The United Nations' World Population Prospects 2022 predicts that the proportion of the elderly population will rise to 16% by 2050 (United Nations Department of Economic and Social Affairs, Population Division, 2022). This demographic shift poses significant challenges for aging societies worldwide. Although aging does not inevitably cause severe cognitive deterioration, it does increase the risk of cognitive decline, which, if left unaddressed, can markedly impair older adults' quality of life. Moreover, cognitive decline may progress into serious neurodegenerative conditions such as Alzheimer's disease and dementia (Kazazi et al., 2018).

Physical exercise can mitigate cognitive decline in older adults through mechanisms such as enhanced neurogenesis, increased cerebral blood flow, reduced inflammation, and synaptic plasticity (De Miguel et al., 2021). It also supports brain networks for executive function, attention, and memory(Won et al., 2021), with aerobic exercise improving episodic memory (Johansson et al., 2022) and resistance training enhancing inhibitory control (Liu-Ambrose et al., 2010). However, older adults are prone to exercise-related injuries due to physical vulnerabilities. Aquatic exercise addresses this by leveraging water properties—buoyancy, hydrostatic pressure, and resistance—to enhance neuromuscular function in those with neurological conditions, whether through active forms such as swimming or passive immersion (Deng et al., 2024; Becker, 2009). Immersion at chest level improves blood circulation, venous return, and cerebral blood flow (Carter et al., 2014). while buoyancy reduces joint impact, prolongs balance reaction time, and alleviates fall-related anxiety (Kwok et al., 2024) Recent evidence further indicates that aquatic exercise may particularly benefit episodic memory and overall cognitive health in aging populations(Tang et al., 2022; Waller et al., 2016).

Common aquatic interventions for cognitive enhancement in older adults include: 1) aerobic exercise, involving continuous rhythmic activities (e.g., aquagym or walking) to boost cardiovascular fitness and cerebral blood flow; 2) high-intensity interval training (HIIT), alternating intense bursts with rest to enhance metabolic and vascular responses; 3) strength training, using water resistance or equipment to promote muscle strength, neurotrophic factors, and neuronal connectivity; and 4) cognitive-motor exercise, combining cognitive tasks (e.g., memory exercises) with motor activities to foster neuroplasticity and functional integration (Maneemai et al., 2024; So et al., 2024) Aerobic exercise may facilitate Aβ clearance across the blood-brain barrier by increasing low-density lipoprotein receptor-related protein-1 levels (Khodadadi et al., 2018). Resistance training promotes cerebral blood supply, enhances the expression of neurotrophic factors, and strengthens neuronal connectivity (Sáez de Asteasu et al., 2017). Cognitive-motor exercise simultaneously activates cognitive functions and motor systems, promoting neuroplasticity and functional integration (Cai et al., 2014).

Although various aquatic exercise modalities show promise for cognitive improvement in older adults, no head-to-head randomized controlled trials have compared their relative effectiveness. This gap hinders the selection of optimal interventions for clinicians and researchers. Network meta-analysis (NMA) addresses this by integrating direct and indirect evidence from RCTs, allowing simultaneous multi-intervention comparisons, effect rankings, and precise insights despite sparse direct data—extending pairwise meta-analysis for evidence-based geriatric decisions. Therefore, this review aimed to conduct a NMA to evaluate the effects of different aquatic exercise interventions on cognitive function in older adults.

This study was registered in the International Prospective Register of Systematic Reviews database (PROSPERO) under registration identification number CRD42025641350, in accordance with the Preferred Reporting Items for Systematic Reviews and Network Meta-Analyses (PRISMA-NMA) guidelines (Hutton et al., 2015).

The literature search was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA 2020) guidelines and the PRISMA extension for searching (PRISMA-S). The search aimed to identify randomized controlled trials (RCTs) evaluating the efficacy of aquatic exercise on cognitive function in elderly populations. Electronic databases were searched from inception through January 2026, with no language restrictions. Databases searched included Web of Science, PubMed, Embase, Cochrane Library, and China National Knowledge Infrastructure.

Additionally, hand searches of the reference list from included studies, relevant systematic reviews, and meta-analyses were conducted to identify any additional eligible trials.

Key search term groups were as follows (Supplementary Table 1):

Population (Elderly): "elderly" OR "older adults" OR "aged" OR "seniors" OR "geriatric" OR "aging" OR "elder".

Intervention (Aquatic Exercise): "aquatic exercise" OR "aquatic therapy" OR "hydrotherapy" OR "water-based exercise" OR "aquatic rehabilitation" OR "water aerobics" OR "swimming therapy" OR"swim".

Outcome (Cognitive Function): "cognitive function" OR "cognition" OR "cognitive impairment" OR "cognitive performance" OR "neuropsychological tests".

The primary outcome was the pre- to post- intervention difference in global cognition measured by validated cognitive assessments, including but not limited to the Mini-Mental State Examination (MMSE), MMSE-K, Montreal Cognitive Assessment (MoCA) (T-score), DSST, HDS-R, SDMT, Raven's Progressive Matrices test, and the Stroop color–word test. These outcomes were incorporated into the NMA. The categorization of activities included in this network meta-analysis was based on the descriptions provided in the original included studies.

Inclusion criteria were: 1) RCTs involving elderly participants (mean age ≥60 years); 2) interventions involving aquatic exercise therapy; 3) controls consisting of no intervention or other exercise therapies; 4) assessment of cognitive outcomes using validated tools (e.g., MMSE, MoCA, or domain-specific tests); and 5) sufficient data for effect size calculation.

Exclusion criteria included: 1) non-RCT designs; 2) studies on non-elderly populations; 3) interventions without an aquatic component; 4) studies lacking cognitive outcome measures; and 5) case reports, descriptive studies, or conference proceedings.

Reasons for exclusion at the full-text stage were documented. The PRISMA flow diagram details the number of records identified, screened, and included.

Once the search strategy was defined and executed across the databases, all retrieved articles were imported into reference management software (EndNote) to remove duplicates. Two reviewers independently screened all titles and abstracts of the search results for relevant articles. Full texts were then evaluated with respect to the inclusion and exclusion criteria. For any non-English publications identified during this process, two bilingual investigators independently assessed eligibility and extracted data, with relevant sections translated into English to ensure accurate interpretation. Any discrepancies were adjudicated by a third author.

Two researchers extracted data from the included studies using a standardized extraction form. Key extracted fields included first author, publication year, country/region, exercise modality, sample size (experimental and control groups), participant demographics (age and sex), intervention characteristics (type, duration, frequency, and intervention period), and primary outcome measures.

For studies with missing data, corresponding authors were contacted via email. If no response was received after the first attempt, a follow-up email was sent one week later. Studies with unresolved missing data after two contact attempts were excluded from the analysis.

The methodological quality and risk of bias of the 9 included studies were assessed using the Cochrane Risk of Bias tool (RoB 2.0) for randomized trials (Sterne et al., 2019).The RoB 2.0 tool, as applied to network meta-analyses (NMAs), was employed to evaluate individual study bias across five core risk domains, each addressing specific methodological aspects of trial design, conduct, and reporting. These domains are: 1) Bias arising from the randomization process; 2) Bias due to deviations from intended interventions; 3) Bias due to missing outcome data; 4) Bias in measurement of the outcome; and 5) Bias in the selection of the reported result(Sterne et al., 2019). Each domain was rated as "low risk," "some concerns" or "high risk". Disagreements during the assessment were resolved through discussion among reviewers.

The primary objective of this NMA was to investigate the effects of different aquatic exercise modalities on cognitive function in older adults. Given the unique structure of NMAs (Li et al., 2014), all experimental groups were defined as aquatic exercise interventions, while control groups included no intervention or alternative exercise therapies. The experimental and control groups were defined based on the original study's designation for studies comparing two aquatic exercise groups.

NMAs were conducted according to the following steps. First, heterogeneity across studies was evaluated using the I² statistic and the p-value from the Cochran's Q test. Interpretation followed established thresholds. An I² less than 30% was considered nonimportant heterogeneity, between 30% and 50% was considered moderate, between 50% and 75% was considered substantial, and 75% or higher was considered high heterogeneity (Higgins and Thompson, 2002). Funnel plots were used to assess potential publication bias or small-study effects (Hutton et al., 2014).

Second, the assumption of transitivity was examined through comparison of clinical and methodological features across included studies, and no major violations were observed. The network structure was grounded in graph theory, where interventions are represented as nodes and direct trial comparisons as edges, forming a connected evidence network(Li et al., 2014). The core principle lies in visualizing direct and indirect evidence integration through topological mapping. The network must satisfy connectivity, ensuring that at least one path exists between any two nodes. Nodes are weighted according to the number of trials, including the respective treatments. The larger the node's size and the thicker the lines, the more studies are involved(Li et al., 2014).

As a form of meta-analysis, NMA synthesizes both direct and indirect comparisons. Direct comparisons are derived from RCTs that explicitly evaluate two interventions, with effects pooled via meta-analysis to generate posterior distributions of relative efficacy (Thorlund and Mills, 2012). Indirect comparisons are based on the premise that indirect effect estimates are obtained from effect estimates from two comparisons sharing a common comparator (Han et al., 2022), and are enabled by shared comparators and constructed within a frequentist framework by modeling the network as a connected graph where interventions (nodes) are linked via direct or indirect pathways (Hutton et al., 2014). A node-splitting analysis was applied to evaluate potential incoherence between direct and indirect evidence. Thus, the validity and reliability of pairwise comparisons in NMA have been extensively validated across multiple studies (Hutton et al., 2014). The output provides probabilistic rankings of interventions, confidence intervals for pairwise effects, and surface under the cumulative ranking curve (SUCRA) values. The confidence in the evidence for all comparisons was assessed using the CINeMA framework. This framework evaluates six domains (within-study bias, reporting bias, indirectness, imprecision, heterogeneity, and incoherence) to ultimately rate the confidence in the evidence for each pair of interventions as High, Moderate, Low, or Very low (Nikolakopoulou et al., 2020).

Since all outcome measures in this study were continuous variables, the SMD with 95% CIs was selected as the effect size metric (Borenstein et al., 2010). All analyses were conducted using the R language (version 4.2.2) in the RStudio environment, employing the following packages: data manipulation: “tidyr”, “dplyr”; network visualization: “igraph”; meta-analysis: “netmeta”, “meta”, “metafor”. These packages were utilized to perform heterogeneity assessments, consistency evaluations, and comparative effect size analyses for the primary outcomes derived from cognitive assessment tools (MMSE, MMSE-K, MoCA T-score, DSST, HDS-R, SDMT, Raven’s Progressive Matrices test, and Stroop color–word test)(Béliveau et al., 2019).

A total of 903 records were included. After the removal of duplicates, the titles and abstracts of 391 records were screened, and the full texts of 21 articles were evaluated. Nine studies met the inclusion criteria and were included in this review and the network meta-analysis (Carral and Perez, 2007; Dunlap et al., 2024; Jang et al., 2021; Kang et al., 2020; Kooncumchoo et al., 2025; Kwok et al., 2024; Ploydang et al., 2023; Sato et al., 2015; So et al., 2024). The detailed selection process for including these trials is presented (Figure 1).

The results of the risk of bias assessment are reported in Figure 2. The assessment of trial quality was performed independently by two authors, according to the Cochrane Risk of Bias 2.0 (RoB 2) tool for randomized trials. Based on the RoB 2 assessment, none of the included studies were judged to be high risk of bias. Consequently, the pre-specified sensitivity analysis excluding high-risk studies could not be performed. In most studies, researchers were involved in the intervention delivery and outcome assessments, and the participants were not blinded to group allocation. After evaluating the risk of bias in the studies, six studies were considered to have some concerns arising from deviations from intended interventions (Carral and Perez, 2007; Jang et al., 2021; Kang et al., 2020; Kwok et al., 2024; Ploydang et al., 2023; Sato et al., 2015.). Overall, six studies were assessed as having some concerns (Dunlap et al., 2024; Kang et al., 2020; Kwok et al., 2024; Ploydang et al., 2023; Sato et al., 2015; So et al., 2024), and three studies were assessed as having low risk (Carral and Perez, 2007; Jang et al., 2021; Kooncumchoo et al., 2025).

This study included a total of nine randomized controlled trials (RCTs) conducted across diverse regions: two from Hong Kong, China (Kwok et al., 2024; So et al., 2024), two from South Korea(Jang et al., 2021; Kang et al., 2020), two from Thailand (Kooncumchoo et al., 2025; Ploydang et al., 2023), and one each from Spain (Sato et al., 2015.), the United States, and Japan (Carral and Perez, 2007). Among these, three studies enrolled exclusively female participants (Carral and Perez, 2007; Kang et al., 2020; Kwok et al., 2024), while the remaining six included both male and female populations(Dunlap et al., 2024; Jang et al., 2021; Kooncumchoo et al., 2025; Ploydang et al., 2023; Sato et al., 2015; So et al., 2024), with two studies focusing on mild cognitive impairment (Kooncumchoo et al., 2025; Ploydang et al., 2023), and another focusing on type II diabetes (Ploydang et al., 2023), collectively encompassing 324 participants aged 60-90 years.

Aquatic exercise interventions were categorized into six modalities, and some studies involved more than one modality: 1) Aquatic high-intensity interval training, 3 studies (Carral and Perez, 2007; Jang et al., 2021; Kwok et al., 2024), 2) Aquatic aerobic exercise, 4 studies (Kang et al., 2020; Kooncumchoo et al., 2025; Ploydang et al., 2023; So et al., 2024), 3) Aquatic cognitive-motor exercise, 3 studies (Dunlap et al., 2024; Kooncumchoo et al., 2025; Sato et al., 2015.), 4) Aquatic strength training, 2 studies (Carral and Perez, 2007; Sato et al., 2015.), 5) Land-based exercise, 3 studies (Jang et al., 2021; Kwok et al., 2024; So et al., 2024), and 6) Nonexercising control, 3 studies (Dunlap et al., 2024; Kang et al., 2020; Ploydang et al., 2023).

Intervention parameters varied in session duration (20-60 minutes/session, predominantly 45 minutes), frequency (1-3 sessions/week, mostly 2-3 sessions), and intervention length (6-12 weeks, with 12 weeks as the most common duration), incorporating formats such as aquatic-only, land-based, or hybrid training regimens (Table 1).

The heterogeneity analysis results indicated an I² statistic of 47.9% (95% CI: 0.0% - 75.8%), a Q statistic of 15.36 (df = 8), with a corresponding p-value of 0.0525. The results suggest moderate heterogeneity among studies (I² ≈ 48%); however, the 95% confidence interval is wide, ranging from no heterogeneity (0%) to substantial heterogeneity (75.8%). The p-value of the Q test is slightly greater than 0.05, indicating that the heterogeneity does not reach statistical significance.

To assess the efficacy of aquatic exercise for improving cognitive function in older adults, the network characteristics incorporated six interventions with seven direct pairwise comparisons (Figure 3). All trials were two-arm clinical trials. Participant distribution across intervention groups was as follows: Nonexercising control (39 participants), Land-based exercise (57 participants), Aquatic cognitive-motor exercise (concurrent or alternating aquatic exercise with cognitive training; 43 participants), Aquatic aerobic exercise (64 participants), Aquatic high-intensity interval training (repeated cycles of high-intensity aquatic exercise interspersed with lower-intensity recovery periods, typically comprising 6-10 intervals; 83 participants), and Aquatic strength training (strength and endurance training; 38 participants). The most frequently evaluated intervention was Aquatic aerobic exercise, reflecting their prominence in current aquatic exercise research for cognitive enhancement.

The consistency assumption was tested by comparing direct and indirect evidence, looking for overall differences, using a node-splitting model. The pairwise comparisons in the global analysis revealed no significant differences between direct and indirect evidence (all p-values > 0.05) (Table 2). The consistency between direct and indirect evidence across multiple intervention comparisons was further evaluated: the direct estimate ranged from –1.77 to 1.72; the indirect estimate ranged from -1.34 to 1.89; and the network estimate ranged from -1.26 to 1.76, further supporting the consistency between direct and indirect evidence across the examined comparisons (Figure 4).

Although SUCRA rankings indicate relative probability of being the best intervention, they do not quantify absolute clinical effect. Based on the results of the frequentist analysis, we evaluated the relative ranking of all six interventions using two complementary metrics: the P-score, which estimates the average probability of one treatment being better than the others, and the SUCRA (Surface Under the Cumulative Ranking Curve), which reflects the probability of a treatment being among the best (Figure 5). The SUCRA data synthesis showed that Aquatic cognitive-motor exercise reached the highest probability of being the best treatment, with a SUCRA value of 85.36% and a P-score of 86.05%. Aquatic aerobic exercise ranked second, with a SUCRA of 61.44% and a P-score of 65.71%, followed by Aquatic strength training (SUCRA = 51.80%, P-score = 50.72%) and Aquatic high-intensity interval training (SUCRA = 49.12%, P-score = 46.99%). Land-based exercise showed a SUCRA of 48.34% and a P-score of 46.13%. All exercise interventions demonstrated substantially higher probabilities of improving cognitive outcomes compared to the Nonexercising control (SUCRA = 3.94%, P-score = 4.40%). Confidence in the evidence for all comparisons in this network meta-analysis was rated as High using the CINeMA framework (Supplementary Figure 1).

A fixed-effect model was employed in the meta-analysis comparing the effects of exercise interventions versus the non-exercising control on cognition in older adults (Figure 6). The standardized mean difference (SMD) results for the various interventions relative to the control group were as follows: Aquatic strength training (F) [SMD = -0.13, 95% CI (-1.00, 0.74)], Land-based exercise (B) [SMD = 0.24, 95% CI (-0.58, 1.06)], Aquatic high-intensity interval training (E) [SMD = 0.40, 95% CI (-0.46, 1.25)], Aquatic cognitive-motor exercise (C) [SMD = 0.48, 95% CI (0.02, 0.95)], and Aquatic aerobic exercise (D) [SMD = 0.51, 95% CI (-0.08, 1.11)]. The horizontal axis ranged from -1 to 1, representing standardized mean difference values. Notably, only Aquatic cognitive-motor exercise (C) demonstrated a statistically significant difference, as its confidence interval did not include zero. The remaining comparisons showed non-significant effects.

Publication bias was assessed using a funnel plot. Visual inspection revealed that the data points were approximately symmetrically distributed around the vertical midline, indicating a low likelihood of significant publication bias. This was further supported by the statistical results, which showed that the majority of p-values were greater than 0.05. Although the p-value for Study 7 reached statistical significance (p = 0.008), and Study 3 approached it (p = 0.054), all other comparisons were non-significant (p = 0.114-0.919). The absolute t-values across all studies ranged from 0.10 to 2.74, with degrees of freedom varying from 10.10 to 65.55 (Supplementary Figure 2), training day.

The findings of the present study indicate that Aquatic cognitive–motor exercise, which integrates aquatic exercise with cognitive training within a single protocol, appears to be the most effective aquatic exercise modality for improving cognition in older adults. The second most effective intervention was Aquatic aerobic exercise. Although aquatic high-intensity interval training (HIIT) is currently a popular exercise modality among older adults, it appeared less effective for cognitive improvement in the present analysis. These findings have important implications for intervention planning, given that a greater proportion of existing studies have examined aquatic aerobic exercise or aquatic HIIT as strategies to improve cognition in older adults (Figure 3). Accordingly, clinicians and researchers may consider adopting Aquatic cognitive–motor exercise to enhance cognitive function in older adults.

Multiple studies have also confirmed the effectiveness and reliability of combining cognitive and motor exercises in healthy and cognitively impaired populations. Yang et al. (2020) reported that combined interventions can improve cognitive and physical functions in older adults with mild cognitive impairment (MCI). Venegas-Sanabria et al. (2024) comparing 13 different interventions, identified physical-cognitive exercise as one of the most effective nonpharmacological treatments for cognitive impairment. Yu et al. (2025) found that longer-term multi-domain cognitive-motor training may improve cognitive function in patients with amnestic MCI. Moreover, aquatic cognitive-motor exercise demonstrated a statistically significant advantage over the non-exercising control in the present NMA (Figure 6). providing further support for these findings.

The underlying mechanisms driving these observations are theoretically and biologically plausible. Cognitive-motor exercise enhances physical capacities, such as aerobic fitness, through its motor component, thereby improving brain structure and function. Simultaneously, the cognitive training component selectively stimulates higher-order brain functions, primarily improving memory, visuospatial abilities, attention, and other cognitive domains in individuals with mild cognitive impairment (MCI). These two components work synergistically, complementing and reinforcing distinct neural functions to produce a combined effect on cognition, maximizing intervention efficacy, and improving cognitive performance (Anderson-Hanley et al., 2018; Park et al., 2019; F. Yu et al., 2018). Cognitive-motor exercise has been shown to benefit cognitively frail populations by increasing gray matter density, improving functional efficiency in specific brain regions, enhancing functional connectivity within neural networks, modulating dopamine receptor density, and preserving white matter integrity (Toril et al., 2014). Furthermore, the combination of physical and cognitive training can synergistically induce neuroplasticity, promoting neurogenesis and synaptogenesis (Di Benedetto et al., 2017; Venegas-Sanabria et al., 2024). Notably, aquatic immersion, due to hydrostatic pressure, redistributes blood toward the upper body and increases cerebral blood flow. This has been confirmed by elevated blood flow velocity in the middle cerebral artery and posterior cerebral artery, providing a plausible explanation for the cognitive-enhancing mechanisms of aquatic exercise (Carter et al., 2014). Beyond hydrostatic effects, the thermal properties of water (e.g., warmth) may further promote relaxation, reduce psychological stress, and facilitate neural recovery, potentially amplifying cognitive benefits (Shukla, 2025). In addition, the physical properties of water, such as buoyancy and resistance, specifically augment cognitive-motor exercise by reducing joint stress and enabling safer dual-tasking compared to land-based equivalents, allowing greater focus on cognitive elements while minimizing fall risk and fatigue (Dunlap et al., 2024).

Cognitive-motor exercise is particularly relevant to activities of daily living, as many routine activities like crossing streets inherently require simultaneous cognitive and motor processing. Rezola-Pardo et al.'s study with 188 cognitively frail older adults further substantiated these findings, demonstrating that dual-task exercise targeting essential cognitive functions for daily living(such as attention, executive functions, and semantic memory) outperformed exercise alone in improving gait speed, physical function, cognition, and mood (Rezola-Pardo et al., 2019). Compared with conventional physical exercise, cognitive-motor exercise demonstrates superior efficacy in enhancing cognition-related activities, including optimizing postural control and motor performance, thereby elevating quality of life through this dual-benefit approach (Khan et al., 2022; Tao et al., 2022).

Other aquatic exercise methods besides Aquatic cognitive–motor exercise are also worth discussing and analyzing. Previous studies have demonstrated that aerobic exercise enhances cardiorespiratory function in older adults (De Miguel et al., 2021), and that the level of cardiorespiratory fitness is positively correlated with brain plasticity following exercise training (Berchicci et al., 2013; Perissiou et al., 2018). Nevertheless, Aerobic exercise may not represent the optimal intervention for improving cognition in older adults. One possible explanation is that the transition from exercise to cognitive-motor exercise induces significantly greater activation of the prefrontal cortex, thereby requiring greater cognitive engagement than aerobic exercise alone (Ding et al., 2024). This study also found aerobic exercise to be superior to strength training, which is consistent with the 2019 World Health Organization (WHO) guidelines titled "Risk reduction of cognitive decline and dementia"(Bull et al., 2020). In contrast, Wang et al.'s prior network meta-analysis reported that resistance exercise outperformed aerobic exercise—a discrepancy that may be attributable to differences in sample size, population characteristics, and the specific focus of the present analysis on aquatic strength training as a distinct subtype of resistance exercise(Wang et al., 2019).

The SUCRA value for aquatic HIIT in this review was 49.12%. While HIIT has consistently demonstrated favorable outcomes in other literature (Oliveira et al., 2024). The unique physiological and safety considerations of older adults may make it difficult to achieve the requisite training intensities in an aquatic environment. Differences between land and water environments could also contribute to its slightly inferior performance. More research on aquatic high-intensity interval training is needed to verify and further explore its effects. The close proximity between the SUCRA value of aquatic high-intensity interval training (49.12%) and that of the land-based group (48.34%) indicates that aquatic exercise can achieve at least comparable, if not superior, effectiveness to land-based training.

Several limitations of the present review warrant future consideration. First, the evidence base is limited, with only nine studies included and relatively small sample sizes, which may reduce the precision and reliability of SUCRA rankings. Second, the absence of long-term follow-up data precluded assessment of the sustainability of cognitive benefits, as well as subgroup analyses for specific populations such as those with MCI. Third, disparities in post-intervention sample sizes across studies may limit the generalizability and comparability of findings, and variations in exercise parameters (e.g., frequency, intensity, session duration) across trials may compromise the stability and comparability of the results. Clinical heterogeneity among included populations—ranging from healthy older adults to those with MCI or type 2 diabetes—was not formally examined through subgroup analyses, due to the insufficient number of studies within each subgroup. Furthermore, the aggregation of diverse cognitive outcomes into a single composite effect size using (SMD) introduces conceptual heterogeneity, as the included instruments — global screening tools (e.g., MMSE, MMSE-K, MoCA T-score, HDS-R) and domain-specific measures (e.g., DSST and SDMT for processing speed, Raven's Progressive Matrices for fluid reasoning, and the Stroop color-word test for executive function and selective attention)—assess overlapping, but distinct cognitive constructs. This heterogeneity, not fully captured by statistical indices, may compromise the construct validity, interpretability, and reliability of the pooled effect estimates. Additionally, most individual effect estimates were imprecise, as reflected by wide confidence intervals, suggesting insufficient statistical power rather than true null effects. Finally, the funnel plot's ability to detect publication bias is inherently constrained in a small, sparse network. These limitations highlight important priorities for future research without undermine the overall clinical relevance of the present findings.

The finding that all aquatic exercise modalities demonstrated cognitive benefits at least comparable to those of land-based exercise, offering a safe and effective alternative for older adults with mobility limitations, osteoarticular diseases, or fear of falling. In clinical practice, a hybrid "water-land combined" model may also be considered to leverage complementary stimuli, including exercise variety, engagement, and neuroplasticity. This study reveals gaps in the existing evidence regarding long - term effects, standardized protocols, and population-specific outcomes. Accordingly, clinicians are encouraged to systematically document training parameters and individual responses when implementing aquatic exercise interventions, thereby contributing to real-world evidence accumulation and supporting the development of more precise actionable clinical guidelines in the future.

In conclusion, this network meta-analysis is the first to comparatively evaluate multiple aquatic exercise modalities and suggests that aquatic cognitive–motor exercise may represent the most promising aquatic intervention for improving cognitive function in older adults (SUCRA = 85.36%), with aquatic aerobic exercise emerging as a viable alternative (SUCRA = 61.44%). Importantly, all aquatic exercise modalities yielded cognitive benefits at least as great as those of land-based exercise, supporting the integration of aquatic programs into exercise guidelines for aging populations. The finding that popular aquatic HIIT showed modest cognitive effects (SUCRA = 49.12%) suggests that intensity alone may be insufficient and that cognitive engagement appears critical for maximizing cognitive benefits. Future research priorities may include: 1) standardizing aquatic cognitive-motor training protocols with defined cognitive tasks and progression parameters; 2) conducting adequately powered RCTs with long-term follow-up; 3) exploring mechanisms through neuroimaging and biomarker studies; and 4) investigating effectiveness in specific populations including those with mild cognitive impairment and dementia, and 5) Explore the efficacy of aquatic exercise in older adults across the cognitive spectrum (from healthy to impaired).