A systematic review and meta-analysis comparing the effect of aquatic and land exercise on dynamic balance in older adults

A total of 11 studies meeting the eligibility criteria were reviewed and coded in REDCap (https://​www.​project-redcap.​org/​). All relevant information was extracted for each study as follows: (1) report characteristics (2) participants (3) AE settings (4) interventions (5) outcome measures (6) results. The included studies were assessed and coded independently by two reviewers (Y.K. and M.V.) and discussed for consensus. If there was a disagreement, the study was re-evaluated to achieve consensus.

The analysis of the methodological quality and risk of bias of the included studies was conducted using the Cochrane risk of bias tool (RoB 2) [24] independently by two authors (Y.K. and M.V.). The tool can be utilized to assess the impact of each potential source of bias, at the “low”, “high”, and “somewhat concerns” risk level, respectively. The following criteria that potentially affect the risk of bias were addressed: randomization process, deviation from intended interventions, missing outcome data, measurement of outcome, selection of the reported result, and overall bias. Any disagreements were discussed until consensus was reached and additionally arbitrated by the third (E.B.) and fourth (B.W.) reviewers if needed. “Small study effects” is a generic term for the phenomenon that smaller studies sometimes show different, often larger, treatment effects than large studies [25]. In meta-analysis, small study effects are a well-known challenging and critical issue that may threaten the validity of the study results, and the most well-known reason of the small study effects is publication bias [25]. The publication bias can be displayed graphically in funnel plots, thus, a small study effect was examined and interpreted through a test for funnel plot asymmetry [26]. In the absence of publication bias, the plot should be shaped like a symmetrical funnel with small studies scattered widely at the bottom of the graph and larger studies spread narrowly [25].

The purpose of the meta-analyses was to compare the pooled effect size between the AE group and LE group on dynamic balance in older adults. For the post-intervention sample size, when all subjects at the baseline were followed up, assessed, and analyzed regardless of their compliance to the intervention (intention-to-treat), the data including means and standard deviations for each outcome measure were used on the preferential basis [27]. Otherwise, the data of subjects who completed a pre-determined intervention(s) and have measurable data at the primary end point without any major protocol violations (per protocol) were used [27]. When data were not reported in the article as means and standard deviations, we contacted the corresponding authors and requested the data.

Outcome measurements included in the meta-analysis were assigned into three categories: (a) dynamic steady-state balance (e.g., 5-m walk test, 10-m walk test, backward tandem walk), (b) proactive balance (e.g., FRT; Functional Reach Test, TUG; Timed Up and Go test, 8-ft up-and-go test), and (c) balance test batteries (e.g., BBS; Balance Berg Scale and BOOMER; Balance Outcome Measure for Elder Rehabilitation) [28]. Where a trial reported more than one outcome in one of these categories, only one outcome with the highest priority was used for the analysis in line with Lesinski et al. [29]. The highest priority was given to the gait speed in the dynamic steady-state balance, FRT in the proactive balance, and BBS in the balance test battery [29]. When these representative outcomes were not available, the most similar outcomes related to the temporal (duration) and spatial (form of the motion) structure were used [29]. For a crossover RCT study [30], first-phase data were used. Sensitivity analyses were additionally performed to explore the robustness of the results by quantifying the differences in outcomes when removing one trial with a distinctly different direction of change in each category of balance outcome measurements.

The effect sizes between AE and LE groups were described as standardized mean differences (SMD) and 95% confidence intervals (CI). An effect size (SMD) 0.2–0.5, 0.5–0.8, and > 0.8 were considered a small, moderate, and large effect, respectively [31]. In case of a lower score indicating better performance in dynamic balance, scale directions were adjusted by multiplying − 1 to data, which resulted in a positive value indicating an improvement in favor of AE. For all analyses, we used an inverse-variance weighted random-effects model. All meta-analyses were performed using the Cochrane Collaboration’s Review Manager Software (RevMan 5.3.).

The electronic search retrieved a total of 2969 potential studies in the five databases, and no additional studies were identified by hand searching. Of these studies, 1491 duplicates were removed, and 1445 studies were excluded based on title and abstract content. We obtained the full text of the remaining 33 trials, 22 of which were excluded because they did not meet eligibility criteria. Finally, 11 studies were retained for our systematic review, and 10 studies were included in the meta-analysis after excluding one study due to insufficient data [32]. The flow diagram in Fig. 1 schematizes the steps of the selection of the studies.

Post-intervention assessment data for BBS, Dynamic Gait Index, tandem gait, and 10 m gait speed from the study by Avelar et al. [32], data for 5-m walk test, FRT, and TUG from the study by Vivas et al. [38], data for BBS from the study by Arnold et al. [11], and data for 10-m gait speed and BOOMER from the study by Adsett et al. [30] were requested, and all data, except those from the study by Avelar et al. were received. Thus, a total of 10 studies were included in the meta-analysis of dynamic balance outcomes for AE compared with LE [11, 30, 33‐40].

Outcome measurements included in each category were as follows: (a) dynamic steady-state balance: 10-m walk test (speed) [30], 5-m walk test (speed) [38], and backward tandem walk (number of errors) [11], (b) proactive balance: FRT [11, 33, 38, 39], TUG [30, 36, 37, 40], and 8-ft up-and-go test [34], (c) balance test batteries: BBS [11, 35, 37‐40] and BOOMER [30]. When a random-effect analysis was applied using the 10 studies involving 343 participants, AE groups compared with LE groups displayed comparable improvements in dynamic steady-state balance (SMD = − 0.24; 95% CI, −.81 to .34), proactive balance (SMD = − 0.21; 95% CI, −.59 to .17), and balance test batteries (SMD = − 0.24; 95% CI, −.50 to .03) (Fig. 4). The sensitivity analyses after excluding one trial with a distinctly opposite direction of change in each category presented that the point estimates changed by − 0.20 (SMD = − 0.44; 95% CI, −.88 to 0) in dynamic steady-state balance, by − 0.08 (SMD = − 0.29; 95% CI, − 62 to .03) in proactive balance, and by − 0.08 (SMD = − 0.32; 95% CI, −.61 to −.03) in balance test batteries (Fig. 5).

Postural strategies vary in different environments regardless of age and physical fitness [16]. Both older and younger adult populations demonstrated the greatest postural sway and sway velocity with the lowest perceived stability in chest-deep water compared to the same measures made at shallow water depths and on land [16, 22, 47]. However, none of the trials included in this current review provided a rationale for the water depth chosen and considered each participant’s height. Although all studies recruited both male and female participants with different mean height, except for only one trial by Arnold et al. [11], the AEs were conducted in water with the non-adjustable water level. That implies the participants in the AE groups were trained with all different exercise intensities despite the identical location, settings, and exercise types. In addition, movement patterns and mechanical power outputs during the same physical performance are presented differently in water and on land [48]. Thus, although most of the trials included provided the same or similar exercise programs to both AE and LE groups, the subjective exercise intensities can be different due to the environmental factors, which may affect the ultimate training effects. The main reason AE is recommended to the older adults is to utilize the physical properties of water and provide an optimized medium for exercise. Therefore, it is recommended that future studies provide rationales for water depth and exercise intensities in all intervention groups to investigate and compare the effects between AE and LE more accurately.

The intervention dose, duration, intensity, and type of exercise varied considerably in each trial, but there was no justification for the exercise dose chosen. According to ‘The 2018 Department of Health and Human Services’ guideline [49], older adults should get at least 150 min per week of moderate-intensity or 75 min per week of vigorous-intensity aerobic activity with moderate or high-intensity muscle-strengthening activities at least 2 days a week. Specifically, it is recommended for older adults with the risk of falls to participate in balance training three or more times per week to reduce falls. Older adults in three trials participated in AE and LE at least 150 min per week [11, 39, 40], and those in two trials practiced balance training at least 3 times per week [11, 39]. The intensity of the activities can be perceived in different ways according to various factors, such as physical fitness, muscular performance, or level of disorder or degeneration. Only two studies [11, 34] assessed subjective exercise intensity using the Borg rating of perceived exertion scale (RPE scale), and participants were instructed to exercise at a predetermined intensity. However, the optimal dosage, duration, and intensity of AE were not identified as most of the studies demonstrated low-to-moderate effect sizes and both AE and LE groups mostly presented comparable results across all trials.

The outcomes were measured using various dynamic balance tests, but the assessments were performed immediately after the interventions were terminated. Although each measurement contains critical components in daily living activities and indirectly predicts the potential risk of falls, the generalization of the results regarding the reduction of fall risks must be interpreted with caution as these are lacking in regards to the longer effects of the interventions. Therefore, future studies may wish to evaluate dynamic balance in an extended length of time to assess endurance-related muscle functions that are also essential for postural adjustment in daily life. The aim of AE interventions in the older population is to improve physical fitness, functional performance, and postural adjustment to ultimately reduce the risk of falls and fall-related injuries and improve their quality of life. Simmons and Hansen [33] tracked the rate of injuries between 10 and 12 months after the termination of the last session and reported that there were no orthopedic injuries from falls in the AE group, whereas there were two bone fractures (16.7%) in the LE group since the last session. Two trials conducted by Pérez de la Cruz et al. [36, 37] also included second post-intervention assessments, but the time interval (1-month post-intervention) was not sufficient to determine long term effects of AE on dynamic balance or fall reductions. Arnold et al. [11] and Volpe et al. [39] reported adverse events that occurred during the interventions, but none of the included studies reported participants’ feedback for the AE or LE programs. Besides the main outcome measures, supplementary information regarding injuries and psychological effects, such as satisfaction and enjoyment, may be helpful for an in-depth interpretation of the effectiveness of AE.

In consideration of the exercise program components, the results of the meta-analyses that demonstrated AE and LE have equivalent effects on dynamic balance should be interpreted with caution. In general, to improve a specific skill, a completely or nearly identical task is generally included in exercise interventions to induce a practice effect. However, among the ten trials in the meta-analyses, only four trials included at least one balance or gait-related task in the exercise programs [11, 35, 38, 39], and the rest of the ten trials included other types of exercises, such as endurance, strength, mobility, or aerobic exercises, that may contribute to the improvement of dynamic balance. Thus, future research may wish to include a goal-focused exercise program that focuses on balance-related tasks and controls for other variables, such as exercise intensity, to more clearly compare the effectiveness of AE and LE on dynamic balance in the older population.

This study did not identify the statistical superiority of AE over LE programs on dynamic balance. However, these results imply that AE can be an appropriate alternative to LE which leads to clinically meaningful improvements in balance. Both AE and LE have different advantages. Because LE is performed under dryland conditions and is more associated with activities of daily living, these can be more applicable and transferable to enable older adults to successfully improve practical skills. Due to environmental characteristics, muscle activation patterns and movement kinematics are different during aquatic activities compared to those during identical land activities [23, 50], which may lead to less transferability to various functional tasks on dry land, however, this has not been formally tested or observed in previous research. The aquatic environment provides older adults with numerous biological, neurological, and musculoskeletal advantages and helps them perform higher exercise intensities in a safer and supportive training environment without the risk or fear of falling [14, 48, 51‐54]. Therefore, it is suggested that future studies and practitioners select the proper exercise mode that matches each participant’s preference and aim of the intervention to maximize the intervention effectiveness. Further investigations regarding the classification of disorder, disease, or history of falls may provide stronger scientific rationales for future balance training protocols for older adults.

As identified in this review, most of the AE programs were administered by physical therapists in clinical facilities. Because of the limited accessibility of aquatic exercise facilities, availability of experts, and higher medical costs, AEs are not broadly practiced in the senior populations. Thus, more easily accessible and lower-cost AE protocols need to be established so that older adults can participate in various physical activities in a safer environment to improve balance, reduce the risk of falls, and ultimately improve their quality of life.

This systematic review and meta-analysis have several limitations. First, this study was limited to peer-reviewed journal articles published in English and RCT designs only, which may increase the risk of publication bias and potentially exclude appropriate studies with high-quality methodologies. In consideration of the potential small study effects and publication bias, future meta-analyses may want to identify and include unpublished outcomes and unpublished studies to improve the validity of results [55]. Also, we included outcomes using the balance categories instead of using just one measure from each study because we only had 10 studies. Due to the small number of studies included in each category, potential covariates, such as the duration of intervention, exercise type, or exercise intensity, could not be appraised using a moderator analysis. In future reviews, it may be appropriate to use a single measure in each study and conduct a meta-regression to identify the impacts of the potential covariates on the effect sizes in the meta-analyses. In addition, five out of 11 studies in the review presented “somewhat concerns” of risk of bias and four had a “high” risk bias, that potentially cause overestimation of the true effects of AE and LE. The randomization process, missing outcome data, and selection of the reported result were the main causes of bias. Thus, we suggest that future trials make advanced plans for these three categories. Furthermore, as only two outcomes [33, 38] in the proactive balance category demonstrated high effect sizes, we were not able to establish the general guideline with optimal exercise type, intensity, dosage, and duration to improve dynamic balance in older adults.