Introduction

Breast cancer is the most common malignant neoplasm among women in much of the world, even in more developed countries, when cases of non-melanoma skin cancer are excluded. Incidence and mortality make breast cancer a very important health problem. This is associated with the fact that the majority of newly diagnosed cases are in early stages, which are highly curable with surgical and adjuvant treatment. This is the result of significant advances in the diagnosis and treatment of the disease. At the same time, high rates of cure lead to a growing contingent of women survivors of breast cancer [1]. This is a population with different medical and social needs, to which the different health systems are neither trained nor prepared to attend [2]. Caring for breast cancer survivors is therefore a matter of growing importance for health professionals.

Obesity is also a growing problem that requires attention from multiple healthcare professionals. There is evidence that obese women and women who gain weight after breast cancer diagnosis have twice the risk of recurrence and death from breast cancer in 5 years and 60% higher risk of death over 10 years, when compared to women normal weight [3, 4]. More than half of women diagnosed with breast cancer experience an increase in body weight associated with menopause and related to chemotherapy and hormonal treatment [5]. In this setting, regular physical activity can help control body weight and has already been shown to reduce the risk of breast cancer [6, 7]. Recent studies suggest that physical activity can also halve the risk of death in patients with breast cancer [8]. The Nurses’ Health Study, one of the largest cohorts in the field, showed that physically active women (from 2987 patients with early breast cancer) had half the risk of recurrence and death when compared to sedentary women [8].

Unlike studies involving chemotherapeutic treatments, physical exercise studies are consistently smaller, with shorter follow-up and different assessments regarding the type of physical exercises whether aerobic or strength exercises. In addition, the existing randomized clinical trials were inconclusive regarding the role of this intervention in anthropometric measurements or quality-of-life outcomes [9].

In this study, we aim to conduct a systematic review and meta-analysis to assess the effects of physical exercise (with or without dietary interventions) in body composition, quality of life, and survival in women after treatment of early-stage breast cancer.

Methods

Protocol and registration

We conducted this systematic review and meta-analysis using a previously published protocol [10] (PROSPERO registry number CRD42014008743) for the research question related to physical activity. We conducted this systematic review according to the Cochrane Handbook for Systematic Reviews of Interventions [11] and reported the manuscript according to PRISMA recommendations [12].

Data sources and searches

The following electronic databases were used to evaluate the indexed literature: Cochrane Database of Systematic Reviews (CDSR), Cochrane Central Register of Controlled Trials (CENTRAL), MEDLINE, and EMBASE. We conducted the electronic search and up to July 2017; it was limited to papers in English (see Supplementary Material). We have also searched gray literature in annals of major meetings and ongoing studies at ClinicalTrials.gov.

Study selection

We included randomized trials that evaluated physical exercise interventions (counseling or structured programs with supervised/individualized exercise sessions). The intervention should have been compared with usual care in women treated for stage I–III breast cancer. We included studies that performed the intervention after the end of adjuvant treatment (excluding hormone therapy) and excluded studies that applied the intervention after 5 years from the diagnosis.

The following primary outcomes were considered in the evaluation of the studies: overall survival and disease-free survival (5 years after treatment or until the maximum follow-up study). The secondary endpoints were weight loss (kg), BMI (kg/m2), waist–hip ratio, percentage body fat (%), and quality of life. Adverse events, such as exercise-induced lesions, were also been considered.

The evaluation of titles and abstracts for potentially eligible studies was conducted in paired and independently. Inclusion and data extraction were conducted for the full texts in the same manner, using a standardized form. Disagreement was solved through discussion.

Data extraction

The data used for meta-analysis and comparison between usual care and intervention were the final values of the groups after the intervention, since these were the data most frequently found. This method was used to minimize the number of errors and minimize the need for imputations. In the studies that presented only the values of the difference between final results and initial results, the final values were calculated from a simple sum of the variation with the initial value. In this case, the values of the standard deviations used were the same as the initial values of the variables. We preferred the EORTC QLQ-C30 quality-of-life score, since this was the most frequently used instrument among the studies. For studies such as Harrigan 2016 [13], Demark-Wahnefried 2014 [14], and Vallance 2007 [15] with two or more intervention groups, the data from all intervention groups were combined into a single group. Finally, some values of standard deviation of the quality-of-life data were not described by the authors in their works (Herrero 2006 [16], Lee 2014 [17], Heim 2007 [18], Rogers 2009 [19], Baruth 2015 [20] and Fields 2016 [21]); in these cases, the standard deviation was used as a weighted mean of the other studies that used the same scale.

Quality assessment

We also carried out a paired methodological quality evaluation of the individual studies, according to the Cochrane Handbook of Systematic Reviews [11] using the Cochrane risk of bias tool. Disagreement was solved through discussion. The overall quality of evidence was assessed using the Grading of Recommendations, Assessment, Development and Evaluations (GRADE) [22], and verified by a third reviewer. The quality of evidence was classified as ‘high,’ ‘moderate,’ ‘low,’ or ‘very low.’

Data synthesis and analysis

The data were combined using the random-effect meta-analysis model, with DerSimonian–Laird estimator as variance estimator. We estimated the treatment effect using the mean difference (MD) as summary measure for continuous outcomes. For continuous outcomes presented in different scales, we used the standardized mean difference (SMD). Data were presented with 95% confidence intervals (CI). All analyzes were performed using software R, version 3.3.2, meta packages version 4.8-4.

Statistical heterogeneity was assessed in each meta-analysis using the statistics I2. Heterogeneity was considered substantial if the I2 was greater than 50%. Heterogeneity was explored through subgroup analysis. We assessed publication bias using funnel plot and the Egger test. A significant publication bias was considered if p < 0.10. We estimated the effect of publication bias on the interpretation of results by a trim-and-fill computation.

Results

Literature search

In total, 19,987 titles were located, of which 3553 were duplicates. In the review of titles and abstracts 16,434 studies were excluded 291 were fully read for assessment of eligibility. After that, 169 studies were excluded, and 60 studies entered the phase of data extraction, of those 54 studies provided quantitative data for meta-analysis. The flowchart of the search is presented in Fig. 1, and the initial characteristics of the studies are presented in Table 1.

Fig. 1
figure 1

Prisma flow diagram

Full size image
Table 1 Characteristics of the included studies
Full size table

Study characteristics

The total number of patients in all of the studies was 6303. The follow-up ranged from 1 to 101 months and the duration of the intervention ranged from 4 weeks to 24 months. The mean age of the patients across the trials was 47.4 years, and the time from breast cancer surgery or from finishing the adjuvant treatment to intervention ranged from 1 month to 4 years. We found multiple types of physical exercise interventions but the most frequent modality was structured or individualized programs (found in 41 trials).

Outcomes

We found only one trial that reported disease-free survival in an abstract. This was an Australian trial that randomized 337 women 6 weeks after surgery for breast cancer to 8 months of physical exercise counseling (either in person or by telephone) or to usual care. After a follow-up of 101 months, the HR for overall survival in the intervention groups compared to placebo was 0.45 (95% CI 0.21–0.97); there was no difference in disease-free survival between the groups (HR 0.66 [95% CI 0.38–1.17]). Despite not establishing causality, the study suggests the potential of physical activity to influence survival.

For the secondary outcomes we found that physical exercise interventions, whether by orientation or by structured programs, promote statistically significant weight reduction in breast cancer patients (− 1.36 kg; 95% CI − 2.51 to − 0.21; p = 0.02; I2 = 17%; very low quality of evidence) (Fig. 2) and BMI reduction (− 0.89; CI 95% − 1.50 to − 0.28; p < 0.01; I2 = 51%; low quality of evidence) especially for the diet and exercise counseling group (Fig. 3). Physical exercise also showed a statistically significant decrease of − 1.60 in body fat percentage points (95% CI − 2.31 to − 0.88; p < 0.01; I2 = 28%; low quality of evidence) especially in the structured or individualized exercise program (Fig. 4).

Fig. 2
figure 2

Forest plot for weight reduction

Full size image
Fig. 3
figure 3

Forest plot for BMI reduction

Full size image
Fig. 4
figure 4

Forest plot for fat percentage reduction

Full size image

The overall quality of life was significantly modified by physical exercise interventions showing a standardized mean difference of 0.45 (95% CI 0.20–0.69, p < 0.01; I2 = 83%; very low quality of evidence) (Fig. 5). The same can be said about the effect on the physical domain of quality of life, showing a standardized mean difference of 0.51 (95% CI 0.23–0.79; p < 0.01; I2 = 86%; very low quality of evidence) (Supplementary Material) and the mental domain, showing a standardized mean difference of 0.28 (95% CI 0.06–0.50; p = 0.013; I2 = 71%; very low quality of evidence) in favor of the intervention groups (Supplementary Material). No significant effect of physical exercise could be seen on HOMA-IR (− 0.03; 95% CI − 0.20 to 0.13; p = 0.68) (see Supplementary Material).

Fig. 5
figure 5

Forest plot for quality of life (general) for different scales

Full size image

We performed subgroup analysis for all the anthropometric outcomes according to type of intervention and for the different quality-of-life scales (as seen in Figs. 2, 3, 4, 5, and in the Supplementary Material). Meta-regression analysis showed no association between duration of intervention and all outcomes.

The Egger test showed significant publication bias for all outcomes, except for fat percentage reduction or the mental domain of quality of life. Funnel plots for all the outcomes can be seen in the Supplementary Material. Trim-and-fill computation resulted in loss of statistical significance when publication bias was corrected. After this correction, the effect of the intervention on weight was − 0.18 (95% CI − 1.52 to 1.15), on BMI was − 0.04 (95% CI − 0.67 to 0.60), on the general quality of life was 0.08 (95% CI − 0.2 to 0.36), and on physical quality of life was 0.09 (95% CI − 0.22 to 0.40).

Quality evaluation

The risk of bias according to Cochrane risk of bias tool is presented in Figs. 6 and 7. Serious methodological flaws were found such as poor randomization methods and poor outcome assessments. Noteworthy, most of the studies were relatively small and just one study, published as an abstract, evaluated outcomes such as overall mortality or disease-free survival [23]. The GRADE summary of findings table is shown in Fig. 8.

Fig. 6
figure 6

Risk of bias graph

Full size image
Fig. 7
figure 7

Risk of bias summary

Full size image
Fig. 8
figure 8

GRADE summary

Full size image

Discussion

The diagnosis of breast cancer affects patients’ quality of life and has long-lasting consequences. Both the diagnosis itself and the treatments that those patients are submitted to have significant influence in mental and physical health for years to come. The incidence of breast cancer in its earlier stages is increasing due to improved screening. In the same way, treatments for the disease are becoming more effective, and so the number of survivors is expected to increase in the next years [24]. Many of these women might live for many years and have time to develop chronic conditions just as the general population [25]. Therefore, high-quality evidence for survivorship guidance will be necessary. Physical activity is key to improve health and quality of life in any population, and here it is not an exception.

Our goal was to collect the best evidence regarding the impact of physical activity on the body composition and quality of life of patients who had been treated for breast cancer with curative intent. We have searched for randomized clinical trials only, considering them to be the best source of evidence for the question. To avoid confusion regarding quality of life, we exclude trials that assessed patients during treatment with chemotherapy or radiation; the included patients may have been receiving adjuvant hormone therapy. Also, trials evaluating patients with metastatic disease were not included, since treatment goals and prognosis are significantly different in more advanced stages of the disease. We may say that we have collected the best evidence about this issue.

The results of this meta-analysis demonstrated a significant decrease in body weight and BMI and an improvement in quality-of-life outcomes, but a serious bias effect was demonstrated. Publication bias was found in almost all outcomes and is a concern regarding physical activity trials, since all of them lost statistical significance after correction.

We found only one study that described overall survival and progression-free survival: it was from an Australian population of 337 newly diagnosed breast cancer women showing lower mortality rates in the physical activity group [23].

Nonetheless, our meta-analysis has several limitations. First, most of the included studies were small and many of them had significant methodological flaws. In some studies reducing pain was the main outcome in aromatase inhibitors users, such as Nyrop 2017 [26] and Fields 2016 [21], so some selection bias was expected from this population. Second, the interventions were very different among the studies, making it very difficult to select any one of them as the best. The interventions varied in duration and quality, since they comprised from exercise counseling to structured and supervised exercise programs. Because of this we chose to compare exercise interventions as a single group. It was also not possible to determine if any of the interventions should have been considered ineffective at all. Third, quality of life was assessed using different scales, which may have influenced the results, since these scales correspond to different metrics assessing the same underlying outcome. Finally, long-term data are seldom available, and conclusions about the possible impact in outcomes such as overall survival or disease-free survival were not possible. High-quality randomized trials with larger numbers of patients are required, and patients should be followed for longer periods of time to assess if the benefits of the interventions are long lasting or temporary.

It is also worth commenting that the studies included were broadly heterogeneous, with variations in the type of intervention (counseling, face-to-face orientation, practicing, phone call orientation), therefore we cannot classify all these interventions under the same label, once they are not equivalent. This study shows that any intervention towards encouraging physical activity is valid and promotes good life habits that might affect the quality of life, the metabolic profile, and mortality.

With all the limitations considered, clinical trials of physical activity as an intervention for breast cancer survivors are highly justifiable, especially because exercise is a cheap intervention, is easy to apply, and practically has no contraindications or adverse effects. Even though a strong and solid evidence relating physical activity and reduced risk of recurrence and death from breast cancer is lacking, survivors are oriented to avoid a sedentary lifestyle and seek to exercise regularly. The potential benefits of this practice can help maintaining a healthy weight, improve fitness and tolerance to treatment, stimulate healthy lifestyles, and prevent other chronic diseases for which this population is particularly vulnerable.