The dose-response relationship of aquatic exercise for musculoskeletal pain relief: a systematic review and meta-analysis of randomized controlled trials Authors: Xuefeng Deng, Yiqi Li, Zhifang Wu, Ziyang Wang Published: 2026 ABSTRACT BackgroundMusculoskeletal pain constitutes a pervasive global health burden and represents a primary driver of disability worldwide. While aquatic exercise (AQE) is a promising intervention, its clinical application remains limited by the lack of quantitative dose guidance and standardized exercise prescription parameters. Therefore, this systematic review and meta-analysis aimed to evaluate the independent analgesic effect of AQE compared with passive controls and to explore the dose-response relationship between exercise dose and pain relief.MethodsFour major databases were searched from inception to January 2026 for randomized controlled trials (RCTs) comparing AQE with passive controls in adults with musculoskeletal pain. Standardized mean differences (SMDs) were calculated using random-effects models. A model-based network meta-analysis using Metabolic Equivalents of Task (METs) was conducted to model the dose-response relationship.ResultsTwenty-nine RCTs involving 2,210 participants were included. AQE yielded a moderate, statistically significant reduction in pain compared to passive controls (SMD = -0.55; 95% CI: -0.69 to -0.42; p < 0.001), with benefits persisting during the follow-up period (SMD = -0.51; 95% CI: -0.80 to -0.23; P < 0.001). Subgroup analyses indicated that middle-aged patients (45–59 years) and those with chronic disease durations (> 10 years) demonstrated larger pooled effect sizes. Among intervention modalities, pool-based multicomponent exercise yielded the largest effect size, followed by resistance and aerobic training. Regarding protocols, a frequency of 3 sessions/week, session duration > 60 min, and intervention periods ≥ 13 weeks were associated with more pronounced improvements. Crucially, dose-response analysis revealed an inverted U-shaped relationship, suggesting an effective therapeutic window of 450–1700 METs-min/week, with the peak analgesic effect observed at approximately 1200 METs-min/week.ConclusionsAQE appears to provide meaningful pain relief for musculoskeletal disorders, with benefits that may persist during follow-up. By identifying a peak therapeutic effect at approximately 1200 METs-min/week, this study aims to advance AQE from a qualitative recommendation toward quantitative, evidence-based guidance. These findings provide rigorous evidence base for understanding the physiological adaptations to aquatic therapy, informing tailored clinical management and optimized rehabilitation strategies.Trial registrationThe review protocol was registered in PROSPERO (CRD420261303731).Graphical Abstract Supplementary InformationThe online version contains supplementary material available at 10.1186/s12891-026-09953-0. Introduction Musculoskeletal pain is broadly defined by the International Association for the Study of Pain (IASP) as acute or chronic pain affecting bones, joints, ligaments, muscles, or tendons [1]. As the primary clinical manifestation of musculoskeletal disorders, it represents a paramount global health challenge, serving as a primary driver of disability and diminished health-related quality of life worldwide [2, 3]. Characterized by persistent pain and functional limitation, these conditions impose a substantial economic and societal burden [4, 5]. According to the Global Burden of Disease study, musculoskeletal disorders account for approximately 7% of total disability-adjusted life years, with low back pain, osteoarthritis, and fibromyalgia emerging as the predominant contributors [6]. Against the backdrop of a rapidly aging global population, the prevalence of these conditions is projected to escalate, rendering the identification of sustainable, safe, and effective management strategies an urgent public health priority [7]. Given the adverse reactions associated with drug therapy and its limited long-term efficacy, non-pharmacological interventions are receiving increasing attention in the management of musculoskeletal pain [8, 9]. Current recommendations from the Osteoarthritis Research Society International (OARSI), the American College of Rheumatology (ACR), and the European Alliance of Associations for Rheumatology (EULAR) generally emphasize exercise-based and other non-pharmacological interventions as key components of musculoskeletal pain management [10–12]. Among these approaches, aquatic exercise (AQE) has emerged as a particularly promising modality [13, 14]. AQE harnesses water’s unique hydrodynamic properties to maximize relief from musculoskeletal pain while improving physical function. Multiple randomized controlled trials (RCTs) support the efficacy of AQE in alleviating pain and restoring function across various musculoskeletal disorders [15–17]. These therapeutic benefits stem from the aquatic environment’s distinctive mechanical and thermal characteristics, making it a valuable rehabilitation modality for patients [18, 19]. However, while previous meta-analyses generally support the benefits of AQE, their clinical utility is compromised by significant methodological limitations [20–22]. Specifically, prior reviews frequently pooled land-based exercise with passive controls [21, 23, 24]. In clinical decision-making, establishing the fundamental, independent efficacy of an intervention against a non-active baseline is a prerequisite. Mixing these comparators artificially dilutes the analgesic effect size of aquatic exercise, leaving clinicians uncertain whether the observed benefits stem specifically from the unique properties of water or merely from general physical activity. To date, no study has employed Metabolic Equivalents of Task (METs) to model the dose-response relationship of AQE [25]. This metric is particularly meaningful in this context because it integrates intensity, duration, and frequency into a unified measure of exercise dose, enabling heterogeneous AQE protocols to be compared on a common scale [26]. Clarifying these two issues is essential for determining not only whether AQE is effective, but also the dose range at which it may provide clinically meaningful benefit. To address these gaps, this study performs a systematic review and meta-analysis designed to isolate the net effect of AQE by strictly comparing it against non-active controls, thereby minimizing potential confounding factors. To ensure clinical homogeneity and the reliability of the dose-response modeling, we explicitly defined the scope of “musculoskeletal pain” in this review to encompass low back pain, osteoarthritis, and fibromyalgia. These three conditions were strategically selected as they represent the primary global burden of chronic, non-malignant pain and constitute the core clinical indications for aquatic therapy. Furthermore, by employing METs to standardize exercise intensity, we aim to quantitatively elucidate the dose-response relationship between AQE and musculoskeletal pain relief. Ultimately, this study seeks to provide a solid scientific foundation for developing evidence-based standardized clinical exercise prescriptions. Methods To ensure methodological transparency and reporting completeness, the entire research process and reporting strictly adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement and its specific extension for dose-response meta-analyses [27, 28]. The review protocol was registered in PROSPERO (CRD420261303731). Search strategy We systematically searched four core databases: PubMed, Embase, the Cochrane Library, and Web of Science, from their inception to January 2026. The search was restricted to studies published in English, with no limitations on the date of publication. Developed based on the PICOS framework, the search strategy employed a combination of subject headings and free-text terms related to “musculoskeletal pain,” “aquatic exercise,” and “randomized controlled trials.” The detailed search strings are provided in Supplementary data 1. Additionally, supplementary searches were conducted by manually screening the reference lists of the included studies and relevant systematic reviews to identify further eligible trials. Inclusion and exclusion criteria Inclusion Criteria Studies were selected based on the following PICOS criteria: a) Participants: Adult patients with a clinical diagnosis of musculoskeletal pain disorders, such as osteoarthritis (OA), low back pain (LBP), or fibromyalgia (FM) [10, 29, 30]. b) Intervention: Structured AQE programs designed for therapeutic purposes. c) Comparison: To isolate the independent physiological effects of the aquatic intervention, comparators were strictly limited to passive control groups. Operationally, a ‘passive control’ was defined as the absence of any newly introduced, structured exercise or physical therapy regimens. Eligible comparators included waiting lists, no intervention, or standard usual care. d) Outcomes: The primary outcome was the change in pain intensity, measured by validated scales such as the Visual Analog Scale (VAS), Numeric Rating Scale (NRS), or disease-specific subscales. e) Study Design: Randomized controlled trials that provided sufficient data to extract or calculate key dose parameters. Exclusion Criteria Studies were excluded if they met any of the following conditions: (a) non-randomized designs; (b) Interventions that did not primarily involve active AQE; (c) Usage of active control groups; (d) Full text unavailable or insufficient data for effect size calculation; (e) Duplicate publications or overlapping datasets. Data extraction Literature screening was conducted independently by two reviewers. After removing duplicates, titles, abstracts, and full texts were sequentially screened to determine eligibility. The selection process strictly adhered to the PRISMA statement, with the flow of study selection illustrated in a flow diagram [28]. Data were extracted using a standardized piloting form, capturing the following domains: (a) Study Characteristics: First author, publication year, country, and sample size. (b) Participant Demographics: Disease type, mean age, sex distribution, and disease duration. (c) Intervention Specifications: Specific type of AQE, water temperature, and control group type. (d) Dose Parameters: Session duration, weekly frequency, and total intervention period. (e) Outcome Data: Means, standard deviations, and sample sizes for pain intensity at post-intervention and follow-up time points. Crucially, we extracted data only from specified pain subscales or standalone pain severity items (e.g., WOMAC, FIQ, or VAS). To ensure the accuracy of the pain outcomes, total composite scores and functional subscales were strictly excluded from the analysis. Any discrepancies that arose during the data extraction process were resolved through discussion or arbitration by a third reviewer. To harmonize the quantification of exercise intensity across studies, METs were employed. Based on the specific descriptions of exercise type and intensity reported in each trial, two reviewers independently assigned a standard METs value to each intervention group, referencing the 2024 Compendium of Physical Activities and its expansion [26]. All extracted and converted data were cross-checked to ensure accuracy, with discrepancies resolved by consensus with a third reviewer. Methodological quality assessment The risk of bias for the included RCTs was independently assessed by two reviewers using the Cochrane Risk of Bias tool, version 2 (RoB2) [31]. Each domain was graded as “low risk,” “some concerns,” or “high risk.” The overall risk of bias judgment for each study was then derived using the pre-defined algorithms within the tool. Given the specific focus on dose-response analysis in this study, particular attention was paid to potential biases related to the fidelity of exercise dose implementation and reporting during the assessment process. Any discrepancies between reviewers were resolved through discussion or by consultation with the corresponding author. Quality of the evidence We assessed the certainty of evidence for each outcome and subgroup using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) framework [32]. The initial evidence was rated as high quality but was downgraded by one level for each of the following criteria met: (i) publication bias, identified when 10 or more studies were pooled and Egger’s test was significant (p < 0.05) [33]; (ii) imprecision, defined as having a total pooled sample size of fewer than 400 participants [34]; (iii) risk of bias, indicated when over 25% of the included studies exhibited a high risk of bias (PEDro score < 6) [35] ; and (iv) inconsistency, denoted by substantial statistical heterogeneity (I² > 50%) [36]. Based on these assessments, the overall confidence in the evidence was classified into one of four levels: high, moderate, low, or very low [37]. Statistical analysis Meta analysis To leverage the distinct strengths of different analytical platforms, our workflow integrated both R and Stata. Meta-analyses and the complex dose-response modeling were primarily performed using the R statistical software environment (version 4.5.2) utilizing the meta and MBNMAdose packages, chosen for their advanced capabilities in handling Bayesian non-linear spline models [38]. Conversely, sensitivity analyses and the assessment of publication bias were conducted using Stata 14.0 (StataCorp, College Station, TX) due to its highly standardized and robust built-in macro commands for these diagnostic tests [39]. To ensure absolute cross-platform reproducibility, the exact same datasets were utilized in both software environments. All statistical tests were two-sided, with the significance level set at P < 0.05. Given the variability in pain assessment tools across the included studies, the therapeutic efficacy of AQE was synthesized using Standardized Mean Differences (SMDs) and their 95% Confidence Intervals (CIs). For continuous data where only standard errors (SEs) were reported, standard deviations (SDs) were estimated using standard transformation formulas outlined in the Cochrane Handbook for Systematic Reviews of Interventions [40–42]. Inter-study heterogeneity was assessed using the I2 statistic, with values of < 25%, 25%–50%, 50%–75%, and > 75% interpreted as negligible, low, moderate, and high heterogeneity, respectively [42, 43]. Given the inherent clinical and methodological heterogeneity across the included trials (e.g., encompassing varying musculoskeletal conditions, diverse aquatic protocols, and different intervention durations), a random-effects model was universally applied a priori for all quantitative syntheses, regardless of the observed I2 statistic. This approach aligns with current methodological guidelines, as the assumption of a single common true effect across such diverse clinical settings is implausible. In instances of substantial statistical heterogeneity (I2 ≥75%), potential sources were further investigated through subgroup and sensitivity analyses [36]. Dose-response analysis To quantitatively elucidate the mathematical dose–response relationship between AQE and musculoskeletal pain relief, we employed a Model-Based Network Meta-Analysis (MBNMA) framework using the “MBNMAdose” package within the R statistical environment (version 4.5.2) [43]. This approach was selected because the included trials evaluated aquatic exercise at varying dose levels, and most comparisons were made against shared passive control groups, making a dose-based network framework suitable for synthesizing relative treatment effects across a continuous dose scale. In this framework, intervention arms were characterized by their corresponding exercise dose (expressed as METs-min/week), while passive controls were treated as the reference condition, allowing different dose levels to be connected through shared comparators. A Bayesian random-effects model was constructed to account for between-study heterogeneity, incorporating natural cubic splines to capture potential non-linear trajectories in the dose–response curve [44]. (Supplementary data 3) Model consistency and fit were evaluated using the Deviance Information Criterion (DIC), and multiple candidate dose–response models were compared to determine the optimal functional form. (Supplementary data 3) Final estimates were reported as posterior medians with 95% Credible Intervals (CrIs) to reflect the probabilistic distribution of the treatment effects [45]. Subgroup analysis To explore potential sources of heterogeneity and identify the impact of specific clinical characteristics on treatment efficacy, subgroup analyses were performed based on pre-specified categorical variables. Differences between subgroups were assessed using the interaction P-value derived from the Chi-square test, with statistical significance defined as P < 0.05. The analyses were stratified by the following covariates: (a) Type of Intervention: Pool-based Aerobic Exercises vs. Pool-based Resistance Training vs. Pool-based Multicomponent Exercises; (b) Disease duration: 5–10 years vs. >10 years [46]; (c) Age groups: 18–44 years vs. 45–59 years vs. 60–74 years [47]. (d) Intervention Duration: ≤ 6 weeks vs. 7–12 weeks vs. ≥ 13 weeks [48]; (e) Weekly Frequency: 2 sessions/week vs. 3 sessions/week vs. ≥ 4 sessions/week, referencing guidelines [49]; (f) Session Duration: ≤ 45 min vs. 50–60 min vs. > 60 min [50]. Sensitivity analysis To examine the robustness of the pooled estimates and determine the influence of individual studies on the overall effect size, a leave-one-out sensitivity analysis was conducted. This involved iteratively excluding one study at a time and recalculating the summary effect size to ensure that the primary findings were not disproportionately driven by any single outlier [51, 52]. Meta-regression analysis To explore potential sources of between-study heterogeneity, univariable meta-regression analyses were performed. This approach was utilized to evaluate whether specific study-level continuous covariates—including water temperature, mean age, disease duration, intervention duration, weekly frequency, and session duration—significantly explained the observed variance in the analgesic effects of aquatic exercise [53]. Results As illustrated in Fig. 1, the initial search yielded 3,650 records. Following the removal of duplicates and preliminary screening, potentially eligible records were subjected to full-text review. After excluding studies due to unavailable full texts, ineligible control groups, or incomplete data, 29 studies ultimately met the inclusion criteria and were included in this systematic review. Fig. 1PRISMA flow diagram of the search process for studies PRISMA flow diagram of the search process for studies Study characteristics A total of 29 RCTs involving an aggregate of 2,210 participants with musculoskeletal pain were included in this review [14–19, 54–76]. The study population was predominantly female (80.8%), with a mean age of approximately 57.63 years. Geographically, the included trials originated from 14 different countries, with the highest concentration of studies conducted in Spain (n = 7; 24.1%), followed by Brazil (n = 4; 13.8%) and Australia (n = 3; 10.3%). Regarding intervention modalities, the protocols were categorized into pool-based aerobic exercises (n = 11), pool-based resistance training (n = 12), and pool-based multicomponent exercises (n = 6). In terms of exercise parameters, session duration ranged from 30 to 90 min, with a frequency of 2–5 sessions per week and a total intervention period of 4–52 weeks. Additionally, post-intervention follow-up data ranging from 3 to 18 months were reported in 8 studies. Additional details are provided in Table 1. Table 1Characteristics of included studiesStudyCountryDiseaseIntervention GroupSample size (Female/Man)Mean ageDuration of diagnosis (years)Setting (°C)Program Duration (weeks) / Frequency (sessions/week) / Session Duration (min)Follow-upBelza et al. (2002) [54]USAOA (Osteoarthritis)Pool-based Resistance trainingIG: 125 (107/18)CG: 124 (108/16)65.98 ± 5.9466.09 ± 6.1611.34 ± 9.1310.15 ± 8.4729.4 ~ 33.3 °C20/2/60NACochrane et al. (2005) [14]UKOA (Osteoarthritis)Pool-based Aerobic ExercisesIG: 153 (97/65)CG: 159 (99/60)69.86 ± 6.8269.63 ± 6.2611.3 ± 10.8913. 6 ± 11.1829 °C52/2/60YesFoley et al. (2003) [55]AustraliaOA (Osteoarthritis)Pool-based Multicomponent ExercisesIG: 35 (15/20)CG: 35 (15/20)73.0 ± 8.269.8 ± 9.0NANA6/3/60NAFransen et al. (2007) [56]AustraliaOA (Osteoarthritis)Pool-based Aerobic ExercisesIG: 55 (40/15)CG: 41 (34/7)70.0 ± 6.369.6 ± 6.18.28 ± 3.1134 °C12/2/60NAHale et al. (2012) [19]New ZealandOA (Osteoarthritis)Pool-based Multicomponent ExercisesIG: 23 (6/17)CG: 16 (4/12)73.6 ± 1.575.7 ± 1.1NA28 °C12/2/40NAHall et al. (1996) [57]UKRA (Rheumatoid Arthritis)Pool-based Resistance trainingIG: 35 (14/21)CG: 35 (10/25)55.8 ± 12.559.8 ± 9.39.7 ± 7.712.2 ± 9.636 °C4/2/30YesHan et al. (2011) [15]South KoreaLow Back PainPool-based Aerobic ExerciseIG: 9CG: 1061.2 ± 3.360.8 ± 5.0NANA10/5/50NAHinman et al. (2007) [16]AustraliaOA (Osteoarthritis)Pool-based Resistance trainingIG: 36(24/12)CG: 35(24/11)63.3 ± 9.561.5 ± 7.88.0 ± 9.38.0 ± 10.634 °C6/2/50YesLim et al. (2010) [58]South KoreaOA (Osteoarthritis)Pool-based Multicomponent ExercisesIG: 26 (23/3)CG: 24 (21/3)65.7 ± 8.963.3 ± 5.3NA34 °C8/3/40NALund et al. (2008) [18]DenmarkOA (Osteoarthritis)Pool-based Resistance trainingIG: 27 (22/5)CG: 27 (18/9)65 ± 12.670 ± 9.98.5 ± 5.484.5 ± 5.3333.5 °C8/2/50YesPatrick et al. (2001) [59]USAOA (Osteoarthritis)Pool-based Resistance trainingIG: 125(107/18)CG: 124(108/16)65.766.1NA29.4 ~ 33.3 °C20/2/45NAStener-Victorin et al. (2004) [61]SwedenOA (Osteoarthritis)Pool-based Resistance trainingIG: 15(8/7)CG: 15(9/6)70.365.5NA34 °C5/2/30YesWang et al. (2007) [60]ChinaOA (Osteoarthritis)Pool-based Aerobic ExercisesIG: 20(16/4)CG: 18(16/2)69.3 ± 13.362.7 ± 10.714.1 ± 11.812.9 ± 11.730 ~ 32 °C12/3/50NAWang et al. (2011) [62]ChinaOA (Osteoarthritis)Pool-based Aerobic ExercisesIG: 26 (22/4)CG: 26 (22/4)66.7 ± 5.667.9 ± 5.97.1 ± 6.36.2 ± 5.430 °C12/3/60NAAndrade et al. (2019) [63]BrazilFMS (Fibromyalgia Syndrome)Pool-based Multicomponent ExercisesIG: 27(22/0)CG: 27(24/0)48 ± 847 ± 89 ± 510 ± 530 ° C16/2/45YesCuesta-Vargas et al. (2011) [64]SpainFMS (Fibromyalgia Syndrome)Pool-based Aerobic ExercisesIG: 22(21/0)CG: 22(20/0)54.07 ± 8.5855.06 ± 9.865.12 ± 5.064.42 ± 4.1227.5 °C8/3/30NALetieri et al. (2013) [65]BrazilFMS (Fibromyalgia Syndrome)Pool-based Resistance trainingIG: 33 (33/0)CG: 33 (31/0)58.2 ± 10.659.6 ± 9.4NA33 °C15/2/45NAMunguia-Izquierdo et al. (2008) [66]SpainFMS (Fibromyalgia Syndrome)Pool-based Multicomponent ExercisesIG: 35 (29/0)CG: 25 (24/0)50 ± 746 ± 814 ± 1014±932 °C16/3/60NATomas-Carus et al. (2007) [67]SpainFMS (Fibromyalgia Syndrome)Pool-based Multicomponent ExercisesIG: 17 (17/0)CG: 17 (17/0)52.2 ± 6.651.9 ± 6.124 ± 919 ± 833 °C12/3/60YesMunguía-Izquierdo et al. (2007) [68]SpainFMS (Fibromyalgia Syndrome)Pool-based Multicomponent ExercisesIG: 35 (29/0)CG: 25 (24/0)50 ± 746 ± 814 ± 1014 ± 932 °C16/3/75NATomas-Carus et al. (2008) [69]SpainFMS (Fibromyalgia Syndrome)Pool-based Multicomponent ExercisesIG: 17 (15/0)CG: 16 (15/0)50.7 ± 10.650.9 ± 6.720.1 ± 8.019.4 ± 6.933 °C32/3/60NACedraschi et al. (2004) [70]SwitzerlandFMS (Fibromyalgia Syndrome)Pool-based Aerobic ExercisesIG: 84 (61/0)CG: 80 (68/0)48.9 ± 9.749.8 ± 9.88.4 ± 8.29.5 ± 9.634 °C6/2/90NAZijlstra et al. (2005) [71]NetherlandsFMS (Fibromyalgia Syndrome)Pool-based Aerobic ExercisesIG: 58 (55/3)CG: 76 (73/3)48 ± 2647± 23NANA3/3/180NALatorre et al. (2013) [72]SpainFMS (Fibromyalgia Syndrome)Pool-based Multicomponent ExercisesIG: 42 (39/0)CG: 30 (25/0)51.79 ± 7.8751.79 ± 7.87NANA24/3/60NASalm et al. (2019) [73]BrazilFMS (Fibromyalgia Syndrome)Pool-based MulticomponentExercisesIG: 14 (14/0)CG: 14 (14/0)50.4 ± 7.952.8 ± 10.0NA32 °C6/3/50NAIde et al. (2008) [74]BrazilFMS (Fibromyalgia Syndrome)Pool-based Multicomponent ExercisesIG: 20 (18/0)CG: 20 (17/0)46.61 ± 9.8045.47 ± 8.65NA32 °C4/4/60NAAltan et al. (2004) [75]TurkeyFMS (Fibromyalgia Syndrome)Pool-based Aerobic ExercisesIG: 24 (24/0)CG: 22 (22/0)43.14 ± 6.3943.91 ± 6.26NA37 °C12/3/25YesAlikhajeh et al. (2021) [76]IranLow Back PainPool-based Aerobic ExercisesIG: 12CG: 1251.6 ± 3.850.7 ± 3.3NA33 °C6/3/60NAKhodaveisi et al. (2018) [17]IranLow Back PainPool-based Multicomponent ExercisesIG: 15 (15/0)CG: 15 (15/0)35.16 ± 1.8036.25 ± 6.57NANA12/3/90NAIG Intervention group, CG Control group, NA Not applicable Table 1Characteristics of included studiesStudyCountryDiseaseIntervention GroupSample size (Female/Man)Mean ageDuration of diagnosis (years)Setting (°C)Program Duration (weeks) / Frequency (sessions/week) / Session Duration (min)Follow-upBelza et al. (2002) [54]USAOA (Osteoarthritis)Pool-based Resistance trainingIG: 125 (107/18)CG: 124 (108/16)65.98 ± 5.9466.09 ± 6.1611.34 ± 9.1310.15 ± 8.4729.4 ~ 33.3 °C20/2/60NACochrane et al. (2005) [14]UKOA (Osteoarthritis)Pool-based Aerobic ExercisesIG: 153 (97/65)CG: 159 (99/60)69.86 ± 6.8269.63 ± 6.2611.3 ± 10.8913. 6 ± 11.1829 °C52/2/60YesFoley et al. (2003) [55]AustraliaOA (Osteoarthritis)Pool-based Multicomponent ExercisesIG: 35 (15/20)CG: 35 (15/20)73.0 ± 8.269.8 ± 9.0NANA6/3/60NAFransen et al. (2007) [56]AustraliaOA (Osteoarthritis)Pool-based Aerobic ExercisesIG: 55 (40/15)CG: 41 (34/7)70.0 ± 6.369.6 ± 6.18.28 ± 3.1134 °C12/2/60NAHale et al. (2012) [19]New ZealandOA (Osteoarthritis)Pool-based Multicomponent ExercisesIG: 23 (6/17)CG: 16 (4/12)73.6 ± 1.575.7 ± 1.1NA28 °C12/2/40NAHall et al. (1996) [57]UKRA (Rheumatoid Arthritis)Pool-based Resistance trainingIG: 35 (14/21)CG: 35 (10/25)55.8 ± 12.559.8 ± 9.39.7 ± 7.712.2 ± 9.636 °C4/2/30YesHan et al. (2011) [15]South KoreaLow Back PainPool-based Aerobic ExerciseIG: 9CG: 1061.2 ± 3.360.8 ± 5.0NANA10/5/50NAHinman et al. (2007) [16]AustraliaOA (Osteoarthritis)Pool-based Resistance trainingIG: 36(24/12)CG: 35(24/11)63.3 ± 9.561.5 ± 7.88.0 ± 9.38.0 ± 10.634 °C6/2/50YesLim et al. (2010) [58]South KoreaOA (Osteoarthritis)Pool-based Multicomponent ExercisesIG: 26 (23/3)CG: 24 (21/3)65.7 ± 8.963.3 ± 5.3NA34 °C8/3/40NALund et al. (2008) [18]DenmarkOA (Osteoarthritis)Pool-based Resistance trainingIG: 27 (22/5)CG: 27 (18/9)65 ± 12.670 ± 9.98.5 ± 5.484.5 ± 5.3333.5 °C8/2/50YesPatrick et al. (2001) [59]USAOA (Osteoarthritis)Pool-based Resistance trainingIG: 125(107/18)CG: 124(108/16)65.766.1NA29.4 ~ 33.3 °C20/2/45NAStener-Victorin et al. (2004) [61]SwedenOA (Osteoarthritis)Pool-based Resistance trainingIG: 15(8/7)CG: 15(9/6)70.365.5NA34 °C5/2/30YesWang et al. (2007) [60]ChinaOA (Osteoarthritis)Pool-based Aerobic ExercisesIG: 20(16/4)CG: 18(16/2)69.3 ± 13.362.7 ± 10.714.1 ± 11.812.9 ± 11.730 ~ 32 °C12/3/50NAWang et al. (2011) [62]ChinaOA (Osteoarthritis)Pool-based Aerobic ExercisesIG: 26 (22/4)CG: 26 (22/4)66.7 ± 5.667.9 ± 5.97.1 ± 6.36.2 ± 5.430 °C12/3/60NAAndrade et al. (2019) [63]BrazilFMS (Fibromyalgia Syndrome)Pool-based Multicomponent ExercisesIG: 27(22/0)CG: 27(24/0)48 ± 847 ± 89 ± 510 ± 530 ° C16/2/45YesCuesta-Vargas et al. (2011) [64]SpainFMS (Fibromyalgia Syndrome)Pool-based Aerobic ExercisesIG: 22(21/0)CG: 22(20/0)54.07 ± 8.5855.06 ± 9.865.12 ± 5.064.42 ± 4.1227.5 °C8/3/30NALetieri et al. (2013) [65]BrazilFMS (Fibromyalgia Syndrome)Pool-based Resistance trainingIG: 33 (33/0)CG: 33 (31/0)58.2 ± 10.659.6 ± 9.4NA33 °C15/2/45NAMunguia-Izquierdo et al. (2008) [66]SpainFMS (Fibromyalgia Syndrome)Pool-based Multicomponent ExercisesIG: 35 (29/0)CG: 25 (24/0)50 ± 746 ± 814 ± 1014±932 °C16/3/60NATomas-Carus et al. (2007) [67]SpainFMS (Fibromyalgia Syndrome)Pool-based Multicomponent ExercisesIG: 17 (17/0)CG: 17 (17/0)52.2 ± 6.651.9 ± 6.124 ± 919 ± 833 °C12/3/60YesMunguía-Izquierdo et al. (2007) [68]SpainFMS (Fibromyalgia Syndrome)Pool-based Multicomponent ExercisesIG: 35 (29/0)CG: 25 (24/0)50 ± 746 ± 814 ± 1014 ± 932 °C16/3/75NATomas-Carus et al. (2008) [69]SpainFMS (Fibromyalgia Syndrome)Pool-based Multicomponent ExercisesIG: 17 (15/0)CG: 16 (15/0)50.7 ± 10.650.9 ± 6.720.1 ± 8.019.4 ± 6.933 °C32/3/60NACedraschi et al. (2004) [70]SwitzerlandFMS (Fibromyalgia Syndrome)Pool-based Aerobic ExercisesIG: 84 (61/0)CG: 80 (68/0)48.9 ± 9.749.8 ± 9.88.4 ± 8.29.5 ± 9.634 °C6/2/90NAZijlstra et al. (2005) [71]NetherlandsFMS (Fibromyalgia Syndrome)Pool-based Aerobic ExercisesIG: 58 (55/3)CG: 76 (73/3)48 ± 2647± 23NANA3/3/180NALatorre et al. (2013) [72]SpainFMS (Fibromyalgia Syndrome)Pool-based Multicomponent ExercisesIG: 42 (39/0)CG: 30 (25/0)51.79 ± 7.8751.79 ± 7.87NANA24/3/60NASalm et al. (2019) [73]BrazilFMS (Fibromyalgia Syndrome)Pool-based MulticomponentExercisesIG: 14 (14/0)CG: 14 (14/0)50.4 ± 7.952.8 ± 10.0NA32 °C6/3/50NAIde et al. (2008) [74]BrazilFMS (Fibromyalgia Syndrome)Pool-based Multicomponent ExercisesIG: 20 (18/0)CG: 20 (17/0)46.61 ± 9.8045.47 ± 8.65NA32 °C4/4/60NAAltan et al. (2004) [75]TurkeyFMS (Fibromyalgia Syndrome)Pool-based Aerobic ExercisesIG: 24 (24/0)CG: 22 (22/0)43.14 ± 6.3943.91 ± 6.26NA37 °C12/3/25YesAlikhajeh et al. (2021) [76]IranLow Back PainPool-based Aerobic ExercisesIG: 12CG: 1251.6 ± 3.850.7 ± 3.3NA33 °C6/3/60NAKhodaveisi et al. (2018) [17]IranLow Back PainPool-based Multicomponent ExercisesIG: 15 (15/0)CG: 15 (15/0)35.16 ± 1.8036.25 ± 6.57NANA12/3/90NAIG Intervention group, CG Control group, NA Not applicable Characteristics of included studies IG: 125 (107/18) CG: 124 (108/16) 65.98 ± 5.94 66.09 ± 6.16 11.34 ± 9.13 10.15 ± 8.47 IG: 153 (97/65) CG: 159 (99/60) 69.86 ± 6.82 69.63 ± 6.26 11.3 ± 10.89 13. 6 ± 11.18 IG: 35 (15/20) CG: 35 (15/20) 73.0 ± 8.2 69.8 ± 9.0 IG: 55 (40/15) CG: 41 (34/7) 70.0 ± 6.3 69.6 ± 6.1 IG: 23 (6/17) CG: 16 (4/12) 73.6 ± 1.5 75.7 ± 1.1 IG: 35 (14/21) CG: 35 (10/25) 55.8 ± 12.5 59.8 ± 9.3 9.7 ± 7.7 12.2 ± 9.6 IG: 9 CG: 10 61.2 ± 3.3 60.8 ± 5.0 IG: 36(24/12) CG: 35(24/11) 63.3 ± 9.5 61.5 ± 7.8 8.0 ± 9.3 8.0 ± 10.6 IG: 26 (23/3) CG: 24 (21/3) 65.7 ± 8.9 63.3 ± 5.3 IG: 27 (22/5) CG: 27 (18/9) 65 ± 12.6 70 ± 9.9 8.5 ± 5.48 4.5 ± 5.33 IG: 125(107/18) CG: 124(108/16) 65.7 66.1 IG: 15(8/7) CG: 15(9/6) 70.3 65.5 IG: 20(16/4) CG: 18(16/2) 69.3 ± 13.3 62.7 ± 10.7 14.1 ± 11.8 12.9 ± 11.7 IG: 26 (22/4) CG: 26 (22/4) 66.7 ± 5.6 67.9 ± 5.9 7.1 ± 6.3 6.2 ± 5.4 IG: 27(22/0) CG: 27(24/0) 48 ± 8 47 ± 8 9 ± 5 10 ± 5 IG: 22(21/0) CG: 22(20/0) 54.07 ± 8.58 55.06 ± 9.86 5.12 ± 5.06 4.42 ± 4.12 IG: 33 (33/0) CG: 33 (31/0) 58.2 ± 10.6 59.6 ± 9.4 IG: 35 (29/0) CG: 25 (24/0) 50 ± 7 46 ± 8 14 ± 10 14±9 IG: 17 (17/0) CG: 17 (17/0) 52.2 ± 6.6 51.9 ± 6.1 24 ± 9 19 ± 8 IG: 35 (29/0) CG: 25 (24/0) 50 ± 7 46 ± 8 14 ± 10 14 ± 9 IG: 17 (15/0) CG: 16 (15/0) 50.7 ± 10.6 50.9 ± 6.7 20.1 ± 8.0 19.4 ± 6.9 IG: 84 (61/0) CG: 80 (68/0) 48.9 ± 9.7 49.8 ± 9.8 8.4 ± 8.2 9.5 ± 9.6 IG: 58 (55/3) CG: 76 (73/3) 48 ± 26 47± 23 IG: 42 (39/0) CG: 30 (25/0) 51.79 ± 7.87 51.79 ± 7.87 Pool-based Multicomponent Exercises IG: 14 (14/0) CG: 14 (14/0) 50.4 ± 7.9 52.8 ± 10.0 IG: 20 (18/0) CG: 20 (17/0) 46.61 ± 9.80 45.47 ± 8.65 IG: 24 (24/0) CG: 22 (22/0) 43.14 ± 6.39 43.91 ± 6.26 IG: 12 CG: 12 51.6 ± 3.8 50.7 ± 3.3 IG: 15 (15/0) CG: 15 (15/0) 35.16 ± 1.80 36.25 ± 6.57 IG Intervention group, CG Control group, NA Not applicable Risk of bias Risk of bias assessment using the Cochrane RoB2 tool indicated moderate overall methodological quality among the 29 included RCTs. Specifically, 16 studies (55.2%) were classified as low risk, 9 (31.0%) raised some concerns, and 4 (13.8%) were deemed high risk. High-risk judgments were primarily driven by deficiencies in outcome measurement and the randomization process, while “some concerns” were most frequently associated with the selection of the reported result and deviations from intended interventions. (Supplementary Figs. 4.3) Meta-analysis results As depicted in Fig. 2a, AQE was associated with a statistically significant reduction in musculoskeletal pain compared to passive controls, yielding a moderate effect size (SMD = -0.55; 95% CI: -0.69 to -0.42; P < 0.001). However, moderate heterogeneity was observed among the studies (I2 = 55%). Regarding sustained benefits, Fig. 2b indicates that this analgesic effect persisted during the follow-up period. The pooled result demonstrated a consistent and significant reduction in pain (SMD = -0.51; 95% CI: -0.80 to -0.23; P < 0.001), with substantial heterogeneity detected (I2 = 61%). Fig. 2Forest plot of the overall and follow-up effects of aquatic exercise on pain in patients with musculoskeletal conditions. a Overall effect; (b) Follow-up effect Forest plot of the overall and follow-up effects of aquatic exercise on pain in patients with musculoskeletal conditions. a Overall effect; (b) Follow-up effect Subgroup analysis results To explore potential sources of heterogeneity and potentially beneficial intervention characteristics, subgroup analyses were conducted based on participant demographics and intervention characteristics, as illustrated in Fig. 3. Fig. 3Subgroup analysis of the effect of aquatic exercise on pain in musculoskeletal conditions Subgroup analysis of the effect of aquatic exercise on pain in musculoskeletal conditions Regarding intervention modalities, the subgroup analysis revealed that all three types of aquatic exercise yielded significant pain relief. Pool-based multicomponent exercise demonstrated the largest effect size among the subgroups (SMD = -0.74, 95% CI: -1.05 to -0.44, P < 0.001), followed by pool-based resistance training (SMD = -0.51, 95% CI: -0.83 to -0.19, P = 0.002) and pool-based aerobic exercise (SMD = -0.44, 95% CI: -0.58 to -0.30, P < 0.001). Participant Characteristics Regarding participant demographics, the analysis revealed variations in analgesic efficacy across age and disease duration. In terms of age, the magnitude of pain reduction appeared to be highest in the 18–44 age group (SMD = -0.91), although this finding was accompanied by substantial heterogeneity (I2 = 90%). In contrast, the 45–59 age group demonstrated a robust and significant effect (SMD = -0.69; 95% CI: -0.90 to -0.48; P < 0.001) with moderate heterogeneity (I2 = 50%), while the effect size was lowest in the 60–74 age group (SMD = -0.42). Additionally, disease duration influenced outcomes, with patients having a duration of > 10 years experiencing demonstrating a larger pooled effect size (SMD = -0.62; 95% CI: -0.91 to -0.34; P < 0.001) compared to those with a duration of 5–10 years (SMD = -0.44). In terms of intervention duration, long-term programs (≥ 13 weeks) demonstrated superior analgesic effects (SMD = -0.60; 95% CI: -0.80 to -0.40; p < 0.001; I2 = 54%), compared to medium-term (7–12 weeks; SMD = -0.58) and short-term interventions (≤ 6 weeks; SMD = -0.50). Regarding frequency, although interventions performed ≥ 4 times/week yielded the largest effect size (SMD = -0.90), this finding relies on a limited number of studies and may lack stability. Consequently, the 3 times/week protocol emerged as a highly effective observed frequency, showing substantial and significant pain relief (SMD = -0.67; 95% CI: -0.92 to -0.42; P < 0.001) that significantly exceeded the 2 times/week regimen (SMD = -0.44). Furthermore, regarding session duration, sessions lasting > 60 min (SMD = -0.68; 95% CI: -1.11 to -0.25; p < 0.001; I2 = 69%) exhibited greater efficacy compared to durations of 50–60 min (SMD = -0.63) and ≤ 45 min (SMD = -0.40). Dose-response relationship The results of the meta-analysis indicated that AQE exerts a clear analgesic effect on musculoskeletal pain. To further quantify the relationship between exercise dose and pain relief, a dose-response analysis was conducted. As shown in the Fig. 4, the results indicate that an exercise volume ranging from 450 to 1700 METs-min/week yields significant analgesic effects. Within this effective range, the peak analgesic effect was observed at approximately 1200 METs-min/week (SMD: -0.66; 95% CrI: -1.01 to -0.30). Fig. 4Dose-response relationship between aquatic exercise and pain reduction in musculoskeletal conditions Dose-response relationship between aquatic exercise and pain reduction in musculoskeletal conditions Heterogeneity and sensitivity analysis To address the moderate heterogeneity (I2 = 55%), we assessed the robustness of the results via a leave-one-out sensitivity analysis. By sequentially omitting individual studies and recalculating the pooled effect size, the SMD fluctuated narrowly from − 0.58 to -0.52 without significant deviation, while maintaining statistical significance in all instances (p < 0.001). (see Supplementary Figs. 4.2) This confirms that the primary findings are robust and not attributable to the influence of any single study. Meta regression Univariable meta-regression analyses were performed to explore potential sources of between-study heterogeneity. Water temperature was not significantly associated with pain relief (β = 0.02, 95% CI: -0.03 to 0.08, P = 0.394; R² = 0.2%). Likewise, weekly frequency, intervention duration, and session duration were not significant moderators of pain relief, with regression coefficients of β = -0.19 (95% CI: -0.41 to 0.03, P = 0.088; R² = 10.9%), β = -0.00 (95% CI: -0.01 to 0.01, P = 0.790; R² = 0.0%), and β = -0.00 (95% CI: -0.01 to 0.00, P = 0.554; R² = 0.0%), respectively. In contrast, disease duration was significantly associated with pain relief (β = -0.04, 95% CI: -0.09 to -0.00, P = 0.046; R² = 35.8%), indicating that a longer disease duration was associated with a smaller analgesic effect of aquatic exercise. Age was also identified as a significant moderator (β = 0.02, 95% CI: 0.00 to 0.03, P = 0.013; R² = 20.7%), suggesting that studies with older participants tended to report greater pain-relieving effects of aquatic exercise. (see Supplementary Figs. 4.1) Publication bias Potential publication bias was assessed using visual inspection of funnel plots and Egger’s linear regression test. As shown in Fig. 5, the funnel plot exhibited a generally symmetrical distribution of scatter points, with no apparent asymmetry observed. This observation was corroborated by Egger’s test (p = 0.147), which provided no statistical evidence of significant publication bias. Fig. 5Funnel plot assessing publication bias in studies of aquatic exercise for musculoskeletal pain Funnel plot assessing publication bias in studies of aquatic exercise for musculoskeletal pain Discussion This systematic review included 29 RCTs from 14 countries, systematically evaluating the analgesic effects of AQE in patients with musculoskeletal pain. Our meta-analysis results demonstrated that AQE significantly reduced pain compared to passive controls. Notably, this analgesic benefit was sustained, with significant pain relief persisting during the post-intervention follow-up period. Furthermore, subgroup analyses revealed that specific participant characteristics and intervention parameters were associated with treatment outcomes. Specifically, patients aged 45–59 years and those with a disease duration of > 10 years were associated with larger pooled effect sizes. Regarding intervention protocols, regimens consisting of 3 sessions/week, session durations > 60 min, and intervention periods ≥ 13 weeks yielded more consistent improvements. To further quantify exercise prescriptions, this study constructed a dose-response relationship model for the first time. We found that an exercise volume ranging from 450 to 1700 METs-min/week yielded significant analgesic effects, with the peak therapeutic effect observed at approximately 1200 METs-min/week. Despite the overall positive findings, moderate heterogeneity was observed in the primary outcome (I² = 55%). To investigate the potential sources of this heterogeneity, rigorous statistical analyses were performed. Sensitivity analysis confirmed the robustness of the pooled results, and Egger’s test revealed no significant publication bias, suggesting that the findings were not driven by individual outlier studies or publication tendencies. Concurrently, meta-regression analysis ruled out water temperature as a potential moderator. Consequently, the heterogeneity that could not be fully explained statistically may be attributed to unquantified variations across studies. Given that the included studies spanned 14 different countries (e.g., Spain, Brazil, Australia), inherent differences in cultural backgrounds, healthcare systems, and subjective pain perception thresholds across geographic regions may exist. Furthermore, variations in instructional styles, patients’ baseline fitness levels, and adherence rates during implementation may also serve as significant potential factors contributing to the observed variability in therapeutic efficacy [52]. In terms of intervention type, our analysis indicated that pool-based multicomponent, resistance, and aerobic exercises are all associated with effective pain relief. Crucially, addressing potential sources of variance, we evaluated whether intervention modality contributed to the overall heterogeneity. Although clinically distinct and showing a gradient in effect size (with multicomponent exercise yielding the largest reduction), the between-subgroup difference did not reach statistical significance (P = 0.21), and moderate heterogeneity persisted within the individual subgroups. This suggests that while intervention type is clinically meaningful, it does not fully explain the overarching statistical heterogeneity. Clinically, the comparable yet graded efficacy across modalities suggests that while aerobic exercise provides foundational cardiovascular and analgesic benefits, the incorporation of resistance training (especially within a multicomponent protocol) may offer supplementary advantages by enhancing local joint stability and muscular endurance. Ultimately, the choice of intervention should be tailored to the patient’s specific functional deficits. The analgesic efficacy of AQE in alleviating musculoskeletal pain can be attributed to the unique physiological mechanisms inherent to the aquatic environment. Primarily, the buoyancy of water counteracts gravity, significantly reducing joint mechanical load.1 As noted by Fransen et al., this allows patients to perform functional movements with minimized nociceptive input, thereby breaking the “pain-immobility” cycle [56]. Secondly, Khodaveisiet al. demonstrated that hydrostatic pressure exerts a compressive effect that facilitates venous return and lymphatic drainage, improving the local microenvironment by reducing tissue edema and accelerating metabolite clearance [17]. Furthermore, Sawant et al. highlighted that the thermal effects and fluid resistance of water not only induce muscle relaxation but also inhibit nociceptive signal transmission via sensory modulation, yielding a synergistic analgesic effect [77]. Collectively, these physical properties constitute the physiological basis for AQE in the pain management of patients with musculoskeletal disorders [18, 65, 78]. Dose-response analysis revealed a clear quantitative relationship between AQE and analgesic efficacy in musculoskeletal pain. An effective therapeutic window was identified ranging from 450 to 1700 METs-min/week, with 1200 METs-min/week identified as the peak observed dose. This pattern exhibits a distinct inverted U-shaped trend, suggesting that analgesic benefits increase with dosage up to a peak, beyond which benefits plateau, likely due to physiological adaptation limits or cumulative fatigue. Mechanistically, this effect may be attributed to the synergistic interplay of abundance-induced unloading, thermal effects, and neuroendocrine modulation, which is consistent with the findings of Lima et al. and Waller et al. [79, 80]. These findings provide a strong empirical basis for developing evidence-based exercise references and personalized intervention strategies in clinical practice. This study possesses several distinct strengths. Primarily, distinct from the earlier work by Verhagen et al., we evaluated the independent net effect of AQE by strictly excluding active comparators, thereby providing a purer estimate of its analgesic efficacy and suggesting that these benefits are sustained during follow-up [22]. Furthermore, building upon the broader findings of Ma et al. and Barker et al., our detailed subgroup analyses identified potentially beneficial exercise protocols based on specific participant profiles and intervention parameters, offering more granular guidance for clinical practice [20, 21, 24]. Most critically, by constructing a dose-response model, this study is the first to quantify the dose associated with the peak analgesic dose of AQE. This pivotal contribution helps to bridge the gap between qualitative recommendations and quantitative prescription, serving as a vital reference for the advancement of precision medicine and personalized rehabilitation. Limitations Nevertheless, several limitations should be acknowledged alongside corresponding recommendations for future research. First, evidence regarding long-term efficacy remains limited; with only eight studies providing follow-up data, the strength of inferences concerning sustained analgesic effects is restricted. Future trials should incorporate extended follow-up periods to thoroughly evaluate the durability of analgesia and identify appropriate maintenance doses. Second, despite rigorous statistical adjustments, moderate heterogeneity persisted across certain outcomes. This likely reflects inherent clinical variations in instructional styles, aquatic environments, and baseline patient characteristics. To mitigate this in future syntheses, primary studies are encouraged to report aquatic environmental parameters and specific instructional details with greater granularity to facilitate the precise exploration of heterogeneity sources. Third, the subgroup analyses present certain methodological constraints. The limited number of studies within specific sub-cohorts inherently reduces the statistical power of those pooled estimates. Furthermore, to facilitate clinical interpretability and provide actionable reference ranges for practitioners, we relied on categorizing several continuous variables (e.g., age, disease duration, and exercise parameters). We explicitly acknowledge that this approach may lead to a loss of information and potentially unstable findings driven by categorical cut-points. Consequently, the variations observed in these subgroup analyses are strictly exploratory and should be interpreted cautiously as potentially beneficial clinical trends, rather than absolute or definitive thresholds. Fourth, the dose-response model was constructed using study-level aggregate data rather than individual patient data, which may limit precision in capturing individual-level physiological variations. Where feasible, future meta-analyses should aim to synthesize individual patient data to generate more precise predictive models, advancing the field toward true personalized medicine. Finally, while this review investigated the umbrella term ‘musculoskeletal pain’, our inclusion criteria were deliberately restricted to low back pain, osteoarthritis, and fibromyalgia to maintain strict clinical homogeneity and ensure the reliability of the dose-response modeling. Consequently, the findings and potentially beneficial dosage protocols identified herein may not be fully generalizable to other specific musculoskeletal conditions, such as acute localized trauma or systemic inflammatory arthropathies. Conclusion This study supports the sustained, independent analgesic efficacy of aquatic exercise. Crucially, by constructing a dose-response model and identifying potentially beneficial intervention characteristics, we aim to advance aquatic therapy from qualitative recommendations toward quantitative, evidence-based guidance. These findings provide vital evidence base for informing tailored rehabilitation strategies in musculoskeletal pain management. Supplementary Information Supplementary Material 1. Supplementary Material 1.