1 BACKGROUND

Breast cancer (BC) is the second malignancy most commonly diagnosed in the world (first among women) (Bray et al., 2018). In 2018, one every four malignancies diagnosed in women were BC diagnosis (Bray et al., 2018). Thanks to the new advances especially in early diagnosis and targeted therapies, BC has an overall 5-year survival rate of 83.7% (Bray et al., 2018), which is one of the highest among all solid tumours. However, it is well-known that BC survivors experience a high burden due to the anti-cancer therapies received, leading to several impairments that can affect patients' functional capacity even years after treatment completion. Cancer-related persisting symptoms such as fatigue, musculoskeletal pain and feelings of anxiety or depression which commonly present as a cluster in survivors of BC can dramatically compromise patients' engagement in physical and social activities and decrease health-related quality of life (HRQoL) (Blaney et al., 2013; Bloom et al., 2007; Minton & Stone, 2008).

The effects of physical activity and exercise have been extensively studied in patients with cancer including BC survivors and are currently regarded as the best non-pharmacological treatment for several side-effect cancer-related therapies (CRT) (Ehlers et al., 2020). Both endurance and strength training have shown to be effective and safe in patients with cancer (during and after treatment) to improve not only physical capacity and symptom management but also to increase uptake of anti-cancer therapies (Bland et al., 2019) and to reduce the risk of recurrence and extend survival (Ammitzbøll et al., 2020; Hanson et al., 2016; Lopez et al., 2021; Madzima et al., 2017; Mctiernan et al., 2019; Meneses-Echávez et al., 2015; Sweegers et al., 2018).

Traditional physical exercise programmes in patients with cancer commonly include both strength and endurance training performed using fitness equipment (cycloergometers, treadmills, dumbbells, etc.) which is not always the patients' preference and frequently results in a reduction in overall attendance. For example, in a study published in 2011 among patients with breast and prostate cancer, almost 40% reported disliking exercise as a barrier to engage in physical activity (Ottenbacher et al., 2011). Adherence in patients with cancer is challenging and is influenced by cancer-related (anti-cancer therapies received, symptoms, stage of disease, etc.) and patient-related factors (co-morbidities, physical condition, age, socioeconomic status, education level etc.). In addition, in most cancer studies, adherence is poorly reported and/or achieves insufficient levels to yield expected results (Turner et al., 2018). Kim et al. (2020) pointed out the need to offer attractive interventions which could be more interesting or enjoyable for patients to ensure long-term adherence, especially among those who experience CRF who face greater barriers to exercise (Kim et al., 2020). In this sense, some studies have shown that water-based exercise can increase adherence comparing to other interventions because it is a comfortable environment, accessible in the community and has the extra beneficial effects of water immersion such as buoyancy (Cantarero-Villanueva et al., 2012; Dionne et al., 2018; Torres-Ronda & del Alcázar, 2014) which decreases joint stress and thus can especially benefit those with shoulder-neck mobility restrictions, pain and/or overweight (Kutzner et al., 2017; McIlroy et al., 2017; Torres-Ronda & del Alcázar, 2014). Furthermore, some studies have also found that water-based exercise can be equally effective or even superior to land-based exercise in improving muscle strength, balance, flexibility, symptom control and functional exercise capacity in different clinical populations (Bergamin et al., 2013; Bocalini et al., 2008; Siqueira et al., 2020; Zoheiry et al., 2017). Despite these initial investigations, there is still a lack of knowledge on the effects of water-based exercise compared to land-based interventions particularly in BC patients. Because there is large variation in the application of aquatic therapy among the studies published and its specific effects on different outcomes remain unknown, in this systematic review, we aimed to synthetize the existing evidence related to the effects of aquatic therapeutic exercise on the following outcomes associated with BC side effects: pain, shoulder mobility, lymphedema, cardiorespiratory fitness, muscle strength, body composition, pulmonary function, cancer-related fatigue and health-related quality of life, both compared to usual care or land-based exercise interventions. As secondary objectives, we aimed to (1) examine adherence to the aquatic therapeutic exercise interventions and (2) explore which exercise parameters (intensity, modality, frequency of training, etc.) may be more effective to improve the aforementioned outcomes in this population.

2 METHODS

The protocol of the systematic review was registered prospectively in the PROSPERO International Prospective Register of Systematic Reviews (CRD42021236555) and can be consulted online. No changes to the protocol were made after registration. This systematic review follows the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement (Liberati et al., 2009), and following this guideline, a specific question is proposed: ‘Is an aquatic intervention effective to improve side effects compared to usual care or other physical interventions for patients who had breast cancer?’

2.1 Eligibility criteria

To retrieve relevant papers, the search was organised using a PICO strategy (acronym for Patient, Intervention, Comparison and Outcome). The population were women diagnosed with BC, the intervention was any type of aquatic therapeutic exercise and there were no initial restrictions regarding the comparison group (usual care, other type of physical interventions) and the outcomes. The eligibility criteria for the inclusion of the studies are detailed in the Table 1. Records were excluded in the screening of titles and abstract if they were protocols with no outcomes, guidelines, reviews, case–control studies and cross-sectional studies books; if they had no aquatic intervention and the patients had other types of cancer or the intervention was not performed on humans or the studies were not relevant for the purpose of the review. In the full-text screening, records were excluded if they fulfilled the previously mentioned characteristics, the intervention was of less than 1 session per week and the patients had other type of cancer or articles were not available.

TABLE 1. Eligibility criteria Inclusion criteria Acronym Definition P Patients

Women that had BC.

>18 years

I Intervention Aquatic exercise intervention C Comparator No restrictions O Outcomes No restrictions S Study design and characteristics

RTCs

Non-randomised trials.

Single-arm pre-post studies.

- Language English or Spanish Exclusion criteria

Guidelines, reviews, opinions, editorials, commentaries, letters, conference abstracts, case–control and case series studies.

Studies including patients with other types of cancer.

Studies conducted in animals.

Patients with secondary lymphedema.

Abbreviations: BC, breast cancer; RCT, randomised controlled trials. 2.2 Search strategy

The following databases were used for conducting the search: Medline (via PubMed), Web of Science, Scopus, Cumulative Index to Nursing and Allied Health Literature (CINAHL) and the Cochrane Library. MeSH terms and keywords were used to make the query. Databases were accessed via The University of Granada, Spain and TecnoCampus University – UPF, Spain. No publication date was imposed or any other additional filters to the search, and the last search was conducted on 14 September 2020. Also, cross-references were additionally searched in eligible full-text articles. Table 2 shows the chosen search terms which were used for Pubmed that were subsequently modified to fit each database. Titles and abstract were identified and selected by two independent reviewers that were added to the set of eligible papers. To update the search, alerts in each database were set up.

TABLE 2. Terms that compose the search strategy for PubMed Concept Terms #1 Intervention ((aquatic AND (exercise[MeSH Terms] OR exercis*[Text Word] OR physical activit*[tiab] OR physical exercis*[tiab] OR acute exercis*[tiab] OR isometric exercis*[tiab] OR aerobic exercis*[tiab] OR exercis* training[tiab] OR resistance training[MeSH Terms] OR resistance training[Text Word] OR strength* training[tiab] OR weight-lifting strength*[tiab] OR weight-lifting exercis* program[tiab] OR weight-bear* strength* program[tiab] OR weight bear* exercis* program[tiab] OR weight bear* exercis* program[tiab] OR exercise therapy[MeSH Terms] OR exercis* therap*[Text Word] OR remedial exercis*[tiab] OR rehabilitation exercis*[tiab])) OR aquatic exercis*[tiab] OR aquatic physiotherapy[tiab] OR aquatic therapy[tiab] OR aquatic physical therapy[tiab]) #2 Condition (neoplasms[MeSH Terms] OR neoplasia*[tiab] OR neoplasm*[tiab] OR tumor*[tiab] OR cancer*[tiab] OR malignanc*[tiab] OR malignant neoplasm*[tiab] OR benign neoplasm*[tiab]) #1 AND #2 2.3 Databases' sensitivity and precision

To determine the quality of databases, an analysis of their sensitivity and precision was calculated.

2.4 Selection of studies

The duplicates were removed using Mendeley (Version 1.13.8, Windows, Elsevier). Titles and abstracts were scanned by the independent reviewers to determine eligibility of the papers using Microsoft Office Excel 2019 (Version 2019 Windows, Microsoft Corporation) template to review the full-text studies and determine inclusion. The reasons for exclusion were recorded, and differences between reviewers were discussed, and where no agreement was reach, a third reviewer was consulted.

2.5 Data extraction

One reviewer collected the data of the different studies covering the following aspects: characteristics of included studies (authors and year of publication), population characteristics (sample size and type of population included) intervention, comparison group, outcomes, assessment times and main results (Table 3). These items were discussed and agreed previously by all authors of the study. The detailed intervention from each study is summarised in Supporting Information S1.

TABLE 3. Characteristics of studies Author/Year publication Participants/Comparison or control group Outcomes Measured time points Adverse effects Main results RCT studies Cantarero-Villanueva et al. (2012) Women with BC. Aquatic exercise intervention (n = 33) Control group (n = 33).

Visual analogue scale

Electronic algometer

Presence of trigger points

Criteria described by Simons et al. (Sterne et al., 2019)

Baseline Post-intervention (8 wk) Transient increase of edema (n = 3), increased fatigue in the first session (n = 4). The aquatic exercise group showed benefits with large intergroup effect size for neck pain (d = 1.1, 95% CI 0.81–1.75), and moderate for shoulder-axillary (d = 0.70, 95% CI 0.14–1.40). For PPT in levels over C5-C6, the intergroup effect size was large (d = 1.5, 95% CI 0.21–2.77) and small for the rest PPT levels (d < 0.11). Also, a greater reduction of active TrPs than the control group (p < 0.05). Cantarero-Villanueva, Fernández-Lao, Cuesta-Vargas, et al. (2013) Women with BC. Aquatic exercise intervention (n = 34) Control group (n = 34). Strength (abdominal and leg)

Trunk curl static endurance test

Multiple sit-to-stand test

Piper fatigue scale

Profile of mood states

Baseline Post-intervention (8 wk) 6-month follow-up Discomfort or low-intensity pain/stiffness.

The aquatic exercise group showed benefits with a large intergroup size effect in trunk endurance (d = 0.92, CI 1.97–3.83), leg strength (d = 1.10; 0.55–2.76), and cancer-related fatigue (d = 0.87, CI 0.48–1.26). At the follow-up period of 6 months, small to large effect sizes regarding cancer-related fatigue, trunk and leg strength (0.25 > d > 0.90).

The effect size for mood state were negligible (d < 0.25).

Johansson et al. (2013) Women with BC Aquatic exercise intervention (n = 15) Control group (n = 14) Range of motion (shoulder)

Goniometer

Bioimpedance spectroscopy

Baseline Post intervention No adverse effects The aquatic exercise intervention had significant effect on the range of motion of the shoulder (p < 0.05) but had no effect on lymphedema compared with the control group. No effect sizes were reported Odynets et al. (2018)

Women with BC with post mastectomy pain syndrome.

Intervention (n = 115):

Aquatic exercise intervention (n = 45)

Pilates intervention (n = 40)

Yoga intervention (n = 30)

Healthy women (n = 50)

Forced spirometry

Baseline 12-months follow-up Not reported All groups improved pulmonary function compared with the group of healthy women (p < 0.05) (except for expiratory reserve volume and maximal voluntary ventilation). Odynets, Briskin, and Todorova (2019) Women with BC Aquatic exercise intervention (n = 45) Pilates intervention (n = 40) Yoga intervention (n = 30)

Functional assessment of Cancer therapy questionnaire, specific module for breast cancer patients (FACT-B).

Baseline 6 months follow-up 12 months follow-up Not reported

All groups had significant improvement of quality of life at 6 months program (p < 0.01).

After 12 months, the aquatic exercise intervention showed benefits in emotional (d = 0.67), physical (d = 0.38), social (d = 0.42) and functional well-being (d = 0.10) compared to the Pilates intervention as well as to a yoga intervention (d = 0.10 to 0.80). The BR subscale also showed moderate improvements (d = 0.27)

Odynets, Briskin, Zakharina, et al. (2019) Women with BC Aquatic exercise intervention (n = 34) Pilates intervention (n = 34)

Spirometry

Baseline 3 months follow-up No adverse effects The aquatic exercise intervention was more effective improving pulmonary function parameters (vital capacity, forced vital capacity, maximal voluntary ventilation, and expiratory reserve volume) (p < 0.01) than a Pilates exercise program. No effect sizes were reported. Tidhar and Katz-Leurer (2010) Women with BC and lymphedema. Aquatic exercise intervention (n = 16) Control group (n = 32)

Limb volume by a water displacement device.

Upper limb lymphedema questionnaire (ULL27)

Before and after the first (baseline) and last session (12 wks). No adverse effects Aqua lymphatic therapy had a significant immediate (p < 0.02) but not long-term effect on limb volume. Quality of life improved in the emotional dimension (p = 0.03) and social dimension (p = 0.001) but not in the physical dimension (p = 0.39). Non-RCT studies Broach and Norrell (2019)

Women with BC.

Aquatic exercise intervention (n = 12)

Usual care (n = 7)

12-minute walk test

Body mass index and bioimpedance

Volumometer

Multidimensional fatigue inventory short form

Impact of event scale

Health-related quality of life (HRQoL) scale developed by the Centers for Disease Control and Prevention

Baseline Post intervention (8wk) Not reported An aquatic exercise program improved body composition (p = 0.03), cardiovascular fitness (p = 0.001), emotional distress (p = 0.002) and fatigue (p = 0.038). No effects were found in arm edema or healthy day scores. No significant differences occurred in the control group. Effect sizes were not reported. Cantarero-Villanueva, Fernández-Lao, Caro-Morán, et al. (2013)

Women with BC.

Aquatic exercise intervention (n = 20)

Control group (n = 20)

Body mass index and waist circumference.

Pressure pain threshold (neck, shoulder, hand and leg)

Electronic algometer

Piper fatigue scale

Baseline Post-intervention (8 wk) No adverse effects An aquatic exercise intervention showed benefits with large intergroup effect size for pressure pain threshold in cervical area (d = 1.49), unaffected hand area (d = 1.15) and leg area (d > 1.15); and moderate for affected shoulder (d = 0.82). For waist circumference the intergroup effect size was moderate (d = 0.580). No effects were found in cancer-related fatigue and body mass index. Fernández-Lao et al. (2013)

Women with BC.

Aquatic exercise intervention (n = 33)

Land-based intervention (n = 31)

Control group (n = 34)

Bioimpedance (weight, body mass index, fat and lean mass) and waist circumference.

Circumferential at 5 and 10 cm below the distal border of the lateral epicondyle of the humerus.

Spanish version of European Organization for Research and Treatment of Cancer Breast Cancer-Specific Quality of Life (EORTC QLQ-BR23) questionnaire

Baseline Post-intervention 6 months follow-up Not reported

The land-based intervention showed greater decrease in fat percentage than the aquatic exercise intervention (p < 0.001), and shower a greater increase in lean body mass (p = 0.008). Both intervention groups showed a decrease in waist circumference compared with the control group (p ≤ 0.003). There were no effects in body weight and body mass index.

The land-based intervention also showed greater improvement in 5-cm forearm circumference than the aquatic exercise intervention (p = 0.024).

The aquatic exercise intervention significantly improved breast symptoms compared to the land and control groups (p < 0.001).

Effect sizes were not reported.

Single-arm pre-post study Siqueira et al. (2020)

Women with BC.

Aquatic exercise intervention (n = 21)

Range of motion (shoulder)

Computerised, photogoniometer with image analysis.

Baseline After 7 wks. Post-intervention (12 wks) No adverse events The aquatic exercise program improved significantly the range of shoulder motion of abduction, flexion, external rotation, adduction (p < 0.05), but no extension nor internal rotation. Abbreviations: BC, breast cancer; CI, confidence interval; PPT, pressure pain threshold; RCT, randomised controlled trial; ROM, range of motion; wk, week. 2.6 Assessment of the risk of bias

Two independent reviewers assessed the risk of bias. The instruments used were the Risk of Bias tool (RoB) 2.0 (Sterne et al., 2019) to assess a RCT design studies and the Risk Of Bias In Non-randomised Studies–of Interventions (ROBINS-I) (Sterne et al., 2016) to assess non-randomised studies. The kappa statistic index was used to determine interrater agreement (values of 0.4 or less, 0.4 to 0.75 and over 0.75, correlates with poor, fair or excellent agreement, respectively) (Fleiss, 1971). The score of the two reviewers was compared, and differences were discussed. If no agreement was reached, a third reviewer intervened. According to study quality, the studies were rated from low to high risk of bias in the ROB 2.0 scale and from low to critical bias in the ROBINS-I scale. There was not a cut point that studies had to reach to be included in the review, but this assessment was used for the solely purpose of classifying the quality of the evidence collected from each study.

2.7 Data analysis

Due to heterogeneity in the type of aquatic therapeutic exercises, the outcomes included in the review and instruments used to assess them, a meta-analysis on the effect of the interventions was deemed inappropriate. Instead, a qualitative synthesis of the studies included was performed, and results from the interventions on each specific outcome of interest were summarised narratively in tables.

When available, results on each outcome of interest were expressed in terms of mean change or effect size. Statistical significance was also reported for each comparison. Details on the exercise interventions were presented in a table according to the FITT principle when reported by the studies. If available, results on adherence were expressed in percentage of sessions attended of those prescribed.

3 RESULTS 3.1 Search results

In the initial search, 145 potential articles were identified. After excluding duplicates and the ones which did not meet the inclusion criteria by screening titles and abstract, 22 studies were identified for full-text assessment. Finally, 11 articles were included in this review after excluding studies with patients with secondary lymphedema (n = 1), case-report (n = 1), cross-sectional (n = 1), reviews (n = 1), studies where a water intervention was a small part of the intervention (n = 1), the intervention combined land and water-based exercises (n = 3), a pilot study without comparison group (n = 1), intervention only consisted of two sessions (n = 1) or studies with outcomes not related (n = 1). The selection process appears in (Figure 1).

Diagram flow

3.2 Sensitivity and precision analysis

The result of the analysis is detailed in Table 4.

TABLE 4. Sensitivity, precision, NNR and unique hits of the search Databases Total hits retrieved Relevant hits retrieved NNR Unique hits Sensitivity Precision Pubmed 26 8 3 0 73 31 Web of science 26 8 3 0 73 31 Scopus 36 11 3 1 100 31 CINAHL 46 10 5 0 91 22 Cochrane 11 2 6 0 18 18 TOTAL 145 11 The number of hits is the remaining after duplicates were removed. NNR: Number Needed to Read (total hits retrieved/relevant hits on a database). Unique paper: relevant study retrieved from one database only. Sensitivity: relevant hits retrieved/relevant hits retrieved TOTAL (%). Note: Precision: relevant hits retrieved/total retrieved (%). 3.3 Participants, study characteristics and design

The studies included in total 737 women (308 in the aquatic exercise intervention, 205 in a comparison intervention group with other intervention programmes and 224 in the usual care group) with a sample ranging from 19 to 165 participants.

Among the studies, six of them were RCTs (Cantarero-Villanueva et al., 2012; Cantarero-Villanueva, Fernández-Lao, Cuesta-Vargas, et al., 2013; Odynets et al., 2018; Odynets, Briskin, & Todorova, 2019; Odynets, Briskin, Zakharina, et al., 2019; Tidhar & Katz-Leurer, 2010), one was a randomised controlled pilot trial (Johansson et al., 2013) and three of them were controlled trials (Broach & Norrell, 2019; Cantarero-Villanueva, Fernández-Lao, Caro-Morán, et al., 2013; Fernández-Lao et al., 2013) and a single-arm pre-post trial (Siqueira et al., 2020). Studies were published from 2010 to 2020. Participants and study main characteristics are in Table 3. The intervention details are gathered in Supporting Information S1. The aquatic intervention was compared to usual care (Broach & Norrell, 2019; Cantarero-Villanueva et al., 2012; Cantarero-Villanueva, Fernández-Lao, Caro-Morán, et al., 2013; Cantarero-Villanueva, Fernández-Lao, Cuesta-Vargas, et al., 2013; Johansson et al., 2013; Tidhar & Katz-Leurer, 2010), to land exercise or usual care (Fernández-Lao et al., 2013), to a Pilates (Odynets et al., 2018; Odynets, Briskin, & Todorova, 2019; Odynets, Briskin, Zakharina, et al., 2019) or a Yoga programme (Odynets et al., 2018; Odynets, Briskin, & Todorova, 2019) and a group of healthy women (Odynets et al., 2018) or had no comparison group (Siqueira et al., 2020).

3.4 Risk of bias of the studies

The summary of the assessment of risk of bias is included in Figure 2 and Table 5. The assessment was made by two independent assessors, and the interrater agreement was excellent (0.75 kappa index). The agreement of 100% was reached through discussion. The majority of the RCTs presented some methodological issues, especially in the randomisation process, deviations from intended interventions and selection of the reported results. As an overall risk of bias, most studies presented some concerns. In non-RCTs, the major methodological issue was bias due to confounding with a moderate to a serious risk, and the overall risk was moderate to critical.

Assessment of the risk of bias scale risk of bias assessment of RCTs

TABLE 5. Robins-I scale for assessment of non-randomised studies' risk of bias D1 D2 D3 D4 D5