Differential Impact of Exercises on Quality-of-Life Improvement in Breast Cancer Survivors: A Network Meta-Analysis of Randomized Controlled Trials

Simple Summary This study aims to find out which types of exercise can help improve the quality of life for people who have survived breast cancer. Researchers analyzed data from different studies to see how various exercises, such as aerobic and strength training, aerobic activity, yoga, and strength exercise, affected these individuals after 12 weeks. The results show that combining aerobic and strength training is the most effective way to improve their quality of life without causing more people to drop out of the exercise programs compared to regular care. This research may help doctors and patients make better decisions about exercise plans for breast cancer survivors. Abstract This study aimed to assess the effectiveness of various exercise interventions in enhancing the quality of life for breast cancer survivors. To achieve this, randomized controlled trials were identified from major electronic databases, focusing on the relationship between exercise and quality of life in breast cancer survivors. The primary outcome was the impact of exercise on quality of life 12 weeks after the intervention, with a secondary outcome comparing dropout rates between intervention groups and a regular care control group. The study protocol was registered with INPLASY (INPLASY202340007). A network meta-analysis of nine randomized controlled trials involving 725 participants was conducted, examining aerobic and strength training, aerobic activity, yoga, and strength exercise. Results showed that aerobic and strength training was the most effective intervention, significantly improving the quality of life of breast cancer survivors (1.31; 95% confidence interval: 0.49 to 2.12). Aerobic activity had a borderline effect (0.83; 0.03 to 1.63), while no exercise interventions were associated with an increased dropout risk compared to the control group (regular care). The study concluded that concurrent aerobic and strength training can improve breast cancer survivors’ quality of life after 12 weeks of intervention without increasing dropout risk compared to regular care.


Introduction
Breast cancer is the most common female malignancy worldwide and has the fifth highest mortality rate of all cancers [1]. Owing to progress in cancer screening and advancements in cancer treatments, the number of breast cancer survivors in the United States exceeds 3.8 million, and it is estimated to rise by more than 30% in the next ten years [2]. Even after completing treatment, long-lasting and severe treatment-related side review and network meta-analysis spanned the duration from the first available entry in each database up to the most recent search date (7 April 2023).
In the initial stage, two authors were tasked with evaluating the titles and abstracts of identified studies for their eligibility using a consensus process. The search was conducted in the aforementioned databases to scrutinize eligible trials. Additionally, the reference lists of various review articles [5,7,[16][17][18][19][20][21][22][23][24] were examined and manual searches were performed. In situations where the two initial reviewers were unable to reach a consensus, a third reviewer and study author (PLC) was consulted. No restrictions on language were imposed on this search.

Inclusion and Exclusion Criteria
The network meta-analysis employed the PICO model (population, intervention, comparison, outcome), featuring the subsequent criteria: (1) P: human participants with breast cancer and completed treatment, including surgery, chemotherapy, and/or radiotherapy; (2) I: exercise interventions; (3) C: control group without intervention; and (4) O: changes in quality of life. The definition of breast cancer survivor was based on the joint guideline provided by the American Cancer Society and the American Society of Clinical Oncology [25].
The study applied the following inclusion criteria: (1) randomized controlled trials that recruited breast cancer survivors who had completed treatments, including surgery, chemotherapy, and/or radiation therapy, (2) randomized controlled trials that investigated the quantitative assessment of quality of life after exercise intervention, (3) the control group that received no intervention or regular care, and (4) trials that had available data on quality of life pre-and post-intervention at 12 weeks.
The selection of the 12-week evaluation duration was based on the initial literature review, which indicated that it was the most commonly used assessment period in the included studies. Several previous large-scale literature analyses have also found that the onset of exercise effects occurred at 12 weeks for patients undergoing rehabilitation [26] after stroke or a transient ischemic attack [27]. In order to compare the effectiveness of various exercise interventions, a standardized time frame is necessary to establish a benchmark for comparison. Therefore, this study focuses specifically on a 12-week duration and excludes other time frames with fewer studies available.
Exclusion criteria for this review and network meta-analysis included: (1) nonrandomized controlled trials, (2) studies without comparisons of exercise vs. exercise or exercise vs. regular care comparison, (3) studies lacking quantitative assessments of quality of life, (4) studies quantitatively assessed quality of life but only reported subscale data and did not provide a total score, (5) incomplete or unavailable data, even after attempts to contact the authors via email, and (6) studies enrolling participants overlapped with a published trial already enrolled in our analysis.

Modeling for Network Meta-Analysis
In the present network meta-analysis, we adhered to the following principles during the construction of the model. To prevent excessive heterogeneity, we restricted the paired comparisons to only exercise vs. exercise or exercise vs. regular care. Comparisons between exercise and cognitive behavioral therapy, eurythmy therapy, and various nutritional supplements were thus excluded. Inclusion of additional treatments might result in disparate network geometries, owing to the variation in the therapies being considered, leading to inconsistent outcomes in the network meta-analysis [28].
When categorizing the exercise type in our study, they were grouped based on the actual exercise prescription content discussion between two authors (TCW, ICT). If there is any disagreement in the categorization, consensus will be reached through discussion with the third author (PLC).

Methodological Quality Appraisal
To assess the methodological quality of the studies included in our analysis, we utilized the Cochrane risk of bias tool for randomized trials (version 2, RoB 2, London, UK) [29]. This tool appraises six principal components for assessing the quality of a study, including the randomization process, adherence to the intervention, missing outcome data, outcome measurement, selective reporting, and overall risk of bias.
2.5. Primary Outcome: Quality-of-Life Improvement, Standardized Mean Difference The primary outcomes evaluated in this study were changes in quality of life measured by quantitative scales. If the study utilized a breast-cancer-specific quality-of-life scale, such as the Functional Assessment of Cancer Therapy-Breast [30,31] or the International Breast Cancer Study Group Quality of Life [32], data extraction was prioritized from these scales. If the study did not use a breast-cancer-specific quality-of-life scale, data extraction was prioritized in the following order: cancer-specific quality-of-life assessment tools, such as the European Organization for Research and Treatment of Cancer Quality-of-Life Questionnaire [33], followed by general quality-of-life assessment tools, such as the Functional Assessment of Cancer Therapy-General [34].

Secondary Outcome: Risk Difference of Dropout Rates
The secondary outcome measure was the risk difference of dropout rates at the 12th week, which provides an intuitive indicator. For example, if an individual chooses a specific exercise regimen to improve their quality of life and experiences a dropout rate of 12%, while the control group, which only receives regular care, has a dropout rate of 7% (which may result in some of them starting an exercise routine on their own), the risk difference in dropout rates would be 5%.

Data Extraction, Management and Conversion
Two authors (TCW and ICT) performed the data extraction process independently, including demographic information, study design, exercise protocol details, and primary and secondary outcomes from the evaluated studies. In situations where the necessary data were not available in the published articles, we reached out to the corresponding authors to obtain the primary data.
Data extraction, conversion, and result merging were conducted in accordance with the recommendations outlined in the Cochrane Handbook for Systematic Reviews of Interventions and relevant medical literature [12,[35][36][37][38].

Statistical Analyses
Due to the inclusion of various exercise types, a random-effects model was utilized for the network meta-analysis [39]. The analysis was performed using MetaInsight (version 4.0.2, Complex Reviews Support Unit, National Institute for Health Research, London, UK) under a frequentist framework. MetaInsight represents a web-based platform for network meta-analysis that leverages the netmeta package in R software for conducting frequentist statistical calculations [40].
Initially, a forest plot and network plot were generated to display all pairwise comparisons from individual studies. Subsequently, forest plots were created for standardized mean differences in the change of quality of life at 12 weeks and the risk differences of dropout rates for each exercise type compared to the control group to provide an overall summary of the effects [41]. The effect sizes were presented as point estimates with a 95% confidence interval (95% CI) [41]. The exercise types were ranked, and numerical values for both direct and indirect comparisons were presented in tables. Inconsistency tests were conducted to detect any data disparities. Statistical significance was defined as a two-tailed p value of less than 0.05.

Sensitivity Analyses
Two sensitivity analyses were conducted to strengthen the robustness of the study findings. The first analysis employed a one-study removal method, which was performed to ensure that the effect estimates of individual studies did not excessively influence the overall results. Sequentially removing one study at a time from the analysis of qualityof-life changes at 12 weeks allowed us to determine whether the study conclusions and ranking remained consistent.
The second sensitivity analysis performed in this study involved the pre-post correlation coefficient. When transforming baseline and post-intervention quality-of-life measurements into mean and standard deviation of changes, it is necessary to assume a prepost correlation coefficient. In this study, a coefficient of 0.8 was utilized, as recommended by the Cochrane handbook [35]. However, different scholars may hold varying opinions on this coefficient with commonly used values being 0.5, 0.7, and 0.8 [42]. To examine whether the selected coefficient would impact the study results, a sensitivity analysis was conducted by calculating the effect sizes of quality-of-life changes at 12 weeks with a coefficient of 0.5 [42]. The direction, size of the effect, statistical significance, and ranking of the results were assessed.

Publication Bias
Potential publication bias was assessed in accordance with the Cochrane Handbook for Systematic Reviews of Interventions [12]. The funnel plot was generated using Comprehensive Meta-Analysis software, version 4 (Biostat, Englewood, NJ, USA), based on the comparison with the control group. Additionally, an Egger's regression test was conducted to quantify the presence of significant publication bias.

Study Identification and Network Model Formation
The PRISMA flowchart detailing the literature search is presented in Figure 1. The PRISMA NMA extension's checklist is provided in Table S1. The number of articles retrieved from various databases is presented in Table S2. After removing duplicate articles and excluding non-relevant articles by screening titles and abstracts, we ultimately included nine randomized controlled trials [6,[8][9][10][11][43][44][45][46]. The articles excluded in the final stage [4, along with their respective reasons for exclusion are listed in Table S3.

Sensitivity Analyses
Two sensitivity analyses were conducted to strengthen the robustness of the study findings. The first analysis employed a one-study removal method, which was performed to ensure that the effect estimates of individual studies did not excessively influence the overall results. Sequentially removing one study at a time from the analysis of quality-oflife changes at 12 weeks allowed us to determine whether the study conclusions and ranking remained consistent.
The second sensitivity analysis performed in this study involved the pre-post correlation coefficient. When transforming baseline and post-intervention quality-of-life measurements into mean and standard deviation of changes, it is necessary to assume a prepost correlation coefficient. In this study, a coefficient of 0.8 was utilized, as recommended by the Cochrane handbook [35]. However, different scholars may hold varying opinions on this coefficient with commonly used values being 0.5, 0.7, and 0.8 [42]. To examine whether the selected coefficient would impact the study results, a sensitivity analysis was conducted by calculating the effect sizes of quality-of-life changes at 12 weeks with a coefficient of 0.5 [42]. The direction, size of the effect, statistical significance, and ranking of the results were assessed.

Publication Bias
Potential publication bias was assessed in accordance with the Cochrane Handbook for Systematic Reviews of Interventions [12]. The funnel plot was generated using Comprehensive Meta-Analysis software, version 4 (Biostat, Englewood, NJ, USA), based on the comparison with the control group. Additionally, an Egger's regression test was conducted to quantify the presence of significant publication bias.

Study Identification and Network Model Formation
The PRISMA flowchart detailing the literature search is presented in Figure 1. The PRISMA NMA extension's checklist is provided in Table S1. The number of articles retrieved from various databases is presented in Table S2. After removing duplicate articles and excluding non-relevant articles by screening titles and abstracts, we ultimately included nine randomized controlled trials [6,[8][9][10][11][43][44][45][46]. The articles excluded in the final stage [4, along with their respective reasons for exclusion are listed in Table S3.  Our analysis included a total of nine randomized controlled trials, involving 725 individuals. Based on the included studies, the exercise types were categorized as follows: aerobic and strength training (concurrent), aerobic activity, yoga, and strength exercise. The network model for the exercise interventions is displayed in Figure 2. Our analysis included a total of nine randomized controlled trials, involving 725 individuals. Based on the included studies, the exercise types were categorized as follows: aerobic and strength training (concurrent), aerobic activity, yoga, and strength exercise. The network model for the exercise interventions is displayed in Figure 2. Among the nine studies included in our analysis, three studies exclusively recruited postmenopausal women [6,10,44], and two studies only enrolled patients with fatigue [11,45]. For further details on the inclusion criteria, the country where the study was conducted, the mean age and standard deviation of the participants, exercise intervention details, quality-of-life assessment scales, and dropout rates, please refer to Table 1. Among the nine studies included in our analysis, three studies exclusively recruited postmenopausal women [6,10,44], and two studies only enrolled patients with fatigue [11,45]. For further details on the inclusion criteria, the country where the study was conducted, the mean age and standard deviation of the participants, exercise intervention details, quality-of-life assessment scales, and dropout rates, please refer to Table 1. The intervention involved gradually increasing aerobic exercise over 12 weeks, starting with 15-20 min, 3 days a week at 40-59% of heart rate reserve and progressing to moderate intensity (>3 times per week, 30-50 min, 40-59% heart rate reserve). Group 0: JME (a 15-min exercise at 60-80% of HRmax, 3 times a day). Group 1: JME with follow-up. Group 2: JME with aerobic activity (30 min, 5 times per week). Groups 0, 1, and 2 were combined as the aerobic exercise intervention. Group 3: JME with resistance training (8 movements with progressive loads, 2-3 times per week). Group 3 was categorized as the aerobic + strength intervention.

Methodological Quality of the Included Studies
Regarding the overall methodological quality of the studies, we observed that 44.4% (4/9) of the studies had a low risk of bias, while 55.6% (5/9) had some risk of bias (refer to Figure S1). The studies with some risk of bias had differences in their protocols between study arms, which could potentially impact the adherence and outcomes of the interventions. The details of the risk of bias assessment are provided in Table S4.

Primary Outcome: Aerobic and Strength Concurrent Training Most Effective
After a 12-week intervention, aerobic and strength training showed a significant improvement in quality of life (effect size: 1.31; 95% CI: 0.49 to 2.12), while aerobic activity demonstrated a borderline effect (effect size: 0.83; 95% CI: 0.03 to 1.63). On the other hand, yoga (effect size: 0.63; 95% CI: −0.67 to 1.92) and strength training (effect size: 0.19; 95% CI: −1.08 to 1.46) did not show a significant difference compared to the control group ( Figure 3). Please refer to Figure S2 for the detailed pair-wise comparisons between study arms as reported in individual studies.

Methodological Quality of the Included Studies
Regarding the overall methodological quality of the studies, we observed that 44.4% (4/9) of the studies had a low risk of bias, while 55.6% (5/9) had some risk of bias (refer to Figure S1). The studies with some risk of bias had differences in their protocols between study arms, which could potentially impact the adherence and outcomes of the interventions. The details of the risk of bias assessment are provided in Table S4.

Primary Outcome: Aerobic and Strength Concurrent Training Most Effective
After a 12-week intervention, aerobic and strength training showed a significant improvement in quality of life (effect size: 1.31; 95% CI: 0.49 to 2.12), while aerobic activity demonstrated a borderline effect (effect size: 0.83; 95% CI: 0.03 to 1.63). On the other hand, yoga (effect size: 0.63; 95% CI: −0.67 to 1.92) and strength training (effect size: 0.19; 95% CI: −1.08 to 1.46) did not show a significant difference compared to the control group (Figure 3). Please refer to Figure S2 for the detailed pair-wise comparisons between study arms as reported in individual studies. The exercise interventions were ranked based on their effect sizes on quality of life, with aerobic and strength training (concurrent) being the most effective, followed by aerobic activity, yoga, and strength exercise in that order. Please see Table 2 for a detailed comparison and ranking of the exercise types. The estimates from pairwise meta-analyses are located above the diagonal line, while the estimates from network meta-analyses are located below the diagonal line.

Secondary Outcome: Dropout Rates Statistically Similar
After 12 weeks of intervention, there was no significant difference in dropout rates between the various exercise types and the control group with all risk differences with The exercise interventions were ranked based on their effect sizes on quality of life, with aerobic and strength training (concurrent) being the most effective, followed by aerobic activity, yoga, and strength exercise in that order. Please see Table 2 for a detailed comparison and ranking of the exercise types. The estimates from pairwise meta-analyses are located above the diagonal line, while the estimates from network meta-analyses are located below the diagonal line.

Secondary Outcome: Dropout Rates Statistically Similar
After 12 weeks of intervention, there was no significant difference in dropout rates between the various exercise types and the control group with all risk differences with their 95% CIs overlapped with 0 (see Figure 4). For a detailed analysis of the pair-wise comparisons between study arms as reported in individual studies, please consult Figure S3. their 95% CIs overlapped with 0 (see Figure 4). For a detailed analysis of the pair-wise comparisons between study arms as reported in individual studies, please consult Figure S3.

Inconsistency Test
The network was constructed by creating nodes and performing direct and indirect comparisons to determine consistency. The results of the quality-of-life inconsistency tests are presented in Table S5, while the dropout rate results are presented in Table S6. All available comparisons had p values greater than 0.05, indicating no evidence of inconsistency between direct and indirect comparisons.

Sensitivity Analyses
The results of the one-study removal analysis showed consistent rankings and clinical significance for all exercise types. The aerobic and strength-training intervention consistently demonstrated a significant improvement in the quality of life of breast cancer survivors, while the aerobic activity intervention remained at borderline significance. Yoga and strength exercise interventions consistently showed no significant effect on quality of life. (See Figure S4a-i) In the second sensitivity analysis, we adjusted the pre-post correlation coefficient from 0.8 to 0.5 and conducted a new network comparison ( Figure S5). Our results showed that the direction of effect sizes, ranking, and interpretation of the results remained consistent with those obtained using a coefficient of 0.8 ( Figure 3).
The above analyses indicate that the results of our study are consistent and not influenced by the inclusion or removal of individual studies as well as the adjustment of assumed values in the calculation process.

Publication Bias
Please see Figure S6 for the funnel plot. The Egger's test yielded a p value of 0.25, indicating no significant publication bias.

Main Findings and Clinical Implications
Our network meta-analysis revealed that among breast cancer survivors, aerobic and strength training was the most effective type of 12-week exercise intervention in improving quality of life (effect size: 1.31; 95% CI: 0.49 to 2.12). Aerobic activity had a borderline effect (effect size: 0.83; 95% CI: 0.03 to 1.63), while yoga and strength exercise showed no significant difference compared to the control group. In terms of dropout rates, there was no significant risk difference between the different types of exercise and the control group. For breast cancer survivors and caregivers, our network meta-analysis provides valuable

Inconsistency Test
The network was constructed by creating nodes and performing direct and indirect comparisons to determine consistency. The results of the quality-of-life inconsistency tests are presented in Table S5, while the dropout rate results are presented in Table S6. All available comparisons had p values greater than 0.05, indicating no evidence of inconsistency between direct and indirect comparisons.

Sensitivity Analyses
The results of the one-study removal analysis showed consistent rankings and clinical significance for all exercise types. The aerobic and strength-training intervention consistently demonstrated a significant improvement in the quality of life of breast cancer survivors, while the aerobic activity intervention remained at borderline significance. Yoga and strength exercise interventions consistently showed no significant effect on quality of life (See Figure S4a-i).
In the second sensitivity analysis, we adjusted the pre-post correlation coefficient from 0.8 to 0.5 and conducted a new network comparison ( Figure S5). Our results showed that the direction of effect sizes, ranking, and interpretation of the results remained consistent with those obtained using a coefficient of 0.8 ( Figure 3).
The above analyses indicate that the results of our study are consistent and not influenced by the inclusion or removal of individual studies as well as the adjustment of assumed values in the calculation process.

Publication Bias
Please see Figure S6 for the funnel plot. The Egger's test yielded a p value of 0.25, indicating no significant publication bias.

Main Findings and Clinical Implications
Our network meta-analysis revealed that among breast cancer survivors, aerobic and strength training was the most effective type of 12-week exercise intervention in improving quality of life (effect size: 1.31; 95% CI: 0.49 to 2.12). Aerobic activity had a borderline effect (effect size: 0.83; 95% CI: 0.03 to 1.63), while yoga and strength exercise showed no significant difference compared to the control group. In terms of dropout rates, there was no significant risk difference between the different types of exercise and the control group. For breast cancer survivors and caregivers, our network meta-analysis provides valuable information for exercise prescription. The data can be used to support the benefits of exercise and encourage patients to adhere to the exercise program for at least three months to achieve a significant improvement in quality of life.

Significance of the Findings Compared to Existing Literature
Aune et al. published a comprehensive pairwise meta-analysis in JNCI Cancer Spectrum in 2022 [7], which collected 79 randomized controlled trials and 14,554 breast cancer patients before 2019, including various exercise protocols and intervention durations. The study concluded that physical activity, compared to regular care, can effectively improve global health-related quality of life. However, the authors also stated that based on their analysis, the evidence regarding the dose and type of physical activity is still insufficient to draw conclusions.
Our study utilized network meta-analysis to compare various exercise interventions and concluded that within a 12-week timeframe, (concurrent) aerobic and strength training is the most effective type of exercise for improving quality of life in breast cancer survivors, followed by aerobic activity with a borderline effect. This study is the first in the literature to provide answers to questions regarding the effectiveness of different types of exercise, their comparison, and the ranking of exercise benefits.
Previously, studies often mentioned that yoga is beneficial for breast cancer survivors [98,99]. However, some of these studies relied on self-reported surveys and lacked prospective designs with specific intervention durations. They included patients with different frequencies and durations of yoga interventions [98]. Some systematic reviews also incorporated breast cancer patients during and after treatment without specifying the exact duration of yoga intervention [99]. In our study, we directly used a 12-week timeframe as the research benchmark and compared and ranked the effects of yoga on quality of life among various exercises. In other words, we are not answering whether yoga is effective for breast cancer survivors, but rather, within the 12-week timeframe, we assessed the varying impact on quality of life from different exercises performed by breast cancer survivors with yoga being part of the ranking results.

Possible Explanations for the Observed Results
Regarding the ranking of the effectiveness of different types of exercise in improving quality of life, we hypothesize that the intensity of the exercise may play a role. Ostman et al. found that the improvement in quality of life is more pronounced with increasing exercise intensity in patients with heart failure [100]. In the exercise protocols designed for breast cancer survivors in our included studies [6,8,11,43,44,46], aerobic exercise is easier to perform, can be sustained for longer durations, and is more likely to achieve moderate or even vigorous intensity. This may suggest that exercise interventions incorporating aerobic activity, such as concurrent aerobic and strength training and aerobic activity only, tend to result in better outcomes. In Dysart et al.'s study, yoga has been found to achieve moderate intensity only 32.75% of the time on average, and most of the time, it only achieves low intensity [101]. As for the strength-exercise-only protocols, our included studies consisted of a home-based exercise program without additional weight bearing [45] and a program based on 40% of one repetition maximum (1RM), gradually increasing to 70% 1RM based on the participant's capacity with the help of a professional trainer [9]. Day et al.'s previous research on the correspondence between resistance training and exercise intensity suggests that 40%, 70%, and 90% 1RM correspond to low, moderate, and vigorous intensity, respectively [102]. Thus, a 40-70% 1RM training protocol [9] corresponds only to low-to-moderate intensity. Moreover, even at 70% 1RM, the actual exercise time of 12 lifts is shorter than that of aerobic exercise.
The lack of significant differences in dropout rates between the exercise interventions and regular care may be attributed to the design of the exercise protocols, which were easily followed. For instance, the yoga classes were led by professional instructors and provided a social component, lasting for 60-90 min [10,45]. The strength-training intervention was facilitated by professional trainers and included progressive overload, leading to a sense of accomplishment after each session [9]. Even the self-administered aerobic activities were completed within an hour, preventing excessive difficulty [44].

Limitations
Our study has limitations. Among the included studies, three studies enrolled only postmenopausal women and two studies enrolled breast cancer survivors with fatigue, which may violate the transitivity assumption due to the heterogeneous study population. However, based on the age distribution of the included participants, it was noted that the age range of participants in studies without specific menopause inclusion criteria was mostly in the postmenopausal phase (Table 1). Additionally, a previous study conducted by Álvarez-Bustos et al. investigated the prevalence of fatigue in breast cancer survivors and reported that only 9% of participants reported no fatigue at all [103]. These findings suggest that the actual participants included in these nine studies were not significantly different from each other, which supports the assumption of transitivity in network meta-analysis. As a confirmation, our study passed the inconsistency test and the sensitivity analysis of one-study removal, indicating that no specific study or study group caused inconsistency or instability in the results.
Furthermore, our study only investigated the effect of 12 weeks of exercise on quality of life, and it is unknown whether exercise types that did not show significant effects at 12 weeks might lead to improvements with longer duration of exercise (e.g., 24 or 48 weeks). Future network meta-analyses with longer follow-up periods are needed to investigate this question. However, we found that studies with longer intervention periods, such as 24-week ones [55][56][57][58][59][60][61], were less abundant in our literature review, and their results may not be directly comparable or applicable to our 12-week study.

Conclusions
In summary, for breast cancer survivors, aerobic and strength concurrent training for 12 weeks is the exercise of choice to improve quality of life, with dropout rates comparable to the control group.
Supplementary Materials: The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/cancers15133380/s1, Table S1: PRISMA for network meta-analysis checklist; Table S2: Keywords and search results in different databases; Table S3: Excluded studies and reasons; Table S4: Detailed quality assessment of included studies; Table S5: Inconsistency test results of quality of life improvement for breast cancer survivors; Table S6: Inconsistency test results for risk difference of dropout rates; Figure S1: Summary of quality assessment for the studies included; Figure S2: The forest plot of pair-wise comparisons for quality of life improvement; Figure S3: The forest plot of pair-wise comparisons for the risk difference of dropout rates; Figure S4: Sensitivity analysis with the one-study removal method; Figure S5: Sensitivity analysis with pre-post correlation coefficient changed from 0.8 to 0.5; Figure S6: Funnel plot of all paired comparisons involving the common comparator, control group. Funding: This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Institutional Review Board Statement:
This network meta-analysis did not intervene or interact with humans or collect identifiable private information and thus, does not require institutional review board approval.

Informed Consent Statement: Not applicable.
Data Availability Statement: Data are contained within the article and Supplementary Files.

Conflicts of Interest:
The authors declare no conflict of interest.