Professional Documents
Culture Documents
2013
CRE271210.1177/0269215513488898Clinical RehabilitationTanaka et al.
CLINICAL
Article REHABILITATION
Clinical Rehabilitation
Abstract
Objective: We performed a systematic review and meta-analysis of randomized controlled trials to
investigate the differences in the efficacies between strengthening and aerobic exercises for pain relief in
people with knee osteoarthritis.
Data sources: This search was applied to Medline, Cochrane Central Register of Controlled Trials, the
Physiotherapy Evidence Database, and the Cumulative Index to Nursing and Allied Health Literature. All
literature published from each source’s earliest date to March 2013 was included.
Review methods: Trials comparing the effects of exercise intervention with those of either non-
intervention or psycho-educational intervention were collected. Meta-analysis was performed for trials in
which therapeutic exercise was carried out with more than three sessions per week up to eight weeks,
for pain in people with knee osteoarthritis. All trials were categorised into three subgroups (non-weight-
bearing strengthening exercise, weight-bearing strengthening exercise, and aerobic exercise). Subgroup
analyses were also performed.
Results: Data from eight studies were integrated. Overall effect of exercise was significant with a large
effect size (standardised mean difference (SMD): −0.94; 95% confidence interval −1.31 to −0.57). Subgroup
analyses showed a larger SMD for non-weight-bearing strengthening exercise (−1.42 [−2.09 to −0.75])
compared with weight-bearing strengthening exercise (−0.70 [−1.05 to −0.35]), and aerobic exercise
(−0.45 [−0.77 to −0.13]).
Conclusion: Muscle strengthening exercises with or without weight-bearing and aerobic exercises are
effective for pain relief in people with knee osteoarthritis. In particular, for pain relief by short-term exercise
intervention, the most effective exercise among the three types is non-weight–bearing strengthening exercise.
Keywords
Exercise, knee osteoarthritis, pain, systematic review, meta-analysis
week). In addition, Fransen and McConnell12 con- Two authors (RT and JO) read the full-text arti-
cluded in their the systematic review, which was cles and independently decided whether the
published in Cochrane library, that land-based ther- retrieved trials met the inclusion criteria or not. Two
apeutic exercise has at least short-term benefit in authors (RT and JO) extracted the data using special
terms of reduced knee pain, but there is a clinical standardised forms developed for this review. We
impression that the beneficial effects of exercise on contacted no author for further information. The
osteoarthritis of the knee disappear over time.13 primary outcome data were entered into the Review
Therefore, if there was heterogeneity in exercise Manager 5.1 software (The Cochrane Collaboration,
frequency or duration among trials, we could not Copenhagen, Denmark) for further statistical analy-
determine the exact difference in the efficacy sis. All differences resulting from discrepant assess-
between strengthening and aerobic exercises. ments during data extraction and analysis were
Trials comparing therapeutic exercises with resolved by consulting a third author (NK) or dis-
either non-intervention or psycho-educational inter- cussion among the authors. Information was
ventions in people with knee osteoarthritis were extracted from each included trial on: (a) the char-
included. Jamtvedt et al.14, in their systematic acteristics of trial participants; (b) the contents of
review, wrote that psycho-educational interventions exercise intervention and regimen; and (c) the out-
improve psychological outcomes, but no clinically come scale.
important difference was found for pain. The quality of the trials that were included in our
The primary outcome measure was pain. Trials review was determined by using the PEDro scale.15
using self-reported scales of pain were selected. The PEDro scale is based on the Delphi list devel-
Studies were identified by searching electronic oped by Verhagen and colleagues at the Department
databases. This search was applied to Medline, of Epidemiology, University of Maastricht.16 The
Cochrane Central Register of Controlled Trials, PEDro scale contains 10 check items: ‘random allo-
the Physiotherapy Evidence Database, and the cation’, ‘concealed allocation’, ‘baseline compara-
Cumulative Index to Nursing and Allied Health bility’, ‘blind subjects’, ‘blind therapists’, ‘blind
Literature. Publication data was through 20 March assessors’, ‘adequate follow-up’, ‘intention-to-treat
2013. Publication language was not restricted. analysis’, ‘between-group comparisons’, and ‘point
The first author (RT) developed and conducted estimates and variability’.
the search. Since the outcome consisted of continuous data,
We used the following search terms to search all which were standardised to the same scale, the stan-
trial registers and databases: osteoarthritis knee, dardised mean difference was used as the effect
exercise, exercise therapy, and pain. The search measure. The standardised mean difference is the
strategy consisted of a combination of free text difference in mean outcome between groups divided
words and medical subject headings terms. Article by the standard deviation of the outcome.17 The
searches in the Medline database were limited to standardised mean difference and their correspond-
clinical trials and randomized controlled trials. ing 95% confidence interval were calculated for the
Search strategies were peer reviewed as part of the continuous outcome data. The selection of using
systematic review process. The search strategy is either a fixed-effect model17 or a random-effect
available in the appendix, which is available online. model17 was decided after examination of statistical
Two authors (RT and JO) scanned the articles heterogeneity between trials.
retrieved by the initial search to exclude obvi- The inverse variance method17 was used to syn-
ously irrelevant studies. Study eligibility was thesise the results by means of the Review Manager
determined by reading the title and abstracts, and 5 software. If a standard deviation was not given, it
obviously irrelevant trials were excluded at this could be calculated from the 95% confidence inter-
stage. Inclusion and exclusion criteria were estab- val or the standard error of the mean values. Trials
lished a priori. Disagreements between reviewers that did not give means and standard deviation,
were resolved by consensus. 95% confidence interval, or the standard error of
the mean values were excluded from the synthesis A list of the included studies is shown in Table 1.
in our study. Positive values indicate that the control Of 11 exercise groups in eight studies, six exercise
group improved on average more than the exercise groups in four studies22–25 examined the effect of
group. Effect sizes of 0.2–0.5 can be interpreted as non-weight-bearing muscle strengthening only (i.e.
small, 0.5–0.8 as medium, and greater than 0.8 as concentric-eccentric, isometric, or isokinetic exer-
large effects.18 Heterogeneity between trials was cise under non-weight-bearing conditions). Two
examined by the χ2 test and the l2 test.17 The l2 test exercise groups in one study26 examined the effect
does not inherently depend on the number of studies of weight-bearing muscle strengthening (i.e. exer-
and is accompanied by an uncertainty interval. A cise with a leg press machine) with or without non-
rough guide to interpretation is as follows:17 0%– weight-bearing muscle strengthening, but did not
40% might not be important; 30%–60% may repre- examine the effect of aerobic exercise. Three exer-
sent moderate heterogeneity; 50%–90% may cise groups27–29 examined the effect of aerobic exer-
represent substantial heterogeneity; 75%–100% cise (i.e. walking, Baduanjin, or Tai Chi exercise).
considerable heterogeneity. The result of PEDro scoring is available in the
We stratified the trials according to the type of appendix, available online.
exercise and subgroup analyses were performed. The The means of original data, standard deviation,
trials stratified according to type of exercise were sample sizes, and the standardised mean difference
classified as non-weight-bearing strengthening exer- (95% confidence interval) of each type of exercise
cise category, weight-bearing strengthening exercise is seen in the forest plots (Figure 2). All eight stud-
category, and aerobic exercise category. In our study, ies showed that the effect of the exercise group was
the stratification according to the delivery mode (e.g. better than that of the control group. Five of the
individual treatments, exercise classes, or home pro- eight studies suggested significant differences
gramme) was not performed, because the mode of between the exercise and control groups. No studies
treatment delivery was not significantly associated indicate that the effect of the control group was sig-
with the magnitude of treatment benefit.19 nificantly better than that of the exercise group. The
For each trial, we plotted the effect by the inverse effect integrated with all trials was significant (P <
of its standard error. The symmetry of such ‘funnel 0.00001), and the pooled standardised mean differ-
plots’ was assessed both visually and formally. As a ence was large (−0.94 [−1.31 to −0.57]).
subsequent check of publication bias, the test of All trials were stratified according to the type of
Egger et al.20 was incorporated. In the Egger test the exercise and subgroup analyses were performed.
standard normal deviation is regressed on precision, The results according to type of exercise are shown
defined as the inverse of the standard error.21 This is in Table 2. Statistical substantial heterogeneity
a formal statistical analysis that is intended to assess between the trials for non-weight-bearing strength-
the same assumption as the funnel plot, and may be ening exercise and the trials for weight-bearing
used as a ‘cross-check’ to the physical inspection of strengthening exercise or aerobic exercise was
the data. observed (l2 = 71.1% and 84.6%, respectively). The
pooled standardised mean difference of trials for
non-weight-bearing strengthening exercise was
larger than that of trials for weight-bearing strength-
Results
ening exercise or aerobic exercise. There was not
The search of Medline, Cochrane Central Register important heterogeneity between the trials for
of Controlled Trials, Physiotherapy Evidence weight-bearing strengthening exercise and the trials
Database, and Cumulative Index to Nursing and for aerobic exercise (l2 = 7.5%).
Allied Health Literature provided a total of 559 cita- Analysis of funnel plots (Figure 3) as well as
tions. We excluded 551 studies as shown in Figure 1 results from the Egger’s test suggested negative evi-
and performed the meta-analysis for eight studies dence of publication bias for all trials (P = 0.123).
(11 exercise groups) consequently. The number of studies in each subgroup is four at
Bilaterally
Both at knee flexion and extension
20-second rest between velocity break
3-minute rest between knee breaks
Lin et al. E: 36 E: 61.6 E: 67% Concentric-eccentric exercise/NWBSE 3 8 WOMAC 8
(2009) C: 36 C: 62.2 C: 72% Concentric quadriceps action, followed by an eccentric
quadriceps action
Bilaterally
50% of 1-repetition maximum
Progressive increment of 5% of the original 1-repetition
maximum every 2 weeks
4 sets with 6 repetitions per set
1-minute rest between sets
5-minute rest between sides
Schilke E: 10 E: 64.5 85 Isokinetic strength training/NWBSE 3 8 OASI 4
et al. C: 10 C: 68.4
(2006) Knee extension and flexion
(Continued)
Table 1. (Continued)
Study Participants Exercise contents/subgroup Frequency Duration Outcome PEDro 1066
(Year) (per week) (week) scale score
NWBSE, non-weight-bearing strengthening exercise; WBSE, weight-bearing strengthening exercise; AB, aerobic exercise; E, exercise group; C, Control group; NRS, Numerical
Rating Scale;VAS,Visual Analogue Scale; WOMAC, Western Ontario and McMaster Universities Arthritis Index; OASI, Osteoarthritis Screening Index; AIMS, Arthritis Impact
Measurement Scale.
Some studies include multiple exercise, their groups are shown as E1, E2 in this table.
Clinical Rehabilitation 27(12)
Tanaka et al. 1067
The type of exercise The SMD of trials [95% CI] Heterogeneity (l2)
NWBSE WBSE AE
NWBSE —1.42 — 71.1% 84.6%
(6 exercise groups) [—2.09, —0.75]
WBSE —0.70 — 7.5%
(2 exercise groups) [—1.05, —0.35]
AE —0.45 —
(3 exercise groups ) [—0.77, —0.13]
NWBSE, non—weight—bearing strengthening exercise; WBSE, weight—bearing strengthening exercise; AE, aerobic exercise; SMD,
standardised mean difference; CI, confidence interval.
maximum. According to the Cochrane handbook there are at least 10 studies included in the meta-
for systematic reviews of interventions,17 tests for analysis, because when there are fewer studies the
funnel plot asymmetry should be used only when power of the tests is too low to distinguish chance
SE(SMD)
0
0.2
0.4
0.6
0.8
SMD
1
-4 -2 0 2 4
Subgroups
Non-weight-bearing strengthening exercise Aerobic exercise
Weight-bearing strengthening exercise
from real asymmetry. Consequently, we could not not included as an outcome measure. To our knowl-
assess the risk of publication bias for subgroups. edge, no studies have reported the comparative effec-
tiveness of weight-bearing positions in muscle
strengthening exercise for pain relief. Exercise in the
standing or weight-bearing positions for participants
Discussion with painful knee osteoarthritis might aggravate
Our study involved a meta-analysis and subgroup symptoms such as pain, swelling, and inflammation
analysis to compare the exact effect of strengthen- if the knee joint is overloaded.31, 32 Overloaded
ing and aerobic exercises on pain relief in people weight-bearing positions should not be suggested for
with knee osteoarthritis. The synthesised trials data strengthening exercise to relieve pain.
suggested statistical heterogeneity among trials. The pooled effect size of non-weight-bearing
Subgroup analysis was performed after the stratifi- strengthening exercise was also larger than that of
cation of trials according to the type of exercise. aerobic exercise. Roddy et al.9 indirectly compared
Both types of strengthening exercise were effec- the effect of strengthening exercise and aerobic exer-
tive for pain relief, regardless of whether the weight- cise on pain relief and concluded that there were no
bearing condition was used. Moreover, we observed differences between them. Although they performed
that the pooled effect size of non-weight-bearing a systematic review and meta-analysis, non-weight-
strengthening exercise was larger than that of weight- bearing strengthening exercise in people with knee
bearing strengthening exercise. In only one trial osteoarthritis was used in only three of 11 exercise
reported by Jan et al.30 that compared the effects of groups categorised as those involving strengthening
non-weight-bearing versus weight-bearing exercise exercise. Another eight trials included subjects with
in participants with knee osteoarthritis, but pain was hip osteoarthritis33 and not knee osteoarthritis,34 an
exercise group involving weight-bearing strengthen- The major strength of this study is its design,
ing exercise,35–38 and a control group with exercise which controls for the effects of confounding fac-
(e.g. stretching or range of movement exercise).39 tors on the efficacy of exercise. Although previ-
Accordingly, their findings did not always reflect a ously published systematic reviews have reported
lack of difference between non-weight-bearing differences in efficacy between various types of
strengthening exercise and aerobic exercise in pain exercise,9,19 no systematic reviews including meta-
relief in people with knee osteoarthritis. analyses have been performed that control for the
Our findings suggest that the advantage of effects of exercise frequency and exercise duration
strengthening exercise exists only for non-weight- of the efficacy of exercise. In our study, almost all
bearing regimens. Ettinger et al.35 showed that both of the included trials used frequencies of three exer-
resistance exercise and aerobic exercise reduced cise sessions per week for a duration of eight weeks.
pain significantly compared with health education However, it should be pointed out that we excluded
over the course of the study. However, they did not studies that were very effective but fell outside the
reveal an advantage of resistance exercise over aer- scope of our study because of their lower exercise
obic exercise. Our findings seem to be inconsistent frequency or longer term than that seen in other
with their results; this contradiction might be studies. Accordingly, we would like to emphasise
explained by the difference of exercise content in our interpretation that the advantage of non-weight–
the resistance group, which included weight-bear- bearing strengthening exercise for pain relief com-
ing strengthening exercise. As mentioned above, pared with weight-bearing strengthening or aerobic
exercise in weight-bearing positions for people with exercise would be restricted to the first eight weeks
painful knee osteoarthritis might aggravate pain.31,32 following the start of the exercise intervention with
Our results might suggest that exercise therapy over three sessions per week.
for pain relief should be performed under reduced Our study has several limitations; one of these is
loading conditions in people with knee osteoarthri- the quality of trials included in our study. The influ-
tis. However, persistent reductions in joint loading ence of the quality of trials on the results of meta-
may have harmful effects on cartilage metabolism. analysis has been reported.19 This meta-analysis
Hinterwimmer et al.40 showed that even during showed that all three study categories (low, medium,
short-term reduced loading conditions (partial or high risk of bias) achieved significant mean treat-
weight-bearing at the knee joint for seven weeks ment benefits in terms of pain, and that studies at a
after a surgical intervention at the ankle), a signifi- low risk of bias showed small mean treatment
cant degree of cartilage thinning was observed. effects for pain. The studies at a high risk of bias in
These data indicate that the cartilage undergoes this review included those that were not blinded to
some process of atrophy in the absence of mechani- outcome assessment, had no appropriate handling
cal stimulation. Furthermore, Jan et al.30 reported of participant attrition, and lacked adequate alloca-
that equally significant improvements were appar- tion concealment. Studies with a high risk of bias
ent on both the Western Ontario and McMaster showed mostly moderate mean treatment effects. In
Universities Arthritis Index function scale and mus- our study, we performed the meta-analysis on trials
cle torque after weight-bearing and non-weight- including those with low quality,22,25,27,28 which
bearing exercises, except for reposition error, for means a high risk of bias in past meta-analyses.19
which improvement was greater in the weight-bear- Therefore, the quality of trials may be considered to
ing exercise group. If people with knee osteoarthri- have been incorporated in our study.
tis performed non-weight-bearing exercises for pain Evidence of publication bias for all trials was not
relief, exercise with a loading condition at the knee detected by Egger’s test in our results. This result
joint should be performed immediately after suffi- suggests that the effect size of overall exercise ther-
cient pain relief is achieved in order to inhibit carti- apy has not been influenced by publication bias.
lage thinning and improve position sense. However, an estimation of publication bias for each
trials for conducting systematic reviews developed by Del- 29. Brismee JM, Paige RL, Chyu MC, et al. Group and home-
phi consensus. J Clin Epidemiol 1998; 51: 1235–1241. based tai chi in elderly subjects with knee osteoarthritis: a
17. Deeks JJ, Higgins JPT and Altman DG. Analysing data and randomized controlled trial. Clin Rehabil 2007; 21: 99–111.
undertaking meta-analyses. In: Higgins JPT and Green S 30. Jan MH, Lin CH, Lin YF, Lin JJ and Lin DH. Effects of
(eds) Cochrane Handbook for Systematic Reviews of Inter- weight-bearing versus nonweight-bearing exercise on func-
ventions. West Sussex: Wiley-Blackwell, 2008, pp.243–296. tion, walking speed, and position sense in participants with
18. Cohen J. Quantitative methods in psychology. Psychologi- knee osteoarthritis: a randomized controlled trial. Arch Phys
cal Bulletin 1992; 112: 155–159. Med Rehabil 2009; 90: 897–904.
19. Fransen M and McConnell S. Land-based exercise for 31. Baliunas AJ, Hurwitz DE, Ryals AB, et al. Increased knee
osteoarthritis of the knee: a metaanalysis of randomized joint loads during walking are present in subjects with
controlled trials. J Rheumatol 2009; 36: 1109–1117. knee osteoarthritis. Osteoarthritis Cartilage 2002; 10:
20. Egger M, Davey Smith G, Schneider M and Minder C. Bias 573–579.
in meta-analysis detected by a simple, graphical test. BMJ 32. Lucchinetti E, Adams CS, Horton WE, Jr. and Torzilli PA.
1997; 315: 629–634. Cartilage viability after repetitive loading: a preliminary
21. Borenstein M. Software for Publication Bias. In: Rothstein report. Osteoarthritis Cartilage 2002; 10: 71–81.
H, Sutton A and Borenstein M (eds) Publication Bias in 33. Hopman-Rock M and Westhoff MH. The effects of a health
Meta-analysis: Prevention, Assessment and Adjustments. educational and exercise program for older adults with
Hoboken (NJ): Wiley, 2005, pp.391–403. osteoarthritis for the hip or knee. J Rheumatol 2000; 27:
22. Gur H, Cakin N, Akova B, Okay E and Kucukoglu S. Con- 1947–1954.
centric versus combined concentric-eccentric isokinetic 34. Thomas KS, Muir KR, Doherty M, Jones AC, O’Reilly SC
training: effects on functional capacity and symptoms in and Bassey EJ. Home based exercise programme for knee
patients with osteoarthrosis of the knee. Arch Phys Med pain and knee osteoarthritis: randomised controlled trial.
Rehabil 2002; 83: 308–316. BMJ 2002; 325: 752.
23. Lin DH, Lin CH, Lin YF and Jan MH. Efficacy of 2 non- 35. Ettinger WH, Jr, Burns R, Messier SP, et al. A random-
weight-bearing interventions, proprioception training ver- ized trial comparing aerobic exercise and resistance exer-
sus strength training, for patients with knee osteoarthritis: a cise with a health education program in older adults with
randomized clinical trial. J Orthop Sports Phys Ther 2009; knee osteoarthritis. The Fitness Arthritis and Seniors Trial
39: 450–457. (FAST). JAMA 1997; 277: 25–31.
24. Salli A, Sahin N, Baskent A and Ugurlu H. The effect of 36. Baker KR, Nelson ME, Felson DT, Layne JE, Sarno R
two exercise programs on various functional outcome and Roubenoff R. The efficacy of home based progres-
measures in patients with osteoarthritis of the knee: A ran- sive strength training in older adults with knee osteoar-
domized controlled clinical trial. Isokinetics and Exercise thritis: a randomized controlled trial. J Rheumatol 2001;
Science 2010; 18: 201–209. 28: 1655–1665.
25. Schilke JM, Johnson GO, Housh TJ and O’Dell JR. Effects 37. Fransen M, Crosbie J and Edmonds J. Physical therapy is
of muscle-strength training on the functional status of effective for patients with osteoarthritis of the knee: a ran-
patients with osteoarthritis of the knee joint. Nurs Res 1996; domized controlled clinical trial. J Rheumatol 2001; 28:
45: 68–72. 156–164.
26. Jan MH, Lin JJ, Liau JJ, Lin YF and Lin DH. Investigation 38. Quilty B, Tucker M, Campbell R and Dieppe P. Physio-
of clinical effects of high- and low-resistance training for therapy, including quadriceps exercises and patellar taping,
patients with knee osteoarthritis: a randomized controlled for knee osteoarthritis with predominant patello-femoral
trial. Phys Ther 2008; 88: 427–436. joint involvement: randomized controlled trial. J Rheumatol
27. An B, Dai K, Zhu Z, et al. Baduanjin alleviates the symp- 2003; 30: 1311–1317.
toms of knee osteoarthritis. J Altern Complement Med 39. Petrella RJ and Bartha C. Home based exercise therapy for
2008; 14: 167–174. older patients with knee osteoarthritis: a randomized clini-
28. Kovar PA, Allegrante JP, MacKenzie CR, Peterson MG, cal trial. J Rheumatol 2000; 27: 2215–2221.
Gutin B and Charlson ME. Supervised fitness walking in 40. Hinterwimmer S, Krammer M, Krotz M, et al. Cartilage
patients with osteoarthritis of the knee. A randomized, con- atrophy in the knees of patients after seven weeks of par-
trolled trial. Ann Intern Med 1992; 116: 529–534. tial load bearing. Arthritis Rheum 2004; 50: 2516–2520.