Background

Osteoarthritis (OA) is the most prevalent rheumatic disease, causing degenerative changes in cartilage and the periarticular area [1]; the knee is the most commonly affected joint. The most prominent symptom of knee OA is pain, while reduction in quality of life (QOL), loss of physical function, and muscle weakness are also often associated [2, 3]. In addition, loss of productivity and personal economic strain are associated with ongoing care and disease management [4]. Therapeutic exercise is recommended in numerous guidelines as a nonpharmacologic treatment for knee OA [5, 6, 7].

Aquatic exercise, which utilizes the characteristics of water to promote health, has a long history and is becoming more popular. Motion against water resistance results in increased muscle tone, power development, and improved endurance [8]. Water also reduces weight bearing due to the property of buoyancy [9]. The heating effect of water temperature has been reported to ease soft tissue contracture, reduce pain, and relieve muscle spasms and fatigue [8, 10, 11]. Since aquatic exercise is easier on the body, the practice of exercise feels better and is perceived to be more enjoyable—and pleasurable exercise appears to improve QOL [12, 13].

Several previous systematic reviews have summarized the effects of aquatic exercise, but they either included mixed populations (e.g., including individuals with chronic diseases, such as hip OA) [14, 15, 16, 17] or were nonrandomized controlled trials [18]. However, existing clinical practice guidelines uniformly recommend aquatic exercise for treatment of knee OA [5, 19].

It is important to verify the evidence for aquatic exercise’s effect on improving physical function, QOL, and pain in individuals with knee OA. A systematic review of all randomized controlled trials (RCTs) to date would determine whether aquatic exercise is effective in improving outcomes in individuals with knee OA.

Methods

Criteria for considering studies for this review

Types of studies

Studies were eligible if they were RCTs.

Types of participants

Patients with primary knee OA were eligible. Diagnosis had to meet the classification criteria of the American College of Rheumatology [20, 21]. No further restrictions were made regarding disease duration or intensity.

Types of interventions

Studies that compared aquatic exercise to no treatment, usual care, or any other active treatment were eligible. All types of exercise developed in a therapeutic/heated indoor pool were eligible. Co-interventions other than exercise in a pool were allowed.

Types of outcome measures

According to the core set of outcome measures defined by Outcome Measures in Rheumatology Clinical Trials [22], studies were eligible if they assessed at least one of the following outcome measures: pain, physical function, or joint stiffness. If available, data on QOL and safety served as secondary outcome measures.

Search methods for identification of studies

Two authors (Meili Lu and Wenting Wang) independently completed a search of electronic databases. The following electronic databases were searched from their commencement through to June 2014: PubMed, the Cochrane Library, Embase, CAMbase, and the Web of Science. The literature search was conducted around search terms for aquatic exercise and knee OA, and adapted for each database as necessary. For PubMed, the search strategy was as follows: “(balneology [MeSH Terms] OR balneology [All Fields] OR balneotherapy [All Fields]) OR (hydrotherapy [MeSH Terms] OR hydrotherapy [All Fields] OR aquatic exercise [All Fields] OR pool exercise [All Fields] OR water exercise [All Fields]) AND (knee osteoarthritides [All Fields] OR knee osteoarthritis [All Fields] OR osteoarthritides of knees [All Fields] OR osteoarthritis of knees [All Fields] OR osteoarthritis, knee [MeSH Terms])”.

For further articles, the reference lists of articles were searched. There was no restriction on language.

Data collection and analysis

Selection of studies

After removal of duplicate records, two authors (Yingjie Zhang and Zhen He) independently screened the title, abstract, keywords, and publication type of all records obtained from the described searches. Disagreement or uncertainty was resolved by discussion with a third author (Youxin Su). The potentially eligible studies were obtained by hardcopy and read in detail, and those deemed eligible were included in the systematic review and meta-analysis. Studies in which characteristics were not clearly described or data were missing, the authors of the study were contacted for clarification.

Assessment of risk of bias in included studies

In order to ensure that variation was not caused by the study design or execution, risk of bias was assessed independently by two authors (Lu Sheng and Changyan Liu) using the Cochrane Handbook for Systematic Reviews of Interventions [23]. The domains recommended for assessing risk of bias in studies included selection bias, performance bias, detection bias, attrition bias, reporting bias, and other bias. Studies that met low risk of bias in all key domains were rated as having low risk of bias, those that met unclear risk of bias in one or more key domains were rated as having unclear risk of bias, and those that met high risk of bias in one or more key domains were rated as having high risk of bias. Where study data were inconclusive, trial authors were contacted for further details. Uncertainty or disagreement was resolved by discussion with a third author (Feiwen Liu).

Analyses and presentation

Studies were stratified in subgroups according to:

  1. 1.

    type of intervention (e.g., aerobic exercise, range of motion (ROM) exercise, strength exercise, and balance exercise),

  2. 2.

    duration of follow-up (e.g., at the end of treatment and 3, 6, and 12 months after treatment) and

  3. 3.

    primary outcome measures, such as pain and physical function.

In each group, the analysis was divided into aquatic exercise versus nonexercise, or aquatic exercise versus another active type of exercise.

Meta-analysis focused on outcome measures concerning improvement in the following aspects: pain, physical function, stiffness, QOL, and mental health.

Data extraction

Two authors (Yanan Li and Yiru Wang) independently extracted data on study characteristics, such as participants, interventions, cointerventions, control conditions, outcome measures, and results. Uncertainty or disagreement was resolved by discussion with a third author (Ziyi Zhang).

When choosing outcome measures for analysis, we decided on the following priority lists if more than one measured parameter in a category was present in the study:

  • The list of pain measures was as follows (in descending order): visual analog scale (VAS pain), Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC pain), Knee injury and Osteoarthritis Outcome Score (KOOS pain), Short Form-36 (SF-36 pain), Short Form-12 (SF-12 pain), and Arthritis Impact Measurement Scale (AIMS pain).

  • The list of physical function measures was as follows (in descending order): WOMAC function, Health Assessment Questionnaire (HAQ function), KOOS activities of daily living (ADL), SF-36 physical function, and Arthritis Self-Efficacy Questionnaire (ASEQ function).

  • The list of stiffness measures was as follows (in descending order): WOMAC stiffness, ROM, and KOOS stiffness.

  • The list of QOL measures was as follows (in descending order): SF-36 QOL, KOOS QOL, AIMS-2 affect, and Quality of Well-Being Scale (QWB).

  • The list of mental health measures was as follows (in descending order): SF-36 mental, SF-12 mental, AIMS-2 satisfaction, and ASEQ mental.

Measures of treatment effect

If at least two trials of comparable aquatic exercise protocols and outcome measures existed, meta-analysis was conducted using Review Manager 5.1 software (The Cochrane Collaboration, Software Update, Oxford) [24]. Standardized mean differences (SMD) with 95 % confidence intervals (CI) were calculated to assess intervention effects. Judgment of overall effect size was based on Cohen’s categories: SMD of 0.2–0.5 was considered a small effect, SMD of 0.5–0.8 a moderate effect, and SMD > 0.8 a large effect [25].

Grades of evidence were judged using criteria from the Cochrane Back Review Group as follows [26]: Strong evidence: consistent findings among multiple RCTs with low risk of bias; moderate evidence: consistent findings among multiple high-risk RCTs and/or one low-risk RCT; limited evidence: one RCT with high risk of bias; conflicting evidence: inconsistent findings among multiple RCTs; or no evidence: no RCTs.

Assessment of heterogeneity

Statistical heterogeneity between studies was tested by performing a χ-squared test. I2  > 25 %, I2  > 50 %, and I2  > 75 % were defined to indicate moderate, substantial, and considerable heterogeneity, respectively [23]. If the P-value of this test was < 0.1, an I2 test was performed. If the I2 test showed a value > 50 %, we considered this to indicate substantial heterogeneity and a random effects model was performed.

Results

Study selection

A total of 1048 papers were identified from the database searches, 255 of which were duplicates (Fig. 1). Of the remaining 793 papers, 760 were excluded based on title or abstract; therefore, 33 full-text articles were assessed for eligibility [8, 11, 13, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56]. Twenty-seven full-text articles were excluded because they involved mixed patient samples [8, 13, 33, 44, 46, 48, 49], were not RCTs [27, 28, 30, 35, 40, 45, 50, 55], had no clinical outcomes [31, 42, 53], did not involve exercise (only water immersion) [32, 39, 41, 47, 51, 52, 54], or were only a protocol [29] or abstract [34]. Six studies involving 398 participants were included in qualitative and quantitative analyses [11, 36, 37, 38, 43, 56].

Fig. 1
figure 1

Flowchart of the results of the literature search

Full size image

Study characteristics

Characteristics of the included studies, including samples, interventions, outcome measures, and results, are presented in Tab. 1.

Tab. 1 Characteristics of the included studies
Full size table

Setting and participant characteristics

Trials originated from the United States [56], Denmark [43], Taiwan [36], Brazil [11], Thailand [38], and Korea [37]. Patients were recruited from outpatient clinics [11, 37, 38, 43, 56] or local community centers [36]. Subjects were diagnosed according to criteria of the American College of Rheumatology (ARC) [11, 43] with clinical [11] and radiographic [11, 36, 37] confirmation of primary [38, 43] and moderate [38, 56] knee OA. In one study, knee pain ranging from 30 to 90 mm on a VAS [11] was also an inclusion criterion. Patients were excluded if they underwent arthroscopic surgery within 1 year [38], had inflammatory joint disease [37, 43, 56], had skin disease [11, 37], received knee joint replacement [36, 43], had received intra-articular corticosteroid injection in the past 30 days [36] or 3 months [11, 38], practiced regular physical activity [11, 36], or had received physical therapy intervention for their knee in the preceding 3 [38] or 6 months [11].

On average, patients were aged in their 60s or 70s, and the majority were female. Two studies reported adverse events [36, 43]. Data on ethnicity were available in two studies [36, 43].

Intervention characteristics

Aquatic exercise lasted 6 [38, 56], 8 [37, 43], 12 [36], or 18 weeks [11], with sessions offered two- [43] or three-times [11, 36, 37, 56] per week. Aquatic exercise included stretching [11, 38, 43], fast walking [37, 38], strengthening [11, 37, 43], and/or aerobic training [36, 37].

Control interventions included land-based exercise [11, 36, 37, 38, 43, 56], following treatment as usual [37], or home-based exercise [36].

Patients received nonsteroidal inflammatory drugs as a cointervention in two studies [11, 38].

Outcome measures

Pain was assessed in five studies; four used a VAS [11, 38, 43, 56] and one used the KOOS pain scale [36]. Physical function was assessed in three studies; two used the KOOS ADL scale [36, 43] and one used the WOMAC function scale. Stiffness was assessed in three studies; two used ROM and one used the WOMAC stiffness scale. Three studies measured QOL using the KOOS QOL scale [36, 38, 43].

Risk of bias in included studies

All studies had a high risk of bias due to nonblinding of participants (Fig. 2 and Fig. 3). All studies had a low risk of reporting bias and attrition bias. All studies also had a low risk of detection bias, with only Yennan et al. [38] not reporting further details regarding assessor blinding. The risk of selection bias was mixed; only one study reported allocation concealment [43] and four studies conducted random sequence generation [11, 36, 37, 43].

Fig. 2
figure 2

Risk of bias graph: review authors’ judgments on each risk of bias item presented as percentages across all included studies

Full size image
Fig. 3
figure 3

Risk of bias summary: review authors’ judgments on each risk of bias item for each included studies

Full size image

Effect of interventions

Aquatic exercise vs. land-based exercise: measurements immediately after exercise intervention

Meta-analysis revealed there was no significant effect on physical function (SMD 0.31; 95 % CI − 0.01–0.63), pain intensity (SMD − 0.25; 95 % CI − 0.74–0.24), stiffness (SMD − 0.15; 95 % CI − 0.47–0.17), or QOL (SMD 0.26; 95 % CI − 0.05–0.58) in favor of aquatic exercise immediately after the intervention (Fig. 4). Only one study reported mental health measurement and there was no evidence for an effect [37].

Fig. 4
figure 4figure 4

Forest plot for effects of aquatic exercise on pain, physical function, quality of life, and stiffness

Full size image

Aquatic exercise vs. land-based exercise: follow-up measurements

Only one study reported 3 months of follow-up measurements [43]. No evidence for physical function or QOL was found for aquatic exercise after 3 months of treatment. There was limited evidence for pain improvement based on follow-up measurements for land-based exercise when compared with aquatic exercise.

Aquatic exercise vs. nonexercise: measurements immediately after exercise intervention

There was moderate evidence for a moderate effect on physical function (SMD − 0.55; 95 % CI − 0.94 to − 0.16) in favor of aquatic exercise immediately after the intervention. No significant effect was found for pain (SMD − 1.16; 95 % CI − 3.03–0.71) or QOL (SMD − 0.21; 95 % CI − 0.59–0.18; Fig. 4). No evidence was found for stiffness [36] or mental health [37].

Aquatic exercise vs. nonexercise: follow-up measurements

Only one study reported 3 months of follow-up measurements [43]. No evidence for physical function or QOL was found for aquatic exercise after 3 months of treatment. There was limited evidence for pain improvement based on follow-up measurements for aquatic exercise when compared with nonexercise strategies.

Discussion

The purpose of this review was to evaluate the curative efficacy of aquatic exercise in patients with knee OA. Knee OA is more prevalent in the elderly, and is associated with large societal and economic burdens. According to previous reports, elderly Chinese women have a higher prevalence of knee OA than Caucasian women [57, 58]. Interventions for knee OA that can stop or slow disease progression are of great importance—from both economic and patient QOL-related viewpoints.

Because of the intervention properties of aquatic exercise, blinding of subjects and executors is impossible. The awareness of being treated may provide a bias when compared to a control group not exposed to treatment. Since both aquatic and land-based exercises in this review involved active treatment and attention from executors, one must assume there is no such bias effect. We identified all studies with a high risk of performance bias (Fig. 2 and Fig. 3) and found very few high-quality studies acceptable for meta-analysis. Furthermore, descriptions of adverse events and withdrawals were generally insufficient. All studies had more than 80 % attendance, however, which is very good for a therapy that demands out-of-house treatment several times a week.

At the end of treatment, meta-analysis revealed there was no significant difference in effects on physical function, pain intensity, stiffness, or QOL between aquatic and land-based exercises immediately after the interventions. The same observation was made for physical function and QOL based on follow-up measurements in one included study; however, pain improvement was superior with land-based exercise compared with aquatic exercise [43].

Aquatic exercise appears to have considerable short-term benefits compared with land-based exercise in patients with knee OA.

Our data are consistent with findings from another systematic review of RCTs of aquatic exercise for hip or knee OA, which was performed to identify function, mobility, and other health outcomes [14].

There was moderate evidence of a moderate effect on physical function (SMD − 0.55; 95 % CI − 0.94 to − 0.16) in favor of aquatic exercise immediately after the intervention, when comparing aquatic exercise with nonexercise [36, 43]. This result is consistent with findings from a previous systematic review [16]. We conducted a further meta-analysis between land-based exercise and nonexercise and their effect on physical function, and determined there was no significant effect (Fig. 4).

Based on these results, physicians may consider advising patients with knee OA to choose aquatic exercise to help maintain function.

When patients are unable to exercise on land, or find land-based exercise difficult, aquatic programs provide an enabling alternative strategy, especially in patients with greater disability. On the other hand, exercise on land may be arranged more easily and at a lower cost.

There was no evidence for stiffness, QOL, or mental health with aquatic exercise. The same observation was made for pain improvement with aquatic exercise immediately after the intervention when comparing aquatic exercise with nonexercise [36, 43]. This finding is different from two studies included in this review [11, 56], which reported aquatic exercise was better for pain reduction when compared with land-based exercise immediately after the interventions. The different pain measures used in these studies may partially explain the different results; a VAS was used to measure pain in the latter two studies. It is possible that the VAS was more sensitive to changes compared with the pain dimension of the KOOS or WOMAC. This may be seen in one study included in this review [43], which found no group differences in the pain dimension of the KOOS. Therefore, we suggest that future studies should use a VAS for pain measurement.

The effects on physical function did not last up to a 3-month follow-up according to the only study that reported such follow-up measurements [43]. There was limited evidence for pain improvement with aquatic exercise when comparing aquatic exercise with nonexercise. In addition, no evidence was found for physical function or QOL. Studies with long-term outcomes are necessary to determine further use of this therapy.

There is a lack of information on patient satisfaction and adherence to the exercise intervention, despite the importance of patient engagement in exercise programs. One potential limitation of the present meta-analysis is the relatively small number of included studies, which makes it difficult to draw firm conclusions. A second limitation is the substantial heterogeneity and variety of exercise strategies among studies. A third limitation is whether the positive effects of aquatic exercise might only be temporary. We are unable to clarify this issue, since the longest follow-up was only 3 months and was reported in only one study [43].

Conclusion

Overall, aquatic and land-based exercises appear to result in comparable benefits for participants. Meta-analysis did not provide confidence that either aquatic or land-based exercise provides greater improvements in physical function, QOL, or pain. Variability in study parameters, study quality, and exercise strategies may have confounded perception of the effects. Meanwhile, aquatic exercise has some short-term benefits compared with nonexercise. Studies in this area are still too few to provide further recommendations on how to apply this therapy. More research is required to determine if the positive effects of aquatic exercise can be supported by appropriately designed studies with medium- and long-term follow-ups.