Osteoarthritis and Cartilage Open 5 (2023) 100338

Contents lists available at ScienceDirect

Osteoarthritis and Cartilage Open

journal homepage: www.elsevier.com/journals/osteoarthritis-and-cartilage-open/2665-9131

Review

Effect of exercise therapy in patients with hip osteoarthritis: A systematic

review and cumulative meta-analysis
Carolien H. Teirlinck a, *, Arianne P. Verhagen a, b, Leontien M. van Ravesteyn b,
Elja A.E. Reijneveld-van de Vendel a, Jos Runhaar a, Marienke van Middelkoop a,
Manuela L. Ferreira c, Sita MA. Bierma-Zeinstra a
a
Dept. General Practice, Erasmus MC University Medical Center Rotterdam, the Netherlands
b
Discipline of Physiotherapy, Graduate School of Health, University of Technology Sydney, Australia
c
Sydney Musculoskeletal Health, The Kolling Institute, School of Health Sciences, Faculty of Medicine and Health, The University of Sydney, Australia

A R T I C L E I N F O A B S T R A C T

Handling Editor: Professor H Madry Objective: To evaluate the existing evidence on the effect of exercise therapy in patients with hip osteoarthritis
(OA) compared to no treatment and explore whether a further trial will change the current evidence.
Keywords: Design: Systematic review and cumulative meta-analysis using randomized controlled trials (RCT) to determine
Exercise therapy the effect on pain and function post-treatment, and at 6–9 months after treatment. Standardized mean difference
Osteoarthritis of the hip
(SMD)
0.37 was considered clinically worthwhile. Extended funnel plots were used to simulate the impact of a
Cumulative meta-analysis
new trial on the pooled effect size of pain and function.
Clinical relevance
Results: 18 RCTs were included. Post-treatment we found a beneficial effect of exercise therapy on pain (SMD
-0.38, 95% Confidence Interval (CI): 0.55 to 0.22) and function (SMD -0.31, 95% CI -0.49 to 0.11). A bene-
ficial effect of exercise therapy on pain (SMD -0.23, 95% CI: 0.41 to 0.05) and function (SMD -0.29, 95% CI:
0.45 to 0.12) was found 6–9 months after treatment. Most effect estimates were small, and it is unclear whether
these are clinically meaningful. Extended funnel plots and a simulation of a new trial showed that only a new trial
with a larger effect than the current pooled effect or a trial including 74,843 participants would change the pooled
effect estimate from an unclear to a clearly clinically worthwhile effect.
Conclusions: We found a beneficial effect of exercise therapy on pain and function in hip OA. It is unlikely a new
trial added to current evidence will change the conclusion.

1. Introduction available. Therefore, Cochrane reviews are updated every few years. In
the case of an inconclusive review, new trials can be of great value.
Exercise therapy is recommended as a conservative treatment of However, the beneficial effect of exercise therapy on pain and disability
osteoarthritis (OA) based on multiple meta-analyses and randomized in patients with hip OA was well established in 2014 [6]. Will new trials
trials [1–3]. Especially for knee osteoarthritis, the volume of evidence of be able to change the conclusion? For knee OA, we recently published a
its effectiveness is overwhelming [4]. A recent Cochrane review included cumulative meta-analysis of the effect of exercise therapy [7]. We
54 studies in their review of exercise therapy for knee OA [5]. Fewer showed that no further trials conducted comparing exercise therapy to
trials have been conducted for the effect of exercise therapy on hip OA; a minimal or no treatment would likely change the current conclusions.
Cochrane review included 10 studies, published in 2014 [6]. Despite this Although fewer trials for this comparison in hip OA have been conducted,
lower number of trials, they found a statistically significant and clinically the Cochrane review already stated in 2014 that further research is un-
relevant reduction in pain and disability immediately after treatment and likely to change the confidence in the estimate of effect but still new trials
these beneficial effects were still present 3–6 months after treatment. are conducted [6]. Evidently, we do not want to waste research resources
Ideally, we practice medicine based on the most current evidence or time spent on unnecessary trials. Moreover, it is unethical to

* Corresponding author. Carolien. H Teirlinck. Department of General Practice Na-19-02, Erasmus MC, P.O. Box 2040, 3000 CA Rotterdam, the Netherlands.
E-mail addresses: [email protected] (C.H. Teirlinck), [email protected] (A.P. Verhagen), [email protected] (L.M. van Ravesteyn),
[email protected] (E.A.E. Reijneveld-van de Vendel), [email protected] (J. Runhaar), [email protected] (M. van Middelkoop), manuela.
[email protected] (M.L. Ferreira), [email protected] (S.MA. Bierma-Zeinstra).

https://doi.org/10.1016/j.ocarto.2023.100338
Received 16 September 2022; Received in revised form 3 January 2023; Accepted 12 January 2023
2665-9131/© 2023 The Authors. Published by Elsevier Ltd on behalf of Osteoarthritis Research Society International (OARSI). This is an open access article under the
CC BY license (http://creativecommons.org/licenses/by/4.0/).
C.H. Teirlinck et al. Osteoarthritis and Cartilage Open 5 (2023) 100338

randomize patients to an ineffective treatment. A cumulative 2.4. Data extraction

meta-analysis and extended funnel plots provides more insight into
whether further trials are needed and if so, what sample size should be Data extraction was done by two review authors (CHT, LMvR or APV)
recommended. independently of each other using a standardized form. Disagreement
Therefore, our aim is to evaluate the existing evidence of the effect of was solved by consensus. The following data were collected: patient
exercise therapy compared to no or minimal treatment and to explore, by population (radiologic and/or clinical hip OA, OA severity), type of
using a cumulative meta-analysis, if more research is needed. intervention (land-based, water-based, individual or group treatment,
duration, and intensity), control group (usual care, education, no treat-
2. Methods ment), results (means and standard deviations) on pain and function
post-treatment and at 6–9 months after the intervention. Standard errors
Our department performed an update of the three Cochrane reviews or 95% confidence intervals (95% CI) were converted to standard de-
[5,6,8] on the effect of exercise therapy in patients with hip and knee OA, viations. If only change data were presented, these were extracted. If
commissioned by the National Health Care Institute of the Netherlands. multiple instruments were used to measure pain or function, we used the
This, to evaluate if exercise therapy for hip and knee OA patients should instrument that was used by most studies in the analysis. If a trial
be covered by the basic health insurance in the Netherlands. For the included hip and knee OA patients and no data for hip OA patients
current systematic review, we only used the studies on hip OA. separately were given, we contacted the first author to provide us the
data for the analysis. Alternatively, data provided in the Cochrane Re-
2.1. Main outcomes views were used.

Main outcomes were pain and function post-treatment and at 6–9 2.5. Data analysis
months after treatment.
Statistical pooling was done using Review Manager 5.3. We used a
2.2. Selection random-effect model assuming clinical heterogeneity. Outcomes were
presented as standardized mean differences (SMD) with a 95% CI. We
A literature search was conducted in Cochrane Central Register of used a SMD of 0.37 as the threshold for a ‘clinically worthwhile effect”
Controlled Trials (CENTRAL), MEDLINE, EMBASE, CINAHL, PEDro [9] where the negative value indicates outcomes in favor of exercise (less
(Physiotherapy Evidence Database) and Web of Science from the date of pain and better function). A funnel plot was created to observe possible
last search in the Cochrane reviews until September 2021. We used the publication bias. For the cumulative meta-analysis, we ordered the
same search terms as the Cochrane reviews. No limits were set for lan- studies by publication date and added each study for a pooled result.
guage. References were searched for relevant studies. The full search can Forest plots were created in SPSS version 24.
be found in the supplement. We used the GRADE approach to grade the quality (or certainty) of
We selected randomized controlled trials with the following charac- evidence of each outcome [10]. Quality of evidence was considered high
teristics: (P) adult patients (>18 years old) with clinical and/or radio- if at least 2 studies were included. Subsequently, this could be lowered to
logical hip osteoarthritis, (I) the intervention was an active form of moderate, low or very low quality of evidence if one or more of the
exercise therapy under supervision of a (physical) therapist, the inter- following occurred: study limitations (>25% of patients are from studies
vention was not part of a multidisciplinary or multimodal program and with an overall high risk of bias), inconsistency of effect (statistical
was evaluated as a standalone intervention, (C) the intervention in the heterogeneity I2 > 40% or < 75% of patients show result in the same
control group was usual care (like medication and/or education), no direction), indirectness (results are not suitable to extrapolate to the
treatment or waiting list, and (O) outcomes were pain and/or function target population according to expert authors JR and SMAB-Z), impre-
and were measured at short term (directly after end of treatment) and/or cision (<400 patients in the analysis), other like publication bias or ‘fatal
at long term (6–9 months after end of treatment). Studies evaluating flaw’ (for example selective loss of follow-up) [11–13].
interventions as hot packs, transcutaneous electrical nerve stimulation, Stata version 15.1 was used to create extended funnel plots for the
ultrasound or likewise were excluded. outcomes pain and function. In these funnel plots regions are calculated
Selection of studies was done by two authors independently of each to evaluate the influence of a new study on the overall effect estimate,
other (CHT, ERvdV or LMvR). First, titles and abstract were screened and calculated by the meta-analysis. These regions indicate how large a new
selected for full-text reading. Secondly, the full texts were screened for study and the effect estimate should be to significantly influence the
inclusion. In case of disagreement, a consensus was reached, or overall effect estimate [14,15]. The red region means that the new study
disagreement was solved by a third author (APV). added to the current pooled effect would generate a new overall effect
estimate and 95% CI that were clearly clinically worthwhile effect (new
2.3. Risk of bias assessment estimate and 95% CI
0.37). The blue region indicates that a study
added to the current pooled effect would generate a new overall effect
Two authors assessed risk of bias independently (ERvdV, CHT, LMvR, estimate and 95% CI that were not a clinically worthwhile effect (new
MvM or APV). A third author (MvM or APV) assessed risk of bias in case estimate and 95% CI > -0.37). The green region means a new study
of disagreement and no consensus. We used the same Cochrane risk of would yield a final pooled effect of uncertain clinical significance.
bias tools as was used in the original Cochrane reviews [5,6], in which on Finally, we simulated an extra cumulative meta-analysis, using the
7 domains a judgement of low, high or unclear risk was given. The 7 included trials and added a fictional new trial. The effect estimate of this
domains are random sequence generation, allocation concealment, new trial is equal to the current overall effect estimate of pain directly
blinding of participants and personnel, blinding of outcome assessment, after treatment of the included trials. Step by step, we raised the number
incomplete outcome data, selective outcome reporting and other bias. In of participants, until the new overall effect estimates reached the clearly
addition, studies were assigned an overall risk of bias. A study was clinically worthwhile SMD (upper limit of the 95% CI
0.37). In this
considered to have a low risk of bias if the randomization procedure was manner, we estimated the sample size of this fictional new trial that
done with a random sequence generation, proper allocation conceal- would ensure a clearly clinically worthwhile effect, given the current
ment, and intention-to-treat analysis was used. Studies were assigned a pooled effect.
high risk of bias if less than 3 domains were assigned a low risk of bias. All
other studies were assigned a moderate risk of bias [5,6,8].

2
C.H. Teirlinck et al. Osteoarthritis and Cartilage Open 5 (2023) 100338

3. Results removal of duplicates. The titles and abstracts of these references were
screened using the selection criteria and 297 references were selected for
3.1. Selection full-text reading. 28 references were included of which nine references
included patients with hip OA (Fig. 1). Three studies from the 12
The literature search and selection procedure were initially done for included studies in the Cochrane reviews were excluded for following
hip OA and knee OA together. For the present study, we only used the reasons: no data available separately for hip OA patients, only an abstract
references of hip OA. In total 4548 references were found after the was provided, exercise was not physical therapy but Tai Chi. In total nine

Fig. 1. Flowchart.

3
C.H. Teirlinck et al. Osteoarthritis and Cartilage Open 5 (2023) 100338

studies of the Cochrane reviews could be included in our present study Both effect estimates were regarded as unclear clinically worthwhile
[16–24], together with the nine references of the literature search effects.
[25–33], so in total 18 studies were included. The quality of evidence was moderate for function post treatment
(downgrading for inconsistency) and high for pain post treatment, pain,
3.2. Characteristics of included studies and function at 6–9 months after treatment (no downgrading).

Patient population. Number of patients per group ranged between 5 3.4. Extended funnel plots
and 102 patients. In 7 studies, the smallest group included less than 25
patients. In most trials, patients were diagnosed using the ACR criteria for We conducted extended funnel plots for pain and function post-
hip OA: clinical (n ¼ 8), radiological (n ¼ 3), clinical and radiological (n treatment (Fig. 6 and figure B in supplement). These plots show that a
¼ 3), or unclear (n ¼ 1) OA. new study need to have a large effect estimate to change the pooled effect
Interventions. All studies evaluated a land-based exercise, except for estimate (and 95% CI) from an unclear (green region) to a clearly clini-
one study [22], which evaluated aquatic exercises. The duration of a cally worthwhile effect estimate (red region).
treatment session varied from 30 to 120 min, frequency from 1 to 3 times To estimate how large the sample size of a new study therefore needs
a week and duration of the intervention from 5 to 16 weeks. Twelve to be, we simulated a cumulative meta-analysis by adding a fictional new
studies were group-based, and six studies were individual based exer- trial using the current overall effect estimate (0.38) of pain post-
cises. Control interventions were education (n ¼ 6), waiting list (n ¼ 5), treatment. We calculated that the new trial should include 74,843 par-
usual care by general practitioner (n ¼ 6) and no intervention (n ¼ 1). ticipants to change the overall effect estimate to a clearly clinically
Outcomes. Pain was measured with the following instruments: worthwhile effect.
Western Ontario and McMaster Universities Osteoarthritis Index
(WOMAC, n ¼ 6), Hip disability and Osteoarthritis Outcome Score 4. Discussion
(HOOS, n ¼ 4), Visual Analogue Scale (VAS, n ¼ 4), Numeric Rating Scale
(NRS, n ¼ 1), Brief Pain Inventory (BPI, n ¼ 1) and Impact of Rheumatic We found a clinically worthwhile effect on pain on short term and
diseases on General Health and Lifestyle (IRGL, n ¼ 1). Function was unclear clinically worthwhile effects on function on short term and pain
measured with the following instruments: WOMAC (n ¼ 8), HOOS (n ¼ and function on long term in our cumulative meta-analysis of exercise
4), IRGL (n ¼ 2), Disability Rating Index (DRI, n ¼ 1), Harris Hip Score (n therapy compared to no or minimal treatment for patients with hip OA.
¼ 1), Health Assessment Questionnaire (HAQ, n ¼ 1) and 6-min walking Although these effects estimates were already statistically significant
test (n ¼ 1). More characteristics of the included studies can be found in after the first study (pain short term) or after multiple studies, this effect
Table 1. is not yet a clearly clinically worthwhile effect. This is because the effect
estimates and 95% CI of pain and function are respectively under but
3.3. Risk of bias assessment close, and above our predetermined value of a clinically worthwhile ef-
fect (SMD
0.37). By simulating the effect of a new trial on current
The risk of bias of each domain of each study can be found in Table 2. evidence, we concluded only an unrealistically large trial would result in
Overall, 13 studies scored a low of risk of bias, two studies a moderate an overall pooled effect estimate and confidence interval of clinically
risk of bias and three studies scored a high risk of bias. Due to the nature significance. This means, that the studies done so far show us that ex-
of the intervention, none of the studies was able to blind their partici- ercise therapy in hip OA patients has a modest effect on pain and func-
pants, personnel, or outcome assessors (which were the participants for tion, possibly just clinically worthwhile. If we assume that this effect is
most outcomes). Therefore, all studies scored a high risk of bias on the the true effect, performing more studies to proof this to be clearly clini-
two items of blinding, even if we had little or no information on these cally worthwhile, will cost a lot of effort (and money) from researchers
items. Only one study [25] reported that patients did not have a treat- and patients while the value of this effort is questionable. Therefore, we
ment preference and was therefore scored as low risk of bias on the would consider not to perform new trials on the effect of exercise therapy
blinding items. on hip osteoarthritis (compared to no or minimal treatment) but instead,
focus on which type of exercises are most effective or which patients
3.3.1. Cumulative meta-analysis benefit most of exercise therapy. Earlier systematic reviews already
A funnel plot was created using function post-treatment as outcome, concluded that there is little evidence on moderators [34] or type of
because most studies reported this outcome (15 studies). The funnel plot exercises [35] on the effect of exercise therapy for hip OA. Recently
did not show apparent evidence of publication bias, see figure A in though, some of the results of a large individual participant data
supplement. meta-analysis on moderators of exercise therapy in knee and hip OA were
Post-treatment, 14 studies reported on pain and 15 studies reported presented and showed that patients with more pain or functional limi-
on function. We found a clinically worthwhile effect of exercise therapy tations at baseline respond slightly better to exercise therapy [36]. Un-
on pain (SMD -0.38, 95% CI: 0.55 to 0.22) and this effect was already fortunately, this analysis was not done for hip OA patients alone. A recent
statistically significant in the first study in 1998 (Fig. 2). The effect could study in a large cohort of patients with hip OA following a program with
not be classified as clearly clinically worthwhile since the 95% CI did education and supervised exercise found that patients with better mental
cross the threshold of SMD -0.37. Further studies showed that the di- well-being and fewer comorbidities are more likely to be a responder
rection of the effect estimate is consistent and only resulted in a smaller (improvement of pain 30% from baseline) to this program [37]. Also, in
and more precise effect estimate in the cumulative meta-analysis. Over- females they found that attending the education lecture and more su-
all, exercise therapy showed an unclear clinical worthwhile effect on pervised exercise made it more likely to respond to the program and that
function post-treatment (SMD -0.31, 95% CI -0.49 to 0.11), which females with symptoms at other joints were less likely to respond.
became statistically significant in 2014 (Fig. 4). Overall, there is not enough evidence to advice health care providers
Long-term outcome, six and seven studies respectively, reported on which hip OA patients benefit from which exercises, what duration, at
pain and function at 6–9 months after treatment. We found an overall what intensity and frequency.
effect on pain in favor of exercise therapy (SMD -0.23, 95% CI: 0.41 to
0.05) (Fig. 3), which became statistically significant in 2013. Exercise 4.1. Strength and limitations
therapy showed an effect on function (SMD -0.29, 95% CI: 0.45 to
0.12), and this effect became statistically significant in 2010 (Fig. 5). Our results are consistent with earlier systematic reviews [6,8,38],

4
C.H. Teirlinck et al. Osteoarthritis and Cartilage Open 5 (2023) 100338

Table 1
Characteristics of included studies.
Study Population Intervention Control Measurementsþ

V Baar 1998 Clinical ACR Exercise (n ¼ 35). Individual physiotherapy program (12 GP education þ medication if necessary After treatment:
[24] weeks, 1–3x/week, 30-min sessions) þ GP education þ (n ¼ 33). Function: IRGL
medication if necessary. Pain: VAS pain past week
Hopman-Rock ACR Exercise (n ¼ 11). Group sessions (6 weeks, 1x/week, Waiting list (n ¼ 13). After treatment:
2000 [20] 60-min classes) þ1x/week education. Function: IRGL-mobility
Pain: IRGL-pain
Foley 2003 Radiological ACR Exercise (n ¼ 6). Group sessions (6 weeks, 2x/week 30- Waiting list (n ¼ 12). Telephone call After treatment:
[18] min classes) every 2 weeks. Function: WOMAC
Pain: WOMAC
Stener- Radiological ACR Aquatic exercise (n ¼ 15). Group session (5 weeks, 2x/ Education (n ¼ 15). Two group meetings At 6 months after treatment
Victorin Patients on waiting list week, 30 min) þ education. Two group meetings lasting lasting 2 h each concerning hip anatomy, Function: Disability Rating
2004 [22] for hip replacement 2 h each concerning hip anatomy, disease process, and disease process, and advice on physical Index
advice on physical activities. activities. Pain: VAS
Tak 2005 [23] Clinical ACR Exercise (n ¼ 55). Group session (8 weeks, 1x/week GP care (n ¼ 54). After treatment:
strengthening þ home program, 60-min) þ education. Function: Harris hip score
Pain: VAS
Fernandes Radiological ACR and Exercise (n ¼ 55). Individually based (12 weeks, 2x/ Patient education (n ¼ 54). After treatment and at 6
2010 [17] symptoms (Harris Hip week) þ patient education months after treatment
Score 60–95) Function: WOMAC
Pain: WOMAC
Juhakoski Radiological and clinical Exercise (n ¼ 60). Group sessions (12 weeks,1x/week, GP-care (n ¼ 60). After treatment and at 9
2011 [21] ACR, K-L grade >1 45 min, þ4 booster sessions 1 year later) þ GP-care months after treatment:
Function: WOMAC
Pain: WOMAC
French 2013 Radiological and clinical Exercise (n ¼ 45). Individually provided ‘standardized’ Waiting list (n ¼ 43). After treatment:
[19] ACR exercise program (8 weeks, 6–8 sessions, 30-min) þ Function: WOMAC-PF
daily home exercise program (aerobic walking/cycling/ Pain: NRS
swimming 30 min)
Abbott 2013 Clinical ACR Exercise (n ¼ 22). Individually provided by GP care (n ¼ 23). At 8 months after
[16] physiotherapist, 50 min (9 weeks, 7 sessions þ2 booster treatment:
sessions week 16). Function: WOMAC
Pain: WOMAC
Villadsen Scheduled for hip Exercise (n ¼ 43). Group sessions of neuromuscular Education (n ¼ 41). Written information, After treatment:
2014 [32] replacement because of training (8 weeks, 2x/week, 60 min) þ education also on various exercises. Function: HOOS
symptomatic OA (written information, also on various exercises) Pain: HOOS
Kraus 2014 Clinical ACR Exercise (n ¼ 71). Group sessions (12 weeks, 1x/week, Control (n ¼ 69). No intervention. After treatment:
[28] 60–90 min, 2x/week home exercises, 30–40 min). Function: WOMAC
Pain: WOMAC
Svege 2013/ Radiological and clinical Exercise (n ¼ 55). Individually provided by Education (n ¼ 54). After treatment and at 6
2015 [30] ACR physiotherapist (12 weeks, 2–3x/week) þ education months after treatment
Function: WOMAC
Pain: WOMAC
Teirlinck 2016 Clinical ACR Exercise (n ¼ 101). Individual therapy (12 weeks, 12 GP care (n ¼ 102). After treatment and 6
[31] sessions, 3 booster sessions in 5th, 7th and 9th month) þ months after treatment:
GP care. Function: HOOS*
Pain: HOOS*
*scores are reversed
Hermann Scheduled for hip Exercise (n ¼ 40). Group sessions of pre-operative Usual care (n ¼ 40). After treatment:
2016 [27] replacement progressive explosive resistance training (10 weeks, 2x/ Function: HOOS
week, 60 min). Pain: HOOS
Saw 2016 [29] Waiting list for hip Exercise (n ¼ 14). Group sessions by physiotherapist (6 Usual care (n ¼ 16). After treatment and 6
replacement, weeks, 1x/week, 120 min) þ education. months after treatment:
radiological and clinical Function: Health Assessment
ACR Questionnaire - functional
disability index
Pain: Brief Pain Inventory
Bieler 2016 Clinical ACR, age> 60 Exercise (n ¼ 50). Group sessions, strengthening/ Counseling þ education (n ¼ 52). After treatment and 8
[25] resistance exercises (16 weeks, 3x/week, 60 min). months after treatment
Function:6MWT
Pain: no data available
Chopp-Hurley Clinical ACR Exercise (n ¼ 5). Group sessions (12 weeks, 3x/week, 60 Waiting list (n ¼ 5). After treatment:
2017 [26] min). Function: HOOS
Pain: HOOS
Thompson Radiological hip OA and Exercise (n ¼ 21). Groups sessions, strengthening/ Waiting list (n ¼ 10). After treatment:
2020 [33] pain and loss of function flexibility/endurance exercises (12 weeks, 3x per week, Function: WOMAC
60 min) Pain: VAS

þ In this table we indicate the measurements and time of measurements that we used in the results. Most studies mentioned more outcomes or times of measurements.
Abbreviations: GP ¼ general practitioner, ACR ¼ American College of Rheumatology, IRGL ¼ invloed van Reuma op Gezondheid en Leefwijze (Influence of rheumatic
diseases on Health and lifestyle), VAS ¼ visual analogue scale, WOMAC¼ Western Ontario and McMaster Universities Osteoarthritis Index, PF¼Physical Function
subscale, NRS ¼ numeric rating scale. HOOS¼ Hip disability and Osteoarthritis Outcome Score.

5
C.H. Teirlinck et al. Osteoarthritis and Cartilage Open 5 (2023) 100338

Table 2
Risk of bias assessment.
Study Random Allocation Blinding of Blinding of Incomplete Selective Other Overall risk
sequence concealment participants and outcome outcome data reporting bias of bias#
generation personnel assessment

V Baar 1998 [24] þ þ - - þ ? þ Low

Hopman-Rock 2000 [20] ? ? - - ? ? þ High
Foley 2003 [18] þ þ - - þ ? ? Low
Stener-Victorin 2004 þ ? - - - þ þ High
[22]
Tak 2005 [23] þ ? - - þ ? þ High
Fernandes 2010 [17] þ þ - - þ þ þ Low
Juhakoski 2011 [21] þ þ - - þ ? þ Low
French 2013 [19] þ þ - - þ þ þ Low
Abbott 2013 [16] þ þ - - þ þ ? Low
Villadsen 2014 [32] þ þ - - þ þ þ Low
Kraus 2014 [28] þ þ - - þ þ þ Low
Svege 2013/2015 [30] þ þ - - þ þ þ Low
Teirlinck 2016 [31] þ þ - - þ þ þ Low
Hermann 2016 [27] þ þ - - þ þ ? Low
Saw 2016 [29] þ ? - - ? þ þ Moderate
Bieler 2016 [25] þ þ þ þ þ þ þ Low
Chopp-Hurley 2017 [26] þ þ - - ? þ þ Moderate
Thompson 2021 [33] þ ? - - þ þ þ Low

# Low RoB: randomization appropriate þ concealed þ ITT analysis; high RoB: <3 items low risk; moderate RoB: all else.

Fig. 2. Cumulative meta-analysis on pain post-treatment, Footnote: each line represents the number of all participants and the pooled effect of the named study and
studies of lines above (cumulative). For example, the line Foley 2003, shows the pooled effect (SMD and 95% CI) and number of participants (N) of the studies: van
Baar 1998, Hopman-Rock 2000 and Foley 2003. The bottom line (Thompson 2020), is the pooled effect of all included studies.

but we are not aware of another cumulative meta-analysis on exercise lower in the exercise group) since pain at baseline is a possible moderator
therapy in patients with hip OA. Looking at the consistency within our of the effect of exercise [36].
research and our results, two studies were diverging. Firstly, one study Nonetheless, all other studies found an effect on both pain and
did not found an effect of exercise therapy on function at short term [24]. function and quality of evidence was considered high (with exception of
A possible explanation could be that the included patients had an early function short term). Furthermore, we followed the international
phase of OA, since they had complaints for less than 1 year. Secondly, recognized guidelines of Cochrane to perform this systematic review. By
another study did not found a difference in WOMAC pain and function adding the cumulative meta-analysis, we tried to give more insight in the
post-treatment [21]. At baseline the control group had a higher WOMAC effort already done by researchers and the value of adding more research.
pain score than the exercise group and, in addition, it seems that patients Other limitations of this review are the differences in intensity and
in the exercise group had more pain and worse function directly duration of the exercises between all studies, which makes it more
post-treatment than at baseline, while the control group had less pain and difficult to compare. In addition, in the included studies, blinding was not
better function (although not statistically significant). These findings can possible because of the intervention and patient reported outcomes.
possibly be explained by a low pain score at baseline (in both groups, but Thus, none of the studies blinded their participants and therefore all

6
C.H. Teirlinck et al. Osteoarthritis and Cartilage Open 5 (2023) 100338

Fig. 3. Cumulative meta-analysis on pain long-term, Footnote: each line represents the number of all participants and the pooled effect of the named study and studies
of lines above (cumulative). For example, the line Juhakoski 2011, shows the pooled effect (SMD and 95% CI) and number of participants (N) of the studies: Stener-
Victorin 2004, Fernandes 2010 and Juhakoski 2011. The bottom line (Saw 2016), is the pooled effect of all included studies.

Fig. 4. Cumulative meta-analysis on function post-treatment, Footnote: each line represents the number of all participants and the pooled effect of the named study
and studies of lines above (cumulative). For example, the line Foley 2003, shows the pooled effect (SMD and 95% CI) and number of participants (N) of the studies:
van Baar 1998, Hopman-Rock 2000 and Foley 2003. The bottom line (Thompson 2020), is the pooled effect of all included studies.

Fig. 5. Cumulative meta-analysis on function long-term, Footnote: each line represents the number of all participants and the pooled effect of the named study and
studies of lines above (cumulative). For example, the line Juhakoski 2011, shows the pooled effect (SMD and 95% CI) and number of participants (N) of the studies:
Stener-Victorin 2004, Fernandes 2010 and Juhakoski 2011. The bottom line (Bieler 2016), is the pooled effect of all included studies.

7
C.H. Teirlinck et al. Osteoarthritis and Cartilage Open 5 (2023) 100338

References

[1] R.R. Bannuru, M.C. Osani, E.E. Vaysbrot, N.K. Arden, K. Bennell, S.M.A. Bierma-
Zeinstra, et al., OARSI guidelines for the non-surgical management of knee, hip, and
polyarticular osteoarthritis, Osteoarthritis Cartilage 27 (2019) 1578–1589.
[2] L. Fernandes, K.B. Hagen, J.W. Bijlsma, O. Andreassen, P. Christensen,
P.G. Conaghan, et al., EULAR recommendations for the non-pharmacological core
management of hip and knee osteoarthritis, Ann. Rheum. Dis. 72 (2013)
1125–1135.
[3] S.L. Kolasinski, T. Neogi, M.C. Hochberg, C. Oatis, G. Guyatt, J. Block, et al., 2019
American College of rheumatology/arthritis foundation guideline for the
management of osteoarthritis of the hand, hip, and knee, Arthritis Care Res. 72
(2020) 149–162.
[4] W. Zhang, R.W. Moskowitz, G. Nuki, S. Abramson, R.D. Altman, N. Arden, et al.,
OARSI recommendations for the management of hip and knee osteoarthritis, part I:
critical appraisal of existing treatment guidelines and systematic review of current
research evidence, Osteoarthritis Cartilage 15 (2007) 981–1000.
[5] M. Fransen, S. McConnell, A.R. Harmer, M. Van der Esch, M. Simic, K.L. Bennell,
Exercise for osteoarthritis of the knee, Cochrane Database Syst. Rev. 1 (2015)
CD004376.
[6] M. Fransen, S. McConnell, G. Hernandez-Molina, S. Reichenbach, Exercise for
Osteoarthritis of the Hip, Cochrane Database Syst Rev, 2014, p. CD007912.
[7] A.P. Verhagen, M. Ferreira, E.A.E. Reijneveld-van de Vendel, C.H. Teirlinck,
Fig. 6. Extended funnel plot on pain post treatment. J. Runhaar, M. van Middelkoop, et al., Do we need another trial on exercise in
patients with knee osteoarthritis?: No new trials on exercise in knee OA,
Osteoarthritis Cartilage 27 (2019) 1266–1269.
studies scored high risk at this item in the risk of bias assessment. [8] E.M. Bartels, C.B. Juhl, R. Christensen, K.B. Hagen, B. Danneskiold-Samsøe,
H. Dagfinrud, et al., Aquatic exercise for the treatment of knee and hip
Nevertheless, we decided that one study could be scored a low risk on the osteoarthritis, Cochrane Database Syst. Rev. 3 (2016) CD005523.
blinding items because patients were asked about their treatment pref- [9] S. Wandel, P. Jüni, B. Tendal, E. Nüesch, P.M. Villiger, N.J. Welton, et al., Effects of
erence and did not report a treatment preference. This decision is glucosamine, chondroitin, or placebo in patients with osteoarthritis of hip or knee:
network meta-analysis, Br. Med. J. 341 (2010) c4675.
debatable, but since this study would also score an overall low risk bias [10] G.H. Guyatt, A.D. Oxman, G.E. Vist, R. Kunz, Y. Falck-Ytter, P. Alonso-Coello, et al.,
without this item, it did not affect the cumulative meta-analysis. GRADE: an emerging consensus on rating quality of evidence and strength of
recommendations, Br. Med. J. 336 (2008) 924–926.
[11] D.C. Bervoets, P.A. Luijsterburg, J.J. Alessie, M.J. Buijs, A.P. Verhagen, Massage
5. Conclusion therapy has short-term benefits for people with common musculoskeletal disorders
compared to no treatment: a systematic review, J. Physiother. 61 (2015) 106–116.
Exercise therapy for patients with hip OA is effective, but the effect is [12] M. van Middelkoop, S.M. Rubinstein, R. Ostelo, M.W. van Tulder, W. Peul,
B.W. Koes, et al., Surgery versus conservative care for neck pain: a systematic
small and not clearly clinically worthwhile. It is unrealistic that by per- review, Eur. Spine J. 22 (2013) 87–95.
forming more trials we can establish with certainty that the effect will [13] A.P. Verhagen, M. van Middelkoop, S.M. Rubinstein, R. Ostelo, W. Jacobs, W. Peul,
become clearly worthwhile. We therefore recommend future trials to et al., Effect of various kinds of cervical spinal surgery on clinical outcomes: a
systematic review and meta-analysis, Pain 154 (2013) 2388–2396.
focus on which patients benefit most of exercise therapy and/or what [14] M.L. Ferreira, R.D. Herbert, M.J. Crowther, A. Verhagen, A.J. Sutton, When is a
kind of exercise therapy is most effective. further clinical trial justified? Br. Med. J. 345 (2012), e5913.
[15] D. Langan, J.P. Higgins, W. Gregory, A.J. Sutton, Graphical augmentations to the
funnel plot assess the impact of additional evidence on a meta-analysis, J. Clin.
Author contributions Epidemiol. 65 (2012) 511–519.
[16] J.H. Abbott, M.C. Robertson, C. Chapple, D. Pinto, A.A. Wright, S. Leon de la Barra,
et al., Manual therapy, exercise therapy, or both, in addition to usual care, for
Concept, AV, SB.; methodology, AV, SB, CT, MF.; formal analysis, CT,
osteoarthritis of the hip or knee: a randomized controlled trial. 1: clinical
AV, ER, JR, MM, LR, SB, MF.; writing—original draft preparation, CT.; effectiveness, Osteoarthritis Cartilage 21 (2013) 525–534.
writing—review and editing, AV, ER, JR, MM, LR, SB, MF.; supervision, [17] L. Fernandes, K. Storheim, L. Sandvik, L. Nordsletten, M.A. Risberg, Efficacy of
AV, SB; All authors have read and agreed to the submitted version of the patient education and supervised exercise vs patient education alone in patients
with hip osteoarthritis: a single blind randomized clinical trial, Osteoarthritis
manuscript. Cartilage 18 (2010) 1237–1243.
[18] A. Foley, J. Halbert, T. Hewitt, M. Crotty, Does hydrotherapy improve strength and
physical function in patients with osteoarthritis–a randomised controlled trial
Funding comparing a gym based and a hydrotherapy based strengthening programme, Ann.
Rheum. Dis. 62 (2003) 1162–1167.
This research received no external funding. [19] H.P. French, T. Cusack, A. Brennan, A. Caffrey, R. Conroy, V. Cuddy, et al., Exercise
and manual physiotherapy arthritis research trial (EMPART) for osteoarthritis of the
hip: a multicenter randomized controlled trial, Arch. Phys. Med. Rehabil. 94 (2013)
Data statement 302–314.
[20] M. Hopman-Rock, M.H. Westhoff, The effects of a health educational and exercise
program for older adults with osteoarthritis for the hip or knee, J. Rheumatol. 27
All data is available on request. (2000) 1947–1954.
[21] R. Juhakoski, S. Tenhonen, A. Malmivaara, V. Kiviniemi, T. Anttonen, J.P. Arokoski,
A pragmatic randomized controlled study of the effectiveness and cost
Declaration of competing interest consequences of exercise therapy in hip osteoarthritis, Clin. Rehabil. 25 (2011)
370–383.
[22] E. Stener-Victorin, C. Kruse-Smidje, K. Jung, Comparison between electro-
Dr. Bierma-Zeinstra reports grants from Dutch Arthritis Society, the acupuncture and hydrotherapy, both in combination with patient education and
Netherlands Organisation for Health Research and Development, and EU, patient education alone, on the symptomatic treatment of osteoarthritis of the hip,
Clin. J. Pain 20 (2004) 179–185.
personal fees from Osteoarthritis & Cartilage, from Pfizer and Infirst [23] E. Tak, P. Staats, A. Van Hespen, M. Hopman-Rock, The effects of an exercise
health care, outside the submitted work. The other authors certify no program for older adults with osteoarthritis of the hip, J. Rheumatol. 32 (2005)
conflict of interest in connection with the submitted article. 1106–1113.
[24] M.E. van Baar, J. Dekker, R.A. Oostendorp, D. Bijl, T.B. Voorn, J.A. Lemmens, et al.,
The effectiveness of exercise therapy in patients with osteoarthritis of the hip or
Appendix A. Supplementary data knee: a randomized clinical trial, J. Rheumatol. 25 (1998) 2432–2439.
[25] T. Bieler, V. Siersma, S.P. Magnusson, M. Kjaer, H.E. Christensen, N. Beyer, In hip
osteoarthritis, Nordic Walking is superior to strength training and home-based
Supplementary data to this article can be found online at https exercise for improving function, Scand. J. Med. Sci. Sports 27 (2017) 873–886.
://doi.org/10.1016/j.ocarto.2023.100338.

8
C.H. Teirlinck et al. Osteoarthritis and Cartilage Open 5 (2023) 100338

[26] J.N. Chopp-Hurley, E.C. Brenneman, E.G. Wiebenga, B. Bulbrook, P.J. Keir, of the hip or knee: a secondary analysis from a randomized controlled trial,
M.R. Maly, Randomized controlled trial investigating the role of exercise in the J. Rheumatol. 41 (2014) 1385–1394.
workplace to improve work ability, performance, and patient-reported symptoms [33] A.R. Thompson, Z. Christopherson, L.M. Marshall, H.L. Carlson, N.L. Carlson,
among older workers with osteoarthritis, J. Occup. Environ. Med. 59 (2017) A pilot randomized controlled trial for aerobic and strengthening exercises on
550–556. physical function and pain for hip osteoarthritis, P & M (Philos. Med.) R 12 (2020)
[27] A. Hermann, A. Holsgaard-Larsen, B. Zerahn, S. Mejdahl, S. Overgaard, 229–237.
Preoperative progressive explosive-type resistance training is feasible and effective [34] J.G. Quicke, J. Runhaar, D.A. van der Windt, E.L. Healey, N.E. Foster, M.A. Holden,
in patients with hip osteoarthritis scheduled for total hip arthroplasty–a Moderators of the effects of therapeutic exercise for people with knee and hip
randomized controlled trial, Osteoarthritis Cartilage 24 (2016) 91–98. osteoarthritis: a systematic review of sub-group analyses from randomised
[28] I. Krauß, B. Steinhilber, G. Haupt, R. Miller, P. Martus, P. Janßen, Exercise therapy controlled trials, Osteoarthritis and Cartilage Open 2 (2020).
in hip osteoarthritis–a randomized controlled trial, Dtsch Arztebl Int 111 (2014) [35] S.-L. Goh, M.S.M. Persson, J. Stocks, Y. Hou, N.J. Welton, J. Lin, et al., Relative
592–599. efficacy of different exercises for pain, function, performance and quality of life in
[29] M.M. Saw, T. Kruger-Jakins, N. Edries, R. Parker, Significant improvements in pain knee and hip osteoarthritis: systematic review and network meta-analysis, Sports
after a six-week physiotherapist-led exercise and education intervention, in patients Med. 49 (2019) 743–761.
with osteoarthritis awaiting arthroplasty, in South Africa: a randomised controlled [36] M.A. Holden, M. Hattle, J. Runhaar, R. Riley, E. Healey, J. Quicke, et al.,
trial, BMC Muscoskel. Disord. 17 (2016) 236. Moderators of the effect of therapeutic exercise for people with knee and/or HIP
[30] I. Svege, L. Nordsletten, L. Fernandes, M.A. Risberg, Exercise therapy may postpone osteoarthritis: an individual participant data meta-analysis, Osteoarthritis Cartilage
total hip replacement surgery in patients with hip osteoarthritis: a long-term follow- 30 (2022) S400.
up of a randomised trial, Ann. Rheum. Dis. 74 (2015) 164–169. [37] A.V. Perruccio, E.M. Roos, S.T. Skou, D.T. Grønne, A.M. Davis, Factors influencing
[31] C.H. Teirlinck, P.A. Luijsterburg, J. Dekker, A.M. Bohnen, J.A. Verhaar, pain response following GLA:D® patient education and supervised exercise in males
M.A. Koopmanschap, et al., Effectiveness of exercise therapy added to general and females with hip osteoarthritis, Arthritis Care Res. (2022) (Published ahead of
practitioner care in patients with hip osteoarthritis: a pragmatic randomized print).
controlled trial, Osteoarthritis Cartilage 24 (2016) 82–90. [38] L. Beumer, J. Wong, S.J. Warden, J.L. Kemp, P. Foster, K.M. Crossley, Effects of
[32] A. Villadsen, S. Overgaard, A. Holsgaard-Larsen, R. Christensen, E.M. Roos, exercise and manual therapy on pain associated with hip osteoarthritis: a systematic
Immediate efficacy of neuromuscular exercise in patients with severe osteoarthritis review and meta-analysis, Br. J. Sports Med. 50 (2016) 458–463.