Multimodal Interventions Including Rehabilitation Exercise for Older Adults With Chronic Musculoskeletal Pain: A Systematic Review and Meta-analyses of Randomized Controlled Trials

Abstract

Background and Purpose: 

Musculoskeletal disorders (MSKDs) are the most common causes of disabilities for older adults. The aim of this systematic review and meta-analysis is to assess the effectiveness of multimodal interventions including exercise rehabilitation for older adults with chronic MSKDs.

Methods: 

A literature search was conducted up to February 2019 in 5 bibliographical databases to identify randomized controlled trials (RCTs) that compared multimodal interventions including exercise rehabilitation with usual medical care or no intervention. Randomized controlled trials were assessed with the Cochrane risk-of-bias tool. Meta-analyses were performed and pooled mean differences (MDs) or standardized mean differences (SMDs) were calculated.

Results: 

Sixteen RCTs (n = 2322 participants) were included. One RCT was considered at low risk of bias, 8 had some concerns of bias, and 7 had a high risk of bias. Participants suffered from hip or knee osteoarthritis (OA) (n = 12 RCTs), low back pain (LBP) (n = 2 RCTs) and generalized chronic pain (GCP) (n = 2 RCTs). Multimodal interventions were significantly more effective than usual care to decrease pain (visual analog scale, out of 10 points) in the short term, MD: −0.71 (95% confidence interval [CI] −1.08 to −0.34, n = 900), and in the long term: MD: −0.52 (95% CI −0.98 to −0.05, n = 575), but these differences are not considered clinically important. In terms of disabilities, multimodal interventions were also significantly more effective than usual care. The SMDs were −0.47 (95% CI −0.61 to −0.34, n = 903) and −0.29 (95% CI −0.46 to −0.13, n = 568) for OA trials in the short and long terms, respectively, and −0.47 (95% CI −0.81 to −0.12, n = 211) for LBP and GCP trials in the short term. The magnitude of these effects may be considered as small to moderate.

Conclusion: 

Multimodal intervention including exercise rehabilitation combined with usual medical care is an efficacious therapeutic option to reduce disabilities in older adults with chronic MSKDs. A significant but not clinically important effect was observed for pain. The most beneficial component of the multimodal interventions in terms of education, exercises, or medication remains to be determined.

CLINICAL HIGHLIGHTS

• Are multimodal (“exercise-plus”) interventions effective for pain and disability reduction in older adults with chronic musculoskeletal pain?
• Compared to control and usual care groups, multimodal exercise participants reported statistically significantly less pain and disability.
• The magnitudes of the improved pain outcomes were not clinically significant, while the disability outcomes were clinically significant.
• Multimodal intervention approaches that include exercise are likely to have a positive effect on pain and disability, with small to moderate effect sizes only for disability.

INTRODUCTION

The worldwide population is progressively aging, and this translates into an increased prevalence of chronic disorders.¹ Musculoskeletal disorders (MSKDs) affect hundreds of millions of people around the world and are the most common cause of disabilities in older adults.^2,3 According to a recent report published in June 2019 by the Canadian Government, approximately 1 in 3 Canadians, 65 years and older, report suffering from MSKD-related pain for more than 6 months.^4,5 The occurrence of MSKDs among this more vulnerable population is associated with impacts such as disabilities and loss of independence. It also leads to a decrease in physical activity, which is the most prevalent risk factor associated with functional decline and increased mortality.^6,7 Musculoskeletal disorders also have an impact on psychological and cognitive health, and are associated with the occurrence of anxiety, depression, sleep disruption, and increased social isolation.⁴ Therefore, MSKDs in older people can have a significant impact on their general health and overall quality of life.⁸ Furthermore, the economic impact of MSKD-related pain is considerable, notably because it is associated with an increased use of health care resources.⁹ Musculoskeletal disorders in older adults are also expected to rise substantially in the coming decades and therefore more efficient therapeutic options are needed to face this public health concern.¹⁰

The use of analgesic drugs is the most common strategy to manage pain associated with MSKDs.¹¹ One in 5 older Americans takes analgesic medications on a regular basis and opioids are often prescribed for severe MSKD pain.¹¹ However, studies suggest that opioid use is associated with only small reductions in pain and disability, especially for chronic MSKDs, while it puts patients at risk of addiction and even death.^12–15 Recent clinical guidelines highlight that best practice therapeutic approaches for the management of MSKD-related pain should combine pharmacological and nonpharmacological treatments such as exercise rehabilitation interventions and education while limiting opioids use to specifically selected persons refractory to other medications.^4,5

Several systematic reviews underlined the efficacy of physical therapy and exercises, pain education programs, and multidisciplinary biopsychosocial interventions for adults experiencing MSKD-related chronic pain.^16–20 Exercises have shown important benefits in reducing the severity of pain, and improving physical functioning and overall physical and mental health.¹⁶

Specific to older populations, systematic reviews have also been published about the efficacy of education and rehabilitation to manage chronic MSKDs. Two systematic reviews have demonstrated that self-management and education programs may be efficient in reducing pain and improving function for MSKD-related pain among older adults, and 2 other systematic reviews reported high-quality evidence in favor of exercise and physical activity to reduce pain among older adults with hip or knee osteoarthritis (OA).^9,21–23 But, to date, existing systematic reviews about nonpharmacological pain treatment among older adults focused either on self-management interventions or on specific disorders like OA.^9,21–24 Until now, no systematic review and meta-analysis has been conducted on the effectiveness of multimodal interventions, including exercise rehabilitation, education, and medical care regarding common MSKD-related pain at large, including OA, low back pain, and more generalized chronic pain, in older adults. Evidence of the efficacy of such multimodal interventions therefore remains to be determined. Consequently, the aim of this systematic review and meta-analysis is to review the evidence on the efficacy of multimodal interventions including rehabilitation exercise for older patients with chronic musculoskeletal pain.

METHODS

Protocol and Registration

This systematic review conforms to PRISMA methodological guidelines.^25,26 The protocol was registered on PROSPERO (number CRD42019133625, available at: https://www.crd.york.ac.uk/prospero/display_record.php?RecordID=133625).

Literature Search

An electronic bibliographic search was performed in 5 databases—MEDLINE, Embase, CINAHL, Cochrane Central, and PEDro—up to February 2019. Search strategy terms used were related to population, intervention, comparison groups, and outcomes, and are available in the Supplemental Material (available at: https://links.lww.com/JGPT/A45).

Study Selection

Two authors (A.K. and S.L.) independently reviewed titles and abstracts to identify trials of interest. Full texts were retrieved for all potentially relevant studies and were screened independently by the same authors for inclusion. Disagreements regarding inclusion were resolved via consensus. A third reviewer (F.D.) was available if consensus was not achieved by the 2 initial reviewers. A bibliographic search was also performed in the reference lists of the included articles and previous systematic reviews related to the current review.

Trials were included if they met the following inclusion criteria: (1) participants were adults with a mean age of 65 years or above with chronic musculoskeletal pain in any body site (for at least 3 months, according to the definition of chronic pain from the International Association for the Study of Pain)²⁷; (2) the study design was a randomized controlled trial (RCT); (3) the study evaluated a multimodal intervention including an active exercise rehabilitation program, and at least one other medical, educational or biopsychosocial intervention; (4) the intervention was compared with usual medical care including medication prescription or to no intervention; (5) at least one outcome measure was related to pain, disability, function, quality of life, safety, psychological function, health care services utilization, or patient satisfaction and (6) the trial was published either in English or in French.

Randomized controlled trials with participants suffering from OA were included even if the pain symptom duration was not presented since OA is considered a chronic condition. Trials were excluded if they included participants with cancer pain, dementia, mental health disorders, history of neurological or inflammatory disorders, pre- or postoperative interventions, or exercise as part of an educational program (with only demonstration and advice, but no active treatments offered to patients).

Data Extraction

Data extraction was conducted by one author (A.K.) using a standardized form and verified by a second author (S.L.). Extracted data included: the number of participants, their characteristics, the type of intervention, outcome measures, length of the follow-up, and results. When data were missing or incomplete, original authors were contacted. Outcomes were categorized into 3 groups according to the follow-up time after randomization: (1) short term: 6 to 12 weeks; (2) medium term: between 3 and 12 months; and (3) long term: 12 months or longer.

Risk of Bias in Individual Studies

The methodological quality of included RCTs was assessed by 2 independent evaluators (A.K. and S.L.) using the revised Cochrane risk-of-bias tool for RCTs (RoB2).²⁸ This tool is structured into a fixed set of domains of bias, focused on different aspects of trial design, conduct, and reporting. Within each domain, a series of questions (“signaling questions”) aim to elicit information about features of the trial that are relevant to risk of bias. A proposed judgment about the risk of bias arising from each domain is generated by an algorithm, based on answers to the signaling questions. Judgment can be “low” or “high” risk of bias, or can express “some concerns.” The final score was obtained through consensus. In cases of disagreement, a third reviewer (F.D.) was consulted to achieve a consensus.

Data Synthesis

Results from studies with similar comparators and outcome measures such as pain, function, and quality of life were considered for pooling into separate meta-analyses. Mean changes from baseline were pooled, and if changes from baseline standard deviation were not available, mean scores were pooled. Only meta-analyses without a significant degree of heterogeneity (χ²: P > .10 and I²< 60%) were kept and reported. Pooled mean differences (MDs) were calculated with 95% confidence intervals (95% CI) using Review Manager (RevMan 5.3, the Cochrane Collaboration, Copenhagen, Denmark). The α level was set to .05. When different scales were used for a similar outcome, standardized mean differences (SMDs) were calculated. The inverse variance method was used to weigh each study and was calculated with random effect. Visual inspection of the funnel plots was performed to assess publication bias. Funnel plots were not generated if the number of trials included in the meta-analysis was smaller than 5.²⁹ For studies not included in the meta-analyses, a qualitative analysis was performed. If necessary, data were imputed according to strategies recommended the Cochrane Collaboration. Clinical significance of pooled MDs was interpreted using relevant minimal clinically important difference (MCID) extracted from the literature. Standardized mean differences were interpreted regarding effect sizes: 0.2 was considered small, 0.5 moderate, and 0.8 large.

RESULTS

Literature Search and Included Studies

From the 77 potentially relevant articles identified through titles and abstracts review, 16 RCTs (including 2322 patients) met the eligibility criteria after full-text review.^30–44 Three other articles were secondary analyses of already included RCTs.^45–47 The study selection process is illustrated in Figure 1 and characteristics of included studies are presented in the Table.

Figure 1.: Flowchart of literature search results.

Table. - Characteristics of Included Studies

Study	Country	Participants	Intervention 1	N	Outcome	Follow-up	Main Results (95% CI)	Conflict of Interest	Risk of Bias
Abbott et al30	New Zealand	Hip or knee OA Symptom duration: 2.5 (1.3) to 2.9 (1.4) y Mean age: 66 (10.7)	Intervention 1: Supervised exercise programs including warm-up, muscle strengthening, muscle stretching, and neuromuscular control exercises (7 × 50 min within 9 wk, 2 × 50 min at wk 16) + home exercise program (3 times/wk) + usual medical care Intervention 2: Exercises + manual therapy delivered by physiotherapists + usual medical care Intervention 3: Manual therapy + usual medical care Control: Usual medical care offered by the GP and other health care provider of the participants	206 Intervention 1: 51 Control: 51	Pain (NRS) Disability (WOMAC total)	3 mo 6 mo 12 mo	Mean difference in score, exercise program and usual medical care (intervention 1) vs usual medical care at3 mo: -NRS: −1.22 (−2.03 to −0.41), P = .004 -WOMAC: 18.29 (3.91 to 32.67), P = .014 At 6 mo: -NRS: −0.94 (−1.80 to −0.08), P = .036 -WOMAC: 17.97 (2.81 to 33.13), P = .021 Mean difference in change from baseline, exercise program and usual medical care (intervention 1) vs usual medical care at 12 mo: -NRS: −0.80 (−1.69 to 0.09), P = .082 -WOMAC: 8.90 (−4.95 to 22.75), P = .213	None	Some concern
Bearne et al31	UK	Hip OA Mean age: 65-67 (range 52-78) Symptom duration: 4.4-5.6 y (range 1-40)	Intervention: Rehabilitation program of 75-min group exercise (strengthening and stretching exercises, cycling, resistance bands, functional and balance/coordination exercises), and self-management sessions (education, coping strategies, self-care, pain control, joint protection, lifestyle changes, handbook) (2 times/wk for 5 wk) + routine care Control: Routine management prescribed by the GPs	48 Intervention: 24 Control: 24	Pain (WOMAC pain) Disability (WOMAC function and WOMAC total)	6 wk 6 mo	Mean difference in score, rehabilitation program and routine medical care vs routine medical care at 6 wk: -WOMAC pain: 1.00 (−0.51 to 2.51), P = .201 -WOMAC function: 0.28 (−0.28 to 0.85), P = .322 -WOMAC total: 5.20 (−1.97 to 12.37), P = .162 At 6 mo: -WOMAC pain: −0.60 (−2.44 to 1.24), P = .526 -WOMAC function: 0.00 (−6.31 to 6.31), P = 1.00 -WOMAC total: 2.40 (−6.41 to 11.21), P = .596	No information	Some concern
Bergland et al32	Norway	History of vertebral fracture Symptom duration is not indicated Mean age: 70.8 (5.9)	Intervention: 60-min supervised exercise sessions (warm-up, functional, balance and coordination exercises, posture, trunk and chest exercises, stretching) (2 times/wk for 3 mo) and 3-h education (self-management with osteoporosis) Control: No treatment	89 Intervention: 47 Control: 42	Pain (Qualeffo pain) Disability (Qualeffo physical function)	3 mo 12 mo	Mean difference in change from baseline, exercises and education sessions vs no treatment at 3 mo: -Qualeffo (pain): −6.60 (−14.81 to 1.61), P = .123 - Qualeffo (physical function): −1.50 (−4.85 to 1.85), P = .392 At 12 mo: - Qualeffo pain: −11.80 (−19.77 to −3.83), P = .005 - Qualeffo (physical function): −1.50 (−3.98 to 0.98), P = .264	None	High risk
Cadmus et al33	US	Hip or knee OA Symptom duration is not indicated Mean age: 65.7 (5.9) to 66.0 (6.1)	Intervention: Community-based program with 45- to 60-min classes in pools (range of motion, strengthening, endurance exercises) (2-5 times/wk) + usual pain medication Control: Usual care	249	Pain (VAS) Disability (DISINX)	10 wk 20 wk	Results for pain and disability outcomes were not available	No information	High risk
Goode et al34	US	Chronic low back pain Symptom duration: >6 mo Mean age: 70.3 (4.9)	Intervention 1: Telephone-based intervention with specific exercises (stretching and strengthening, regular aerobic activity) and cognitive and behavioral therapy (managing pain associated with activity) and written and audio materials (activity pacing, breathing, relaxation, cognitive restructuring) (15-min calls every week for 10 wk) + usual pain medication Intervention 2: Telephone-based intervention including specific exercises only (without CBT) Control: Usual care	60 Intervention 1: 20 Control: 20	Disability (RMDQ)	3 mo	Mean difference in change from baseline, exercises and CBT (intervention 1) vs no treatment at 3 mo: -RMDQ: −4.10 (−8.13 to −0.07), P = .040	The authors completed the ICMJE Form for Disclosure of Potential Conflicts of Interest. S. Taylor reported employment/money received from the Department of Veterans Affairs, Health Services Research & Development.	High risk
Hay et al35	UK	Knee OA Symptom duration: >3mo for 78% Mean age: 67.9 (8.5) to 68.2 (8.0)	Intervention: 20-min sessions of individualized exercise program (general aerobic, muscle strengthening and stretching) and education (safety and importance of exercises, pacing, pain relief, coping strategies) (3-6 sessions/wk over 10 wk) + GP treatment on a standardized format Intervention 2: Pharmacy reviewing by an experienced pharmacist to optimize pharmacological pain control and GP treatment Control: GP treatment (advice on analgesia), information leaflet and one telephone call from a rheumatology nurse	325 Intervention 1: 109 Control: 108	Pain (NRS) Disability (WOMAC function)	3 mo 6 mo 12 mo	Mean difference in change from baseline, exercises and education vs usual medical care at 3 mo: -NRS: −0.86 (−1.46 to −0.26), P = .005 -WOMAC function: 3.99 (1.41 to 6.57), P = .003 At 6 mo: -NRS: −0.38 (−0.86 to 0.10), P = .131 -WOMAC function: 1.82 (−1.32 to 4.96), P = .258 At 12 mo: -NRS: −0.22 (−0.97 to 0.53), P = .563 -WOMAC function: 2.85 (−0.77 to 6.47), P = .125	None	Low risk
Hopman-Rock and Westhoff36	Netherlands	Hip or knee OA Symptom duration: >1 y for 97% Mean age: 65.3 (5.5)	Intervention: 2-h sessions including exercises (warm-up, dynamic and static strengthening) and education (pathophysiology, lifestyle, physical activity, pain management, weight reduction, ergonomic aspects) (1 session/wk over 6 wk) Control: No treatment	105 Intervention: 55 Control: 45	Pain (VAS) Disability (IRGL mobility)	6 wk 6 mo	Mean difference in score, exercise and education intervention vs no treatment at 6 wk: -VAS: 0.00 (−0.94 to 0.94), P = 1.000 -IRGL mobility: 1.50 (−0.70 to 3.70), P = .181 At 6 mo: -VAS: −0.32 (−1.14 to 0.50), P = .444 -IRGL mobility: 0.80 (−1.36 to 2.96), P = .466	No information	High risk
Hurleyet al37	UK	Knee OA Symptom duration: 5-7 y Mean age: 66-68 (range: 51-91)	Intervention: 12 supervised sessions that combined discussion regarding self-management, coping and individualized, progressive exercise regimen (2 times/wk for 6 wk) Control: Usual medical care by participants' physicians	418 Intervention: 278 Control: 140	Pain (WOMAC pain) Disability (WOMAC function and WOMAC total)	6 wk 6 mo	Mean difference in change from baseline, exercises and discussion vs usual medical care at 6 wk: -WOMAC pain: 1.72 (1.06 to 2.38), P < .001 - WOMAC total: 8.38 (5.68 to 11.08), P < .001 At 6 mo: - WOMAC pain: 1.04 (0.30 to 1.78), P = .007 -WOMAC total: 4.78 (1.51 to 8.05), P = .004	No information	Some concern
Kovar et al38	US	Knee OA Symptom duration: >4 mo Mean age: 68.5 (11.32)	Intervention: 90-min supervised fitness sessions (walking, stretching and strengthening) and education sessions (medical aspects of osteoarthritis, barriers and benefits of walking, supportive encouragement, guidebook) (24 sessions, 3 times/wk) + medical supervision by a physician member of the investigator team Control: Usual medical care	92 Intervention: 47 Control: 45	Pain and disability (AIMS)	8 wk	Mean difference in score, exercise and education intervention vs routine medical care at 8 wk: -AIMS pain: −1.00 (−1.79 to −0.21), P = .015 -AIMS physical activity: −2.22 (−3.25 to −1.19), P = .001	No information	High risk
Mat et al39	Malaysia	Knee OA Symptom duration is not indicated Mean age: 71.9 (5.06) to 76.3 (5.86)	Intervention: Multifactorial intervention including a home-based balance and strength retraining program (30-min sessions, 3 times/wk during 6 mo), fall education, cardiovascular and visual intervention, medical review Control: Usual medical care	46 Intervention: 22 Control: 24	Pain (KOOS pain) Disability (KOOS function)	6 mo	Mean difference in change from baseline, multifactorial intervention vs usual medical care at 6 mo: -KOOS pain: 8.18 (−4.43 to 20.79), P = .213 -KOOS function: 9.16 (−6.78 to 25.10), P = .269	No information	Some concern
Nicholas et al40,45	Australia	Back and leg pain: 38% Generalized pain: 45% Symptom duration: 178 (208) mo Mean age: 73.9 (6.5)	Intervention 1: 8 2-h supervised sessions (2 times/wk for 4 wk), based on psychological intervention (cognitive behavioral pain management) and exercises sessions (stretching, aerobic, strengthening, functional tasks repetitions of step-up and walking) + home exercises (charts, encouragement) + usual care Intervention 2: Same exercise sessions but only with group discussion, no psychological intervention Control: Usual care	141 Intervention 1: 49 Control: 39	Pain (NRS) Disability (RMDQ)	1 mo 2 mo 6 mo 12 mo	Mean difference in change from baseline, exercise and psychological sessions (intervention 1) vs no treatment: -NRS: −1.09 (−1.72 to −0.46), P = .001 -RMDQ: −2.40 (−3.89 to −0.91), P = .003 Results for control group were not available at 6 and 12 mo	None	Some concern
Saraboon et al41	Thailand	Knee OA Symptom duration is not indicated Mean age: 67.5 (7.32)	Intervention: Multimodal intervention including health education program, quadriceps muscle exercise program, home-visit program Control: Usual care. Participants were given the osteoarthritis knee booklet and a video compact disc	80 Intervention: 40 Control: 40	8 wk	Pain (NRS)	Mean difference in score, multimodal intervention including exercises vs no treatment group at 8 wk: -NRS: −4.48 (−5.19 to −3.77), P < .001	None	High risk
Skou et al42,46	Denmark	Knee OA Symptom duration: >6 mo for 90% of participants Mean age: 67.1 (9.1)	Intervention: Multimodal intervention including exercises (60-min sessions, 2 times/wk for 3 mo, neuromuscular training program followed by 8 wk to gradually accustom the patient to continue exercises at home), education (2 × 60-min sessions on disease characteristics, osteoarthritis pain, pain management during exercise), dietary advice (if BMI >25: 60-min sessions of dietary weight loss program based on motivational interviewing, with the aim of reducing body weight by 5%) + prescribed pain medication Control: Usual care group (standardized information leaflets about knee OA)	100 Intervention: 50 Control: 50	3 mo 12 mo	Pain (VAS) Disability (KOOS)	Mean difference in change from baseline, multimodal intervention including exercises vs usual care group at 3 mo: -VAS: −1.45 (−2.73 to −0.17), P = .029 -KOOS: 4.94 (−0.08 to 9.96), P = .057 At 1 y: -VAS: −1.29 (−2.35 to −0.23), P = .019 -KOOS: 11.10 (3.57 to 18.63), P = .005	None	Some concern
Tak et al43	Netherlands	Hip OA Symptom duration is not indicated Mean age: 67.4 (7.6) to 68.9	Intervention: 60-min supervised sessions (1 time/wk for 8 wk) of strength training (warm-up, fitness exercises with equipment, walking), information sessions (health aspect of knee OA) and dietary education (in relation with BMI) + home visit for individualized counseling about activity restrictions and home-exercise program Control: Usual medical care (self-initiated visit to own GP)	109 Intervention: 55 Control: 54	8 wk 5 mo	Pain (VAS) Disability (Harris Hip Score)	Mean difference in score, multimodal intervention including exercises vs usual medical care at 8 wk: -VAS: −0.50 (−1.56 to 0.56), P = .353 -HHS: 5.80 (1.13 to 10.47), P = .016 At 5 mo: -VAS: −1.60 (−2.60 to −0.60), P = .003 -HHS: 4.30 (−0.56 to 9.16), P = .087	No information	High risk
Teirlinck et al44 Tan et al47	Netherlands	Hip OA Symptom duration: 365 (810) Mean age: 64 (8.5) to 67 (9.6)	Intervention: Exercise therapy (12 × 30-min treatment sessions of strengthening and stretching, aerobic exercises for 3 mo and 3 booster sessions later), home exercise program (booklet), advices (lifestyle adaptations, walking aids, pain management) and usual GP care Control: Usual GP care and brochure about hip OA	203 Intervention: 101 Control: 102	6 wk 3 mo 6 mo 9 mo 12 mo	Pain (NRS) Disability (HOOS function)	Mean difference in change from baseline, exercise therapy and usual GP care vs usual GP care at 3 mo: -NRS: −0.09 (−0.68 to 0.50), P = .765 -HOOS function: 4.96 (1.20 to 8.72), P = .011 At 6 mo: -NRS: −0.08 (−0.76 to 0.60), P = .818 -HOOS function: 2.57 (−1.83 to 6.97), P = .254 At 12 mo: -NRS: −0.21 (−0.88 to 0.46), P = .537 -HOOS function: 4.02 (−1.14 to 9.18), P = .126	All authors have completed the Unified Competing Interest form at www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author).	Some concern
Tse et al48	China (Hong Kong)	Musculoskeletal pain Symptom duration: >3 mo Mean age: 76.5 (5.9)	Intervention: 8-wk program (90-min session, 1 time/wk) composed of pain education and physical exercises (45 min each, taught by classes within a community center) + home exercises (4 times/wk, 30-min sessions) Control: Usual care	56 Intervention: 31 Control: 25	8 wk	Pain (NRS) Disability (Elderly mobility Scale)	Mean difference in change from baseline, education and exercises program vs no treatment, at 8 wk: -NRS: −0.77 (−2.27 to 0.73), P = .290 -Elderly Mobility Scale: 1.61 (−0.06 to 3.28), P = .074	No information	Some concern

Abbreviations: AIMS, Arthritis Impact Measurement Scales; BMI, body mass index; CBT, cognitive behavioral therapy; CI, confidence interval; DISINX, 19-item Disability Index; GP, general practitioner; HHS, Harris Hip Score; HOOS, Hip Injury Osteoarthritis Outcome Score; IRGL, Rheumatic Disease on General Health and Lifestyle; KOOS, Knee Injury Osteoarthritis Outcome Score; NRS, Numeric Rating Scale; OA, osteoarthritis; RMDQ, Roland Morris Disability Questionnaire; VAS, visual analog scale; WOMAC, Western Ontario and McMaster Universities Arthritis Index.

The mean age range of participants was between 65 and 76 years. Mean duration of symptoms ranged from 3 months to 15 years. Twelve studies included participants with lower limb OA.^{30,31,33,35–39,41–44} Two studies included participants with low back pain.^32,34 Two studies included participants with generalized chronic pain.^40,48 Follow-up duration of the included studies ranged from 2 to 12 months.

Interventions

Multimodal interventions consisted of rehabilitation exercise combined with usual medical care,^30,33 with an education intervention^{32,36,37,41,43} or with both education intervention and usual medical care.^{31,34,35,38,40,44,48} Two other studies assessed exercise rehabilitation with a medical intervention, which consisted of specific prescriptions for analgesic medication.^39,42 The last 2 studies also included a diet program for OA participants.^41,42 Usual medical care consisted of nonstandardized care in 14 studies. It included self-initiated primary care physician consultation^{30,31,35,37–40,43,44} or only taking usual pain medication.^{33,34,41,42,44} For 2 studies, control groups experienced a no-treatment approach.^32,36

Main Outcome Measures

To assess pain, 6 RCTs used a Numeric Rating Scale (NRS) or a visual analog scale (VAS).^{33,36,40,41,43,48} Three RCTs used pain subscales of the Western Ontario and McMaster Universities Arthritis Index (WOMAC), the Knee Injury Osteoarthritis Outcome Score (KOOS), and the Hip Injury Osteoarthritis Outcome Score (HOOS).^31,37,39 Four RCTs used both an NRS and the WOMAC subscale.^{31,37,39,42,44} Two RCTs used different scales that were not pooled into meta-analyses: the pain subscale of the Arthritis Impact Measurement Scale,³⁸ and the Qualeffo pain subscale.³² One RCT did not include an assessment of pain.³⁴

Disability outcomes were assessed in 7 RCTs with the use of the WOMAC, KOOS or HOOS scales.^{30,31,35,37,39,42,44} Three RCTs evaluated back pain-related disability with the Roland Morris Disability Questionnaire^34,40 and the Qualeffo physical function subscale.³² Four studies used other scales including the Arthritis Impact Measurement Scale (AIMS),³⁸ the Influence of Rheumatic Disease on General Health and Lifestyle (IRGL),³⁶ the Harris Hip Score,⁴³ and the Elderly Mobility Scale.⁴⁸ These outcomes could not be pooled into meta-analyses. Two RCTs did not assess disability outcomes.^33,41

Risk of Bias in the Included Studies

Only one RCT was considered at low risk of bias by the Cochrane risk-of-bias tool v2.³⁵ Eight RCTs presented unclear risk of bias.^{30,31,37,39,40,42,44,48} Seven RCTs were considered at high risk of bias (Figure 2).^{32–34,36,38,41,43} Blinding of the participants and providers delivering the intervention was not possible in 15 trials due to the nature of the intervention, leading to an unclear risk of bias in the second domain of the tool (related to the effect of assignment to an intervention). One RCT however reported enough information to be considered at low risk of bias (2.3 in the Cochrane RoB tool v2) despite having a nonblinded intervention.³⁵ Four studies did not perform an intention-to-treat analysis.^33,34,36,39 Missing data were likely to influence outcomes in 1 RCT.³³ Significant differences between baseline scores were observed in 1 study.³⁹

Figure 2.: Detailed methodological assessment of included studies using the Cochrane risk-of-bias tool v2. + indicates low risk of bias; −, high risk of bias; ?, unclear risk of bias. This figure is available in color online (www.jgeript.org).

Effects of Interventions

Funnel plot analyses were performed for all meta-analyses, and they did not show substantial asymmetry that could suggest a publication bias. None of the meta-analyses was excluded due to substantial heterogeneity.

Primary Outcomes

Pain

Eight RCTs reported NRS or VAS scores for pain and were pooled into a meta-analysis (Figure 3). The MD shows a statistically significant effect in favor of a multimodal intervention of −0.71 out of 10 points (95% CI −1.08 to −0.34, n = 900) at 6 to 12 weeks, −0.58 (95% CI −1.06 to −0.10, n = 651) at 6 months, and −0.52 (95% CI −0.98 to −0.05, n = 575) at 12 months. However, the magnitudes of these differences are below MCID, of 1 to 2 points reduction for patients with chronic musculoskeletal pain as reported in the literature.^49,50

Figure 3.: Forest plots for NRS/VAS pain at 6 to 12 weeks, 3 to 6 months, and 1 year comparing multimodal versus control interventions. NRS indicates Numeric Rating Scale; VAS, visual analog scale.

Pain outcomes were also reported with the WOMAC, KOOS and HOOS subscales in 7 studies and pooled into a meta-analysis specific to OA studies. The SMD showed a statistically significant difference in favor of a multimodal intervention at each time point to reduce pain. The SMD was −0.42 (95% CI −0.55 to −0.30, n = 1013) at 6 to 12 weeks, −0.19 (95% CI −0.32 to −0.06, n = 949) at 6 months, and −0.22 (95% CI −0.41 to −0.03, n = 575) at 12 months (see Figure S1, available at: https://links.lww.com/JGPT/A45).

One other trial included 18- and 30-month follow-up outcomes. The results showed a statistically significant difference in favor of the multimodal intervention for the WOMAC pain subscale of −1.73 points (95% CI −2.22 to −1.24) at 18 months and −1.47 points (95% CI −1.99 to −0.95) at 30 months.³⁷ These results represent a change of respectively 8.6% and 7.4% of the maximal value. The MCID was estimated at 7.5% of the maximal value for comparable population and intervention.⁵¹

Disability

Seven RCTs provided data for disability outcomes using the WOMAC, KOOS, and HOOS function subscales (Figure 4). The SMD ranged from −0.47 (95% CI −0.61 to −0.34, n = 903) in the short term to −0.29 (95% CI −0.46 to −0.13, n = 568) in the long term. Four RCTs also reported the WOMAC, KOOS, and HOOS total scores at each time point. The pooled SMD ranged from −0.56 (95% CI −0.73 to −0.39, n = 1013) in the short term to −0.44 (95% CI −0.73 to −0.14, n = 575) and in the long term (see Figure S2, available at: https://links.lww.com/JGPT/A45). The difference was statistically significant in favor of the multimodal interventions at every time point.

Figure 4.: Forest plot for WOMAC/KOOS/HOOS function at 6 to 12 weeks, 3 to 6 months, and 1 year comparing multimodal versus control intervention. HOOS indicates Hip Injury Osteoarthritis Outcome Score; KOOS, Knee Injury Osteoarthritis Outcome Score; WOMAC, Western Ontario and McMaster Universities Arthritis Index.

One study included 18- and 30-month follow-up. The intervention group showed a statistically significant difference of 7.6% at 18 months and 6.9% at 30 months on the WOMAC total score compared with usual care.³⁷ The MCID was evaluated at 6.5% of the possible maximal value.⁵¹

For back pain, disability was assessed in 3 studies with the RMDQ and Qualeffo questionnaires and were pooled in a meta-analysis (n = 211) (Figure 5). The pooled effect size was −0.47 (95% CI −0.81 to −0.12) at a 6- to 12-week follow-up, in favor of the intervention group.

Figure 5.: Forest plot for back function at 6 to 12 weeks comparing multimodal versus control intervention.

Secondary Outcomes

Forest plots for all secondary outcomes are available in the Supplementary Material (available at: https://links.lww.com/JGPT/A45).

Functional performance

Eight RCTs evaluated functional performance with the Timed Up and Go (TUG) test in adults suffering from hip or knee OA or back pain. The overall effect estimate was −1.01 seconds (95% CI −1.55 to −0.48, n = 614) in the short term, −1.30 seconds (95% CI −2.18 to −0.43, n = 276) in the medium term, and −0.66 seconds (95% CI −1.06 to −0.26, n = 472) in the long term in favor of the multimodal interventions compared with usual care (see Figure S3, available at: https://links.lww.com/JGPT/A45). A reduction greater than or equal to 0.8 to 1.2 seconds in the TUG test is considered a clinically important improvement in patients with lower limb OA.⁵²

Functional performance was also assessed with the 20-m walking test in 3 studies. A statistically significant MD of −1.65 seconds (95% CI −2.52 to −0.78, n = 256) on the 20-m walking test was detected in the short term in favor of the multimodal interventions. In the long term, the MD was −0.88 seconds (95% CI −1.69 to −0.06, n = 189) (see Figure S4, available at: https://links.lww.com/JGPT/A45). This translates to an increased walking speed of approximately 0.08 m/s at short term, and 0.05 m/s at long term. A small but clinically meaningful change for a geriatric population has been established at 0.05 m/s.⁵³

Two other studies showed a statistically significant mean reduction of −3.86 cm (95% CI −5.61 to −2.12, n = 183) on the Functional Reach Test supporting the short-term effect of multimodal interventions including exercises for participants suffering from back pain (see Figure S5, available at: https://links.lww.com/JGPT/A45).^32,40

General health

Three studies reported on general health outcomes using the EuroQol 5 Dimensions (EQ-5D) questionnaire in the short, medium, and long terms. These analyses did not show a significant effect of multimodal interventions compared with usual care, as pooled SMDs were not statistically significant for the EQ-5D, 12-item Short Form Survey (SF-12), and Qualeffo-41 scales (see Figure S6, available at: https://links.lww.com/JGPT/A45).

Psychosocial Outcomes

A qualitative analysis was performed for psychosocial outcomes. Three RCTs assessed the effect of multimodal interventions on anxiety, using the Hospital Anxiety and Depression Scale (HADS). Only 1 RCT reported a statistically significant between-group difference at 6 months in favor of a multimodal intervention (P = .043).³⁷ The other 2 trials did not show any effect of multimodal intervention on anxiety.

Five studies evaluated the effect of multimodal interventions on depression among the participants, using various scales. None of these studies reported a significant difference from the control intervention.^{31,35,37,40,48} A statistically significant effect in favor of an active multimodal intervention on pain self-efficacy of the participants was however found in 3 RCTs (P < .05),^36,37,48 whereas 3 others did not find any significant effects.^31,35,40

Health care use

A qualitative analysis was performed. Three RCTs evaluated the impact of a multimodal program on the frequency of pain medication use. Kovar et al³⁸ reported that pain medication use was less frequent in the multimodal group than in the usual medical care group at a 2-month follow-up (P = .08). The RCT of Teirlinck et al⁴⁴ showed a statistically significant difference in pain medication use between the 2 approaches in favor of multimodal intervention at short term (P = .034). Skou et al⁴² did not find any significant difference between groups for pain medication use.

DISCUSSION

The aim of this systematic review and meta-analyses was to assess the efficacy of multimodal interventions including rehabilitation exercise for older patients with chronic musculoskeletal pain. Sixteen RCTs were included. Only 1 RCT was considered at low risk of bias, 8 RCTs had some concerns of bias, and 7 were considered at high risk of bias.

Main Findings and Implications for Practice

Based on these meta-analyses of RCTs of low to high risk of bias, multimodal intervention including exercise rehabilitation appears more effective to reduce pain and disabilities and to improve functional performance in older adults, compared with usual medical care or no treatment. For pain reduction, the magnitude of the effect may not be considered as clinically important.^54,55 However this result must likely be interpreted in the context of the target population in these trials: older adults with chronic symptoms. As a long duration of symptoms is a risk factor for poor prognosis, a modest improvement in symptom severity, especially in pain intensity, may still favorably impact the well-being of this population.⁵⁶ The heterogeneity of the included studies regarding pain location could also affect the effect we measured on pain reduction.⁵⁷

For disability, the magnitude of the effect is considered to be small to moderate. This effect is clinically important in the short and medium terms. Multimodal intervention including active rehabilitation can be, in this way, an appropriate therapeutic option to help older adults maintain their functional abilities and potentially improve their independence.

Regarding general health and psychosocial wellness, a multimodal intervention may or may not be effective to improve general health and reduce anxiety and depression among older adults suffering from chronic pain, compared with usual medical care or no treatment. The responsiveness to change of the EQ-5D scale has, however, been reported to be not as high as other more specific outcome measures for chronic pain populations, which could explain these results.⁵⁸ Regarding participants' pain self-efficacy, the effect was found to be statistically significant. That could suggest some benefit of a multimodal approach including exercises to improve confidence level of older adults suffering from chronic pain.

Results for medication use were not significant and further RCTs including economic analyses are warranted to formally conclude on the benefits of a multimodal approach to reduce health care resources use.

Finally, the general benefit of rehabilitation exercise is not restricted to the outcomes we assessed in this review. There is clear evidence that exercise can improve cardiovascular function, muscular strength, balance, and reduce the risk of falling on older adults, all of which are outcomes that were not addressed in this review.^59,60 Furthermore, active rehabilitation is also rarely associated with adverse events and can be, in this way, a suitable therapeutic option for older adults who are at risk of medication abuse.¹¹

Comparison With Previous Reviews

In a Cochrane overview of reviews published in 2017, Geneen et al¹⁶ reported a statistically significant impact of a physical activity and rehabilitation exercise intervention to improve physical function for adults with chronic pain, with small to moderate effect sizes. For psychological function and quality of life, results either were favorable to exercise, with small to moderate effect sizes, or showed no difference between groups. The results of this review are consistent with ours.

In another Cochrane review on multidisciplinary biopsychosocial rehabilitation for adults with chronic low back pain, the authors reported low- to moderate-quality evidence that multidisciplinary biopsychosocial rehabilitation is more effective than usual medical care to reduce pain and disability in the long term. Effect sizes were also moderate.¹⁹ Other previous reviews also evaluated the impact of multimodal interventions, including exercise rehabilitation and education, on older participants with OA of the knee or hip.^22,24 An active multimodal intervention was superior to usual care or no treatment to reduce pain and to improve function and physical health in older adults with MSKDs. Effect sizes ranged from low to moderate for all outcomes. The authors did not find any effect on psychological outcomes. Our results are in accordance with these previous reviews in term of improvement for disability but not for other outcomes. Differences are however found between these reviews and ours, including age, pathology and symptom duration of the included populations. That could explain the difference we found for several outcomes in our review, including the lack of effect for pain outcome.

Strengths and Limitations of the Review

Strengths of this review include the use of 5 databases, a thorough search strategy, a total of 16 included RCTs with large sample sizes in the meta-analyses, and the use of the validated Cochrane risk-of-bias tool (v2) to evaluate the quality of the evidence.

Nevertheless, our systematic review with meta-analyses also has limitations. First, one of our inclusion criteria was a mean age of at least 65 years, such that a proportion of participants were younger than 65 years in the included studies. We made this choice to avoid missing some studies of interest. Still, the vast majority of participants included in our review can be considered to be representative of a population of older adults. This review and meta-analysis included interventions consisted of exercise and education programs that were heterogeneous. The intensity, duration, frequency, settings, and nature of interventions varied. Therefore, the exact composition of an efficient intervention remains to be defined, even if evidence tends to support these multimodal approaches including rehabilitation exercise. Heterogeneity was also found in control interventions because of nonstandardized usual medical care. This may have affected the comparability of our studies, but is representative of the clinical reality. By including nonspecific interventions and populations, our review highlights global benefits of multimodal interventions for patients with OA, low back pain, and generalized chronic pain.

Future RCTs could be conducted including patients older than 65 years suffering from back pain and generalized chronic pain. Standardized medical care could even be used as control intervention. Moreover, complementary interventions promoting adherence, enhancement of social support, and encouraging integration of physical activity into daily life for older adults could also be proposed to emphasize the effect of exercise and education programs, as part of a supporting multimodal approach.

CONCLUSION

Choosing a multimodal approach including rehabilitation exercise over usual medical care for older adults with chronic musculoskeletal pain is likely to result in a positive effect on pain and disability, with small to moderate effect sizes only for disability. The benefits on quality of life, psychosocial function, and health care use need to be further investigated. The exact components of exercise programs remains to be determined.