Introduction

The Global Burden of Disease Study 2020 [1] indicated that breast cancer remains the leading cancer diagnosis among females [2]. Although survival rates of breast cancer are improving, patients still experience a series of adverse effects caused by cancer and its related treatment, such as depression, fatigue, sleep disturbance, and bone marrow suppression [3, 4]. Cancer-related fatigue (CRF) is one of the most familiar and often overlooked symptoms [5], referring to a general, persistent, and subjective feeling of fatigue caused by cancer or relevant treatment that cannot be improved by sleep or rest [6] and may persist for months or even years [7]. The incidence of CRF in breast cancer patients is higher than that in other types of cancers [1], among which up to 33% of patients experience fatigue five years after the end of breast cancer treatment [8]. CRF adversely affects breast cancer patients in multiple aspects [9], which severely not only affects their quality of sleep but also prolongs their length of hospital stay and can result in a reduction of physical, mental, and emotional function and poor quality of life [10].

Currently, some pharmaceutical agents such as stimulants, antidepressants, acetylcholinesterase inhibitors, and corticosteroids have been recommended for CRF management in breast cancer patients [10] However, those pharmacological approaches were reported to be associated with a range of undesirable side effects such as tumor protection [11], decreased appetite [12] and venthrombotic events [13, 14]. The unclear pathophysiological mechanism of CRF also makes it difficult to develop tailored pharmacological interventions for CRF management [6]. Non-pharmacological interventions such as exercise interventions [10], mindfulness-based decompression therapy [15], and cognitive behavioral therapy [16] have been explored as adjuvant approaches to pharmacological interventions to alleviate CRF. Exercise interventions refer to a physical activity treatment that is planned, structured, and repetitive and have a final or an intermediate objective of improving or maintaining physical fitness, which includes running, aerobics, tai chi, yoga, and resistance exercise [17]. Exercise interventions have been commonly utilized and recommended as an effective intervention for the alleviation of CRF by the American Society of Clinical Oncology (ASCO) [18] and Exercise and Sports Science Australia (ESSA) [19]. In addition, the Japan Breast Cancer Society (JBCS) [20], the German Gynecological Oncology Group (AGO) [21], and a previous systematic review [22] found that exercise interventions are an effective, low-risk modality for breast cancer patients in reducing morbidity and improving body functions and quality of life. However, most of the literature on exercise interventions [18, 19, 21] have not clearly stated the type, frequency, and duration of exercise for practice in breast cancer patients with CRF, leading to a gap in developing personalized and evidence-based exercise intervention protocols tailored to patients’ health conditions and needs.

With the rapid development of evidence-based medicine in the field of cancer supportive care, an increasing body of systematic reviews (SRs)/meta-analyses have provided much evidence on using exercise interventions for CRF management in breast cancer patients, but their conclusions were inconsistent [23, 24] and the methodological quality varied across studies, which are barriers to the transformation of research evidence to practice and the application of clinical decision-making. To our knowledge, no overviews of systematic reviews on the effects of exercise interventions on CRF in breast cancer patients have been conducted so far. Thus, the aim of this overview was to critically appraise and consolidate evidence from current SRs/meta-analyses on the effects of exercise interventions on CRF in breast cancer patients. Specifically, the study objectives were as follows: (1) to identify the effects of exercise interventions on relieving CRF in breast cancer patients; (2) to assess the methodological quality of as well as the level of evidence from current SRs/meta-analyses on the effects of exercise interventions for breast cancer patients with CRF; and (3) to identify the optimal modality, duration, and frequency of exercise interventions for CRF management in breast cancer patients.

Methods

This overview of systematic reviews was reported in accordance with the Preferred Reporting Items for OoSRs (PRIO-harms) checklist and the Preferred Reporting Items for OoSRs (PRIO) checklist. The protocol has been registered with PROSPERO (CRD42020219866). A pre-print version of this manuscript is also available at https://www.researchsquare.com/article/rs-1376171/v1

Data sources and searches

This overview included SRs/meta-analyses that focused on the effects of exercise therapy on CRF in breast cancer patients. Relevant SRs/meta-analyses were comprehensively searched until September 2021 through the following data sources: (1) PubMed, Cochrane Library, Excerpta Medica Database (EMBASE), Web of Science, China National Knowledge Infrastructure (CNKI), China Biology Medicine Disc (CBMdisc), Wan Fang Data, and China Science and Technology Journal Database, and The Lancet; (2) references of the included SRs/meta-analyses; and (3) grey literature from the National Institute for Health Research (NIHR) Centre, such as unpublished manuscripts and published reports. The search terms included “breast neoplasms”, “exercise therapies”, “fatigue”, “systematic review”, and “meta-analysis”. The search procedure in the databases above followed the text string “((Breast neoplasms) OR (Breast tumor) OR (Mammary cancer) OR (Breast cancer) OR (Carcinoma breast)) AND (Exercise OR (Physical activity) OR (Physical exercise) OR (Exercise training) OR (Exercise therapies)) AND (Fatigue OR CRF) AND ((Systematic review) OR meta-analysis)”. Taking PUBMED and EMBASE as examples, a full search strategy was summarized in Supplementary file A.

Inclusion and exclusion criteria

The inclusion criteria were developed in accordance with the Population, Intervention, Comparator, Outcome, Study (PICOS) framework: (1) types of studies: SRs/meta-analyses of randomized controlled trials (RCTs) that were published in either English or Chinese; (2) types of populations: adult breast cancer patients [5] with CRF [25], regardless of stages of cancer, age, gender, and nationality; (3) types of interventions: exercise interventions [17], such as aerobic exercise, tai chi, yoga, resistance training, dancing, and walking; (4) types of comparison: routine methods of treatment and care with no active exercise components or any other types of active treatments; and (5) types of outcomes: CRF as the primary outcome as measured by valid assessment tools, such as the Functional Assessment of Chronic Illness Therapy-Fatigue (FACIT-F) Scale, the Multidimensional Fatigue Inventory (MFI), or the Brief Fatigue Inventory (BFI). Exclusion criteria were the following: (1) proposals of SRs or meta-analyses; (2) study population was breast cancer mixed with other diseases or complications; (3) conference abstracts; (4) full text was not available after multiple search methods, including contacting the author.

Literature screening and data extraction

Duplications were identified and removed via reference management software (NoteExpress). The titles and abstracts of the rest of the SRs/meta-analyses were screened by two reviewers (HJZ and YZX) independently to determine the potentially eligible SRs/meta-analyses. Full texts of the potentially eligible SRs/meta-analyses were further screened and examined by the same two reviewers. If there were duplications, the latest version of the SR or meta-analysis was selected. Eligible SRs/meta-analyses were finally included after discussion between the reviewers. Any contradiction regarding study inclusion was resolved through consultation or arbitration by an experienced third reviewer (TW). Data from the included SRs/meta-analyses were extracted using a data extraction form predesigned by one reviewer (HJZ), which was verified by another reviewer (YZX). Disagreements between the two reviewers regarding data extraction were discussed by involving a third reviewer (TW). The extracted data included the author, publication year and country, number of studies and sample size of the participants, types of intervention and control, quality assessment (whether the included SRs/ meta-analyses evaluated the quality of their included studies and the tools used for the quality appraisal), measurement tools, main conclusion, and whether it included a meta-analysis. Moreover, relevant data for subgroup analysis including the exercise type, frequency, duration, and inclusion/absence of supervision were extracted and verified. The study selection process is presented in Fig. 1.

Fig. 1
figure 1

Flowchart of the study selection process (PRISMA diagram)

Full size image

Quality appraisal of the included reviews

The methodological quality and the level of evidence of the included SRs/meta-analyses were independently assessed by two reviewers (HJZ and YZX) with two tools (see the “Methodological quality” and the “Evidence quality” sections). The final assessment results were cross-checked. Any disapprovals were discussed and decided by involving a third reviewer (JYT).

Methodological quality

A Measurement Tool to Assess systematic Reviews II (AMSTAR II) was used to comprehensively assess the methodological quality of the included SRs/meta-analyses [26], which is presently the most widely used methodological quality assessment tool [27, 28]. The AMSTAR II includes 16 items (www.amstar.ca), each of which can be answered “yes” or “no”, and some of the items can be answered “partially yes” [27, 28]. Seven items, including items 2, 4, 7, 9, 11, 13, and 15, that are considered to critically affect the validity of the included reviews and its conclusions are generally recommended as critical items [26]. The methodological quality of the included reviews was rated using the following criteria: (1) high quality: no or only one non-critical item flaw; (2) moderate quality: more than one non-critical item flaw but no critical item flaws; (3) low quality: one critical item flaw, with or without a non-critical item flaw; and (4) critically low quality: more than one critical item flaw, with or without a non-critical item flaw [26, 27, 29].

Evidence quality

Two reviewers (HJZ and YZX) used the Grading of Recommendations Assessment, Development and Evaluation (GRADE) to rate the level of evidence of the included SRs/meta-analyses in five aspects, including limitations, inconsistencies, indirectness, imprecision, and publication bias [30]. Disagreements were addressed by involving a third author (TW) until consensus was achieved. For each aspect, the evidence was graded as high, moderate, low, or extremely low. Detailed grading criteria were as follows [31]: (1) high-level evidence: not downgraded, which represents the true effect estimates; (2) moderate-level evidence: downgraded one grade, which indicates that the true value is possible to come near to the estimate but is substantially different; (3) low-level evidence: downgraded two grades, which indicates that there is a significant difference between the actual and estimated values; and (4) extremely low-level evidence: downgraded three grades, which indicates that the true value is likely to be very different from the estimated value.

Data analysis

The characteristics of the included SRs, including author, publication year and country, number of studies and sample size, types of intervention and control, and main findings (i.e., effects of the exercise on CRF), are summarized in Table 1. The overlap across the included studies (only RCTs) of the analyzed SRs/meta-analyses was estimated using the corrected covered area (CCA) [32]. A lower CCA value indicated a lower likelihood of overlaps [32]. A CCA value of 5% or below was regarded as a “slight overlap”, 6–10% as a “moderate overlap”, 11–15% as a “high overlap”, while above 15% was regarded as a “very high overlap” [33]. For continuous variables, mean differences (MD) or standardized mean differences (SMD) with 95% confidence intervals (CI) was used for meta-analysis and effect size calculation (a P value of ≤ 0.05 was considered statistically significant). For continuous variables, the random-effects model was used to calculate the number of participants and RCTs included in the meta-analyses and to summarize the effect size [with 95% confidence intervals (CI) and P values ≤ 0.05 considered significant]. According to Cohen [33], 0.2 is considered a small effect, 0.2 to 0.8 a medium effect, and 0.8 or above a large effect. We also extracted and analyzed the data of included meta-analyses to better illustrate the effects of the duration of the interventions, exercise type, frequency, and duration of each session on CRF of breast cancer patients. Because of the lack of relevant data, direct comparisons between different interventions were impossible. I-square (I2) statistics were used to measure the heterogeneity of the included SRs/meta-analyses and explain the various thresholds by effect size and direction and the P-value from Cochran’s Q test [34]. An I2 value > 50% is regarded as a substantial level of heterogeneity [34]. Sub-group analyses are planned based on exercise type, frequency, and duration, and inclusion/absence of supervision. Statistical analysis was conducted with the Stata version 16.0 software.

Table 1 Characteristics of the 29 SRs/meta-analyses
Full size table

Adapted from: Preferred reporting items for overviews of systematic reviews [35].

Results

Identification of the included reviews

A total of 369 records were searched, of which 135 were excluded due to duplication and 193 were screened by title or abstract and deemed irrelevant to the topic. Of the remaining 41 records, 12 were excluded after assessing the full text for eligibility. Twenty-nine SRs/meta-analyses [23, 24, 36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62] were finally included in the overview. The literature retrieval and selection process are shown in Fig. 1. A list of excluded reviews from full-text analysis with reasons is provided in supplementary file B.

Characteristics of the included reviews

The 29 SRs/meta-analyses included 402 studies, with a total of 33,655 patients, published between 2006 and 2021. Meta-analyses were carried out for all the included reviews. A total of 252 RCTs were included and analyzed across the 29 reviews, with a CCA of 3% indicating a slight overlap rate that reflected a low level of unnecessary duplications in the reviews and less biased results. Sixteen reviews [23, 24, 36,37,38,39,40,41,42,43,44,45,46,47,48,49] explored the effects of exercise therapy on CRF by including studies with different types of exercise, including aerobic exercise, yoga, resistance exercise, and Pilates. For the other 13 reviews, six [50,51,52,53,54,55] focused on yoga, five [56,57,58,59,60] on aerobic exercise, and two [61, 62] on tai chi. Routine methods of care and/or health education without any active exercise components were commonly utilized as the study comparisons. Seventeen reviews [23, 24, 36, 37, 39, 43, 44, 46, 50,51,52,53,54, 56,57,58, 61] used the Cochrane risk of bias (RoB) criteria. Other reviews were assessed using the Quality Assessment of Controlled Intervention Studies [38], the Norwegian Knowledge Centre for the Health Services (NOKC) Handbook for Systematic Reviews [41], Homemade Standard [42], the Revised Risk-of-Bias Tool for Randomized Trials (RoB 2.0) [55], Jadad Scores [59], the Newcastle–Ottawa Scale (NOS) [60], the Physiotherapy Evidence Databases (PEDro) Scale [40, 45, 62], and the Joanna Briggs Institute-Critical Appraisal for Randomized Controlled Trials (JBI-MAStARI) tool [47], respectively. Two reviews [48, 49] did not describe its methodological quality assessment process. The FACIT-F Scale [23, 36, 37, 39,40,41,42,43, 45,46,47,48,49, 51,52,53,54,55,56,57,58,59,60,61,62], the BFI [36, 37, 39, 40, 45, 46, 53,54,55, 58], the MFI [23, 24, 36, 38, 39, 41, 44, 45, 48, 54, 55, 58, 59], and the Piper Fatigue Scale (PFS) [24, 36, 39,40,41,42, 47, 48, 56,57,58,59,60] were the most commonly used instruments for CRF assessment in the 29 SRs/meta-analyses, the characteristics of which are presented in Table 1.

Quality appraisal of included reviews

Methodological quality

Regarding the methodological quality of the included SRs/meta-analyses, three reviews [40, 45, 53] were evaluated as high quality, 21 reviews [23, 24, 36, 37, 39, 41, 43, 44, 46,47,48, 50,51,52, 54, 56,57,58, 60,61,62] were rated as low quality, and the remaining five reviews [38, 42, 49, 55, 59] were assessed as critically low quality. Specifically, the critical items that had an effect on the quality of the reviews were item 2 (only five reviews [45, 46, 50, 55, 61] were evaluated as “yes” due to registered proposals in the early stage, and the remaining reviews only provided the research methods so they were assessed as “partly yes”, which means that the research methods could not be compared with the registered proposals approved by official organizations and may have caused a risk of bias), item 4 (whether to search for grey literature and counsel experts in the relevant field was not mentioned in 18 reviews[23, 24, 36, 37, 39, 43,44,45, 48,49,50,51, 54,55,56,57, 59, 60], suggesting that there may have been incomplete retrievals in the above, which may have led to results and conclusion errors), and item 7 (apart from three reviews[40, 45, 53], the list of excluded references and the causes for their exclusion were not provided and illustrated in the other reviews, which reduced the rigor of the study and the reliability of the results). In addition, non-critical item 10 also affected the methodological quality results since none of the 29 reviews reported the funding of their included RCTs, which indicated uncertainty about the possibility of commercial funding interference that might have made study results favorable to the commercial funder. All the reviews described the basic characteristics of and were able to scientifically discuss and analyze the included studies. Specific methodological quality assessment results are shown in Table 2.

Table 2 Methodological quality of the 29 SRs/meta-analyses
Full size table

Evidence quality

Eleven reviews [37, 38, 40, 43, 49,50,51,52, 59, 60, 62] were evaluated as having an extremely low level of evidence, 13 reviews [23, 24, 42, 44, 46,47,48, 53, 54, 56,57,58, 61] had a low level of evidence, and the remaining five [36, 38, 39, 41, 45] had a moderate level of evidence. Inconsistency (n = 22, 68.75%) was the most common reason for downgrading levels in the included reviews, followed by publication bias (n = 17, 53.12%), limitations (n = 15, 46.8%), imprecision (n = 10, 31.25%). Elaborating on the reasons for downgrading levels, the most common reason was the significant heterogeneity of the results [23, 36, 37, 39,40,41,42,43, 45,46,47, 49,50,51,52,53,54,55,56,57, 59, 61, 62] (n = 23) and the missing grey literature and manual retrieval [24, 37, 40, 42, 44, 50, 57,58,59, 61, 62] (n = 11). Other reasons included an unclear description of the blinding procedures [36, 40, 50, 52, 53, 55, 61, 62] (n = 8), inclusion of invalid values (RR = 1.0) within the confidence intervals [23, 42, 49, 51, 52, 59, 60, 62] (n = 8), failure to report publication bias [38, 42, 48, 49, 51, 62] (n = 6), unsatisfactory methodological quality of the included RCTs [24, 43, 44, 52, 53] (n = 5), unreported or incomplete report of outcomes such as adverse reactions [58] (n = 1), and the inclusion of only one RCT resulting in an inability to measure heterogeneity [42] (n = 1). The results of the evidence assessment are presented in Table 3.

Table 3 Evidence quality of the 29 SRs/meta-analyses
Full size table

Data synthesis and meta-analysis

The overall effects of the exercise interventions in the 29 SRs/meta-analyses indicated that exercise had a moderate effect on the reduction of fatigue in the breast cancer patients in the intervention groups (SMD = − 0.40 [95%CI − 0.58, − 0.22]; P = 0.0001) (see Fig. 2). However, due to the high heterogeneity among the 29 reviews (I2 = 95.4%), the overall effect size might be affected by various existing moderating variables. Therefore, the standardized mean differences of the moderating variables were further analyzed to determine the most effective intervention modalities based on the exercise type, frequency, intervention duration, and inclusion/absence of supervision in the intervention protocol. These results are shown in Table 4.

Fig. 2
figure 2

Effect of exercise intervention vs. no exercise or routine methods of care of CRF in breast neoplasms

Full size image
Table 4 Quantitative evidence synthesis for fatigue included in the 29 SRs/meta-analyses
Full size table

Exercise type

Seven exercise types were reported in the 29 SRs/meta-analyses, including aerobic exercise, resistance exercise, yoga, mind–body exercise (Pilates and gymnastics), combination exercise (home-based exercise and combined aerobic-resistance exercise), tai chi, and other exercises such as periodic rehabilitation exercise. The subgroup analyses of exercise type showed that both yoga (SMD = − 0.30 [95%CI − 0.56, − 0.05]; I2 = 28.7%; P = 0.021) and aerobic exercise (SMD = − 0.29 [95%CI − 0.56, − 0.02]; I2 = 16%; P = 0.048) had positive effects on improving CRF in breast cancer patients in the intervention groups compared with those in the control groups receiving routine methods of care. The remaining five exercise types—resistance exercise (SMD = − 0.01 [95%CI − 0.31, 0.29]; I2 = 48%; P = 0.958), tai chi (SMD = − 0.19 [95%CI − 1.13, 0.76]; I2 = 75.2%; P = 0.702), combination exercise (SMD = − 0.35 [95%CI − 0.76, 0.05]; I2 = 38.8%; P = 0.086), mind–body exercise (SMD = − 0.21 [95%CI − 0.63, 0.22]; I2 = 45.1%; P = 0.336), and other exercises (SMD = − 0.09 [95%CI − 0.45, 0.26]; I2 = 93.2%; P = 0.599)—showed no statistically significant differences between the intervention groups and the control groups using routine methods of care.

Duration of the intervention

Intervention duration among the included studies were categorized as: less than 2 months, 2 to 6 months, and more than 6 months. The effect magnitude of the moderating variable of duration had high heterogeneity (I2 = 94.8%; P < 0.000), indicating that the intervention duration could have affected the results of exercise for CRF management in breast cancer patients. The subgroup data indicated that only the intervention duration of more than 6 months had a beneficial effect (SMD = − 0.88 [95%CI − 1.59, − 0.17]; I2 = 42.7%; P = 0.000) on CRF compared with routine methods of care. The other intervention duration categories— less than 2 months (SMD = − 0.55 [95%CI − 1.15, 0.66]; I2 = 34.2%; P = 0.078) and 2 to 6 months (SMD = − 0.23 [95%CI − 0.65, 0.19]; I2 = 59.1%; P = 0.288)—resulted in no statistically significant improvement of CRF.

Frequency of exercise

The frequency of exercise included in the 29 SRs/meta-analyses can be categorized as 3 ≤ times per week and > 3 times per week, and the heterogeneity of the combined effect size between the two types was 95.3% (P = 0.0001), suggesting that the breast cancer patients’ CRF was affected by the frequency of exercise. The subgroup results of exercise frequency ≤ 3 times per week (SMD = − 0.75 [95%CI − 1.58, 0.08]; I2 = 91.6%; P = 0.076) indicated that its effect on CRF in the intervention groups was no statistically different from that in the control groups using routine methods of care, while exercise frequency > 3 times per week (SMD = − 0.77 [95%CI − 1.04, − 0.05]; I2 = 0%; P = 0.0001) indicated that the improvement effect was better in the intervention groups compared with the control groups receiving routine methods of care.

Duration of exercise

Subgroup analysis revealed that the duration of exercise that lasted 30 to 60 min per session (SMD = − 0.81 [95%CI − 1.15, − 0.47]; I2 = 42.3%; P = 0.0001) and 60 min per session (SMD = − 0.77 [95%CI − 1.04, − 0.50]; I2 = 0%; P = 0.0001) showed improvement effects on relieving CRF in breast cancer patients in the intervention groups compared with those in the control groups receiving routine methods of care.

Inclusion or absence of supervision during exercise

Relevant data on supervised and unsupervised exercise interventions were extracted from the included SRs/meta-analyses and the heterogeneity of effect size was 95.9%, indicating that supervision might impact the effects of exercise on CRF in breast cancer patients. Compared with routine methods of care, exercise interventions relieved the breast cancer patients’ CRF regardless of whether they were supervised while exercising; however, supervised exercise (SMD = − 0.48 [95%CI − 0.77, − 0.18]; I2 = 87%; P = 0.001) was shown to produce a larger effect on CRF compared with unsupervised exercise (SMD = − 0.21 [95%CI − 0.40, − 0.01]; I2 = 92.2%; P = 0.023).

Discussion

In this overview, we assessed the methodological and evidence quality of the included SRs/meta-analyses, and additional meta-analyses were performed for the 29 reviews to identify the effects of exercise on CRF in breast cancer patients. The intervention duration, exercise type, duration, and frequency, and whether the exercise intervention was supervised had varying degrees of influence on the effects of exercise on CRF. However, the unsatisfactory methodological quality and level of evidence of the included reviews might affect the reliability of the overview findings on the effects of exercise interventions on CRF in breast cancer survivors, which warrants a prudent interpretation of the study results.

Findings from this study suggested that exercise could be introduced as an effective intervention for CRF management in breast cancer patients. The findings supported the recommendations proposed in some clinical practice guidelines [18, 19, 68], in which exercise was rated and recommended as a beneficial approach to alleviating CRF. However, these guidelines [18, 19, 68] did not mention specific exercise plans. To help further detail the recommendations in the guidelines and facilitate healthcare professionals’ decision-making, subgroup analyses based on the type, frequency, duration, and intensity of exercise were conducted in this study. For exercise type, the study findings suggested that aerobic exercise and yoga were commonly recommended as promising approaches to improving CRF, which is consistent with Lin’s study [69], indicating that yoga can relieve patients’ tension and anxiety and help decrease their fatigue. Yoga is a convenient, easy-to-practice, and safe exercise modality that has been recommended as Grade I evidence by the U.S. National Comprehensive Cancer Network Guidelines [70]. The subgroup analyses results on the duration of exercise indicated that patients who had exercised for more than 6 months achieved the best improvement in fatigue. Our study findings showed that exercising more than three times per week for 30 to 60 min per session was beneficial for CRF alleviation, which is in line with previous research findings [71,72,73]. The included SRs/meta-analyses also indicated that patients should be encouraged to participate in supervised exercise when conditions permitted, which might lead to a better CRF outcome.

Although quantitative synthesis indicated that exercise interventions can alleviate CRF in breast cancer patients, the findings should be interpreted with caution given the unsatisfactory methodological quality (e.g., lack of reporting a list of exclusion studies and unclear funding resources) and level of evidence identified in the included reviews. For future studies, a list of excluded studies should be provided as an independent appendix to journals to facilitate readers’ understanding of the data selection process and further improve the reliability of the review findings [74]. Moreover, funding sources should be clearly declared in future publications to help readers determine whether funding bias existed. To achieve a comprehensive literature search, future systematic reviews are suggested to identify potential studies by searching not only the commonly used databases but also gray literature retrieval websites to minimize publication bias, for example, Greynet International (http://greynet.org/), the British Library (http://www.b1.uk), and other free grey literature sites such as PLoS. Conference abstracts, book chapters, academic theses and dissertations should also be sources of gray literature. In order to further improve the level of evidence of the included SRs, more original studies with rigorous study designs and detailed descriptions of the intervention protocols (e.g., type, frequency, intensity, and duration of the exercise) are necessary. Nevertheless, there is a need to acknowledge that the results of the unsatisfactory methodological quality of the included SRs may have been related to the selection of the quality appraisal tool. In this overview, nine of the included SRs were published before 2017, while the tool that we used for the quality appraisal (i.e., AMSTAR II) was also updated in 2017, which is an issue that needs to be considered for future overviews.

Study limitations

This overview has some limitations. Suboptimal methodological quality of some of the includes reviews (e.g., lack of registered protocols, unclear descriptions of data sources) may affect the strength of the evidence. language bias is possible given that only reviews published in Chinese and English were included. Due to the limited number of included reviews, within each current subgroup analysis, further subgroup analyses based on the intervention “dose” (e.g., intervention duration and frequency of each type of exercise intervention) were not conducted, which might, to some extent, limit the generalizability of the review findings to clinical practice.

Conclusion

Findings from this overview suggested that yoga and aerobic exercise with a long-term practice duration (over 6 months) might benefit CRF alleviation in breast cancer patients. Exercising for at least three times per week for 30 to 60 min per session might be an appropriate dose for alleviating CRF in breast cancer patients. Although existing evidence indicated that exercise interventions have a positive impact on CRF in breast cancer patients, the results should be interpreted with caution due to the limited quantity and unsatisfactory methodological quality and level of evidence of the included SRs/meta-analyses. More rigorously designed large-scale RCTs are needed to provide more robust evidence to specify the exact exercise type, duration, frequency, and intensity to have an optimal effect on CRF in breast cancer patients.