The dose–effect relationship of aquatic exercise in improving physical function in older adults

Abstract

Objective

Global aging has increased the risk of falls and functional dependence among elderly people. Aquatic exercise, with its low-impact and multimuscle group activation characteristics, provides a unique physiological stimulation environment for improving the physical functions of the elderly population. However, previous studies have insufficiently integrated multidimensional functional indicators and dose–effect relationships. In this study, a meta-analysis was conducted to systematically evaluate the effects of aquatic exercise on the multidimensional physical functions of elderly people and to determine the differential effects of different intervention programs.

Methods

Databases in both Chinese and English were searched up to April 2025, and 23 RCTs (1179 healthy elderly individuals aged ≥ 60 years) were included. The Cochrane tool was used to assess the risk of bias, and data on intervention programs (type, duration, frequency) and functional indicators were extracted. RevMan 5.4 was used to combine effect sizes (MD and 95% CI), and subgroup analysis was conducted to explore the sources of heterogeneity (I² > 75%).

Results

(1) Balance ability: Static balance improved (single-leg standing time: MD = 3.05 s, P = 0.007); dynamic balance improved (TUG time reduced by 0.59 s, P = 0.028), with resistance training having the greatest effect (MD = -1.80 s, P = 0.006); and a higher frequency intervention (> 2 times/week) was more effective. (2) Muscle strength: Lower limb strength increased (30-s chair stand test increased by 2.77 times, P < 0.001), with combined resistance and aerobic training having the greatest effect (MD = 3.98 times, P < 0.001); upper limb strength increased (arm curl test increased by 3.17 times, P < 0.001). (3) Cardiopulmonary function: The 2MST increased 7.56 times (P < 0.001). (4) Flexibility: Lower limb flexibility improved by 3.21 cm (P = 0.024), whereas upper limb flexibility did not significantly improve. (5) Subgroup analysis revealed that long-term (> 12 weeks) and high-frequency (> 2 times/week) interventions were more effective for improving dynamic balance (MD = -0.87 vs. -0.34 s) and increasing lower limb strength (MD = 5.65 vs. 1.45 times); combined resistance and aerobic training had significant overall benefits.

Conclusion

Aquatic exercise can comprehensively improve the physical function of elderly people. High-frequency (> 2 times/week) and long-term (> 12 weeks) combined resistance and aerobic training programs are recommended. Clinically, task-oriented interventions should be designed on the basis of individual deficits, and future research should standardize exercise parameters and explore its psychological and social benefits.

Introduction

The global population is undergoing an unprecedented structural aging transformation. According to the United Nations, the global population aged 60 and above reached 1 billion in 2019 and is expected to surge to 2.1 billion by 2050, with particularly rapid growth in developing countries [1]. As life expectancy increases, the elderly population generally faces nonpathological physiological declines, such as weakened lower limb muscle strength, reduced balance ability, and unstable gait, which significantly increase the risk of falls and fractures. The gradual decline in physical function is directly related to daily activity capabilities, and functional dependency may lead to psychosocial isolation and economic burdens [2]. Traditional land-based exercises (such as aerobic and resistance training) can improve cardiovascular function and muscle strength, but their high-impact nature increases the joint load [3]. In contrast, aquatic exercise, with its buoyant support, adjustable resistance, and low-impact characteristics, provides a unique physiological stimulus environment for elderly people [4]. For example, the buoyancy of water can reduce joint pressure by 50%–90%, while fluid resistance can activate multimuscle group coordination, improving neuromuscular control and balance ability [5–7].

Aquatic exercise, as a low-intensity, high-safety intervention, has been widely proven to offer multidimensional health benefits for elderly people. However, a systematic integration of evidence has not yet reached a consensus. Previous meta-analyses have often focused on single dimensions (such as improvements in lower limb muscle strength or balance ability) [8], with insufficient attention given to comprehensive functional indicators (such as joint mobility and dynamic balance) and cardiorespiratory endurance parameters. For instance, Ardalan et al. reported that the current evidence for improving balance ability among elderly people shows significant heterogeneity, possibly because of differences in assessment tools and intervention protocols [9]. Moreover, there is ongoing debate regarding the improvement of joint mobility: on the one hand, supporting evidence suggests that aquatic exercise can enhance upper limb flexibility through fluid resistance, such as in the backstroke test, where aquatic resistance training significantly increased the range of motion and extension of the shoulder [10], possibly because of the viscous resistance of water promoting coordinated contraction of the rotator cuff and joint capsule extensibility; on the other hand, opposing viewpoints argue that if the training program lacks a task-oriented design, its effects may be limited to short-term improvements in joint mobility and may not translate into enhanced flexibility for daily movements [11].

Therefore, unlike previous meta-analyses that primarily focused on isolated outcomes such as balance or lower-limb strength, the present study makes two key contributions. First, it systematically integrates multidimensional indicators of physical function—including balance, muscular strength, flexibility, and particularly cardiovascular function and joint mobility—to provide a more comprehensive assessment of the benefits of aquatic exercise in older adults [12]. Second, it is among the few studies to quantitatively explore the dose–response relationship of aquatic exercise (i.e., how intervention frequency, duration, and training modality influence functional outcomes). These innovations allow the study to clarify not only whether aquatic exercise is effective, but also how much and what type of exercise yields optimal benefits for the elderly population.

Materials and methods

Literature search

We systematically searched PubMed (MEDLINE), Web of Science Core Collection, EBSCOhost databases, China National Knowledge Infrastructure (CNKI), and Wanfang Data from database inception to [30 April 2025]. The last search was performed on [30 April 2025]. Searches were limited to publications in English or Chinese.Search strings combined controlled vocabulary and free-text terms and were organized into three concept blocks (#1 aquatic exercise, #2 older adults, #3 physical function), with the final set #4 = #1 AND #2 AND #3. Strategies were adapted to each database’s syntax and indexing (e.g., MeSH in PubMed; Topic field in Web of Science; Subject Headings/TI/AB in EBSCOhost;主题词/篇名/摘要in CNKI/Wanfang). The complete, line-by-line PubMed strategy (MeSH and Title/Abstract fields) is provided in Supplementary Fig. 1.

Fig. 1 PubMed database search strategy.

Note1:#1 Aquatic exercise, #2 older adults, #3 physical function; #4 = #1 AND #2 AND #3. Database-specific fields: PubMed (MeSH + Title/Abstract), Web of Science (Topic), EBSCOhost (Subject Headings + TI/AB), CNKI/Wanfang (主题词/篇名/摘要). Language: English or Chinese. Timeframe: inception–[30 April 2025]; last search [30 April 2025]. Databases: PubMed; Web of Science; EBSCOhost: [CINAHL; SPORTDiscus; APA PsycINFO; Academic Search Complete]; CNKI; Wanfang

The search was structured into three concept blocks and combined as #4 = #1 AND #2 AND #3.

#1 Aquatic exercise exposure: PubMed used MeSH (e.g., “Aquatic Exercise”, “Hydrotherapy”, “Swimming”) plus Title/Abstract synonyms (e.g., aquatic exercise, water-based exercise, aqua aerobics). Web of Science used TS = (aquatic exercise OR hydrotherapy OR “water-based exercise” OR “aqua aerobics”). EBSCOhost used Subject Headings (MH) where applicable plus TI/AB fields (e.g., aquatic exercise, hydrotherapy). CNKI/Wanfang used主题词/关键词与篇名/摘要 (如 “水中运动” “水中有氧操” “水疗训练”等) 。

#2 Older adults: PubMed MeSH (e.g., “Aged”, “Aged, 80 and over”, “Frail Elderly”) and Title/Abstract synonyms (elderly, older adults, seniors); Web of Science TS = (elderly OR “older adults” OR seniors OR aged); EBSCOhost MH/TI/AB; CNKI/万方使用 “老年人/老年人群/60岁及以上/≥ 60岁”等。

#3 Physical function outcomes: PubMed MeSH (如 “Physical Fitness” “Postural Balance” “Muscle Strength” “Gait” “Activities of Daily Living”等) 与 Title/Abstract 同义词 (physical function, functional performance, gait speed, TUG, 6-min walk, handgrip, cardiorespiratory fitness 等); Web of Science “TS = ”对应术语; EBSCOhost MH/TI/AB; CNKI/万方使用 “身体功能/运动能力/肌力/平衡能力/心肺功能”等。

Limits and timeframe: Language limited to English or Chinese; publication date from database inception to [30 April 2025]; last search on [30 April 2025].

Databases: PubMed (MEDLINE); Web of Science Core Collection; EBSCOhost: [CINAHL; SPORTDiscus; APA PsycINFO; Academic Search Complete]; CNKI; Wanfang Data.

Inclusion and exclusion criteria

The inclusion criteria for the literature strictly followed the medical PICOS principles [13]. The study subjects were elderly individuals aged 60 years or older, and those with chronic diseases (such as osteoarthritis or cognitive impairment) were excluded. The intervention group engaged in aquatic exercise, whereas the control group underwent either conventional exercise or no intervention. The comparison type involved a pre- and postintervention effect comparison between the experimental and control groups. The outcome measures included balance ability, muscle strength, joint mobility, and cardiovascular function. The study types included both Chinese and English language randomized controlled trials.

Exclusion criteria:

1. Reviews and nonrandomized controlled studies;

2. Literature not published in English or Chinese;

3. Studies with incomplete outcome data that could not be extracted.

Registration and protocol

The review protocol was registered in the PROSPERO international prospective register of systematic reviews (Registration Number: CRD420251153520). A separate, comprehensive review protocol document was not prepared. The PROSPERO registration record represents the a priori protocol for this systematic review and is publicly available at: (https://www.crd.york.ac.uk/PROSPERO/myprospero).

No amendments were made to the protocol after registration. The review was conducted in accordance with the planned methods as stated in the PROSPERO record.

Data extraction and quality assessment

A systematic deduplication process was implemented using EndNote X9 reference management software, creating a standardized literature database. Two researchers independently performed a blind screening process (Jiahui Wang, Fuyou Li), cross-checking the titles and abstracts against the predefined PICOS inclusion/exclusion criteria in a double-blind manner. Studies that did not meet the research topic criteria were removed in the first round, and potentially relevant studies retained after the initial screening were retrieved for full-text evaluation and in-depth assessment. This assessment focused on the specificity of the intervention measures (e.g., aquatic exercise parameters such as frequency, intensity, and duration), completeness of outcome measures, and data extractability. This process was conducted independently, and in cases of disagreement, a third researcher (Pu Sun) was involved in the evidence-based decision-making process.

Data extraction was carried out using pre-designed standardized forms, which was independently completed by two researchers and then cross-checked. The extracted contents included:

1. Basic information of the study: First author, publication year.

2. Characteristics of the research subjects: Sample size, age.

3. Details of the intervention plan: Intervention measures for the experimental group and the control group, intervention period (weeks), frequency (times/week), single duration (minutes), specific exercise content and intensity.

4. Outcome indicator data: Pre- and post-test data of all relevant physical function indicators (mean, standard deviation). To ensure the reliability of the analyzed data, this study only included indicators that evaluated the physical function of the elderly using standardized testing methods. The specific definitions and improvement judgment criteria are shown in Table 1.

Table 1.

Physical function assessment indicators, testing methods, and improvement judgment criteria for older adults

Test Name	Test Definition	Improvement Judgment
Single-leg Static Balance Test	Single-leg Stand for 30 Seconds (nonsupporting leg bent at 90°), record the duration of the hold	Positive value ↑: Longer hold time or improved score indicates an improvement in static balance ability
Time Up And Go Test (TUG)	Stand up from the chair → Walk 3 m → Return and sit down, record the total time (seconds)	Negative value ↓: Shorter time indicates an improvement in dynamic balance ability
30-Second Arm Curl Test (30 s-AC)	Number of bicep curls completed with dumbbells in 30 s (elbow joint 0° → 90°)	Positive value ↑: Increased repetitions indicate improved upper limb muscle endurance
30-Second Chair Stand Test (30 s-CS)	Number of sit-to-stand repetitions in 30 s (using only lower limbs)	Positive value ↑: Increased repetitions indicate improved lower limb muscle strength
Chair Sit-and-Reach Test	Seated single-leg extension, with both hands reaching forward (distance in cm) Positive value = beyond the toes, negative value = not reaching the toes	Positive value ↑: Increased reach distance indicates improved lower limb flexibility
Back Scratch Test	Standing position with both hands reaching behind, measure the distance between the middle fingers (cm) Positive value = overlapping, negative value = distance between fingers	Positive value ↑: Increased overlap distance (or negative value ↓: decreased distance) indicates an expanded range of motion in the upper limbs
2-Minute Step Test (2MST)	Number of steps taken in 2 min with the right knee raised to the midpoint of the patella and iliac crest (approximately 90° hip flexion)	Positive value ↑: Increased repetitions indicate improved cardiovascular endurance

Quality assessment was performed using the Cochrane Risk of Bias Tool (version 2.0, RoB 2.0) recommended in the Cochrane Handbook for Systematic Reviews of Interventions.

Two reviewers independently assessed each study across the seven domains (random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting, and other biases).

Any discrepancy was resolved by discussion to reach consensus; if disagreement remained, a third reviewer acted as arbitrator.

Two researchers (Jingwen Chen and Baojin Zhao) independently evaluated the included randomized controlled trials using the Cochrane risk of bias assessment tool. The evaluation contents included: random sequence generation, allocation concealment, blinding of researchers and participants, outcome assessment blinding, incomplete outcome data, selective reporting, and other biases. Each item was judged as “low risk”, “high risk” or “unclear risk”. Disagreements during the evaluation process were resolved through discussion or by a third researcher (Bochen Liu). The Cochrane risk of bias assessment results revealed that in terms of random sequence generation, 22 studies were assessed as low risk and 1 as high risk. With respect to allocation concealment, 1 study was considered low risk, and 22 were considered unclear. In terms of the blinding of participants and therapists, 19 studies were assessed as low risk, 2 as unclear, and 2 as high risk. With respect to the outcome assessment, 4 studies were assessed as low risk, 1 as unclear, and 18 as high risk. In terms of incomplete outcome data, 22 studies were considered low risk, and 1 was considered high risk. In terms of selective reporting, 23 studies were assessed as low risk. With respect to other biases, 23 studies were assessed as low risk, as shown in Fig. 3.

Fig. 3.

Literature screening flowchart

Overall, the risk-of-bias assessment indicated that the majority of the 23 included studies were rated as low risk across most domains. More than 80% of studies showed low risk in random sequence generation and selective reporting, whereas allocation concealment and outcome assessor blinding were commonly judged as unclear. The proportion of low-, unclear-, and high-risk assessments for each domain is summarized in Fig. 2.

Fig. 2 Literature quality assessment chart.

Note 2: A represents the risk of bias summary, with “ + ” indicating low risk, “-” indicating high risk, and “?” indicating unclear risk; B represents the risk of bias chart

Statistical analysis

All the included studies, as long as they reported the data of a certain outcome indicator as preset in this review, were included in the corresponding Meta-analysis for that indicator. Statistical analysis in this study was conducted using Review Manager 5.4 software. Since the outcome measures included in the analysis were all continuous variables and the measurement methods and units used in the included studies were consistent, the effect size was combined using the mean difference (MD) and 95% confidence interval (95% CI). Heterogeneity between studies was assessed using the I² value and P value. When I² ≤ 50% and the Q test P ≥ 0.10, a fixed-effects model was used; when I² > 50% or the Q test P < 0.10, a random-effects model was used. When heterogeneity was greater than 50%, subgroup analysis was conducted to explore potential sources of heterogeneity. When heterogeneity exceeded 75%, additional sensitivity analyses were performed. Meta-regression analysis was not performed due to the limited number of studies for each outcome, which should be taken into account when interpreting the findings. The presentation of the research results will be visualized using forest plots, and the results of subgroup analysis and sensitivity analysis will be summarized in tables. All p-values are reported consistently, and extremely small values (P < 0.001) are uniformly expressed as “P < 0.001” for clarity and consistency. All subgroup results (MD, 95% CI, I², and P values) reported in the text were cross-checked and made fully consistent with Table 3 outputs.

Table 3.

Subgroup analysis results of the effects of aquatic exercise on physical function in older adults

Outcome Measures	Moderating Variables		Number of Studies	Heterogeneity Test		Combined Effect Size
				p value	I2(%)	MD (95% CI)	p value
TUG Test	Age	60–70 years	8	< 0.001	85.0	−1.00(−1.47,−0.52)	< 0.001
		> 70 years	7	< 0.001	86.8	−0.91(−1.71,−0.10)	0.028
	Intervention Duration	≤ 12 weeks	9	0.190	34.7	−0.34 (−0.60, −0.08)	0.012
		> 12 weeks	7	0.122	48.2	−0.87 (−1.20, −0.53)	< 0.001
	Intervention Frequency	≤ 2 times/week	8	0.005	75.6	−0.66(−1.12,−0.20)	0.005
		> 2 times/week	8	0.001	94.2	−1.12 (−1.77, −0.46)	0.001
	Intervention Program	Resistance + Aerobic Exercise	11	< 0.001	83.9	−0.68(−1.23, −0.13)	0.015
		Resistance Exercise	5	0.001	84.9	−1.80(−3.07, −0.52)	0.006
		Aerobic Exercise	2	0.447	0.0	−0.19(−0.52,0.13)	0.009
30 s Chair Stand Test	Age	60–70 years	7	< 0.001	70.5	2.27 (0.94, 3.60)	0.001
		> 70 years	5	0.017	89.0	3.05 (0.43, 5.67)	0.022
	Intervention Duration	≤ 12 weeks	9	0.006	66.6	1.72 (0.54, 2.89)	0.004
		> 12 weeks	4	0.023	64.7	1.88 (0.66, 3.09)	0.002
	Intervention Frequency	≤ 2 times/week	8	< 0.001	71.2	1.45 (0.23, 2.68)	0.020
		> 2 times/week	5	< 0.001	99.1	5.65 (2.29, 9.02)	0.001
	Intervention Program	Resistance + Aerobic Exercise	10	< 0.001	98.0	3.98(1.89, 6.06)	< 0.001
		Resistance Exercise	4	0.235	29.2	1.25(0.47, 2.03)	0.002
		Aerobic Exercise	1	< 0.001	0.0	3.2(0.87, 5.53)	0.007
Chair Sit-and-Reach Test	Age	60–70 years	4	< 0.001	97.8	4.97(−3.69, 13.64)	0.261
		> 70 years	6	0.178	34.4	3.81(2.24, 5.38)	< 0.001
	Intervention Duration	≤ 12 weeks	8	< 0.001	98.7	3.05(0.01, 7.31)	0.050
		> 12 weeks	5	0.173	37.2	3.66(0.58, 5.51)	0.016
	Intervention Frequency	≤ 2 times/week	6	< 0.001	98.9	3.39(−0.46, 7.23)	0.084
		> 2 times/week	7	0.233	26.9	3.37(1.14, 5.59)	0.003
	Intervention Program	Resistance + Aerobic Exercise	9	< 0.001	97.1	4.58(0.70, 8.47)	0.021
		Resistance Exercise	3	0.139	49.4	3.21(−0.35, 6.78)	0.077
		Aerobic Exercise	1	< 0.001	0.0	−1.05(−7.12, 5.02)	0.735

Data preparation for synthesis

The data extraction process confirmed that all included studies reported the required outcome data as means and standard deviations. Consequently, no data transformations or conversions were performed. The extracted data were directly suitable for meta-analysis using the mean difference (MD) as the effect measure. Studies with missing or incompletely reported data that could not be obtained through author correspondence were excluded from the meta-analysis for that specific outcome, as pre-specified in our protocol.

To assess the robustness of the synthesis results, we conducted a sensitivity analysis on the primary outcome indicators (TUG test, 30-s sit-to-stand test, forward extension of the chair test, and 2-min stepping test), using the method of sequentially eliminating individual studies to observe the changes in the combined effect size.

Publication bias was assessed using Stata18 software to create funnel plots and perform Egger’s test.

Results

Literature search and screening results

A total of 3,844 articles were retrieved from the databases. After 994 duplicates were removed using EndNote software, 2,557 articles were excluded on the basis of title screening, and 243 were excluded on the basis of abstract screening. After reviewing the full texts, 29 articles were further excluded. Two additional studies were included from the references of the selected articles and relevant reviews. Ultimately, 23 studies were included in the meta-analysis. The literature screening process is shown in Fig. 3.

Basic characteristics of included studies

The time span of the included studies ranged from 2006 to 2024, involving healthy elderly participants aged 60 to 77 years, with a total sample size of 1,179 subjects, including 660 in the experimental group and 519 in the control group. The exercise interventions in the included studies were all conducted under the supervision of trained physical therapists. The types of exercise included resistance + aerobic exercise (14 studies), pure resistance exercise (6 studies), and pure aerobic exercise (3 studies). The basic information of the included studies is shown in Table 2.

Table 2.

Basic characteristics of the included studies

Included Studies	Sample Size (n)		Age (years)		Intervention Program		Intervention Dosage			Outcome Measures
	Experimental Group	Control Group	Experimental Group	Control Group	Experimental Group	Control Group	Period	Frequency	Duration
THOMAI, et al. [14], 2006	12	10	69.3 ± 1.9	68.4 ± 6.7	Aquatic Resistance + Aerobic Exercise	Avoid structured exercises or activities	24	3 times/week; 60 min/session	10 min of warm-up and stretching, 20 min of resistance exercise, 25 min of aerobic exercise; heart rate (HR) at 80% of maximum heart rate; 5 min of stretching	①③
Danilo, et al. [15], 2008	27	20	64.0 ± 1.0	64.0 ± 1.0	Aquatic Resistance + Aerobic Exercise	Land Walking Exercise	12	3 times/week; 60 min/session	10 min of warm-up and stretching, 45 min of resistance + aerobic exercise; heart rate (HR) at 70% of maximum heart rate; 5 min of stretching	①②③
Danilo, et al. [16], 2010	27	18	> 62.0	> 62.0	Aquatic Resistance + Aerobic Exercise	No intervention in normal daily activities	12	3 times/week; 60 min/session	10 min of warm-up and stretching, 45 min of resistance + aerobic exercise; heart rate (HR) at 70% of maximum heart rate; 5 min of stretching	①②③
AliAkbar, et al. [17], 2010	15	15	65.8 ± 4.8	66.60 ± 4.1	Aquatic Resistance Exercise	Land Resistance Exercise	Not mentioned	Not mentioned; 40 min/session	8 walking-related exercises in different conditions + 3 resistance exercises	①②
HSCOTT, et al. [18], 2012	15	11	75.6 ± 4.8	79.6 ± 10.1	Aquatic Resistance + Aerobic Exercise	Land Walking or Aerobic Exercise, Avoid Strength Training	8	2 times/week; 45 min/session	45 min of resistance + aerobic exercise	①②
Paulo, et al. [19], 2012	12	14	65.6 ± 4.2	65.6 ± 4.4	Aquatic Resistance + Aerobic Exercise	No intervention in normal daily activities	12	3 times/week; 60 min/session	10 min of warm-up, 20 min of aerobic exercise, 20 min of resistance exercise; 10 min of stretching	①②③
Yaser, et al. [20], 2012	14	14	75.7 ± 4.0	70.0 ± 13.3	Aquatic Resistance Exercise	No intervention in normal daily activities	8	3 times/week; 60 min/session	Water environment adaptation, stretching, and static and dynamic balance exercises, with low to moderate intensity, constant intensity, frequency, and speed	①
Mohammad, et al. [21], 2013	7	9	63.8 ± 3.3	61.3 ± 1.6	Aquatic Resistance + Aerobic Exercise	No intervention in normal daily activities	8	2 times/week; 75 min/session	Warm-up 10–15 min; aerobic exercise 20–25 min; resistance training 20–30 min; relaxation 5 min	①②
MaryE, et al. [22], 2013	48	18	73.6 ± 13.5	72.8 ± 27.4	Aquatic Resistance Exercise	No intervention in normal daily activities	16	3 times/week; 45 min/session	10 min of warm-up; 35 min of training using the S.W.E.A.T.™ method	①②③
MarcoB, et al. [23], 2013	17	19	71.2 ± 5.4	71.2 ± 5.4	Aquatic Resistance Exercise	No intervention in normal daily activities	24	2 times/week; 60 min/session	8 min of warm-up; 45 min of 10 exercises, 5 upper body exercises, 5 lower body exercises; 8 min of stretching; exercise intensity maintained at around 60% of maximum HR (range 55%−65%)	①②
MarcioR, et al. [24], 2014	28	23	69.0 ± 4.0	69.0 ± 4.0	Aquatic Resistance + Aerobic Exercise	Strength, Endurance, and Flexibility Training on Land	12	2 times/week; 60 min/session	Neurocontrol and motor coordination 15 min, upper and lower body resistance and aerobic exercises 25–30 min; stretching exercises	①
Il-Myeong, et al. [25], 2015	25	25	72.2 ± 4.5	71.0 ± 5.4	Aquatic Resistance + Aerobic Exercise	No intervention in normal daily activities	8	3 times/week; Not mentioned	Aquatic fitness class: flexibility exercises, resistance + aerobic exercise	①②③④
SeJunOh, et al. [26], 2015	34	32	74.7 ± 2.9	68.2 ± 4.4	Aquatic Resistance + Aerobic Exercise	Land Resistance + Aerobic Exercise	10	3 times/week; 60 min/session	10 min of warm-up; upper and lower body resistance, flexibility exercises for 40 min; 10 min of stretching	①②③
Andrew, et al. [27], 2015	27	33	63.2 ± 7.6	65.6 ± 6.6	Aquatic Aerobic Exercise	No intervention in normal daily activities	1	6 times/week; Not mentioned	Moderate to high-intensity aquatic aerobic exercise, reaching 60%−70% of maximum heart rate	④
N.A.PERIGO, et al. [28], 2016	180	119	68.5 ± 5.7	66.8 ± 5.9	Aquatic Resistance + Aerobic Exercise	No intervention in normal daily activities	Not mentioned	Not mentioned; 60 min/session	10 min of warm-up; 40 min of resistance, aerobic, and coordination training; 10 min of stretching and balance exercises	①②④
MESanders, et al. [29], 2016	13	13	70.8 ± 4.0	70.1 ± 3.2	Aquatic Resistance + Aerobic Exercise	No intervention in normal daily activities	12	3 times/week; 60 min/session	10 min of warm-up; 15 min of ADL exercises, 15 min of aerobic exercise, 10 min of resistance training; 10 min of stretching	①②③
Daisy, et al. [30], 2017	9	10	64.3 ± 4.2	64.4 ± 4.2	Aquatic Aerobic Exercise	No intervention in normal daily activities, avoid unusual physical activities	18	2 times/week; 50 min/session	10 min of warm-up; 30 min of deep water running exercises; 10 min of relaxation	①③
Carlos, et al. [31], 2021	48	27	71.9 ± 5.6	73.6 ± 5.2	Aquatic Resistance + Aerobic Exercise	No intervention in normal daily activities	28	2 times/week; 45 min/session	20–25 min of continuous aerobic exercise; 20–25 min of 6–8 upper and lower body resistance exercises; target intensity at 60% to 70% of maximum heart rate	①②③④
AnaVictoria, et al. [32], 2021	13	11	64.4 ± 4.4	67.5 ± 5.6	Aquatic Aerobic Exercise	No intervention in normal daily activities	12	2 times/week; 60 min/session	4 upper and lower body resistance exercises	①②
Alejandro, et al. [33], 2022	17	17	69.6 ± 5.0	67.7 ± 3.6	Aquatic Resistance + Aerobic Exercise	No intervention in normal daily activities	14	3 times/week; 60 min/session	15 min of warm-up (10 min of aerobic and resistance exercise and 5 min of stretching); 30 min of aquatic interval resistance training; 15 min of stretching and relaxation	②
Antonio, et al. [34], 2024	25	26	> 60.0	> 60.0	Aquatic Resistance Exercise	Engaging in recreational exercise programs without specific cycles targeting muscle strength or cardiovascular ability	16	2 times/week; 45 min/session	Aquatic interval training, with progressive overload each week, encouraging participants to exercise at maximum intensity	①②③
Katarzyna, et al. [35], 2024	15	15	> 60.0	> 60.0	Aquatic Resistance + Aerobic Exercise	No intervention in normal daily activities	12	2 times/week; 60 min/session	10 min of warm-up; 40 min of resistance + aerobic exercise (aquatic balance, walking, running, swinging exercises); 10 min of stretching	①②③④
YuChen, et al. [10], 2024	20	20	66.2 ± 2.3	66.2 ± 2.3	Aquatic Aerobic Exercise	No intervention in normal daily activities	16	2 times/week; 60 min/session	10 min of warm-up; 50 min of high-intensity interval training, with freestyle training as follows: weeks 1–5, complete 6 sets of 30 s of maximum-effort front crawl intervals, with 3 min of passive recovery between sets and full ART interval exercises; weeks 6–10, increase the sets to 8 with the same interval and recovery duration; weeks 11–16, perform 10 sets of resistance training	①②③

Note 1: ① represents balance ability; ② represents muscle function; ③ represents joint mobility; and ④ represents cardiovascular function

Assessment of literature quality

The risk-of-bias assessment is summarized in Fig. 2. Most studies were judged as low risk for random sequence generation and selective reporting but unclear for allocation concealment.

Meta-analysis results

Balance ability analysis

Twenty-one studies reported the effects of aquatic exercise on improving balance ability in older adults. Five studies reported changes in single-leg balance ability among 295 participants. It was found that I² = 96.8% (P < 0.001), indicating heterogeneity among the studies. A random-effects model was used for analysis. The results of the meta-analysis (MD = 3.05; 95% CI: 0.85 to 5.25; P = 0.007) are shown in Fig. 4A and indicate that aquatic exercise can improve single-leg static balance ability.

Fig. 4 Effects of aquatic exercise on balance ability among older adults.

Note 3: A represents the forest plot for single-leg static balance ability; B represents the forest plot for dynamic balance ability

In terms of dynamic balance ability, 18 studies reported changes in TUG balance test results for 533 participants. It was found that I² = 95.6% (P < 0.001) indicated heterogeneity among the studies. A random-effects model was used for analysis. The results of the meta-analysis (MD = −0.59; 95% CI: −1.12 to −0.06; P = 0.028) are shown in Fig. 4B and indicate that aquatic exercise can improve dynamic balance ability.

Muscle function analysis

Eighteen studies reported the effects of aquatic exercise on improving muscle function in older adults. Fifteen studies reported changes in the 30-s chair stand test in 501 participants. It was found that I² = 97.4% (P < 0.001), indicating heterogeneity among the studies, so a random-effects model was used for analysis. The results of the meta-analysis (MD = 2.77; 95% CI: 1.19 to 4.35; P < 0.001) are shown in Fig. 5A and indicate that aquatic exercise can improve lower limb muscle strength in older adults.

Fig. 5 Effects of aquatic exercise on muscle function in older adults.

Note 4: A represents the forest plot for the 30 s chair stand test; B represents the forest plot for the 30 s arm curl test

Seven studies reported changes in upper limb muscle strength among 186 participants. It was found that I² = 24.6% (P = 0.241), indicating that there was no heterogeneity among the studies. A fixed-effects model was used for analysis. The results of the meta-analysis revealed that aquatic exercise can improve upper limb muscle strength in older adults (MD = 3.17; 95% CI: 2.08 to 4.26; P < 0.001), as shown in Fig. 5B.

Joint mobility analysis

With respect to lower limb joint mobility, 13 studies reported changes in lower limb flexibility in 335 participants through the chair sit-and-reach test. It was found that I² = 97.8% (P < 0.001), indicating heterogeneity among the studies, so a random-effects model was used for analysis. The results of the meta-analysis revealed that aquatic exercise effectively improved lower limb flexibility in older adults (MD = 3.21; 95% CI: 0.43 to 5.99; P = 0.024), as shown in Fig. 6A.

Fig. 6 Effects of aquatic exercise on joint mobility among older adults.

Note 5: A represents the forest plot for the chair sit-and-reach test; B represents the forest plot for the back scratch test

Eight studies reported changes in upper limb joint mobility in older adults through the back scratch test. One study was excluded because of different measurement units, and seven studies were ultimately included in the meta-analysis. It was found that I² = 27.8% (P = 0.216), indicating that there was no heterogeneity among the studies. A fixed-effects model was used for analysis. The results of the meta-analysis (MD = −0.99; 95% CI: −3.13 to 1.14; P = 0.363) are shown in Fig. 6B and indicate that aquatic exercise did not significantly improve upper limb joint mobility among older adults.

Cardiopulmonary function analysis

Five studies reported changes in cardiopulmonary function in 424 participants through the 2MST test. It was found that I² = 61.6% (P = 0.034), indicating heterogeneity among the studies, so a random-effects model was used for analysis. The results of the meta-analysis (MD = 7.56; 95% CI: 5.04 to 10.08; P < 0.001) are shown in Fig. 7, indicating that aquatic exercise can improve cardiopulmonary function in older adults.

Fig. 7.

Effects of aquatic exercise on cardiopulmonary function in older adults

Subgroup analysis

To explore the differential effects of intervention programs and exercise doses, subgroup analyses were performed on three main outcomes: TUG (dynamic balance), 30-s chair stand test (lower limb muscle strength), and chair sit-and-reach test (lower limb flexibility).

Before subgroup analysis, several studies were excluded due to missing key variables required for stratification. Specifically, three studies lacking clear age information were excluded from the age-based subgroup analysis; two studies that did not report intervention duration and two studies that failed to specify exercise frequency were excluded from the dose-based subgroup analyses.

Additionally, certain outcome indicators were excluded from subgroup analyses for methodological reasons:

1. Single-leg static balance and cardiopulmonary function (2MST test) were excluded because of insufficient sample size and incomplete reporting of duration or frequency parameters;

2. Upper limb muscle strength (30-s arm curl test) was excluded due to low heterogeneity (I² = 24.6%), indicating consistent results across studies;

3. Upper limb joint mobility (back scratch test) was excluded because the pooled effect was non-significant (MD = −0.99, 95% CI: −3.13 to 1.14), suggesting no clear subgroup differences.

These exclusions ensured that only outcomes with sufficient and comparable data (age, intervention duration, frequency, and exercise type) were included in subgroup analyses to maintain analytical robustness and interpretability.

The subgroup classifications were as follows: exercise type: resistance + aerobic combined training, resistance training alone, and aerobic training alone; exercise dose: intervention duration (≤ 12 weeks vs. > 12 weeks); intervention frequency (≤ 2 times/week vs. > 2 times/week); and population characteristics: age stratification (60–70 years vs. > 70 years). The specific analysis results are shown in Table 3.

TUG subgroup analysis

With respect to the TUG dynamic balance ability measure, a subgroup analysis was performed on the basis of age. Since 3 studies did not provide specific age details, 15 studies (466 participants) were included. For the 60–70-year-old age group, the meta-analysis results were MD = −1.00 (95% CI: −1.47 to −0.52; I² = 85.0%; P < 0.001); for the > 70-year-old age group, the meta-analysis results were MD = −0.91 (95% CI: −1.71 to −0.10; I² = 86.8%; P < 0.001). The improvement in dynamic balance ability was slightly better in the 60–70-year-old group. Owing to high heterogeneity in both groups, age was not the source of heterogeneity.

Subgroup analysis by intervention duration included 16 studies since 2 studies did not report detailed intervention periods. For the ≤ 12-week group, the meta-analysis revealed that aquatic exercise improved TUG balance ability in older adults (MD = −0.34; 95% CI: −0.60 to −0.08; I² = 34.7%; P = 0.190); for the > 12-week group, the meta-analysis revealed that aquatic exercise significantly improved TUG balance ability in older adults (MD = −0.87; 95% CI: −1.20 to −0.53; I² = 48.2%; P = 0.122).

Subgroup analysis by intervention frequency included 16 studies since 2 studies did not provide details on intervention frequency. For the ≤ 2 times/week group, the meta-analysis found MD = −0.66 (95% CI: −1.12 to −0.20; I² = 75.6%; P < 0.01); for the > 2 times/week group, the meta-analysis found MD = −1.12 (95% CI: −1.77 to −0.46; I² = 94.2%; P < 0.001), indicating that the > 2 times/week group had a slightly better improvement in dynamic balance ability.

Subgroup analysis by intervention program revealed that for the resistance + aerobic exercise group, the meta-analysis results were MD = −0.68 (95% CI: −1.23 to −0.13; I² = 83.9%; P < 0.001); for the resistance training alone group, the meta-analysis results were MD = −1.80 (95% CI: −3.07 to −0.52; I² = 84.9%; P < 0.001); and for the aerobic exercise alone group, the meta-analysis results were MD = −0.19 (95% CI: −0.52 to 0.13; I² = 0.0%; P = 0.009). These results suggest that resistance training alone in water has the greatest effect on improving dynamic balance among older adults.

30 s Chair stand test subgroup analysis

With respect to the lower limb muscle strength measurements, a subgroup analysis was performed on the basis of age. Since 3 studies did not provide specific age details, 12 studies were included. For the 60–70-year-old age group, the meta-analysis revealed MD = 2.27 (95% CI: 0.94 to 3.60; I² = 70.5%; P < 0.001); for the > 70-year-old age group, the meta-analysis revealed MD = 3.05 (95% CI: 0.43 to 5.67; I² = 89.0%; P = 0.017), indicating that the > 70-year-old age group had slightly better improvement in lower limb muscle strength.

Subgroup analysis by intervention duration included 13 studies since 2 studies did not report detailed intervention durations. For the ≤ 12-week group, the meta-analysis revealed that aquatic exercise improved lower limb muscle strength in older adults (MD = 1.72; 95% CI: 0.54 to 2.89; I² = 66.6%; P = 0.006); for the > 12-week group, the meta-analysis revealed that aquatic exercise improved lower limb muscle strength (MD = 1.88; 95% CI: 0.66 to 3.09; I² = 64.7%; P = 0.023), with the effect size slightly greater in the longer intervention group.

Subgroup analysis by intervention frequency included 13 studies since 2 studies did not provide details on intervention frequency. For the ≤ 2 times/week group, the meta-analysis showed that MD = 1.45 (95% CI: 0.23 to 2.68; I² = 71.2%; P < 0.001); for the > 2 times/week group, the meta-analysis showed that MD = 5.65 (95% CI: 2.29 to 9.02; I² = 99.1%; P < 0.001), indicating that the > 2 times/week group had better improvement in lower limb muscle strength than the ≤ 2 times/week group did.

Analysis based on the intervention program revealed that for the resistance + aerobic exercise group, the results of the meta-analysis were highly significant (MD = 3.98; 95% CI: 1.89 to 6.06; I² = 98.0%; P < 0.001). For the resistance training alone group, the meta-analysis showed MD = 1.25 (95% CI: 0.47 to 2.03; I² = 29.2%; P = 0.235), and for the aerobic exercise alone group, the meta-analysis showed MD = 3.20 (95% CI: 0.87 to 5.53; I² = 0%; P = 0.007), but the sample size was very small, making the reliability questionable.

Chair sit-and-reach test subgroup analysis

In the chair sit-and-reach test measure, a subgroup analysis was performed on the basis of age. Since 3 studies did not provide detailed age information, 10 studies were included. For the 60–70-year-old age group, the meta-analysis revealed MD = 4.97 (95% CI: −3.69 to 13.64; I² = 97.8%; P < 0.001); for the > 70-year-old age group, the meta-analysis revealed MD = 3.81 (95% CI: 2.24 to 5.38; I² = 34.4; P < 0.001), indicating that the > 70-year-old age group had better improvement in lower limb flexibility.

Thirteen studies were included in the subgroup analysis by intervention duration. For the ≤ 12-week group, the meta-analysis revealed that aquatic exercise improved lower limb flexibility in older adults (MD = 3.05; 95% CI: 0.01 to 7.31; I² = 98.7%; P < 0.001); for the > 12-week group, the meta-analysis revealed that aquatic exercise improved lower limb flexibility (MD = 3.66; 95% CI: 0.58 to 5.51; I² = 37.2%; P = 0.173), with a greater effect size than that of the short-term intervention group.

Thirteen studies were included in the subgroup analysis by intervention frequency. For the ≤ 2 times/week group, the meta-analysis revealed MD = 3.39 (95% CI: −0.46 to 7.23; I² = 98.9%; P < 0.001); for the > 2 times/week group, the meta-analysis revealed MD = 3.37 (95% CI: 1.14 to 5.59; I² = 26.9%; P = 0.233), indicating that aquatic exercise improved lower limb flexibility in the > 2 times/week group.

Analysis based on the intervention program revealed that for the resistance + aerobic exercise group, the results of the meta-analysis were highly significant (MD = 4.58; 95% CI: 0.70 to 8.47; I² = 97.1%; P = 0.021). For the resistance training alone group, the meta-analysis revealed MD = 3.21 (95% CI: −0.35 to 6.78; I² = 49.4%; P = 0.077), indicating that resistance training alone had a poor effect on improving lower limb flexibility in older adults. For the aerobic exercise alone group, the meta-analysis revealed MD = −1.05 (95% CI: −7.12 to 5.02; I² = 0%; P = 0.735), but the sample size was very small, making the reliability questionable.

Clinical interpretation of subgroup findings

From a clinical perspective, the subgroup differences observed in this study have practical implications for functional improvement in older adults. For example, a reduction of approximately 0.8 s in the Timed Up and Go (TUG) test exceeds the minimal clinically important difference (MCID) reported in geriatric populations (approximately 0.6–0.8 s), indicating a meaningful enhancement in dynamic balance and fall risk reduction. Similarly, the improvement of 2–3 repetitions in the 30-s chair stand test reflects a notable gain in lower-limb strength and daily activity capacity, such as standing up from a chair or climbing stairs. These results suggest that higher-frequency (> 2 times/week) and longer-duration (> 12 weeks) interventions may yield not only statistically significant but also clinically meaningful benefits for functional independence in elderly individuals.

Sensitivity analysis

To assess the reliability of the meta-analysis results, a sensitivity analysis was conducted for the four main outcomes (TUG test, 30 s chair stand test, sit-and-reach test, and 2MST test) by sequentially excluding each study. For the TUG test, after any study was excluded, the MD results ranged from −0.59 to −0.27 (I² = 91% to 96%, 95% CI: −0.92 to −0.14, P < 0.001 for all), indicating that the overall results of the TUG test are relatively stable. For the 30 s chair stand test after any study was excluded, the MD results ranged from 2.23 to 3.72 (I² = 93% to 98%, 95% CI: 1.99 to 4.01, P < 0.001 for all), indicating that the overall results of the 30 s chair stand test are relatively stable. For the sit-and-reach test, after any study was excluded, the MD results ranged from 1.62 to 5.03 (I² = 95% to 98%, 95% CI: 1.20 to 5.40, P < 0.001 for all), indicating that the overall results of the sit-and-reach test are relatively stable. For the 2MST test, after any study was excluded, the MD results ranged from 6.12 to 9.31 (I² = 70%, 95% CI: 4.02 to 9.73, P < 0.001 for all), indicating that the overall results of the 2MST test are relatively stable.

Publication bias

This study assessed publication bias for the primary outcomes of physical function: TUG balance test, 30 s chair stand muscle strength test, and sit-and-reach test. Egger’s test revealed that for the TUG balance test, P = 0.752 (Fig. 8A), indicating no publication bias; for the 30 s chair stand test, P = 0.829 (Fig. 8B), indicating no publication bias; and for the sit-and-reach test, P = 0.790 (Fig. 8C), indicating no publication bias (Fig. 9).

Fig. 8 Subgroup analysis combined forest plot.

Note 6: A represents the subgroup comparisons for the TUG test; B represents the subgroup comparisons for the 30-Second Chair Stand test; C represents the subgroup comparisons for the Chair Sit-and-Reach test

Fig. 9 Funnel plots of publication bias in the included studies.

Note 6: A represents the funnel plot for the TUG balance test; B represents the funnel plot for the 30 s chair stand test; C represents the funnel plot for the chair sit-and-reach test

GRADE evidence quality assessment

We used the Grading of Recommendations, Assessment, Development, and Evaluations (GRADE) approach to rate the overall quality of evidence for each primary outcome. The evidence quality was categorized into four levels: high, moderate, low, or very low, based on the following domains: risk of bias, inconsistency, indirectness, imprecision, and publication bias.

The GRADE summary of findings is presented in Table 4. Overall, the evidence for the improvements in physical function through aquatic exercise was rated as moderate to high quality. The outcomes of 30-s chair stand test, 30-s arm curl test, and chair sit-and-reach test were assessed as high-quality evidence. The outcomes of single-leg static balance test, time up and go test, back scratch test, and 2-min step test were rated as moderate-quality evidence, primarily due to downgrades for risk of bias (unclear randomization or deviation from interventions) and heterogeneity.

Table 4.

Quality of evidence

Certainty assessment							№ of patients		Effect		Certainty	Importance
№ of studies	Study design	Risk of bias	Inconsistency	Indirectness	Imprecision	Other considerations	Intervention	Control	RR (95% CI)	MD (95% CI)
Single-leg Static Balance Test
5	RCT	seriousb	not serious	not serious	not serious	none	295	191	-	MD 3.05 (0.85 to 5.25)	⨁⨁⨁◯ Moderate	CRITICAL
Time Up and Go Test
18	RCT	seriousa	not serious	not serious	not serious	none	533	407	-	MD −0.59 (−1.12 to −0.06)	⨁⨁⨁◯ Moderate	CRITICAL
30-Second Chair Stand Test
15	RCT	not serious	not serious	not serious	not serious	none	501	358	-	MD 2.77 (1.19 to 4.35)	⨁⨁⨁⨁ High	CRITICAL
30-Second Arm Curl Test
7	RCT	not serious	not serious	not serious	not serious	none	186	124	-	MD 3.17 (2.08 to 4.26)	⨁⨁⨁⨁ High	CRITICAL
Chair Sit-and-Reach Test
13	RCT	not serious	not serious	not serious	not serious	publication bias suspectedc	335	257	-	MD 3.21 (0.43 to 5.99)	⨁⨁⨁◯ Moderated	CRITICAL
Back Scratch Test
7	RCT	not serious	not serious	not serious	seriousa	none	172	152	-	MD −0.99 (−3.13 to 1.14)	⨁⨁⨁◯ Moderate	CRITICAL
2-Minute Step Test
5	RCT	seriousb	not serious	not serious	not serious	publication bias strongly suspectedc	424	219	-	MD 7.56 (5.04 to 10.08)	⨁⨁⨁◯ Moderate	CRITICAL

Overall, evidence grading followed the GRADE Working Group criteria. Each outcome was assessed across five domains: risk of bias, inconsistency, indirectness, imprecision, and publication bias. Downgrading by one level indicates a moderate concern within that domain

Footnotes for downgrading rationale

^aDowngraded one level due to unclear or inadequate randomization and deviations from intended interventions, resulting in potential risk of bias

^bDowngraded one level due to substantial inconsistency among studies (I² ≥ 50%), indicating high heterogeneity across included trials

^cDowngraded one level due to suspected publication bias based on visual asymmetry in the funnel plot or limited number of studies (< 10)

^dFor the “chair sit-and-reach test”, the evidence quality was downgraded from “high” to “moderate” due to extreme heterogeneity (I² = 97.8%), following GRADE methodological recommendations

Discussion

Evidence summary

In this study, a systematic review and meta-analysis, including 23 randomized controlled trials with a total of 1,179 elderly participants, was conducted to investigate the effects of aquatic exercise on the multidimensional physical functions of older adults. The results revealed that aquatic exercise significantly improved static balance ability (MD = 3.05; P = 0.007), dynamic balance ability (MD = −0.59; P = 0.028), upper limb muscle strength (MD = 3.17; P < 0.001), lower limb muscle strength (MD = 2.77; P = 0.001), lower limb joint mobility (MD = 3.21; P = 0.024), and cardiovascular endurance (MD = 7.56; P < 0.001) but had no significant effect on upper limb joint mobility (back scratch test (MD = −0.99; P = 0.363)). Subgroup analysis further revealed that a higher frequency (> 2 times/week) and longer intervention duration (> 12 weeks) produced more favorable effects. Additionally, resistance + aerobic combined training resulted in prominent improvements in lower limb muscle strength (MD = 3.98) and performance on the chair sit-to-stand test (MD = 4.58), whereas the pure resistance exercise group showed significant improvement in dynamic balance (MD = −1.80).

In this study, on the basis of a performance-based physical function assessment framework for older adults, the regulatory effects of aquatic exercise on multidimensional functional performance were systematically analyzed, with a focus on core attributes such as muscle strength, endurance, flexibility, and balance. By integrating evidence from 23 randomized controlled trials, the differential effects of resistance, aerobic, and combined training programs were quantified, and a quantifiable relationship between exercise dosage (e.g., > 2 times/week, > 12 weeks) and functional gains was established. These findings not only validate the physiological benefits of aquatic exercise but also provide evidence-based support for the development of personalized rehabilitation strategies that align with global strategies for healthy aging.

This study revealed that the effect of aquatic exercise on improving dynamic balance (MD = −0.59) was consistent with the findings of a meta-analysis by Kim et al. [36] (SMD = −0.21), but the heterogeneity (I² = 95.6%) was significantly greater than that reported in previous studies [9] (I² = 77%), which may be attributed to differences in the intervention protocols included in the studies. For example, the effect of pure resistance training on improving dynamic balance (MD = −1.80) was significantly better than that of the combined training group (MD = −0.68), which contradicts the conclusions of SeJun Oh et al. [27]. This contradiction may be due to differences in the “task-specific” design of resistance training: some studies have used functional resistance [10, 19] (e.g., aquatic step resistance), whereas others have used machine-based resistance [20, 32, 33] (e.g., water dumbbells), with the latter possibly lacking targeted stimulation for dynamic posture control. In addition, there was no significant improvement in upper limb joint mobility, which may be related to the lack of multiplanar compound movements for the shoulder joint in the training design, supporting Marco B’s proposed necessity for “task-oriented training” [11].

Mechanisms of exercise dosage and intervention plan

Subgroup analysis revealed the relationships between aquatic exercise dosage parameters (duration, frequency, and program) and functional improvements as follows:

Lower limb muscle strength analysis showed that the effect size for intervention duration > 12 weeks (MD = 1.88) was greater than that for ≤ 12 weeks (MD = 1.72). These findings indicate the presence of long-term cumulative adaptations involving neuromuscular remodeling and muscle hypertrophy. Repeated aquatic resistance exposure stimulates type II muscle fiber hypertrophy and enhances α-motor neuron recruitment efficiency, thereby improving coordinated motor unit activation [37–40].

Furthermore, chronic aquatic resistance training has been shown to upregulate sarcoplasmic reticulum Ca²⁺-ATPase activity and increase acetylcholine receptor density at neuromuscular junctions, which collectively improve excitation–contraction coupling and peak power output [41, 42].

High-frequency aquatic exercise (> 2 times/week) further amplifies these adaptations by increasing metabolic stress and lactate accumulation, which activate the mTOR signaling pathway and stimulate muscle protein synthesis, leading to increases in muscle fiber cross-sectional area [43–45].

These physiological mechanisms provide a biological basis for the observed dose–response relationship, whereby higher frequency and longer duration aquatic training yield superior gains in strength, balance, and endurance.

Pure resistance training improved dynamic balance the most (MD = −1.80), which may be related to a task-specific design (such as step resistance) targeting the vestibulospinal pathway. The resistance + aerobic combined program performed exceptionally well in improving lower limb muscle strength (MD = 3.98) and dynamic balance (MD = −0.68), as multijoint exercises (such as squats and leg presses) increase abdominal wall tension under fluid resistance, simultaneously improving power output and postural control [46]. The efficacy of such combined modality training is further corroborated by land-based research. Kuzu et al. (2025) reported that adding aerobic exercise to core stabilization exercises led to superior outcomes in functional capacity and fall risk reduction in geriatric individuals compared to core exercises alone, highlighting the synergistic benefits of targeting both stability and endurance systems [47].

The subgroup analysis for lower limb muscle strength revealed that high-frequency interventions (> 2 times/week) had a significantly greater effect size (MD = 5.65) than low-frequency interventions did (MD = 1.45). This mechanism is closely related to the unique physical properties of water, as increased speed can increase resistance by up to 4 times, forcing fast-twitch muscle fibers to be recruited first to overcome the high viscosity load. High-frequency stimulation, through lactic acid accumulation and metabolic product buildup, activates the mTOR pathway, promoting protein synthesis and increasing muscle fiber cross-sectional area. For indicators not included in the subgroup analysis, although the meta-analysis revealed a significant improvement in cardiovascular function (MD = 7.56), the dose–effect relationship needs further verification because of heterogeneity in the original studies (e.g., differences in the proportion of interval training). Upper limb joint mobility did not reach statistical significance (MD = −0.99), possibly because of the lack of multiplane movements of the shoulder joint or insufficient resistance load, suggesting the need for stronger task-oriented design.

In summary, optimizing aquatic exercise dosage parameters through multilevel physiological mechanisms works synergistically to ultimately enhance the musculoskeletal system, cardiovascular function, and neuromuscular control in older adults.

Heterogeneity and external validity considerations

Despite the application of random-effects models, substantial heterogeneity (I² > 90% in most analyses) was observed across several outcome measures. This high heterogeneity likely reflects the diversity in study designs, participant characteristics (e.g., age range 60–77 years, health status), and variation in aquatic exercise protocols, including exercise intensity, water temperature, depth, and supervision level. Additionally, inconsistency in measurement tools and outcome definitions across studies may have contributed to statistical variability.

From an interpretive perspective, such high heterogeneity suggests that the observed pooled effects should be interpreted with caution when generalizing to broader populations. Although the random-effects model accounts for between-study variability, the wide dispersion of true effects implies that the overall estimate represents an average trend rather than a uniform outcome. Therefore, while the results confirm the overall positive influence of aquatic exercise on functional performance in older adults, the magnitude of effect may vary substantially depending on specific intervention conditions.

To enhance external validity in future research, standardized intervention protocols (e.g., fixed resistance levels, water temperature, and duration) and consistent outcome measurement frameworks are needed. Furthermore, more detailed subgroup analyses and meta-regression based on exercise parameters and participant characteristics should be performed to better elucidate the sources of heterogeneity.

Study strengths and limitations

The strengths of this study include the following: (1) strict adherence to the PRISMA guidelines, including the inclusion of high-quality RCTs; (2) the use of a performance-based framework for assessing the physical function of older adults and analyzing the regulatory mechanisms of aquatic exercise on multidimensional functional performance; and (3) subgroup analyses revealing dose‒effect relationships, providing refined evidence for clinical decision-making.

This study has several limitations that should be acknowledged.

1. Some included trials had inadequate allocation concealment or blinding, which may increase the risk of bias and affect the reliability of pooled estimates.

2. Variability in intervention parameters—such as resistance intensity, water depth, and water temperature—across studies may have contributed to the observed heterogeneity and potential bias.

3. Incomplete reporting of exercise details and limited follow-up data restricted our ability to evaluate the long-term effects of aquatic exercise.

4. Although subgroup analyses were performed, the number of studies for certain outcomes was insufficient to conduct meta-regression or fully explore potential sources of heterogeneity.

5. Potential sample overlap among studies from the same research groups, particularly those conducted in Brazil and China, as well as the restriction to English and Chinese publications, may have introduced selection and publication bias.

Despite these limitations, this meta-analysis provides valuable and relatively high-quality evidence supporting the beneficial effects of aquatic exercise on multidimensional physical function in older adults and highlights the need for greater standardization in future randomized controlled trials.

Conclusion

Aquatic exercise effectively improved static/dynamic balance, upper limb muscle strength, lower limb muscle strength, lower limb joint mobility, and cardiovascular function in older adults. High-frequency (> 2 times/week) and long-term (> 12 weeks) combined resistance + aerobic interventions had better effects. Clinical practice should design task-oriented programs based on individual functional deficits (e.g., balance disorders vs. muscle weakness). For high-risk fall populations, aquatic step training focusing on resistance should be recommended, whereas for those with cardiovascular decline, aquatic intermittent aerobic training may be considered appropriate. This focus on fall prevention aligns with recent findings by Kuzu et al. (2025), where a combined exercise intervention significantly reduced fall risk in geriatric individuals with chronic pain, suggesting that the principles of multimodal training are effective across different exercise modalities and populations [47]. Future research should standardize exercise parameters (e.g., resistance intensity and water temperature) and explore the synergistic effects of aquatic exercise on psychosocial functioning.

Authors’ contributions

JH W plays a role in the following areas: research concepts and research design, literature review, data collection, data analysis and interpretation, statistical analysis, manuscript writing or reviewing/editing the draft of the manuscript. FY L plays roles in the following fields: literature review, data collection, and reviewing the initial draft of the manuscript. JW C plays a role in the following fields: literature review, data collection. HS L plays a role in the following fields: literature review, data collection. BJ Z plays a role in the following fields: literature review, data collection. BC L plays a role in the following fields: literature review, data collection and statistical analysis. P S plays a role in the following fields: literature review, data collection, and reviewing the initial draft of the manuscript.

Funding

The authors declare that no funds, grants, or other support were received during the preparation of this manuscript.

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

Ethical approval and informed consent are not applicable for this study. As a systematic review and meta-analysis, our research synthesizes data from already published articles that have obtained their own ethical approvals. No primary data involving human participants were collected by the authors.

Consent for publication

Not applicable. Since this study is a systematic review and meta-analysis that synthesizes data from already published literature, it does not involve any new data collected from participants that would require consent for publication.

Competing interests

The authors declare no competing interests.