A search strategy was developed by iterative exploration of 18 databases using a variety of search terms that emphasized sensitivity (high recall) over specificity (precision).¹⁵ In the final search strategy, studies were sought from 14 databases, including Medline, AMED, EMBASE, SportDiscus, PEDro, CINAHL, ASSIA, and the Cochrane Library. Over the period January 1980 to June 2006, we used a range of search terms based around the concepts of aquatic exercise (hydrother$, balneo$, aquarobic$, aquatic rehab$, aqua$ and exercise, water and gymnast$, water aerobic$) and pain using the International Association for the Study of Pain terminology.¹⁶ Reference and bibliographic lists of retrieved articles and relevant reviews were also examined. The search was limited to English-language works and those that studied adults (people aged ≥18y).

A 3-stage process was used to select studies for inclusion in the review. In the first stage, the title of each identified article was checked against predetermined criteria by 2 reviewers using a standard form and coding sheet. Abstracts and full articles were similarly reviewed in the next 2 stages. In general, a policy of inclusion was adopted—that is, in the absence of any information to the contrary, each article was forwarded to the next stage of the screening process. Criteria for inclusion were titles, abstracts, and/or articles that suggested some aspect of aquatic exercise as defined in adults with neurologic or musculoskeletal pathology and with pain as an outcome measure. In addition, included studies were limited to full reports of RCTs that examined the effectiveness of aquatic exercise compared with no treatment or other interventions, such as land-based exercise or immersion, in adult patients (≥18y) with any neurologic or musculoskeletal pathology and in which at least 1 outcome measure of “subjective pain experience captured by ratings of pain intensity, sensation, and unpleasantness” was reported.¹⁷ Studies that considered the prevention of pain in healthy conditions (eg, pregnancy) or which incorporated additional interventions (eg, education) were excluded from the review.

Selected studies were subject to unmasked quality assessment¹⁸ by 2 reviewers using the criteria for RCTs recommended by the Scottish Intercollegiate Guidelines Network (SIGN 50).¹⁹ An overall assessment of the 9 questions provides a bias rating of low (++), moderate (+), or high (−). A low bias rating indicates a high-quality study in which all or most of the 9 criteria have been fulfilled and where they have not been fulfilled the conclusions of the study are thought unlikely to alter. One study was authored by a reviewer and was therefore assessed by another independent reviewer.

Seven hundred ninety-three publications were identified and screened against the inclusion and exclusion criteria. Of these publications, 729 were rejected at the title and abstract stage (fig 1). Sixty-nine studies proceeded to the paper screening stage, and 19 of these were accepted for review.^{23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41} Of the 19, 5 studies were of sufficient quality and had adequate raw data to be entered into meta-analyses under one of 2 comparisons: aquatic exercise versus no-treatment controls and aquatic versus dry land exercise. The study selection and validity assessment process was undertaken by 2 independent reviewers. Where anomalies existed, discussion between the reviewers enabled consensus to be achieved.

Fig 1

Trial flow.

 

Nine studies compared aquatic exercise with no-treatment or waiting-list control groups,^{26, 28, 30, 32, 33, 35, 38, 39, 41} 10 compared it with land exercise,^{23, 24, 25, 27, 29, 31, 34, 35, 37, 40} and 2 with immersion^{27, 36} (table 1). Two studies incorporated more than 2 treatment arms.^{27, 35}

Within the 19 studies, 717 patients participated in hydrotherapy, with an average age of 56.0±11.3 years (range, 25–81y) and, based on available data, an overall men-to-women ratio of 1:3 (not all studies reported this). The number of subjects randomized to the aquatic exercise arm ranged from 7 to 153, with 1 study omitting to report how many patients were randomized to the different treatment arms.³⁴ Most patients presented with rheumatology conditions, and only Sutherland et al³³ involved patients with neurologic problems. Of the rheumatology articles, 4 considered fibromyalgia,^{31, 36, 40, 41} 3 chronic low back pain,^{29, 30, 37} and 11 osteoarthritis or rheumatoid arthritis.^{23, 24, 25, 26, 27, 28, 32, 35, 38, 39} The duration of symptoms varied from 2.75 to 24.00 years with an average of 9.98±5.50 years.

Drop-out rates for aquatic exercise patients ranged from 0% to 27.4%. Of the 3 studies with drop-out rates exceeding 20% (SIGN criterion), the time from baseline to posttest varied from 6 to 52 weeks. Extraneous causes such as comorbidity, time issues, or travel difficulties were responsible rather than the deleterious side effects of treatment. Three authors reported that a few patients (n=5) experienced an exacerbation of symptoms, but from the information provided, the treatment group affiliation of these patients is difficult to determine.^{25, 29, 35}

Thirteen articles reported the aims of the aquatic exercise intervention and details, however brief, of the specific activities were mentioned in all. By categorizing the activities, it was possible to identify 3 types of exercise program: a general exercise program that included elements of muscle strengthening, increasing range of movement, and functional activities (n=9)^{24, 25, 26, 27, 29, 30, 32, 34, 36}; aerobic exercise (n=9)^{23, 28, 31, 33, 37, 38, 39, 40, 41}; and strengthening exercise (n=1).³⁵ Of the 9 articles in which aerobic exercise was reported, 5 based the exercise intensity on heart rate.^{23, 28, 31, 40, 41}

The intervention setting in 12 studies indicated that hospital or clinic pools were used. Three used community or public swimming pools, and 4 omitted to report the treatment setting. The overall duration of aquatic exercise treatment ranged from 6 weeks or less^{27, 30} to more than 12 weeks,^{28, 31, 32, 39, 40} with an average of 33.25±19.20 sessions. Treatments took place on an outpatient basis 2 to 3 times a week for a minimum of 30 and a maximum of 75 minutes (mean, 50.7±12.2min); 3 studies did not report session duration.^{25, 34, 37} Total treatment time ranged from 4 to 84 hours (mean, 25.5±20.54; median, 22.5). Where reported, treatment was performed in groups of 4 to 24 patients supervised by physiotherapists (n=8) or trained instructors (n=5) and using programs designed by a physiotherapist or fitness professional (n=5). In 11 studies, the average temperature of the water was 32.4°±2.7°C (range, 28°–36°C). Only 4 studies reported the depth of the water,^{23, 25, 34, 41} which was between waist and chest height.

Pain outcomes were measured before and after intervention in each reviewed study; follow-up data were reported in 8 studies.^{23, 25, 27, 31, 36, 38, 39, 41} A variety of instruments were used to measure sensory pain, with the 10-cm visual analog scale (VAS) being the most common.^{24, 25, 26, 28, 29, 31, 34, 36, 37, 40, 41} Other instruments included the McGill Pain Questionnaire^{27, 30} and pain subscales from a variety of self-report questionnaires (Arthritis Impact Measurement Scale,²³ Health Assessment Questionnaire,³² health-related quality of life,³³ Western Ontario and McMaster Universities Osteoarthritis Index [WOMAC],^{35, 39} 36-Item Short-Form Health Survey³⁸). Only 3 studies included pain as a primary outcome measure,^{39, 40, 41} and 2 of these had used this pain outcome in a power calculation for sample size.

Five studies were judged to be at low risk of bias using the SIGN criteria and are therefore judged as high quality.^{27, 32, 35, 39, 40} Three studies were of moderate^{26, 31, 38} and 11 studies^{23, 24, 25, 28, 29, 30, 33, 34, 36, 37, 41} of low quality (table 2). Although all studies were described as randomized, most (91%) of those that had a high level of bias failed to meet this SIGN criterion. Other criteria that were frequently inadequate included randomization concealment and intention-to-treat analysis (both n=14 [73.6%]). Although a priori power calculations are not included within the SIGN checklist we noted that only 6 studies^{27, 30, 32, 35, 39, 40} included any sample size power calculations and that 3 of these^{35, 39, 40} were based on pain outcomes. Five studies^{23, 27, 30, 32, 39} were independently powered to detect either a clinically meaningful or a conventionally large effect size for pain.^{42, 43, 44}

Table 2Quality Assessment of the 19 Studies Included in the Review Using Modified SIGN Criteria for RCTs

Study	Clear Question	Acceptable Randomization Method	Adequate Concealment Method	Blinding of Assessors	Groups Similar at Baseline	Only Difference Between Groups Is Aquatic Exercise Intervention	Outcomes Measurement: Standard, Valid, and Reliable?	Percentage of Drop-Outs Before Posttest	ITT	Overall Bias Rating
Ahern et al26	WC	WC	NA	AA	PA	NR	AA	NA	NA	+
Altan et al36	WC	PA	NA	WC	WC	PA	WC	Gp1=4 Gp2=12	NA	–
Assis et al40	WC	WC	WC	WC	WC	NA	WC	Gp1=13.3 Gp2=14.8	WC	++
Bilberg et al38	AA	AA	AA	AA	WC	NR	AA	Gp1=0 Gp2=0	NA	+
Cochrane et al39	WC	WC	WC	WC	WC	WC	WC	Gp1=27.4 Gp2=24.5	WC	++
Foley et al35	WC	WC	WC	AA	WC	AA	WC	Gp1=20 Gp2=25.7 Gp3=8.6	WC	++
Green et al25	AA	PA	NA	PA	PA	PA	PA	Overall=25	NA	–
Gusi et al41	AA	NR	NA	NA	AA	NA	AA	Gp1=5.5 Gp2=0	NR	–
Hall et al27	WC	WC	AA	AA	WC	AA	WC	Overall=6	NA	++
Jentoft et al31	WC	AA	NA	WC	WC	WC	WC	Gp1=18 Gp2=27	NA	+
McIlveen and Robertson30	AA	PA	NA	AA	PA	AA	WC	Gp1=19.6 Gp2=5.6	NA	–
Minor et al23	AA	NR	NA	NA	WC	WC	AA	Gp1=14.9 Gp2=22 Gp3=12.5	NA	–
Patrick et al32	WC	WC	NA	NApp⁎	WC	WC	WC	Gp1=16.8 Gp2=3.0	WC	++
Rintala et al28	WC	AA	NA	NA	AA	AA	AA	Gp1=5.5 Gp2=18.7	NA	–
Sjogren et al29	AA	PA	NA	AA	AA	AA	AA	Gp1=6.7 Gp2=6.7	NA	–
Sutherland et al33	AA	NR	NA	PA	AA	AA	AA	NA	WC	–
Sylvester24	AA	NR	NA	AA	AA	AA	AA	Gp1=0 Gp2=0	NA	–
Wyatt et al34	AA	NR	NA	AA	NA	AA	PA	Overall=8.7	NA	–
Yozbatiran et al37	WC	NA	NA	NA	AA	PA	AA	NA	NA	–

Abbreviations: AA, adequately addressed; Gp, group; ITT, intention-to-treat analysis (all subjects are analyzed in the groups to which they were randomly allocated); NA, not addressed; NApp, not applicable; NR, not reported; PA, poorly addressed; WC, well covered.

Legends: +, some of the criteria have been fulfilled. Where they have not been fulfilled the conclusions of the study are thought unlikely to alter; ++, all or most of the criteria have been fulfilled. Where they have not been fulfilled the conclusions of the study are thought very unlikely to alter; −, few or no criteria fulfilled. The conclusions of the study are thought likely or very likely to alter.

The outcomes consisted of self-administered postal questionnaires only.

Of the 9 studies in this comparison,^{26, 28, 30, 32, 33, 35, 38, 39, 41} 5 were of moderate to high quality,^{26, 32, 35, 38, 39} and of those, 3 had data suitable for incorporation into a meta-analysis^{32, 38, 39} (fig 2A). This showed a small posttreatment effect in favor of aquatic exercise (P=.04; SMD=−.17; 95% CI, −.33 to −.01) (see fig 2A). Removing the smallest study³⁸ (n=22) from the analysis altered the SMD slightly from −.17 to −.19 (95% CI, −.35 to −.02) and the level of significance to P equal to .03. Statistical tests for heterogeneity were not significant. Qualitative analysis, based on review of results reported by researchers showed a lack of consistency between study results, whether of high or low quality. In summary, the evidence shows a small beneficial treatment effect in favor of aquatic exercise compared with no treatment. Only 1 study in the meta-analysis³⁹ independently showed an effect size in favor of pain relief (for WOMAC pain) that is clinically meaningful.⁴² However, the small number of good-quality studies that could be included into the meta-analysis and the inconsistency of results across all the studies suggest that, currently, we have insufficient evidence to categorically state that aquatic exercise is an effective modality for relieving pain.

Fig 2

Meta-analysis of trials with moderate to low risk of bias. (A) Aquatic exercise versus no-treatment control group. (B) Aquatic exercise versus land-based exercise. The forest plots show the differences between aquatic exercise and no-treatment control groups and land exercise. Included are the means ± SD for each group, sample size, SMD, 95% CIs and the weighting for each study, and the combined results.

 

Ten studies were identified in this comparison^{23, 24, 25, 27, 29, 31, 34, 35, 37, 40}; one of these was assessed as moderate quality³¹ and 3 as high quality.^{27, 35, 40} Meta-analysis of the moderate- to high-quality studies with available data^{27, 31} showed no differences between aquatic exercise and land-based exercise (P=.56; SMD=.11; 95% CI, −.27 to .50) (see fig 2B). Data in the remaining 2 high-quality studies, reported as median and interquartile ranges for the WOMAC³⁵ and pain VAS,⁴⁰ are consistent with these findings. Of the poor-quality studies, only one³⁴ contradicts the overall consistency of results. In summary, the evidence—based on meta-analysis and overall consistency of results—suggests that aquatic exercise and land-based exercise have comparable pain-relieving effects.

When compared with no-treatment controls, meta-analysis shows a small but significant posttreatment effect in favor of hydrotherapy. The 95% CI for this effect comes close to, but does not cross, the line of no treatment effect. In 2 of the studies in the meta-analysis, exercise was performed in water ranging from 29.5° to 33.5°C^{32, 39}; temperature was described as “temperate” in one.^38(p503) This encompasses a wide range of water temperature from cool to near thermoneutral (usually described as 34.5°–35°C). There is an assumption, by therapists and patients alike, that warmer water is more conducive to pain relief, and the mechanisms whereby heat alters pain perception are well known. Neuromuscular, hemodynamic, and metabolic responses to skin heating have been described^{45, 46}; however, skin heating through whole-body immersion and core body temperature changes during exercise in water await investigation. The contribution of the warmth of the water to pain relief during hydrotherapy therefore remains speculative.

Several researchers have also reported a relationship between water temperature and adherence, which suggests that thermal comfort is an important environmental factor in patients with a similar profile to those in the studies included in this review.^{39, 47} Our review found evidence that aquatic exercise has a small effect in relieving pain even at cooler temperatures below thermoneutral. Potentially, this has important implications in terms of water heating costs. In addition, it suggests that effective aquatic exercise can be practiced in community settings that traditionally maintain lower temperatures than hospital-based pools.

Forty-two percent of studies in this review failed to cite water temperature and, because it is considered a critical variable, this should be reported in future aquatic exercise studies. Also, some studies fail to explicitly state the aim of the water-based exercise program and/or to provide an adequate description of its type and intensity (n=6 [31.5%]). The effect of exercise-induced analgesia on pain suggests that pain intensity is reduced after exercise.⁴⁸ However, in humans, consistent results have been shown only for high-intensity land exercise (ie, ≥70% of maximal aerobic capacity). Exercise prescription is commonly based on predicted maximal heart rate because of the linear relationship between aerobic capacity and heart rate.⁴⁹ However, Tanaka et al⁵⁰ have recently questioned the accuracy of exercise prescription based on predicted maximal heart rate, particularly in older people. Furthermore, the variable effect of water temperature on heart rate makes it uncertain that the exercise intensity in the studies reviewed was above the analgesic threshold.⁵¹

The interaction between exercise intensity and water temperature, as well as having practical considerations,⁵² may affect pain differentially. In addition, water depth alters the nature of exercise through buoyancy and hydrostatic pressure. This in turn may influence the physiologic mechanisms underlying pain relief through the relationship between cardiovascular and pain regulatory pathways.⁵³

The duration of the aquatic exercise program is another variable that may play a significant role in pain relief. In his review Janal⁵⁴ highlights the uncertainty for the optimal duration of exercise-induced analgesia but suggests that the interaction between intensity and duration affects exercise-induced hypoalgesia. We noted that duration of treatment varied from twice a week for 4 weeks²⁷ to 4 times a week for 53 weeks.³⁹ The 3 studies^{32, 38, 39} comparing aquatic exercise and no-treatment controls are homogenous in terms of the type of exercise performed within the pool environment, but they have different durations of intervention, ranging from 12 to 52 weeks on a twice-weekly basis. It is interesting that those studies that do not report significant differences between groups are those of shorter duration. The duration of aquatic exercise programs for maximum pain relief is both clinically and economically important and warrants further investigation by good-quality longitudinal follow-up studies.

There is a general assumption that hydrotherapy will provide better pain relief than either land-based exercise or immersion alone. However, we found no significant differences between hydrotherapy and immersion in the 2 studies available.^{27, 36} In addition, no significant between-group differences were noted between aquatic and land-based exercise, which suggests that for those who find exercise on land challenging or tedious, exercise in water provides a similar effect. This lack of difference has been reported in previous studies,^{23, 31, 40} and currently, given the paucity of evidence, it is difficult to speculate which of the many variables (ie, temperature, exercise intensity and duration, treatment setting) could, either independently or in combination, be critical in pain relief. In contrast to our findings, the popularity of aquatic exercise as a modality for pain relief suggests that any additional benefits compared with land exercise have not yet been captured by the research.

The impetus for this review was our perception of a disparity between anecdotal reports of significant pain relief from aquatic exercisers and our informal reading of the literature. Thus we chose to focus on the outcome, pain, rather than a specific population. In focusing on pain, we made an a priori selection of the primary pain outcome measure based on the availability of raw data and the most frequently occurring measure across all studies when multiple outcomes for this variable were cited. The lack of consistency in pain outcome measurements across the studies validated our decision to limit our definition to pain intensity; future reviews might be able to incorporate evaluations of pain behavior and cognitive coping strategies as prospective studies include such measures. We selected the RCT checklist produced by SIGN, one of many quality assessment tools, which allows overall assessment of individual components and is included in best practice systems reported by the Agency for Healthcare Research and Quality. However, as Katrak et al⁵⁵ point out, there is no criterion standard for quality assessment tools in allied health research, and so our results must be considered in the light of the instrument we used.