <0.1 (two-tailed test)
books. Disagreements about eligibility were resolved by the
principal researcher.
Box 2. Criteria used to downgrade ratings in the Grades of
Assessment of characteristics of trials Recommendation, Assessment, Development and Evaluation
(GRADE) system. See the Methods section for further details.
Quality
Risk of bias
The PEDro score, obtained by searching the PEDro website
 for outcomes where meta-analysis was possible, fewer
(www.pedro.org.au), was used to assess the methodological than half of the trials included in the primary analysis had
quality of each selected trial. For trials that were not originally a PEDro score of 6
listed on the PEDro website, the PEDro team was contacted via  for outcomes where meta-analysis was not possible,
email to request them to examine these trials and provide the fewer than half of the trials included for outcome
PEDro scores. Hence, the PEDro scores of all trials included in this evaluation had a PEDro score of 6

Inconsistency
 for outcomes where meta-analysis was possible, I2  50%
Box 1. Inclusion criteria. in the primary meta-analysis and the meta-analysis that
involved only trials with high methodological quality
Design  for outcomes where meta-analysis was not possible,
 Randomised trial mixed results were reported
 English language
Indirectness
Participants  the participants, intervention, comparator intervention,
 People with a primary diagnosis of mild cognitive outcome measure or study design did not match between
impairment or dementia the included studies and the eligibility criteria for this
review
Intervention
 Physical exercise Imprecision
 insufficient studies for meta-analysis
Outcome measures  the number of subjects included in the primary meta-
 Measures of physical function analysis was less than that required by a conventional
 Measures of quality of life sample size calculation for a single trial
 the 95% CI spanned zero
Comparisons
 Exercise versus no intervention/placebo Publication bias
 Exercise plus other intervention versus other intervention  p < 0.1 on the two-tailed Egger’s regression asymmetry
only test

Please cite this article in press as: Lam FMH, et al. Physical exercise improves strength, balance, mobility, and endurance in people with
cognitive impairment and dementia: a systematic review. J Physiother. (2018), https://doi.org/10.1016/j.jphys.2017.12.001
G Model
JPHYS-391; No. of Pages 12

Research 3

Records identified through electronic database

search (n = 9635)

Duplicates removed (n = 6522)

Records screened by title and abstract (n = 3113)

Records excluded after screening titles and abstracts (n = 2969)

Potentially relevant articles retrieved for evaluation

of full text (n = 144)

Articles excluded after evaluation of full text (n = 106)

• the effect of physical exercise could not be isolated (n = 38)
• not a randomised trial (n = 30)
• participants did not have cognitive impairment (n = 15)
• no target outcome measures were assessed (n = 15)
• not a journal article (n = 8)

Eligible articles identified (n = 38)

Eligible studies identified (n = 32)

Forward citation tracking

Eligible papers identified (n = 16)
Eligible studies identified (n = 11)

Eligible articles included in review (n = 54)

Eligible studies included in review (n = 43)

Figure 1. Flow of studies through the review.

indicated the presence of publication bias. The mean difference Characteristics of included trials
provided a summary measure of the effect of exercise. Sensitivity
analyses were conducted based on different patient subgroups, The methodological quality of the included studies is sum-
outcome mono-dimensionality, and methodological quality, if marised in Table 1. A more detailed version of this table is available
three or more trials remained eligible for the additional analyses. in Appendix 2 on the eAddenda. The characteristics of their study
The level of evidence for each outcome measure was then populations are summarised in Table 2. The physical exercise
calculated according to the Grades of Recommendation, Assess- protocols assessed in the included studies are summarised in
ment, Development and Evaluation (GRADE) system.26 The rating Table 3. More details about the study populations, the intervention
for each level of evidence started with ‘high quality’, as only protocols, and the outcome measures are available in Appendix
randomised, controlled trials were included in this review. It was 3 on the eAddenda.
downgraded by one level for each of the criteria shown in Box 2.
The quality of evidence was downgraded by two levels if less than Quality
half of the trials included in the primary meta-analysis had a PEDro The quality of evidence assessed according to the GRADE is
score of 4. For outcomes for which meta-analysis was not provided in Table 4. The results were organised according to the
possible, the same criteria were adopted with reference to the total three levels of functioning described in the International
number of trials reviewed for that particular outcome. Evidence Classification of Functioning, Disability, and Health (ICF), namely:
quality was upgraded by one level if the effect size was large. After body functions and structures, activities, and participation.81
considering the balance of desirable and undesirable aspects of the Outcomes that were reported by less than three trials are listed in
available evidence, a strong or weak recommendation would be Appendix 4 on the eAddenda.
given according to existing evidence.26

Effect of exercise on body functions and structures

Results Body mass index (moderate-quality evidence)

Body mass index (BMI) was assessed in three trials (173 parti-
Flow of trials through the review cipants).30,35,50 Multimodal exercise30,50 or specific aerobic train-

The database searches identified 6935 records. After screening Table 1

Methodological quality of the included trials (n = 43).
titles and abstracts, 144 potentially relevant articles were obtained
in full text and evaluated for eligibility. The 38 articles that were PEDro total score Trials References
eligible for the review were supplemented by 16 more identified n (%)
through the reference lists searches. Therefore, 54 articles were excellent (9 to 10) 0 (0)
included in the review.27–80 These reported on 43 trials involving a good (6 to 8) 30 (70) 27,28,30–35,37–42,44,45,47,50–53,57,60,61,63–69,71–80

29,36,38,46,48,49,54–56,59,62,70
total of 3988 subjects (Figure 1). Three papers were excluded from fair (4 to 5) 11 (26)
43,58
poor (0 to 3) 2 (5)
some of the meta-analyses, as the required information was
missing despite efforts to contact the original authors.36,48,58 Percentages do not sum to 100, due to the effects of rounding.

Please cite this article in press as: Lam FMH, et al. Physical exercise improves strength, balance, mobility, and endurance in people with
cognitive impairment and dementia: a systematic review. J Physiother. (2018), https://doi.org/10.1016/j.jphys.2017.12.001
G Model
JPHYS-391; No. of Pages 12

4 Lam et al: Exercise in cognitive impairment and dementia

Table 2
Characteristics of the participants in the included trials (n = 43).

Characteristic Trials References

n (%)

Sample size
34,35,44,46–50,52,53,55,56,58,59,61,62,79,80
50 18 (42)
27–33,36–43,45,51,54,57,60,63–78
>50 25 (58)
Cognitive impairment level
27–40,74,79
mild cognitive impairment (mean MMSE > 24) 11 (26)
41,42,45–53,63,64,80
mild dementia (mean MMSE 20 to 24) 11 (26)
43,44,54–62,65,71–73,75–78
moderate dementia (mean MMSE 10 to 20) 16 (37)
66–70
severe dementia (mean MMSE < 10) 5 (12)
Setting of the trial
29,43,44,47,49,51,54,56,59,61,65,66,68–71,75–79
residential care units 20 (47)
27,28,30–40,45,46,48,50,52,53,55,60,62,72–74,80
community 19 (44)
67
hospital respite care 1 (2)
41,42,57,58,63,64
mixed 3 (7)

Percentages may not sum to 100, due to the effects of rounding. MMSE = Mini Mental State Examination.

ing35 (45 to 90 minutes per session, 2 to 7 sessions per week for 4 to training with a resistance exercise component (30 to 120 minutes
12 months) were compared to usual care,50 health promotion,30 or per session, 2 to 4 sessions per week, for 9 weeks to
low-intensity stretching and balance exercise.35 All three trials 4 months).41,49,50,71 The intensity of the resistance training was
reported no significant change in BMI after physical exercise not specified. The training effect appeared to diminish starting at
relative to the control intervention.30,35,50 9 weeks and 3 months after training.42,71
Upper limb strength was reported in four trials.39–42,49 Three
of these trials assessed handgrip strength and one used the
Strength (moderate-quality evidence) arm-curl test. The one trial that incorporated resistance training
Meta-analysis of four trials (278 participants) showed a reported results in favour of the exercise group.49 The other three
significant effect of exercise on improving 30-second sit-to-stand trials, which all lacked specific upper limb resistance training,
performance by 2.1 repetitions (95% CI 0.3 to 3.9).48,49,71,77 See reported no significant effects.39–42
Figure 2, or for a more detailed forest plot see Figure 3 on the
eAddenda. The result remained significant in sensitivity analyses
that included only those trials that involved participants in Flexibility (very-low-quality evidence)
institutionalised settings (three trials, 258 participants). The Flexibility was assessed in three trials.49,58,62 All trials
heterogeneity across trials was high in both the primary and incorporated multimodal exercise with a flexibility training
sensitivity analysis, with I2 values of 82% and 88%, respectively. component (2 to 3 sessions per week, 16 weeks to 12 months),
Together with four other trials that assessed other types of sit-to- and reported significant improvements in the intervention group.
stand performance (eg, five times sit-to-stand test),41,50,52,55 all Of these, two reported that the improvements in the intervention
trials that reported significant training effects adopted multimodal group were significantly greater than no-intervention controls

Table 3
Characteristics of the physical exercise protocols used in the included trials (n = 43).

Characteristic Trials References

n (%)

Mode of exercise a
34–38,43,45,47,48,61,71,75
aerobic exercise 11 (23)
39,65,70
walking exercise 3 (6)
65,69
dual-task walking 2 (4)
27,28,30,31,40–42,46,49–56,58,60,62–64,66,70–74,76–78
multimodal 21 (45)
57,68
ADL/functional training 2 (4)
36,79
strengthening exercise 2 (4)
29,32,33,44,59,67,80
others 6 (13)
Duration of session (min)
34,59,75
<30 3 (7)
29,47,48,53,61,65,67,69,71,80
30 to 40 10 (23)
27,36–38,45,54–58,60,66,72–74,76–79
45 to 60 15 (35)
28,30,31,39–42,49,63,64,68
75 to 150 5 (12)
32,33,35,43,44,62,70
variable (15 to 60) 6 (14)
46,50–52
unclear 4 (9)
b
Frequency (sessions/week)
28,30,31,34,36–42,44,45,48,53,55,57,58,60,63,64,66,72,73,76–78
1 to 2 17 (39)
27,35,43,47,49,51,54,56,59,61,65,71,74,79,80
3 to 4 15 (34)
29,50,52,54,68–70,75
5 to 7 8 (18)
32,33,46,62,67
variable or unclear 4 (9)
Duration c
34
single session 1 (2)
39,41–47,49,55–57,59,63,64,71,76,77,79,80
8 to 15 weeks 16 (36)
27–29,31,32,35,36,40,48,50,52–54,58,61,65,68–70,78
16 to 24 weeks 18 (40)
30,33,37,38,51,60,62,66,72–75
12 months 9 (20)
67
unclear 1 (2)

Percentages may not sum to 100, due to the effects of rounding. ADL = activities of daily living.
a
n = 47. Four of the trials had two exercise groups in addition to the control group: Nagamatsu36 had aerobic and strengthening groups; Roach70 had walking and
multimodal groups; Tappen65 had walking and dual-task walking groups; and Bosser71 had aerobic and multimodal groups.
b
n = 44. Apart from the control group, Christofoletti54 had two exercise groups, which had different numbers of exercise sessions per week: 3 and 5.
c
n = 45. Two trials used interim evaluation: in the 12-month trial by Uemura,30 a 6-month interim evaluation was reported by Doi,28 Makizako40 and Suzuki.31 Similarly, in
the 12-month trial by Lam,32 a 6-month interim evaluation was also reported.33

Please cite this article in press as: Lam FMH, et al. Physical exercise improves strength, balance, mobility, and endurance in people with
cognitive impairment and dementia: a systematic review. J Physiother. (2018), https://doi.org/10.1016/j.jphys.2017.12.001
G Model
JPHYS-391; No. of Pages 12

Research 5

Table 4
Grades of Recommendation, Assessment, Development and Evaluation (GRADE) quality of evidence.

Outcome Risk of Inconsistency Indirectness Imprecision Publication Effect GRADE Direction of

bias bias size quality recommendation

Body mass index 0 0 0 –1f 0 0 Moderate Against

Strength: 30-s sit to stand test 0 –1b 0 0 0 0 Moderate For
Flexibility –2a 0 0 –1f 0 0 Very low For
Step length 0 0 0 0 0 0 High For
Balance: Berg Balance Scale 0 0 0 0 0 0 High For
Balance: Functional reach test 0 0 0 0 0 0 High For
Mobility: Timed Up and Go test 0 0 0 –1g 0 0 Moderate For
Mobility: Walking speed 0 –1b 0 0 0 0 Moderate For
Walking endurance: 6-minute walk test 0 –1b 0 0 –1h +1i Moderate For
Dual-task ability 0 –1c 0 –1f 0 0 Low Against
Activities of daily living: Barthel index 0 –1b 0 0 –1h 0 Low For
Activity level 0 –1c 0 –1f 0 0 Low Against
Quality of life 0 0d 0 –1f 0 0 Moderate Against
Falls 0 0 –1e –1g 0 0 Low No recommendation
a
More than half of the trials included for outcome evaluation had a PEDro score  4.
b
I2  50% in the primary and high methodological quality analysis.
c
Mixed result reported across trial and meta-analysis was not possible.
d
Quality of life is not rated down here, considering that all studies involving multi-modal exercise reported non-significant effect.
e
The method of collecting falls outcome by the trials may undermine the actual effect of exercise.
f
Insufficient studies for meta-analysis.
g
The effect size overlapped zero in the primary or high methodological quality analysis.
h
Publication bias present.
i
Large effect size detected.

as assessed by the chair sit-and-reach test49,58 and back scratch plot see Figure 9 on the eAddenda. Similar results were obtained in
test.49 Between-group analysis was not conducted in one of these sensitivity analyses that included only those trials with: high
trials.62 methodological quality (four trials, 199 participants), participants
with mild cognitive impairment and mild-grade dementia
Step length (high-quality evidence) participants (five trials, 219 participants), or community settings
Meta-analysis of five trials (296 participants) revealed a (five trials, 197 participants). The heterogeneity across trials in all
significant effect of exercise on improving step length by 5 cm analyses was minimal with I2 ranging from 0 to 1%.
(95% CI 2 to 8).28,41,56,64 See Figure 4, or for a more detailed forest Together with other trials that used clinical balance assessment
plot see Figure 5 on the eAddenda. The heterogeneity across trials scales as outcomes (eg, Tinetti balance assessments), it was found
was low (I2 = 29%). Sensitivity analyses that included only trials that various types of exercise interventions were effective in
with high methodological quality (four trials, 265 participants), or improving balance, including multimodal exercise, Tai Chi,
individuals with mild cognitive impairment and mild-grade resistance and functional training, and walking (20 to 120 minutes
dementia (three trials, 183 participants) remained significant per session, at least 2 sessions per week, for a total of 9 weeks to
and the I2 was reduced to 0%. Those trials that reported positive 12 months).32,33,41,45,48–52,54,58,59,71,77,78 Four trials with follow-up
outcomes adopted multimodal exercise with walking, aerobic assessments suggested that the training effect was found to
training, or functional training components, or specific aerobic diminish at 9 weeks,71 3 months,41,42 or 6 months,45,76 although the
training (60 to 120 minutes per session, 2 to 3 sessions per week for outcomes remained significantly better than controls in three of
15 weeks to 6 months).28,41,56,64 One of the three trials performed the trials.41,42,45,76
long-term follow-up assessments and reported that the improve-
ment was diminished and returned to baseline values 9 months Mobility (moderate-quality evidence)
after the intervention.41,42 Meta-analysis (11 trials, 606 participants) showed that exercise
significantly reduced the time required to complete the Timed Up
Effect of exercise on activity and Go (TUG) test by 1 second (95% CI –2 to 0).41,47,50–55,71,75,79
See Figure 10, or for a more detailed forest plot see Figure 11
Balance (high-quality evidence)
The primary analysis of the effect of exercise on the Berg
Balance Scale (six trials, 722 participants) revealed that exercise MD (95% Cl)
Study Random
significantly improved scores by 3.6 points (95% CI 0.3 to
7.0).33,45,54,59,77,78 See Figure 6, or for a more detailed forest plot Arcoverde 2014
see Figure 7 on the eAddenda. Publication bias was noted for this
Bossers 2015
meta-analysis. A sensitivity analysis that included only trials with
high methodological quality (four trials, 673 participants) marked- Santana-Sosa 2008
ly reduced the heterogeneity across trials, with I2 value dropping
Telenius 2015
from 91 to 17%. There was no publication bias in this analysis but
the mean improvement found was reduced to 1.5 points (95% CI Pooled
0.1 to 3.0). Another sensitivity analysis that included only studies
Institutionalised only
involving participants with moderate-grade dementia and living in
institutions (four trials, 376 participants) also yielded significant
results in favour of the exercise group, but the I2 value was high –4 –2 0 2 4 6
(80%).
The primary analysis of the effect of exercise on the functional Favours control (repetitions) Favours excercise
reach test (six trials, 242 participants) showed that exercise Figure 2. Weighted mean difference (95% CI) in 30-second sit-to-stand test
significantly improved the reaching distance by 3.9 cm (95% CI performance (number of repetitions) due to exercise, pooling data from four studies
2.2 to 5.5).45,48,50,52,55,79 See Figure 8, or for a more detailed forest (278 participants).

Please cite this article in press as: Lam FMH, et al. Physical exercise improves strength, balance, mobility, and endurance in people with
cognitive impairment and dementia: a systematic review. J Physiother. (2018), https://doi.org/10.1016/j.jphys.2017.12.001
G Model
JPHYS-391; No. of Pages 12

6 Lam et al: Exercise in cognitive impairment and dementia

MD (95% Cl) 537 participants) yielded a significant but slightly smaller

Study Random improvement of 0.08 m/s (95% CI 0.01 to 0.15). Improvement in
Bossers 2015 walking speed among individuals with mild cognitive impairment
and mild-grade dementia (three trials, 183 participants) was
Doi 2013
not significant (p = 0.19), whereas that in those with moderate-
Hauer 2012 to-severe-grade dementia (four trials, 385 participants) was
marginally significant (mean difference 0.14 m/s, 95% CI –0.01 to
Kemoun 2010
0.29). It should be noted that the trials were highly heterogeneous
Suttanon 2013 in all analyses (I2  74%). The trials that reported significant
training effects all adopted multimodal training, with a balance,
Pooled
walking, or functional training component (45 to 90 minutes per
High methodological quality only day, 2 to 3 days per week for a minimum of 15 weeks).28,40,41,56,58,64
MCI and mild grade dementia only
Walking endurance (moderate-quality evidence)
Meta-analysis (seven trials, 402 participants) revealed that
–0.2 –0.1 0 0.1 0.2
exercise significantly increased the distance covered in the
Favours control (m) Favours excercise 6-minute walk test by 50 m (95% CI 18 to 81).30,45,61,62,65,70,71
See Figure 14, or for a more detailed forest plot see Figure 15 on the
Figure 4. Weighted mean difference (95% CI) in step length (m) due to exercise, eAddenda. Publication bias was noted for this meta-analysis. The
pooling data from five studies (296 participants). MCI = mild cognitive impairment.
results were similar and remained significant in sensitivity
analyses that included only trials with: high methodological
quality (five trials, 318 participants), participants with moderate-
to-severe-grade dementia (five trials, 229 participants), commu-
on the eAddenda. Upon examination of trials that reported
nity settings (three trials, 203 participants), or aerobic and walking
significant benefits,39,41,48–50,58,75,79 most adopted multimodal
exercises only (five trials, 283 participants). The effect of exercise
exercises41,49,50,58 (15 to 120 minutes per day, at least 2 days/week
on the 6-minute walk distance among patients in an institutio-
for a minimum of 12 weeks). The result became marginally
nalised setting (four trials, 199 participants) did not reach
significant when only high methodological quality trials (nine
statistical significance (p = 0.10). It was found that aerobic exercise
trials, 553 participants) were included (p = 0.08). The improvement
(three trials, 187 participants) led to the largest improvement in
found was also reduced to –1 seconds (95% CI –1 to 0). The group
6-minute-walk distance of 75 m (95% CI 25 to 125). The
with moderate-grade dementia (four trials, 249 participants) had a
heterogeneity across trials was high in all analyses (I2  56%).
tendency to benefit more from training with a marginally
Trials that reported positive findings adopted either specific
significant improvement of –2 seconds (95% CI –5 to 0) compared
aerobic training,35,36,38,45,61 walking exercise,65 or multimodal
with individuals with mild cognitive impairment and mild-grade
exercise,30,49,62,71 with a training duration ranging from 30 to
dementia (seven trials, 357 participants) (–1 second, 95% CI –1 to
90 minutes/session, 2 to 4 sessions/week, for a total of 9 weeks to
0). Overall, it was found that aerobic exercise alone (three trials,
12 months. Reported effective training intensity was 30 to 60%
236 participants) did not significantly improve TUG performance
VO2max,62 or 40% of heart rate reserve that gradually progressed to
(p = 0.35). The heterogeneity across trials was low, with I2 ranging
85%.30,35,36
from 0 to 32%.
Meta-analysis (seven trials, 568 participants) revealed that
exercise improved walking speed by 0.13 m/s (95% CI 0.03 to Dual-task ability (low-quality evidence)
0.24).28,41,52,56,66,71,77 See Figure 12, or for a more detailed forest Five trials examined the effect of exercise on dual-task
plot see Figure 13 on the eAddenda. Sensitivity analysis that ability.39,40,48,52,63 A dual-task training component was incorpo-
included only trials with high methodological quality (six trials, rated into the overall exercise program in two trials.40,63 Only one

MD (95% Cl)
Study Random
Christofoletti 2008

Lam 2012

Miu 2008

Telenius 2015

Toots 2016

Yoon 2013

Pooled

High methodological quality only

Moderate grade dementia / institutionalised only

–10 –5 0 5 10

Favours control (points) Favours excercise

Figure 6. Weighted mean difference (95% CI) in Berg Balance Scale performance (0 to 56) due to exercise, pooling data from six studies (722 participants).

Please cite this article in press as: Lam FMH, et al. Physical exercise improves strength, balance, mobility, and endurance in people with
cognitive impairment and dementia: a systematic review. J Physiother. (2018), https://doi.org/10.1016/j.jphys.2017.12.001
G Model
JPHYS-391; No. of Pages 12

Research 7

MD (95% Cl) MD (95% Cl)

Study Random Study Random

Arcoverde 2014 Bossers 2015

Doi 2013
Lü 2015
Hauer 2012
Miu 2008
Kemoun 2010
Netz 2007
Rolland 2007
Suttanon 2013
Suttanon 2013
Vreugdenhil 2012
Telenius 2015
Pooled
Pooled
High methodological quality only High methodological quality only
MCI and mild grade dementia only MCI and mild grade dementia only
Community setting only Moderate to severe grade dementia only

–10 –5 0 5 10 –0.5 –0.25 0 0.25 0.5

Favours control (m/s) Favours excercise
Favours control (cm) Favours excercise
Figure 12. Weighted mean difference (95% CI) in walking speed (m/s) due to
Figure 8. Weighted mean difference (95% CI) in functional reach test result (cm) due exercise, pooling data from seven studies (568 participants). MCI = mild cognitive
to exercise, pooling data from six studies (242 participants). MCI = mild cognitive impairment.
impairment.

one outcome measure in the ADL category.29,39,43,44,46,49,50,59–

of these trials reported results that favoured the training group.63 61,66,68,70,75,77,80
Meta-analysis (four trials, 237 participants) found
Significantly greater improvements in dual-task gait speed, that exercise significantly improved the Barthel Index by 10 points
cadence, stride length, and single support duration were reported (95% CI 3 to 16).49,50,61,77 See Figure 16, or for a more detailed forest
in the dual-task training group when compared with non-specific plot see Figure 17 on the eAddenda. Publication bias was noted for
low-intensity exercise. No significant results were found in the this meta-analysis. Sensitivity analyses that involved only trials
other four trials.39,40,48,52 with high methodological quality (three trials, 221 participants) or
those that were conducted in institutionalised settings (three
Activities of daily living (low-quality evidence) trials, 197 participants) also yielded results in favour of exercise.
Twenty-two trials assessed the effect of exercise on activities of The heterogeneity across trials was high in all analyses (I2  72%). A
daily living (ADL) performance.29,39,43,44,46,49,50,53,57,59–62,66,68– wide selection of exercise protocols (eg, ADL training, multimodal
70,74,75,77,78,80
Sixteen trials reported significant benefits in at least exercise, aerobic exercise) were found to improve ADL perfor-
mance (20 to 150 minutes per session, at least 2 sessions per week,
for 12 weeks to 15 months).29,39,46,49,50,59,61,62,66,68,70,75,80
MD (95% Cl)
Study Random
Aguiar 2014 Participation
Bossers 2015
Activity level (low-quality evidence)
Cancela 2016
Five trials examined the effect of exercise on activity
Christofoletti 2008 level.27,39,41,42,46,52 Four trials reported that the exercise training
Hauer 2012
MD (95% Cl)
Kovács 2013 Study Random
Lü 2015 Bossers 2015

Netz 2007 Kwak 2008

Miu 2008
Suttanon 2013
Roach 2011
Varela 2011
Tappen 2000
Vreugdenhil 2012
Uemura 2012
Pooled
Venturelli 2011
High methodological quality only
Pooled
MCI and mild grade dementia only
High methodological quality only
Moderate grade dementia only
Moderate to severe grade dementia only
Community setting only Community setting only
Institutionalised only Institutionalised only
Aerobic exercise only Aerobic and walking exercise only

Aerobic exercise only

–10 –5 0 5 10
–100 –50 0 50 100
Favours excercise (s) Favours control
Favours control (m) Favours excercise
Figure 10. Weighted mean difference (95% CI) in Timed Up and Go test performance
(s) due to exercise, pooling data from 11 studies (606 participants). MCI = mild Figure 14. Weighted mean difference (95% CI) in 6-minute walk test results (m) due
cognitive impairment. to exercise, pooling data from seven studies (402 participants).

Please cite this article in press as: Lam FMH, et al. Physical exercise improves strength, balance, mobility, and endurance in people with
cognitive impairment and dementia: a systematic review. J Physiother. (2018), https://doi.org/10.1016/j.jphys.2017.12.001
G Model
JPHYS-391; No. of Pages 12

8 Lam et al: Exercise in cognitive impairment and dementia

MD (95% Cl) controls, while the incidence of fractures or hospitalisation did not
Study Random
differ between groups.
Santana-Sosa 2008

Telenius 2015 Attendance

Venturelli 2011 Twenty-three trials reported participants’ attendance to

Vreugdenhil 2012 exercise.27,28,30,31,36–40,46,48,49,52,55,60,61,63–67,71,74–80 Among these,
the mean attendance rate ranged from 33 to 99%. Most
Pooled
trials reported an attendance rate between 70 and
High methodological quality only 90%.27,28,30,31,38,39,46,55,60,65,67,71,74–80 Two trials reported an
attendance rate <70%.65,66 Three trials reported reasons for low
Institutionalised only
attendance, which included acute disease, disagreement or
unwillingness to continue,66 behaviour disorders,66 increased
–20 –10 0 10 20 disability in ADLs,66 and health-related problems.37,52 One trial
Favours control (points) Favours excercise showed that women had a better attendance to exercise training
than men.74 Better baseline cognitive ability also showed a weak
Figure 16. Weighted mean difference (95% CI) in Barthel Index score (0 to 100) due association with better attendance.74
to exercise, pooling data from four studies (237 participants).

Safety
did not result in higher activity level than controls,27,39,46,52 except
for one of the outcome variables (ie, total steps per week) reported In almost all of the reviewed trials, the exercise training was
by Lautenschlager and colleagues.27 One trial reported a signifi- conducted under the supervision of either the caregiver (four
cantly greater improvement in the exercise group immediately trials)46,50,52,61 or professional staff (eg, research staff, certified
after training, but the effect was not maintained at 3 or exercise instructors, therapists) (33 trials).28–45,47–49,51,53–60,62–79
9 months.41,42 The only exception was the trial by Lautenschlager and collea-
gues,27 in which the participants with mild cognitive impairment
were taught the exercises and advised to do them independently at
Quality of life (moderate-quality evidence)
home. Of the four trials in which the caregiver provided
Eight trials measured quality of life.27,37,39,45,46,52,53,76 All six
supervision, three involved participants with mild-grade demen-
trials that used multimodal exercise as the experimental interven-
tia46,50,52 and one involved moderate-grade dementia.61 The
tion (at least two 40 to 60 minute sessions/week for 12 weeks to
remaining trials did not provide information on supervision of
6 months) reported no significant results.27,45,46,52,53,76 In the other
the exercise.44,56,59,62,80
two trials,37,39 a walking program was adopted; one of these trials
Nineteen trials explicitly reported if any adverse effects
reported improvement in satisfaction in daily life.39 While the trial
occurred.27,29,31–33,36,37,41,42,46,48,49,52,53,60,61,63,64,66,72,74,76–79
that reported no significant improvement adopted moderate-
Among these, 10 trials reported that no adverse event occur-
intensity walking exercise (60 minutes/day, 2 days/week, 1 year),37
red.29,37,41,42,46,48,49,61,63,64,74,76,77,79 Four trials reported few ad-
the trial that reported significant improvement in quality of life
verse events in the exercise group (ie, foot pain,27 falls,31,33,60,72
administered a 90-minute walking program once a week for
hospitalisation31) that were considered unlikely to be related to the
3 months. In the latter program, the participants were also
intervention by the original authors.27,33 Five trials reported
encouraged to walk daily and organise walking events with other
adverse events that may be related to exercise.36,52,53,66,78 These
group members.39
adverse events included a higher number of hospitalisations per
patient,66 falls,36,66 shortness of breath,36 erythema,53 and pain or
Falls (low-quality evidence) discomfort, which eased as time progressed or by slight
Four trials examined the effect of exercise on falls.42,51,52,60,72,73 modification of the exercise.52 One trial reported that one of the
Meta-analysis (three trials, 191 participants) showed that exercise participants fell ill one day after an exercise session and later
did not reduce the number of fallers compared to controls (OR 0.98, passed away with a diagnosis of circulatory failure and general
95% CI 0.49 to 1.95).42,51,52 See Figure 18, or for a more detailed atherosclerosis.78
forest plot see Figure 19 on the eAddenda. The heterogeneity across
trials was low in this analysis (I2 = 26%). For other fall-related
Discussion
outcomes, three trials reported that the exercise group did not
differ from the controls in cumulative falls, fall rate, fall risk score
This systematic review summarised the evidence on the effects
and time to first fall.42,51,52 Pitkälä and colleagues60 reported that
of exercise on physical functioning and quality of life in individuals
12 months of multimodal exercise resulted in a significant lower
with mild cognitive impairment and dementia. The large number
number of falls per year when compared with no-intervention
of trials involved in this review made sensitivity analyses possible,
OR (95% Cl) which was missing in previous reviews.13–18,22,82–84 Sensitivity
Study Random analysis allowed the effects of exercise on patient subgroups to be
determined, and the potential factors that may affect training
Kovács 2013 outcomes to be identified.
Overall, very few adverse events were reported with exercise.
Suttanon 2013
Hence, according to the quality and direction of the evidence, there
Zieschang 2013 was strong evidence to support the use of exercise in improving
strength, step length, balance, TUG, walking speed, and endurance
Pooled in people with mild cognitive impairment, or mild-to-moderate-
grade dementia. On the other hand, the evidence for supporting
0.2 0.5 1 2 5 the use of exercise in improving flexibility and ADL was weak.
There was weak evidence against the use of non-specific exercise
Favours excercise Favours control
in improving BMI, dual-tasking ability and activity level, and strong
Figure 18. Odds ratio (95% CI) of number of fallers in the exercise versus control evidence against the use of non-specific exercise in improving
groups, pooling data from three studies (191 participants). quality of life. The review provided no recommendations on the

Please cite this article in press as: Lam FMH, et al. Physical exercise improves strength, balance, mobility, and endurance in people with
cognitive impairment and dementia: a systematic review. J Physiother. (2018), https://doi.org/10.1016/j.jphys.2017.12.001
G Model
JPHYS-391; No. of Pages 12

Research 9

effect of exercise on falls, as most trials collected fall data during 60 minutes;48,55,76,77 significant: 75 to 120 minutes).41,42,49 This
the intervention period.51,52,60 This data collection method is prone suggests that the strength training component cannot be too short
to inaccuracy, as the effect of exercise may not appear in the initial in a multimodal training program if benefits on muscle strength
periods of intervention. The meta-analysis also may not have had are desired. For walking endurance, trials that reported negative
sufficient power to detect a significant between-group difference findings also tended to adopt a shorter training duration per
in fall rate, given the limited sample size. The results of this review session (10 to 30 minutes).67,69,70 A longer exercise duration in each
can be generalised to people with mild cognitive impairment or session (30 minutes of aerobic training) is required to induce a
mild-to-moderate-grade dementia. positive training effect.
Specificity of training appears to be an important factor in With a large number of trials included in our review, it was
obtaining beneficial effects on impairment-level outcomes (eg, possible to conduct sensitivity analyses for people with different
strength, flexibility) and certain activity-level outcomes (ie, levels of cognitive impairment. Indeed, sensitivity analyses that
walking endurance). For example, among all trials that reported involved only people with mild cognitive impairment or mild-
favourable results on lower limb strength and walking endurance, grade dementia did not show superior benefits of exercise in
100% and 71% included specific training for the respective improving step length and functional reach when compared with
outcomes. Indeed, the sensitivity analysis showed that specific the primary analysis. On the contrary, when compared with the
aerobic training led to the greatest improvement in the 6-minute primary analyses, the moderate dementia group tended to have a
walk distance. On the other hand, for lower limb strength, two48,55 slightly better improvement in BBS and the moderate-to-severe
out of five trials46,48,52,55,76,77 that did not identify a treatment dementia group had a better improvement in 6-minute walk
benefit did not incorporate training that specifically targeted lower distance. When compared with the mild cognitive impairment and
limb muscle strength. For endurance, all trials that did not identify mild-grade dementia group, the moderate dementia group also
a treatment benefit did not include an aerobic training component showed more improvement in TUG, while the moderate-to-severe
in their training program.67,69,70 All trials that measured flexibility dementia group showed slightly more improvement in walking
also included flexibility exercise as part of the overall training, and speed. It should be noted that the response in the moderate-to-
reported positive results.49,58,62 For more complex activity-level severe-grade dementia subgroup also tended to be more
outcomes such as balance, the specificity of training seemed to be heterogeneous (eg, TUG, walking speed, 6-minute walk test). This
less important. Sixty-four per cent of the trials that reported finding challenges the previous finding that better cognitive ability
favourable results on balance outcomes incorporated specific is associated with greater improvement in TUG performance
balance training. However, two trials45,48 that adopted aerobic within a group of patients with mild cognitive impairment.86
exercise, and one trial that incorporated aerobic and resistance However, it should be noted that this comparison involved a much
exercise71 also reported positive effects on balance. On the other greater between-group difference in baseline cognitive ability (ie,
hand, although Tai Chi33 and cycling exercise59 were reported to mild cognitive impairment and mild-grade dementia versus
induce significant effects on balance, the improvement was very moderate-to-severe-grade dementia) rather than comparing
small and markedly less than the minimal clinically important individuals within the same diagnostic category (eg, within mild
change.85 cognitive impairment only). As the cognitive decline was found to
For even higher levels of function like ADL, various forms of be associated with poorer physical function, the group with
exercise other than specific ADL training were found to be more advanced cognitive impairment may indeed have a larger
beneficial, including aerobic walking training,61 cycling exer- potential for improvement in physical function with exercise
cise,75,80,59 multimodal training without a specific ADL training training. For example, a previous trial revealed that the effect of
component,18,46,49,50,66 handball training,29and dance and move- exercise on fall reduction is more potent in people with advanced
ment therapy.44 This could be because ADL ability can be dementia due to their inherent high fall rate.73
influenced by many impairment variables. Taken together, these results suggest that people with
Based on the overall evidence, it can be deduced that moderate-grade cognitive impairment are still likely to respond
regular multimodal exercise with a combination of resistance, positively to exercise training. The response to exercise training
aerobic, balance, flexibility and functional training for around among people with severe-grade cognitive impairment remains
60 minutes a day, 2 to 3 days a week is effective in improving largely uncertain, as only five trials66–70 were conducted in this
various aspects of physical functioning (lower limb strength, patient subgroup and only one was included in the meta-analysis
mobility, balance, walking endurance). Considering that specific of walking speed66 and 6-minute walk test distance,70 respectively.
training is more important for impairment level outcomes, a Participants’ physical ability at baseline also appears to
comprehensive assessment of physical functioning is required influence training outcome. For TUG, the sensitivity analysis
to determine the specific areas of deficit, so that the appropriate involving those trials that included only community-dwelling
type of exercise (flexibility, aerobic, resistance) can be individuals yielded non-significant results. This could be because
incorporated into the overall program to address the impair- community-dwelling participants tended to have better baseline
ments identified. The training program duration varied (ie, mobility level (mean TUG: 10 to 18 seconds) and thus have less
8 weeks to 15 months), but the overall evidence indicated that room for improvement,50,52,53,55 when compared with most of
positive training effects on physical function can be obtained in the other trials included in the primary meta-analysis (mean
9 to 16 weeks. Based on the few trials that included follow-up TUG: 15 to 33 seconds).41,47,51,54,71,75 Perttila et al72 also suggested
assessment, the improvements in strength, step length and that while the worsening of ADL performance was attenuated in
balance diminished from 9 weeks to 9 months after the both the prefrail and frail subgroups of people with dementia upon
intervention. Continuous exercise programs may be required to exercise training, a significant effect was achieved at the sixth
maintain the treatment effect. month of training for the advanced frailty group but only at the
With respect to the frequency, intensity and duration of twelfth month for the prefrail group. The fall reduction effect
training, no clear differences in these parameters were identified was also greater in the advanced frailty group (incidence rate
between the trials that yielded larger effect sizes and those with ratio: advanced frailty: 0.43; prefrail: 0.63). This suggested that
smaller effect sizes for all outcomes, except lower limb strength cognitively impaired individuals who have deteriorated physical
and walking endurance. For strength, one trial suggested that performance could potentially benefit more from exercise training.
30 minutes of specific strength training of 12 repetition-maximum Exercise attendance appeared to be generally acceptable
could lead to significant strength improvement.71 However, across trials. Two trials reported an attendance rate <70%.65,66
among the trials that incorporated multimodal exercise, those Upon examining the training protocols and the participants’
that reported no significant results tended to use a shorter characteristics, no obvious relationship could be found between
total training duration per session (non-significant: 30 to the training protocols used and attendance rates. However, in the

Please cite this article in press as: Lam FMH, et al. Physical exercise improves strength, balance, mobility, and endurance in people with
cognitive impairment and dementia: a systematic review. J Physiother. (2018), https://doi.org/10.1016/j.jphys.2017.12.001
G Model
JPHYS-391; No. of Pages 12

10 Lam et al: Exercise in cognitive impairment and dementia

two trials that reported exceptionally low attendance, the way that the different dementia grades were classified (Table 2).
cognitive ability of the participants was also the poorest (mean Finally, only trials reported in English were included in this
MMSE = 8.8 and 11.1, respectively).65,66 However, two other trials review.
that involved participants with relatively low cognitive ability Despite the large number of trials that have been conducted,
(mean MMSE = 13) reported good attendance.55,61 Therefore, no determinants of successful training outcomes are still unclear for
consistent trend could be identified regarding the relationship many of the outcome measures. More effort could be put into
between cognitive ability and exercise attendance. Taken together, identifying key factors (ie, training type, and parameters) that
the difference in attendance rate could possibly be due to complex determine the favourable treatment outcomes. Good quality trials
interaction of factors such as the exercise training protocol are needed to study the effects on exercise on falls and dual-tasking
adopted, cognitive ability, level of physical functioning and ability, which are important concerns in individuals with cognitive
involvement of the family/caregiver. impairments. Mediators or strategies that are required to translate
Unfortunately, other aspects of adherence (eg, intensity level, the improvement in body functions/structures-level outcomes
training duration) were seldom reported in the included trials. into improvement in activity/participation-level outcomes should
Therefore, the influence of patient or intervention-related factors be identified. Long-term effects of exercise training also await
on patients’ adherence to the prescribed protocol could not be further investigations. More economical ways of providing exercise
deduced. The influence of exercise adherence to treatment training (eg, through educating the caregiver) should be further
outcomes also remains uncertain. investigated. More research on people with severe dementia
It is recommended that exercise training be supervised by should be conducted.
professional staff. For those with mild cognitive impairment or In conclusion, this review demonstrated that physical exercise
mild dementia, the training may be supervised by a properly training is a feasible intervention for people with mild cognitive
trained caregiver.46,50,52,61 Less than half of the trials included in impairment and dementia, and that the benefits far outweigh the
this review included information regarding adverse effects. Only risks. Strong evidence was found to support the use of physical
five out of 14 trials reported adverse effects that were related to exercise in improving strength, step length, balance, mobility, and
exercise.36,52,53,66,78 Most adverse events reported were mild and walking endurance. There was weak evidence supporting the use
subsided quickly.36,52 A higher hospitalisation rate in the exercise of exercise for improving flexibility and ADL, and against the use of
group was reported by Rolland et al66 but it remains unclear if non-specific exercise for improving dual-tasking ability and
the between-group difference in hospitalisation rate was already increasing activity levels. The evidence against the use of
present prior to commencement of the exercise intervention. The non-specific exercise alone for improving quality of life was
single death event reported by one trial was not directly related to strong. Taken together, supervised multimodal exercise for about
exercise training.78 In summary, no evidence suggests that exercise 60 minutes a day and 2 to 3 days a week was beneficial in
poses an substantial risk to individuals with mild cognitive improving physical functioning in individuals with mild cognitive
impairment or dementia. impairment or mild-to-moderate-grade dementia. The effect of
Research in people with severe dementia is scarce. The results exercise in severe-grade dementia was less certain, due to paucity
of this review are therefore better generalised to people with mild of research.
cognitive impairment, and mild or moderate dementia. Details on
the training protocol, such as the specific exercises performed,
training intensity and progression rules, and the time devoted to
different types of exercise in a multimodal program are crucial What is already known on this topic: Exercise training
pieces of information, but were seldom reported. Delineating the improves cognitive and physical function in healthy older
factors that are important determinants of successful training adults, and is feasible and beneficial for people with cognitive
outcomes is therefore difficult. Detailed information on adherence impairment. Patients with dementia are less likely to partici-
pate in regular physical exercise when compared with their
to exercise training (ie, degree of deviation between the actual
counterparts with normal cognition.
exercise conducted by the participants and the prescribed exercise
What this study adds: Strong evidence now indicates that
protocol) was also rarely reported in the included trials. Long-term exercise for people with mild cognitive impairment or mild to
follow-up trials were few and so the carry-over effects of exercise moderate dementia improves sit to stand, step length, balance,
were not well evaluated. mobility and walking endurance. People across these levels of
This review had some limitations. We used the mean baseline cognitive impairment are more likely to achieve these benefits
cognitive outcome scores to classify the selected trials into with supervised multi-modal exercise for about 60 minutes a
different patient sub-groups. There may be potential inaccuracy day, 2 to 3 days a week.
with this approach, as heterogeneity of the sample within the
same trial was not considered. Most trials used multimodal
exercise as the experimental intervention, and so it was difficult Footnotes: a Review Manager 5.3, The Nordic Cochrane Centre,
to single out the effect of a particular exercise type. We did not The Cochrane Collaboration, Copenhagen, Denmark; b Compre-
perform any economic analysis and hence could not comment on hensive Meta-analysis version 3, Biostatc, Inc., Englewood, NJ, USA.
the cost-effectiveness of the exercise programs studied. We did eAddenda: Figures 3, 5, 7, 9, 11, 13, 15, 17 and 19 and Appendices
not separate the trials that tended to examine ‘efficacy’ (training 1, 2, 3 and 4 can be found online at https://doi.org/10.1016/j.jphys.
performed in a highly controlled environment) from those that 2017.12.001
tended to examine ‘effectiveness’ (training performed under real- Ethics approval: Not required.
world conditions) on the continuum from efficacy to effective- Competing interests: The authors declare that there is no
ness. To assist readers in judging whether the findings of the conflict of interest.
present review apply to ‘real-world’ conditions in their respective Source of support: Freddy MH Lam and MZ Huang were granted
contexts, other relevant information of the trials (key subject a full-time research scholarship by the Hong Kong Polytechnic
characteristics, attendance, adverse effects, and supervision level) University (RTSF). LR Liao received support from the Research
was provided. Many trials did not contain adequate details Grant Council. Marco YC Pang was provided with a research grant
necessary for meta-analysis. Meta-analysis could not be per- by the Hong Kong Polytechnic University (G-YJ41).
formed for several outcomes, due to the small number of trials. Acknowledgements: Nil.
Publication bias was present in a small number of analyses. A Provenance: Not invited. Peer reviewed.
definite conclusion could not be made on the effect of exercise in Correspondence: Marco YC Pang, Department of Rehabilitation
severe-grade dementia because there is less research in this Sciences, Hong Kong Polytechnic University, Hong Kong. Email:
subgroup of the dementia population. This may also be due to the Marco.Pang@polyu.edu.hk

Please cite this article in press as: Lam FMH, et al. Physical exercise improves strength, balance, mobility, and endurance in people with
cognitive impairment and dementia: a systematic review. J Physiother. (2018), https://doi.org/10.1016/j.jphys.2017.12.001
G Model
JPHYS-391; No. of Pages 12

Research 11

References 31. Suzuki T, Shimada H, Makizako H, Doi T, Yoshida D, Ito K, et al. A randomized
controlled trial of multicomponent exercise in older adults with mild cognitive
1. Alzheimer’s Disease International. World Alzheimer Report 2015, The Global Impact of impairment. PLoS One. 2013;8:e61483.
Dementia: An Analysis of Prevalence, Incidence, Cost and Trends; 2015. 32. Lam LCW, Chau RCM, Wong BML, Fung AW, Lui VW, Tam CC, et al. Interim follow-up
2. Lövheim H, Sandman P-O, Karlsson S, Gustafson Y. Behavioral and psychological of a randomized controlled trial comparing Chinese style mind body (Tai Chi)
symptoms of dementia in relation to level of cognitive impairment. Int and stretching exercises on cognitive function in subjects at risk of progressive
Psychogeriatr. 2008;20:777–789. cognitive decline. Int J Geriatr Psychiatry. 2011;26:733–740.
3. Franssen EH, Souren LE, Torossian CL, Reisberg B. Equilibrium and limb 33. Lam LCW, Chau RCM, Wong BML, Fung AW, Tam CW, Leung GT, et al. A 1-year
coordination in mild cognitive impairment and mild Alzheimer’s disease. J Am randomized controlled trial comparing mind body exercise (Tai Chi) with
Geriatr Soc. 1999;47:463–469. stretching and toning exercise on cognitive function in older Chinese adults at
4. Shaw FE. Falls in cognitive impairment and dementia. Clin Geriatr Med. risk of cognitive decline. J Am Med Dir Assoc. 2012;13:568.e15–20.
2002;18:159–173. 34. Segal SK, Cotman CW, Cahill LF. Exercise-induced noradrenergic activation
5. Lam LCW, Tam CWC, Lui VWC, Chan WC, Chan SS, Wong S, et al. Prevalence of very enhances memory consolidation in both normal aging and patients with amnestic
mild and mild dementia in community-dwelling older Chinese people in Hong Kong. mild cognitive impairment. J Alzheimers Dis. 2012;32:1011–1018.
Int Psychogeriatrics. 2008;20:135–148. 35. Baker LD, Frank LL, Foster-Schubert K, Green PS, Wilkinson CW, McTiernan A, et al.
6. Ferri CP, Prince M, Brayne C, Brodaty H, Fratiglioni L, Ganguli M, et al. Global Effects of aerobic exercise on mild cognitive impairment: a controlled trial. Arch
prevalence of dementia: a Delphi consensus study. Lancet. 2005;366(9503): Neurol. 2010;67:71–79.
2112–2117. 36. Nagamatsu LS, Handy TC, Hsu CL, Voss M, Liu-Ambrose T. Resistance training
7. Baker NL, Cook MN, Arrighi HM, Bullock R. Hip fracture risk and subsequent promotes cognitive and functional brain plasticity in seniors with probable mild
mortality among Alzheimer’s disease patients in the United Kingdom, cognitive impairment. Arch Intern Med. 2012;172:666–668.
1988–2007. Age Ageing. 2011;40:49–54. 37. van Uffelen JGZ, Chinapaw MJM, Hopman-Rock M, vanMechelen W. The effect of
8. Eriksson S, Gustafson Y, Lundin-Olsson L. Risk factors for falls in people with and walking and vitamin B supplementation on quality of life in community-dwelling
without a diagnose of dementia living in residential care facilities: a prospective adults with mild cognitive impairment: a randomized, controlled trial. Qual Life Res.
study. Arch Gerontol Geriatr. 2008;46:293–306. 2007;16:1137–1146.
9. Angevaren M, Aufdemkampe G, Verhaar HJJ, Aleman A, Vanhees L. Physical 38. van Uffelen JGZ, Chinapaw MJM, Hopman-Rock M, vanMechelen W. Feasibility and
activity and enhanced fitness to improve cognitive function in older people effectiveness of a walking program for community-dwelling older adults with mild
without known cognitive impairment. Cochrane Database Syst Rev. 2008;2: cognitive impairment. J Aging Phys Act. 2009;17:398–415.
CD005381. 39. Maki Y, Ura C, Yamaguchi T, Murai T, Isahai M, Kaiho A, et al. Effects of intervention
10. Howe TE, Rochester L, Neil F, Skelton DA, Ballinger C. Exercise for improving using a community-based walking program for prevention of mental decline: a
balance in older people. Cochrane Database Syst Rev. 2011;11:CD004963. randomized controlled trial. J Am Geriatr Soc. 2012;60:505–510.
11. Liu C-J, Latham NK. Progressive resistance strength training for improving physical 40. Makizako H, Doi T, Shimada H, Yoshida D, Tsutsumimoto K, Uemura K, et al. Does a
function in older adults. Cochrane Database Syst Rev. 2009;8:CD002759. multicomponent exercise program improve dual-task performance in amnestic
12. Lam FMH, Lau RWK, Chung RCK, Pang MYC. The effect of whole body vibration on mild cognitive impairment? A randomized controlled trial. Aging Clin Exp Res.
balance, mobility and falls in older adults: A systematic review and meta-analysis. 2012;24:640–646.
Maturitas. 2012;72:206–213. 41. Hauer K, Schwenk M, Zieschang T, Essig M, Becker C, Oster P. Physical training
13. Littbrand H, Stenvall M, Rosendahl E. Applicability and effects of physical exercise improves motor performance in people with dementia: a randomized controlled
on physical and cognitive functions and activities of daily living among people with trial. J Am Geriatr Soc. 2012;60:8–15.
dementia: a systematic review. Am J Phys Med Rehabil. 2011;90:495. 42. Zieschang T, Schwenk M, Oster P, Hauer K, Library M, Library PY. Sustainability of
14. Suttanon P, Hill K, Said C, Dodd K. Can balance exercise programmes improve motor training effects in older people with dementia. J Alzheimers Dis.
balance and related physical performance measures in people with dementia? A 2013;34:191–202.
systematic review. Eur Rev Aging Phys Act. 2010;7:13–25. 43. Stevens J, Killeen M. A randomised controlled trial testing the impact of exercise on
15. Potter R, Ellard D, Rees K, Thorogood M. A systematic review of the effects of cognitive symptoms and disability of residents with dementia. Contemp Nurse.
physical activity on physical functioning, quality of life and depression in older 2006;21:32–40.
people with dementia. Int J Geriatr Psychiatry. 2011;26:1000–1011. 44. Hokkanen L, Rantala L, Remes AM, Härkönen B, Viramo P, Winblad I. Dance and
16. Blankevoort CG, van Heuvelen MJ, Boersma F, Luning H, deJong J, Scherder EJ. movement therapeutic methods in management of dementia: a randomized,
Review of effects of physical activity on strength, balance, mobility and ADL controlled study. J Am Geriatr Soc. 2008;56:771–772.
performance in elderly subjects with dementia. Dement Geriatr Cogn Disord. 45. Miu D, Szeto SLS, Mak YFY. A randomised controlled trial on the effect of exercise
2010;30:392–402. on physical, cognitive and affective function in dementia subjects. Asian J Gerontol
17. Hauer K, Becker C, Lindemann U, Beyer N. Effectiveness of physical training on Geriatr. 2008;3:8–16.
motor performance and fall prevention in cognitively impaired older persons: a 46. Steinberg M, Leoutsakos J-M, Podewils LJ, Lyketsos CG. Evaluation of a home-based
systematic review. Am J Phys Med Rehabil. 2006;85:847–857. exercise program in the treatment of Alzheimer’s disease: the Maximizing
18. Pitkälä K, Raivio M, Strandberg T, Savikko N, Laakkonen M-L. Efficacy of physical Independence in Dementia (MIND) study. Int J Geriatr Psychiatry. 2009;24:
exercise intervention on mobility and physical functioning in older people with 680–685.
dementia: A systematic review. Exp Gerontol. 2013;48(1):85–93. 47. Varela S, Ayán C, Cancela JM, Martín V. Effects of two different intensities of aerobic
19. Farina N, Rusted J, Tabet N. The effect of exercise interventions on cognitive exercise on elderly people with mild cognitive impairment: a randomized pilot
outcome in Alzheimer’s disease: a systematic review. Int Psychogeriatr. study. Clin Rehabil. 2011;26:442–450.
2014;26:9–18. 48. Arcoverde C, Deslandes A, Moraes H, Almeida C, Araujo NB, Vasques PE, et al.
20. Forbes D, Forbes SC, Blake CM, Thiessen EJ, Forbes S. Exercise programs for people Treadmill training as an augmentation treatment for Alzheimer’s disease: a pilot
with dementia. Cochrane Database Syst Rev. 2015;4:CD006489. randomized controlled study. Arq Neuropsiquiatr. 2014;72:190–196.
21. Brett L, Traynor V, Stapley P. Effects of physical exercise on health and well-being of 49. Santana-Sosa E, Barriopedro MI, López-Mojares LM, Pérez M, Lucia A. Exercise
individuals living with a dementia in nursing homes: a systematic review. J Am Med training is beneficial for Alzheimer’s patients. Int J Sports Med. 2008;29:
Dir Assoc. 2016;17:104–116. 845–850.
22. Heyn P, Abreu BC, Ottenbacher KJ. The effects of exercise training on elderly 50. Vreugdenhil A, Cannell J, Davies A, Razay G. A community-based exercise
persons with cognitive impairment and dementia: a meta-analysis. Arch Phys programme to improve functional ability in people with Alzheimer’s disease: a
Med Rehabil. 2004;85:1694–1704. randomized controlled trial. Scand J Caring Sci. 2012;26:12–19.
23. Burton E, Cavalheri V, Adams R, Browne CO, Bovery-Spencer P, Fenton AM, et al. 51. Kovács E, Sztruhár Jónásné I, Karócz IC, Korpos A, Gondos T. Effects of a multimodal
Effectiveness of exercise programs to reduce falls in older people with dementia exercise program on balance, functional mobility and fall risk in older adults with
living in the community: a systematic review and meta-analysis. Clin Interv Aging. cognitive impairment: a randomized controlled single-blind study. Eur J Phys
2015;10:421–434. Rehabiliatation Med. 2013;49:639–648.
24. Lewis M, Peiris CL, Shields N. Long-term home and community-based exercise 52. Suttanon P, Hill KD, Said CM, Williams SB, Byrne KN, LoGiudice D, et al. Feasibility,
programs improve function in community-dwelling older people with cognitive safety and preliminary evidence of the effectiveness of a home-based exercise
impairment: a systematic review. J Physiother. 2017;63:23–29. programme for older people with Alzheimer’s disease: a pilot randomized
25. Borenstein M, Hedges L, Rothstein H. Meta-analysis fixed effect vs. random effects. controlled trial. Clin Rehabil. 2013;27:427–438.
www.meta-analysis.com. https://www.meta-analysis.com/downloads/Meta- 53. Aguiar P, Monteiro L, Feres A, Gomes I, Melo A. Rivastigmine transdermal patch
analysisfixedeffectvsrandomeffects.pdf. Published 2007 [accessed 02/12/2016]. and physical exercises for Alzheimer’s Disease: a randomized clinical trial. Curr
26. Goldet G, Howick J. Understanding GRADE: an introduction. J Evid Based Med. Alzheimer Res. 2014;11:532–537.
2013;6:50–54. 54. Christofoletti G, Oliani MM, Gobbi S, Stella F, Bucken Gobbi LT, Renato Canineu P. A
27. Lautenschlager NTN, Cox KLK, Flicker L, Foster JK, van Bockxmeer FM, Xiao J, et al. controlled clinical trial on the effects of motor intervention on balance and
Effect of physical activity on cognitive function in older adults at risk for Alzheimer cognition in institutionalized elderly patients with dementia. Clin Rehabil.
Disease a randomized trial. JAMA. 2008;300:1027–1037. 2008;22:618–626.
28. Doi T, Makizako H, Shimada H, Yoshida D, Tsutsumimoto K, Sawa R, et al. Effects of 55. Netz Y, Axelrad S, Argov E. Group physical activity for demented older adults—
multicomponent exercise on spatial-temporal gait parameters among the elderly feasibility and effectiveness. Clin Rehabil. 2007;21:977–986.
with amnestic mild cognitive impairment (aMCI): Preliminary results from a 56. Kemoun G, Thibaud M, Roumagne N, Carette P, Albinet C, Toussaint L, et al.
randomized controlled trial (RCT). Arch Gerontol Geriatr. 2013;56:104–108. Effects of a physical training programme on cognitive function and walking
29. Wei X, Ji L. Effect of handball training on cognitive ability in elderly with mild efficiency in elderly persons with dementia. Dement Geriatr Cogn Disord.
cognitive impairment. Neurosci Lett. 2014;566:98–101. 2010;29:109–114.
30. Uemura K, Doi T, Shimada H, Makizako H, Yoshida D, Tsutsumimoto K, et al. Effects 57. Lam LCW, Lui VWC, Luk DNY, Chau R, So C, Poon V, et al. Effectiveness of an
of exercise intervention on vascular risk factors in older adults with mild cognitive individualized functional training program on affective disturbances and
impairment: a randomized controlled trial. Dement Geriatr Cogn Dis Extra. functional skills in mild and moderate dementia?a randomized control trial. Int
2012;2:445–455. J Geriatr Psychiatry. 2010;25:133–141.

Please cite this article in press as: Lam FMH, et al. Physical exercise improves strength, balance, mobility, and endurance in people with
cognitive impairment and dementia: a systematic review. J Physiother. (2018), https://doi.org/10.1016/j.jphys.2017.12.001
G Model
JPHYS-391; No. of Pages 12

12 Lam et al: Exercise in cognitive impairment and dementia

58. Toulotte C, Fabre C, Dangremont B, Lensel G, Thévenon A. Effects of physical advanced Alzheimer’s Disease: Secondary analyses of a randomized controlled
training on the physical capacity of frail, demented patients with a history of study. Dement Geriatr Cogn Disord. 2016;41:233–241.
falling: a randomised controlled trial. Age Ageing. 2003;32:67–73. 74. Lam LC., Chan WC, Leung T, Fung AW., Leung EM.. Would older adults with mild
59. Yoon JE, Lee SM, Lim HS, Kim TH, Jeon JK, Mun MH. The effects of cognitive activity cognitive impairment adhere to and benefit from a structured lifestyle activity
combined with active extremity exercise on balance, walking activity, memory intervention to enhance cognition? A cluster randomized controlled trial. PLoS One.
level and quality of life of an older adult sample with dementia. J Phys Ther Sci. 2015;10:e0118173.
2013;25:1601–1604. 75. Cancela JM, Ayán C, Varela S, Seijo M. Effects of a long-term aerobic exercise
60. Pitkälä KH, Pöysti MM, Laakkonen M-L, Tilvis RS, Savikko N, Kautiainen H, et al. intervention on institutionalized patients with dementia. J Sci Med Sport.
Effects of the Finnish Alzheimer disease exercise trial (FINALEX): a randomized 2016;19:293–298.
controlled trial. JAMA Intern Med. 2013;173:894–901. 76. Telenius EW, Engedal K, Bergland A. Long-term effects of a 12 weeks high-intensity
61. Venturelli M, Scarsini R, Schena F. Six-month walking program changes cognitive functional exercise program on physical function and mental health in nursing
and ADL performance in patients with Alzheimer. Am J Alzheimers Dis Other Demen. home residents with dementia: a single blinded randomized controlled trial. BMC
2011;26:381–388. Geriatr. 2015;15:158.
62. Kwak Y, Um S, Son T, Kim D. Effect of regular exercise on senile dementia patients. 77. Telenius EW, Engedal K, Bergland A. Effect of a high-intensity exercise program on
Int J Sports Med. 2008;29:471–474. physical function and mental health in nursing home residents with dementia: An
63. Schwenk M, Zieschang T, Oster P, Hauer K. Dual-task performances can be assessor blinded randomized controlled trial. PLoS One. 2015;10:1–18.
improved in patients with dementia: a randomized controlled trial. Neurology. 78. Toots A, Lindelöf N, Littbrand H, Wiklund R, Holmberg H, Nordström P, et al. Effects
2010;74(24):1961–1968. of a high-intensity functional exercise program on dependence in activities of
64. Schwenk M, Zieschang T, Englert S, Grewal G, Najafi B, Hauer K. Improvements in daily living and balance in older adults with dementia. J Am Geriatr Soc.
gait characteristics after intensive resistance and functional training in people with 2016;64:55–64.
dementia: a randomised controlled trial. BMC Geriatr. 2014;14:73. 79. Lü J, Sun M, Liang L, Feng Y, Pan X, Liu Y. Effects of momentum-based dumbbell
65. Tappen RM, Roach KE, Brooks Applegate E, Stowell P. Effect of a combined walking training on cognitive function in older adults with mild cognitive impairment: A
and conversation intervention on functional mobility of nursing home residents pilot randomized controlled trial. Clin Interv Aging. 2015;11:9–16.
with Alzheimer disease. Alzheimer Dis Assoc Disord. 2000;14:196–201. 80. Holthoff VA, Marschner K, Scharf M, Steding J, Meyer S, Koch R, et al. Effects of
66. Rolland Y, Pillard F, Klapouszczak A, Reynish E, Thomas D, Andrieu S, et al. Exercise physical activity training in patients with Alzheimer’s dementia: Results of a pilot
program for nursing home residents with Alzheimer’s disease: a 1-year RCT study. PLoS One. 2015;10:1–11.
randomized, controlled trial. J Am Geriatr Soc. 2007;55:158–165. 81. World Health Organization. How to Use the ICF: A Practical Manual for Using the
67. Pomeroy VM, Warren CM, Honeycombe C, Briggs RS, Wilkinson DG, Pickering RM, International Classification of Functioning, Disability and Health (ICF). Exposure Draft
et al. Mobility and dementia: is physiotherapy treatment during respite care for Comment. Geneva: WHO; 2013.
effective? Int J Geriatr Psychiatry. 1999;14:389–397. 82. Heyn PC, Johnsons KE, Kramer AF. Endurance and strength training outcomes on
68. Tappen RM. The effect of skill training on functional abilities of nursing home cognitively impaired and cognitively intact older adults: a meta-analysis. J Nutr
residents with dementia. Res Nurs Health. 1994;17:159–165. Health Aging. 2008;12:401–409.
69. Cott CA, Dawson P, Sidani S, Wells D. The effects of a walking/talking program on 83. Forbes D, Forbes S, Morgan DG, Markle-Reid M, Wood J, Culum I. Physical activity
communication, ambulation, and functional status in residents with Alzheimer programs for persons with dementia. Cochrane Database Syst Rev. 2008;3:
disease. Alzheimer Dis Assoc Disord. 2002;16:81–87. CD006489.
70. Roach KKE, Tappen RMR, Kirk-Sanchez N, Williams CL, Loewenstein D. A random- 84. Rao AK, Chou A, Bursley B, Smulofsky J, Jezequel J. Systematic review of the effects
ized controlled trial of an activity specific exercise program for individuals with of exercise on activities of daily living in people with Alzheimer’s disease. Am J
Alzheimer Disease in long-term care settings. J Geriatr Phys Ther. 2011;34:50–56. Occup Ther. 2014;68:50–56.
71. Bossers WJR, van der Woude LHV, Boersma F, Hortobágyi T, Scherder EJA, van 85. Godi M, Franchignoni F, Caligari M, Giordano A, Turcato AM, Nardone A.
Heuvelen MJG. A 9-week aerobic and strength training program improves cognitive Comparison of reliability, validity, and responsiveness of the Mini-BESTest
and motor function in patients with dementia: a randomized, controlled trial. Am J and Berg Balance Scale in patients with balance Disorders. Phys Ther.
Geriatr Psychiatry. 2015;23:1106–1116. 2013;93:158–167.
72. Perttila NM, Öhman H, Strandberg TE, Kautiainen H, Raivio M, Laakkonen ML, et al. 86. Uemura K, Shimada H, Makizako H, Doi T, Yoshida D, Tsutsumimoto K, et al.
Severity of frailty and the outcome of exercise intervention among participants Cognitive function affects trainability for physical performance in exercise
with Alzheimer disease: A sub-group analysis of a randomized controlled trial. Eur intervention among older adults with mild cognitive impairment. Clin Interv
Geriatr Med. 2016;7:117–121. Aging. 2013;8:97–102.
73. Öhman H, Savikko N, Strandberg TE, Kautiainen H, Raivio M, Laakkonen ML, et al.
Effects of exercise on functional performance and fall rate in subjects with mild or

Please cite this article in press as: Lam FMH, et al. Physical exercise improves strength, balance, mobility, and endurance in people with
cognitive impairment and dementia: a systematic review. J Physiother. (2018), https://doi.org/10.1016/j.jphys.2017.12.001