The Effects of Post-diagnosis Recreational Aerobic Exercise among Breast Cancer Survivors: A Systematic Review/Meta-Analysis

Abstract

We aimed to estimate the incremental effects of post-diagnosis recreational aerobic exercise and possible variations in effects on recurrence and mortality to support individualized breast cancer survivorship care plans in clinical settings. Seven databases were searched to identify observational studies that examined the effects of exercise on recurrence, breast cancer–specific mortality, and all-cause mortality (ACM) among female breast cancer survivors. Fully adjusted hazard ratios (HR) and 95% confidence intervals were extracted for comparisons reported in relation to no/minimal exercise (reference). Dose–response relationships between exercise and events were examined using restricted cubic splines. Less than half of participants (44.3%, n = 50,689) met aerobic exercise guidelines for health (≥∼2.5 hours/week). Meeting guidelines was associated with a ∼50% reduction in the HR for ACM, with further reductions up to ∼4.5 hours/week. A ∼25% reduction in the HR for ACM was associated with ∼1 hour/week. The 5-year (unadjusted) ACM rates were 11% for no/minimal exercise, 4% for insufficient exercise, and 3% for meeting exercise guidelines (n = 5 studies). There were limited data for subgroups. Similar patterns were observed for recurrence and breast cancer–specific mortality. Exercise may lower the risk of recurrence and mortality among breast cancer survivors. Though meeting guidelines for health offers the greatest benefits, exercise below guideline-recommended exercise levels is also beneficial.

Introduction

Aerobic exercise, i.e., exercise that increases the heart rate and the body’s oxygen use (1), may reduce the burden of treatment-related side effects (e.g., fatigue; ref. 2) and lower the risk of comorbidities (3), recurrence (4–8), and mortality among breast cancer survivors (4, 5, 7–12). These benefits have been linked to the impact of exercise on adiposity, adipocytokine profiles, insulin-related pathways, and inflammation in cancer survivors (13–15). Current cancer exercise guidelines recommend survivors build toward meeting existing exercise guidelines for health: ≥2 hours and 30 minutes/week (i.e., ≥150 minutes/week) of aerobic exercise and ≥2 days/week of muscle-strengthening exercise (16). However, many breast cancer survivors experience challenges maintaining exercise levels during and following cancer treatment (17, 18), with 37.7% [95% confidence interval (CI), 36.1%–39.4%] of female breast cancer survivors in the United States meeting aerobic exercise recommendations for health (≥150 minutes/week) compared to 40.9%(95% CI, 40.5%–41.4%) of women without cancer (19).

Communication of exercise information to breast cancer survivors may achieve meaningful increases in exercise participation (20–23). However, discussions about exercise between healthcare providers and breast cancer survivors are not standard practice. Recent studies show that only 54.4% of breast cancer survivors report having discussed exercise with a healthcare provider (24), and only 44.1% of healthcare providers reported giving exercise advice to their patients (25). Barriers to exercise discussions in clinical settings may be related to clinicians’ lack of knowledge, time, self-efficacy, low availability of supportive resources, and limited data to guide individualized discussions about exercise in clinical settings (26).

Current exercise guidelines recommend that clinicians counsel cancer survivors to “avoid inactivity” (16). During treatment, guidelines suggest that survivors should be provided with individualized exercise prescriptions that build up to ≥90 minutes/week of aerobic exercise and ≥2 days/week of muscle-strengthening exercise to manage the common side effects of cancer treatment. Ultimately, survivors should gradually progress toward meeting exercise guidelines for health: ≥150 minutes/week of aerobic exercise and ≥2 days/week of muscle-strengthening exercise (16, 27, 28). As a result, the recurrence and survival benefits of exercise below the guideline-recommended ∼150 minutes/week may be a useful starting point for patient–provider exercise discussions in clinical settings (29, 30). However, to our knowledge, previous studies have provided limited information on the incremental effects of post-diagnosis recreational aerobic exercise (i.e., the effects of exercise relative to a consistent “fixed” comparator) to help identify the benefits of exercise below guideline-recommended exercise levels in relation to recurrence and survival. Furthermore, clinical guidelines recommend providing exercise recommendations that consider the individual demographic, clinical, and treatment characteristics of cancer survivors. Therefore, clinicians may also find information on the variation of these benefits based on individual characteristics to advise patients seen in clinical settings (31). Yet, existing studies (4–7, 9–12) provide limited data on the variations of the relationship between exercise and recurrence/mortality considering breast cancer survivors’ individual, clinical, and/or contextual characteristics. This information may especially be useful because exercise discussions are now a part of survivorship care plans for breast cancer survivors in the United States (29, 30).

In this study, we aimed to address these knowledge gaps by estimating incremental effects and the amount of exercise needed to achieve approximately half of the estimated benefits of meeting exercise guidelines for health. We also aimed to summarize information on the absolute benefits of exercise to support clinical discussions in relation to recurrence and survival. In addition, we summarized the variations of exercise benefits based on the individual, clinical, and/or contextual characteristics of breast cancer survivors. The overarching goal of this study was to ascertain the quantitative relationship of exercise with the risk of recurrence and mortality to support individualized breast cancer survivorship care plans incorporating exercise in clinical practice.

Materials and Methods

This systematic review and meta-analysis was carried out and reported in accordance with the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guidelines (32). The PRISMA Checklist is included in Supplementary Table S1. The review was registered in the Open Science Framework (https://osf.io/dexnh).

Data sources and search strategy

We conducted a search of the published literature that reported on the association between aerobic exercise and at least one of the following events among breast cancer survivors: (i) all-cause mortality (ACM), (ii) breast cancer–specific mortality (BCSM), (iii) non–breast cancer mortality, and/or (iv) recurrence. Breast cancer survivors were defined as women ever diagnosed with breast cancer (33). Our comprehensive search strategy included a combination of terms relating to breast cancer survivors, exercise (aerobic), and recurrence/mortality.

The search strategy was developed using an iterative approach with three rounds of preliminary searches and refinement of the search strategy based on initial search results by a trained librarian (GB) at the NIH. Each preliminary search was performed in the PubMed/MEDLINE database constrained to research published from January 2003 until October 2024 with English-only language restrictions. The preliminary PubMed/MEDLINE search strategy was pilot-tested with 50 records. Two authors screened each title/abstract independently and resolved disagreements via consensus. A subsample of those included after screening at full text were also used to pilot test the data extraction tool.

The comprehensive search strategy included a combination of keywords, synonyms, medical subject headings terms, and Emtree terms related to exercise and events in women diagnosed with breast cancer. Ultimately, we performed the search across the following seven widely used scientific databases: MEDLINE via PubMed (National Library of Medicine), PsycINFO (American Psychological Association), Embase (Elsevier), Scopus (Elsevier), Web of Science Core Collection (Clarivate Analytics), Cochrane CENTRAL (Wiley & Sons), and CINAHL (EBSCO). The search strategies for each database are detailed in Supplementary Table S2. We also screened the references lists of each article to identify sources that might have been missed in the initial search. The date of our last search was October 2024.

Inclusion and exclusion criteria

Inclusion criteria were as follows: (i) female breast cancer survivors; (ii) examination of the effect of post-diagnosis recreational aerobic exercise on risk of recurrence and/or mortality; (iii) primary empirical observational research studies; and (iv) full-text available in English. See Supplementary Table S3 for details. The association of muscle-strengthening exercise with recurrence and mortality is the focus of a separate review (34).

Screening procedure

We first imported the articles obtained from the database searches into EndNote, which identified and excluded most duplicates. Screening was conducted in Covidence (RRID: SCR_016484), a web-based systematic review screening software platform. Next, the article titles and abstracts were independently screened to gauge shared understanding of the selection criteria, discuss disagreements, and further specify the inclusion and exclusion parameters. Additional duplicates identified by authors were removed at this stage. Finally, full texts were screened independently, with conflicts resolved through discussion.

Quality appraisal

Article quality was appraised using the Newcastle–Ottawa scale for assessing the quality of non-randomized studies in a meta-analysis (35). The scale uses a ‘star system’ to judge studies on three broad perspectives: selection of study groups, comparability of groups, and ascertainment of the outcome of interest. Two authors appraised each article independently and resolved disagreements via discussions to reach consensus. We also extracted details on the covariates used as well as model fit statistics of articles included in our main analyses for ACM, BCSM, and recurrence. Model fit statistics were extracted as they are needed to determine the effect of unmeasured confounding.

Data charting and synthesis

Extraction

Data were extracted into Microsoft Excel. Authors were contacted via email for additional information and clarification if articles were missing data. The Hazard Ratios (HR) extracted were those identified as the primary adjusted model by the authors.

Article characteristics

Article characteristics and aerobic exercise measurement and guidelines were summarized descriptively in the text using a narrative approach supported by accompanying Supplementary Materials and Methods. We calculated the proportion of participants meeting approximately the equivalent of exercise guidelines for health from various organizations (e.g., ACS) equivalent to ∼≥2 hours and 30 minutes/week. This was done using 17 articles that reported mutually exclusive data from 18 studies.

Metabolic equivalents

The frequency, intensity, and duration of self-reported exercise are used to calculate the metabolic equivalent of task (MET), which provides a standardized way to compare the energy expenditure (i.e., amount of exercise; ref. 36). Data extracted from each source were converted to MET hours/week so that we had a consistent variable to explore the dose–response relationship with spline analyses. Consistent with previous literature, we used a ratio of four MET minutes for each minute of moderate-intensity exercise (37). For example, 150 minutes/week of moderate-intensity exercise is equivalent to 600 MET minutes/week or 10 MET hours/week. For clinical translation, MET hours/week were converted back into hours/week of moderate-intensity exercise.

Comparators

We extracted data for no/minimal exercise (reference; ∼<45 minutes/week; ∼<3 MET hours/week), engaging in some exercise (i.e., less than meeting guidelines ∼≥45 to ∼<150 minutes/week; ∼≥3 to ∼<10 MET hours/week), or meeting exercise guidelines for health (∼≥150 minutes/week; ∼≥10 MET hours/week).

Long-term events

We extracted and summarized the results for breast cancer recurrence, BCSM, and ACM.

Absolute differences

We extracted 5- and 10-year unadjusted recurrence and survival rates for no/minimal exercise, some exercise, or meeting aerobic exercise guidelines when data were available from Kaplan–Meier curves. There were only sufficient data to conduct a meta-analysis for 5-year ACM rates. Findings for other events were reported separately.

Relative differences

We extracted HRs for breast cancer recurrence, BCSM, and ACM for no/minimal exercise, some exercise, or meeting aerobic exercise guidelines for health when data were available, computable, or requestable from authors. To estimate a pooled effect and corresponding 95% CIs, the HRs were weighted by the inverse of their variance in random effects models (38). For ease of interpretation, the natural log transformations of summary estimates and their 95% CIs were converted back to HRs. Articles that reported a HR that did not approximate the geometric mean of the CIs were excluded from meta-analyses as this may indicate an error in reporting. Details on the variables adjusted for in analyses are reported in Supplementary Materials and Methods.

Spline

We conducted restricted cubic spline analyses to examine the nonlinear dose–response relationship between exercise and recurrence, BCSM, and ACM (39). We extracted HRs from fully adjusted Cox proportional models reported in each study for all available exercise levels (e.g., >0 to <10 MET hours/week) in comparison with “no/minimal exercise” (reference group). Point estimates for each exercise level were assigned using the midpoint of the range of each comparison group. For example, five MET hours/week was assigned for a comparison group with a range of >0 to <10 MET hours/week. If a comparison group did not have an upper limit (e.g., ≥10 MET hours/week), the width of the previous comparison group reported in the same study was applied to determine the midpoint of the subsequent group. These estimates were included in the spline regression analysis, and HRs for two MET hour/week increments were extracted from the regression analysis. MET hours/week were converted to the equivalent hours/week of moderate-intensity aerobic exercise (2 MET hours/week is equivalent to 0.5 hours/week of moderate-intensity aerobic exercise according to the International and Global Physical Questionnaire data processing and analysis guidelines; refs. 37, 40) to facilitate the interpretation of results. Knots were placed at the 10th, 50th, and 90th percentiles. Sensitivity analyses were conducted placing knots and different percentiles, and no differences were found. Splines for recurrence and BCSM relied on a small number of data points. Splines were fit through the estimated HRs and lower and upper CIs which are exponentiated and therefore not symmetrical.

Subgroup analyses

Effects across subgroups identified based on individual (e.g., race and/or ethnicity) and clinical (e.g., comorbidities) characteristics were extracted when reported.

Between-study variance and sensitivity analyses

The impact of heterogeneity between studies was assessed using the I² statistic. I² statistics of 25%, 50%, and 75% are indicative of a variation attributable to low, moderate, and high heterogeneity of underlying target parameters, respectively (41, 42). Leave-one-out (43) sensitivity analyses were conducted for each meta-analysis by sequentially omitting one study at a time to examine their impact on analyses. Small study effects were also assessed.

Software

All analyses were conducted using Stata (Stata/SE 18.0, StataCorp).

Data availability

Data sharing is not applicable to this review as no datasets were analyzed, and all data generated during the review process are reported in the article or Supplementary Material.

Results

Source selection

Initial searches retrieved 5,831 sources following the removal of duplicates. These were screened at the title and abstract levels, followed by a full-text review of 1,308 remaining sources. Thirty articles from 20 studies were identified and proceeded to extraction (see PRISMA flow diagram in Supplementary Fig. S1). Reasons articles were excluded from the meta-analyses are detailed in Supplementary Table S4.

Source characteristics

Details of the study and participant characteristics are reported in Supplementary Tables S5 and S6. All studies combined included more than 50,000 women. Most studies originated from the United States (n = 15), with other studies also including participants from Canada, China, Denmark, Germany, and the Netherlands. In the 17 studies that reported stage at diagnosis, more than 80% of the women were diagnosed with stages I to II or localized breast cancer (44–60). Across the 19 studies that reported age, the median age of the women included in the analyses ranged from 51 to 73 years (44–62). Of the 14 studies with details on chemotherapy, five reported that the majority (>90%) of women had received chemotherapy (47, 53, 61–63). The proportion of women who had received hormone therapy ranged from 22% to 100% (45, 46, 49–51, 53–56, 59, 60, 62). Of the 13 studies with details on radiation treatment, the proportion of women who had received radiation therapy ranged from 27% to 70% (44, 49, 51, 53, 59, 60). The Shanghai Breast Cancer Survival Study (SBCSS) included mainly Chinese Asian women (63–65), whereas all other studies to report race and/or ethnicity were comprised of predominantly White women (range, 52%–99%). Of the 13 studies with details on education status, eight reported that the majority (range, 54%–100%) had an education greater than the equivalent of high school and/or a general education diploma (46, 47, 50, 53, 54, 57, 61, 62).

Source quality appraisal

Quality appraisal details are reported in Supplementary Materials and Methods S1 and Supplementary Table S7. All 20 studies included participants who were somewhat representative of the community from which they were recruited. Only three studies (15%; refs. 47, 55, 63) analyzed exercise measured at multiple time points. Eight studies (40%; refs. 48, 51, 55, 60–63, 66) measured exercise via an interview as opposed to a self-report survey. Eleven studies (55%; refs. 44, 48, 50, 55, 57–61, 67, 68) demonstrated that breast cancer was not present at the start of their study. With respect to comparability, all but two studies (90%; refs. 44, 50) adjusted for age in the design or analysis, whereas all studies adjusted for other individual and/or clinical characteristics in their design or analysis. All studies determined breast cancer diagnosis via medical records or record linkage as opposed to self-report. All but three studies (85%; refs. 49, 50, 66) had a follow-up greater than or equal to 5 years. Seven studies (35%; refs. 49, 52, 53, 63, 66, 69, 70) reported having adequate follow-up (i.e., participant retention and data completeness) for their cohort. It is also worth noting that no study reported model fit statistics (Supplementary Table S8). Details on retention, time from diagnosis to exposure (i.e., exercise measurement), follow-up length, and time points at which exposure was measured are detailed in Supplementary Table S9.

Aerobic exercise measurement and guidelines

The weighted average of breast cancer survivors meeting exercise guidelines for health (∼≥150 minutes/week; ∼≥10 MET hours/week) was 44.3% (n = 50,689). Exercise was measured using a range of methods, details of which are reported in Supplementary Table S10. The Shanghai Women’s Health Study Physical Activity Questionnaire was used by the three articles from the SBCSS (63–65). The European Prospective Investigation into Cancer and Nutrition Physical Activity Questionnaire was used by the Danish and Dutch studies (44, 51). Cannioto and colleagues (47, 71) used recreational items from the Physical Activity Questionnaire in the two articles from the Diet, Exercise, Lifestyle and Cancer Prognosis (DELCaP) study. The study from Kaiser Permanente used the Godin–Shephard Leisure-Time Physical Activity Questionnaire (48). A study using Healthy Eating, Activity, and Lifestyle (HEAL) study data used the Modified Activity Questionnaire (58). Outside of the three articles that included Nurses’ Health Study (NHS) data (50, 67, 70), four other articles from three studies used items from the NHS (52, 57, 62, 72). Articles including Life After Cancer Epidemiology (LACE) or Pathways study data used the Arizona Activity Frequency Questionnaire (53, 54, 69, 70, 73). Outside of the two articles that included Women’s Health Initiative (WHI) data (56, 68), four articles from that used data from the Women's Healthy Heating and Living (WHEL) study also used versions of WHI items (46, 70, 74, 75). The name of instrument used to measure aerobic exercise was unclear in five studies (51, 55, 59–61). Full details on the instruments used to measure exercise, guidelines applied, and the proportion meeting guidelines are reported in Supplementary Table S10.

ACM

Absolute differences

A meta-analysis of five studies (46, 48, 58, 65, 67) comprising 7,293 women of whom 813 (11.1%) died during follow-up found that the 5-year ACM rate was 11% for no/minimal exercise (∼<45 minutes/week; ∼<3 MET hours/week), 4% for some exercise (∼≥45 to <150 minutes/week; ∼≥3 to <10 MET hours/week; absolute difference 7%), and 3% for meeting exercise guidelines for health (∼≥150 minutes/week; ∼≥10 MET hours/week; absolute difference 8%; Table 1). Bertram and colleagues (46) reported 9-year ACM rates of 11% for those not meeting guidelines and 6% for those meeting guidelines. Holmes and colleagues (67) reported 10-year ACM rates of 13% for no/minimal exercise, 10% for some exercise, and 7% for meeting exercise guidelines.

Table 1.

Unadjusted 5-year ACM rates for no/minimal exercise, some exercise, and meeting exercise guidelines.

	Mortality rate			% Absolute differences
	No/minimal exercise	Some exercise	Meeting exercise guidelines	Some exercise – no/minimal exercise	Meeting exercise guidelines –some exercise
Bertram and colleagues (46)	-	-	3 (0.5)	-	-
Chen and colleagues (48)	-	4 (1.6)	1 (1.3)	-	3
Holmes and colleagues (67)	7 (0.8)	3 (0.5)	3 (0.5)	4	4
Irwin and colleagues (58)	29 (2.3)	11 (2.0)	6 (1.4)	18	23
Su and colleagues (74)	14 (1.1)	10 (1.0)	13 (1.1)	4	1
Weighted average	11	4	3	7	8

NOTE: Weighted based on sample size. No/minimal exercise = ∼<45 minutes/week of moderate-intensity exercise (<∼3 MET hours/week). Some exercise = ∼≥45 to <150 minutes/week moderate-intensity exercise (∼≥3 to <10 MET hours/week). Meeting exercise guidelines for health = ∼≥150 minutes/week moderate-intensity exercise (∼≥10 MET hours/week).

Relative differences

A meta-analysis of 12 studies (46, 47, 50, 51, 56–58, 60, 64, 66, 72, 76) comprising 39,963 women of whom 7,247 (18.1%) died of all causes was conducted (Fig. 1). Compared with no/minimal exercise (<∼45 minutes/week; <∼3 MET hours/week), some exercise (∼≥45 to <150 minutes/week; ∼≥3 to <10 MET hours/week; HR = 0.72; 95% CI, 0.62–0.83) and meeting exercise recommendations for health (∼≥150 minutes/week; ∼≥10 MET hours/week). HR = 0.60; 95% CI, 0.51 to 0.70) were associated with a lower risk of ACM. An additional meta-analysis of five studies (44, 47, 48, 52, 77) comprising 14,421 women of whom 2,271 (15.7%) died of all causes was conducted (Supplementary Fig. S2). Meeting exercise guidelines for health (∼≥150 minutes/week; ∼≥10 MET hours/week) was associated with a lower risk of ACM (HR = 0.64; 95% CI, 0.50–0.82) compared with not meeting guidelines (<150 minutes/week; <10 MET hours/week).

Figure 1.

Forest plots for meta-analyses of the association between post-diagnosis aerobic exercise and ACM among breast cancer survivors. It contains summary HR estimates (95% CI) of (A) ACM for some exercise (∼≥45 to <150 minutes/week; ∼≥3 to <10 MET hours/week) compared with no/minimal exercise (∼<45 minutes/week; ∼<3 MET hours/week) and (B) ACM for meeting exercise guidelines for health (∼≥150 minutes/week; ∼≥10 MET hours/week) compared with no/minimal exercise. NHIS, National Health Interview Survey; WHI, Women’s Health Initiative. CWLS, Collaborative Women's Longevity Study; DELCap, Diet, Exercise, Lifesyle, and Cancer Prognosis: HEAL, Healthy, Eating, Activity, and Lifestyle; NHS, Nurses' Health Study; NMWHS, New Mexico Women's Health Study; MARIE, Mammary Carcinoma Risk Factor Investigation; WHEL, Women's Health Eating and Living; WISC, Wisconsin in Situ Cohort.

Spline

Figure 2 shows a nonlinear dose–response association between exercise and ACM estimated using a restricted cubic spline regression. The estimated HRs and CIs reported in Table 2 at incremental values of 2 MET hours/week are based on the restricted cubic spline regression (Table 2). Aerobic exercise equivalent to meeting guidelines for health, ∼≥150 minutes/week (∼≥2.5 hours/week; ∼≥10 MET hours/week), was associated with a ∼50% reduction in the HR for ACM, with steady declines in the HRs up to around 4.5 hours/week. Half of this reduction associated with the equivalent meeting exercise guidelines, i.e., ∼25%, was associated with ∼1 hour/week (∼4 MET hours/week) of exercise. Increasing levels of exercise beyond ∼6 hours/week was associated with a steady increase in the HRs for ACM.

Figure 2.

Spline graph of the association between post-diagnosis aerobic exercise and ACM among breast cancer survivors. It depicts the nonlinear association between exercise and ACM (0 hours/week is the reference). The gray shaded areas denote the range within which women would approximately meet exercise guidelines (left) and participate in approximately double the equivalent of the guidelines (right); ∼0.5 hours/week of moderate-intensity exercise = ∼2 MET hours/week.

Table 2.

Estimated HRs and CIs at incremental 2 MET hour/week intervals in relation to ACM.

MET hours/week	Moderate-intensity exercise (equivalent hours/week)	Estimated HR	95% CI		Interpretation of point estimates and 95% CIs
			Lower limit	Upper limit
2	0.5	0.91	0.86	1.00	Statistically consistent with parameter values ranging from lower to no difference in the risk of ACM
4	1.0	0.76	0.64	0.97	Statistically consistent with a lower risk of ACM
6	1.5	0.65	0.48	0.95
8	2.0	0.56	0.36	0.93
10	2.5	0.49	0.27	0.93
12	3.0	0.44	0.21	0.93
14	3.5	0.40	0.17	0.93
16	4.0	0.38	0.15	0.96
18	4.5	0.37	0.13	0.98
20	5.0	0.37	0.13	1.01	Statistically consistent with parameter values ranging from a considerably lower to slightly higher risk of ACM
22	5.5	0.37	0.13	1.04
24	6.0	0.38	0.13	1.08	Statistically consistent with parameter values ranging from a considerably lower to higher risk of ACM
26	6.5	0.40	0.14	1.12
28	7.0	0.43	0.16	1.17
30	7.5	0.46	0.19	1.22
32	8.0	0.49	0.21	1.27
34	8.5	0.53	0.25	1.32
36	9.0	0.57	0.29	1.38
38	9.5	0.62	0.33	1.44
40	10.0	0.66	0.37	1.50

Subgroup analyses

There were minimal differences in the risk of ACM based on weight status or race and/or ethnicity based on relatively limited evidence (Supplementary Table S11). By contrast, meeting guidelines for health (∼≥150 minutes/week; ∼≥10 MET hours/week) seemed to lower the risk of ACM more among postmenopausal women compared with premenopausal women. Women diagnosed with stages IIb to III breast cancer who met guidelines seemed to have lower risk of ACM compared with similarly active women diagnosed with stages I to IIa breast cancer. For hormone receptor status, two studies reported subgroup analyses for estrogen receptor/progesterone receptor (ER/PR)-negative and ER/PR-positive. Fortner and colleagues (50) reported findings that suggest meeting exercise guidelines offers benefits for ER/PR-positive based on analyses of a sample or predominantly White women. By contrast, Chen and colleagues (64) reported findings that suggest meeting exercise guidelines offers benefits for ER/PR-negative based on analyses of a sample or predominantly Asian women. In addition, Jung and colleagues (51) found that sufficient post-diagnosis exercise may offer survivors who participated in insufficient prediagnosis exercise a reduction in the risk of ACM (HR = 0.43; 95% CI, 0.26–0.72; Supplementary Table S11).

BCSM

Absolute differences

The unadjusted 5-year BCSM was ∼2% to 3% across exercise levels (absolute difference: <∼1%; refs. 57, 69). According to Jones and colleagues (69) the unadjusted 10-year BCSM rates were ∼7% for no/minimal exercise (∼<45 minutes/week; ∼<3 MET hours/week), ∼6% for some exercise (∼≥45 to <150 minutes/week; ∼≥3 to <10 MET hours/week), and ∼5% for meeting exercise guidelines for health (∼≥150 minutes/week; ∼≥10 MET hours/week).

Relative differences

A meta-analysis of eight studies (50, 51, 56–58, 60, 66, 76) comprising 29,900 women of whom 2,292 (7.7%) died of breast cancer was conducted (Fig. 3). Compared with no/minimal exercise (<∼45 minutes/week; <∼3 MET hours/week), meeting exercise guidelines for health (∼≥150 minutes/week; ∼≥10 MET hours/week) was associated with a lower risk of breast cancer mortality (HR = 0.71; 95% CI, 0.59–0.85). Data were too sparse to estimate BCSM rates at incremental levels of exercise, but the spline analysis result for BCSM is reported in Supplementary Fig. S3. In addition, Beasley and colleagues (77) also reported that meeting exercise guidelines for health (∼≥150 minutes/week; ∼≥10 MET hours/week) was associated with a lower risk of BCSM (HR = 0.75, 95% CI, 0.65–0.85) compared with not meeting exercise guidelines (<150 minutes/week; <10 MET hours/week).

Figure 3.

Forest plots for meta-analyses of the association between post-diagnosis aerobic exercise and BCSM among breast cancer survivors. It contains summary HR estimates (95% CI) of (A) BCSM for some exercise (∼≥45 to <150 minutes/week; ∼≥3 to <10 MET hours/week) compared with no/minimal exercise (∼<45 minutes/week; ∼<3 MET hours/week) and (B) BCSM for meeting exercise guidelines for health (∼≥150 minutes/week; ∼≥10 MET hours/week) compared with no/minimal exercise. CWLS, Collaborative Women's Longevity Study; HEAL, Healthy, Eating, Activity, and Lifestyle; MARIE, Mammary Carcinoma Risk Factor Investigation; NHIS, National Health Interview Survey; NHS, Nurses' Health Study; NMWHS, New Mexico Women's Health Study.

Subgroup analyses

Variations in the effects are reported in Supplementary Table S12. In general, the direction of findings were similar to overall findings but with lower precision. In addition, Jung and colleagues (51) found that sufficient post-diagnosis exercise may offer survivors who participated in insufficient post-diagnosis exercise a reduction in the risk of BCSM (HR = 0.48; 95% CI, 0.25–0.91; Supplementary Table S12).

Breast cancer recurrence

Absolute differences

According to Jones and colleagues (69), the unadjusted 5-year recurrence rates were 6% for no/minimal exercise (∼<45 minutes/week; ∼<3 MET hours/week), some exercise (∼≥45 to <150 minutes/week;∼≥3 to <10 MET hours/week), and meeting guidelines for health (∼≥150 minutes/week; ∼≥10 MET hours/week) and the unadjusted 10-year recurrence rates were 11% for no/minimal exercise, some exercise, and meeting guidelines. As such, there were no absolute differences based on exercise levels.

Relative differences

A meta-analysis of five studies (47, 51, 62, 67, 69) comprising 13,977 women of whom 1,884 (13.2%) experienced recurrence (Supplementary Fig. S4) found that some exercise may lower the risk of recurrence; however, the results were statistically consistent, with parameter values ranging from no difference to a small increase in recurrence risk (HR = 0.92, 95% CI, 0.80–1.06). Meeting exercise guidelines for health (∼≥150 minutes/week; ∼≥10 MET hours/week) was associated with a lower recurrence risk (HR = 0.80, 95% CI, 0.65–0.99). Data were too sparse to estimate recurrence rates at incremental levels of exercise, but the spline analysis result for recurrence is reported in Supplementary Fig. S5. In addition, Beasley and colleagues (77) compared the risk of recurrence between those who did and did not meet exercise guidelines for health using data pooled from four studies (LACE, NHS, SBCSS, and WHEL; ref. 78). Meeting guidelines (∼≥150 minutes/week; ∼≥10 MET hours/week) was associated with a lower risk of recurrence (HR = 0.96, 95% CI, 0.86–1.06) compared with not meeting exercise guidelines (<150 minutes/week; <10 MET hours/week). By contrast, Cannioto and colleagues (47) did not find that meeting guidelines was associated with a lower risk of recurrence (HR = 1.00, 95% CI, 0.70–1.42).

Subgroup analyses

Su and colleagues (74) reported that postmenopausal women who met exercise guidelines (≥10 MET hours/week) had a lower recurrence risk than those who did not; however, the results were consistent, with parameter values ranging from no difference to a small increase in recurrence risk (HR = 0.87, 95% CI, 0.70–1.08). Similarly, McLaughlin and colleagues (62) reported results for women diagnosed with ductal carcinoma in situ that were statistically consistent, with parameter values ranging from a considerable decrease in risk to a considerable increase in risk.

Other events

Findings for other recurrence/mortality-related events are reported in Supplementary Table S13.

Between-study variance and sensitivity analyses

Heterogeneity and small study effects for meta-analyses are reported in Supplementary Table S14. These showed that heterogeneity was moderate-to-high for ACM, low-to-moderate for BCSM, and low-to-high for recurrence. Evidence suggesting potential small study effects was only present for ACM. Sensitivity analyses are reported in Supplementary Table S15 and showed that exclusion of no one study shifted findings in a manner that differed significantly from the findings of main analyses

Discussion

The direction of our findings (i.e., meeting aerobic exercise guidelines for health reduces the risk of mortality) is consistent with previous studies (4–12). However, no difference across BCSM and ACM estimates was reported by previous meta-analyses in relation to meeting guidelines (e.g., BCSM HR: 0.86 versus ACM HR: 0.85; refs. 8, 79). In contrast, a difference across BCSM and ACM estimates for meeting guidelines was observed for our meta-analyses (HR: 0.71 vs. HR: 0.60), as well as our spline analyses (HR: ∼0.77 vs. HR: ∼0.51). A larger proportion of breast cancer survivors die due to other causes (e.g., cardiovascular diseases; refs. 80, 81); therefore, ACM estimates are expected to be lower than BCSM estimates. We believe our analysis is an improvement on past studies. In particular, the lack of a fixed, constant, and narrow reference level of exercise in previous meta-analyses may explain the absence of a difference reported across these estimates.

We also built on previous knowledge and estimate the benefits of participating in different levels of exercise below guideline-recommended levels as well as summarize unadjusted absolute benefits of exercise. Accordingly, our analyses indicate that participation in ∼1 hour/week of exercise offers a notable reduction in the HRs associated with ACM but that meeting exercise guidelines for health offers double this benefit. This information may help healthcare providers to discuss an initial exercise goal far less than meeting guidelines. Such an approach may enhance patients' self-efficacy to successfully accomplish more achievable exercise goals (82, 83). Patients can then gradually progress from this starting point to eventually meet guidelines. Further research is needed to refine the precision of these estimates, generate similar estimates for other long-term outcomes (e.g., recurrence and BCSM), estimate the adjusted absolute benefits of exercise, as well as the determine the extent to which these estimates differ based on individual, clinical, and contextual characteristics.

It is worth noting that the relative benefits of exercise on BCSM and ACM were somewhat similar. This could be due to the similar biological pathways via which exercise impacts cancer-specific mortality and ACM. Exercise has been shown to favorably impact adiposity, adipocytokine profiles, insulin-related pathways, and inflammation in cancer survivors (13–15). Therefore, any difference in relative benefits between BCSM and ACM may be attributable to cardiovascular disease–related mortality (see Supplementary Table S13 for findings on other events).

Our review extends prior reviews in several ways. First, we estimated the incremental effects of exercise in comparison with a fixed, constant, and narrow reference level of exercise. None of the previous reviews used a consistent reference group when summarizing the incremental effects of exercise on the risk of recurrence or mortality. For example, Cariolou and colleagues (8) used different exercise levels ranging from 0 up to 21.0 MET hours/week as their reference group. This wide variation in reference group definitions limited our ability to identify the incremental effects of exercise or a minimum level of exercise beneficial for breast cancer survivors. Therefore, we designed our analyses to use a consistent, fixed, and narrowly defined “no/minimal level of exercise (∼<45 minutes/week; ∼<3 MET hours/week)” as the reference group. As a result, we were able to assess the amount of exercise needed to achieve approximately half of the benefits of meeting exercise guidelines. Second, we attempted to estimate the absolute benefits of exercise. Overall, only seven studies (46, 48, 57, 58, 67, 69, 74) provided absolute risks, and there were no data on adjusted absolute risks. However, preliminary evidence suggests that exercise (∼≥150 minutes/week; ∼≥10 MET hours/week) may offer an 8% (unadjusted) absolute reduction in ACM for breast cancer survivors compared with no/minimal exercise (∼<45 minutes/week; ∼<3 MET hours/week). Third, we identified and included studies not considered (46, 60, 66), as well as updated analyses of studies included (50, 54, 56), in prior reviews. This increased the sample size and precision of our estimates. It is also important to note that we also excluded several studies with methodological concerns for reasons detailed in Supplementary Table S4. Fourth, we included a wider range of recurrence- and mortality-related outcomes compared with prior reviews. In addition to recurrence, BCSM, and ACM, we also included cardiovascular disease–related mortality (66, 72), non–breast cancer mortality (60), and any-cancer mortality (72) in our analysis. Fifth, we summarized results on the variations of the relationship between exercise and recurrence/mortality considering a broad range of demographic, clinical, and contextual characteristics (84–93). Our findings indicate the need for individualized exercise prescriptions based on women’s menopausal status, stage of diagnosis, hormone receptor status, and post-diagnosis exercise. Finally, we also conducted a detailed quality appraisal using the Newcastle–Ottawa scale (Supplementary Table S7) and provided additional information on study characteristics and model fit statistics (Supplementary Tables S8 and S9).

Our review and meta-analyses offer a synthesis of the literature about long-term effects of post-diagnosis recreational aerobic exercise on recurrence and mortality among breast cancer survivors. However, our findings must be interpreted within the limitations of the studies included in our analysis. First, in clinical practice, absolute differences are considered clinically meaningful and perhaps more useful in advising patients (31). Though our findings indicate that exercise may offer considerable reductions in HRs associated with ACM, by as much as ∼60%, they should be interpreted with caution as a large reduction in HRs does not necessarily equate to a reduction in the absolute risk of a comparable magnitude. Further research to establish the effect of exercise on the adjusted risks of mortality is needed.

We found that 44.3% of more than 50,000 unique breast cancer survivors met exercise guidelines, which is considerably higher than the 37.7% reported previously based on analyses of two decades of National Health Interview Survey data (19). This difference may be indicative of healthy adherer bias within the mortality-related exercise literature concerning breast cancer survivors, whereby healthier women, who are more able and likely to engage in exercise, are recruited and retained for this research (94). Beyond the healthy adherer bias, our findings should also be interpreted with a degree of caution because of the use of self-report measures of exercise which are susceptible to social desirability and recall biases. Recall and social desirability biases may contribute to potential measurement error as individuals tend to overreport exercise when exercise is measured via self-report, which all of the studies identified did because of our focus on recreational exercise (95, 96). Measurement of exercise across different instruments is likely to differ and somewhat limit comparability. However, we attempted to mitigate this by only including studies that measured recreational aerobic exercise using instruments from which METs could be calculated. This meant that we excluded some studies that focused specifically on walking and/or running (97) and others that lacked details on measurement (98). It also meant that we excluded studies that measured muscle-strengthening exercise, which is worthy of separate investigation in future.

We were restricted to using published data rather than individual subject–level data for our analyses. Therefore, our study is based on information reported within articles or acquired by contacting authors. The thresholds used to determine whether women met exercise guidelines varied greatly across studies, which prevented perfect comparisons across studies. As interventions often lack follow-up to capture long-term events, exercise interventions were excluded. Other limitations include the inability to account for participation in non-recreational exercise or muscle-strengthening exercise, no consideration of differences in effects based on exercise intensity, and potential residual confounding, selection bias, and reverse causation. We included only sources published since 2003 and in English, though we checked previous reviews to identify additional sources. The results extracted from fully adjusted models were not accompanied by any goodness of fit statistics for us to evaluate the effects of unmeasured confounding, multicollinearity, or assumptions of proportional hazards. In future, researchers should consider using guidelines for reporting, such as STROBE (99), to improve transparency and help readers to interpret their findings.

We found less data for all events regarding survival rates and absolute differences compared with relative differences. A meta-analysis was possible for ACM but not BCSM or recurrence. Additionally, all survival results reported were based on unadjusted analyses. Given absolute differences are what are typically used in clinical practice, in the future, researchers should report adjusted and unadjusted 5- and 10-year survival rates to help the field generate a better sense of the association between exercise and survival. In pursuit of consistent comparisons, data from some studies were excluded. Thus, our findings are not representative of all available data that have been collected that could be used to examine the effect of exercise on recurrence and mortality. Finally, we focused solely on pre-diagnosis recreational aerobic exercise because it is more modifiable (27) compared with non-recreational or prediagnosis exercise. However, one study included in our analyses (51) showed that pre-diagnosis exercise may be beneficial for women with lower pre-diagnosis exercise levels.

The findings of this review highlight a number of directions for future research. The limited number of studies reporting variations in the relationship between exercise and recurrence/mortality restricted our analysis of findings to a narrative synthesis. In the future, whenever feasible, researchers should consider reporting the relationship between exercise and recurrence/mortality stratified by demographic, clinical, and contextual factors to support eventual quantitative analyses. In particular, physical, functional, psychologic, and cognitive impairments should be considered, as they have been overlooked as covariates when exploring the relationship between exercise and recurrence/survival among breast cancer survivors (100–103). We focused on pre-diagnosis recreational aerobic exercise. However, future researchers may want to consider different domains, types, and measures of physical activity. Occupation, transport, household physical activity, and muscle-strengthening activity may influence the relationship observed between recreational physical activity and recurrence/survival (34). Given the association of pre-diagnosis exercise with recurrence and survival (4, 5, 7, 9–11, 79), examining the effects of both pre-diagnosis and post-diagnosis exercise warrants consideration in future research (51). Devices such as smart watches may offer more objective and accurate measures of aerobic exercise, though the ability to practically measure muscle-strengthening exercise objectively does not exist.

Conclusions

In summary, survival benefits suggest that some exercise, even less than recommended by the guidelines, may lower the risk of mortality. Aerobic exercise equivalent to meeting guidelines for health, ∼2.5 hours/week, was associated with a ∼50% reduction in the HR for ACM, with further reductions up to ∼4.5 hours/week. A ∼25% reduction in the HR for ACM was associated with ∼1 hour/week. Further research to understand the relative benefits of exercise in relation to breast cancer recurrence, the absolute benefits of exercise in relation to recurrence and mortality, and how these benefits differ based on individual, clinical, and contextual (e.g., access to green spaces and neighborhood safety) characteristics. Conducting multiple larger prospective studies of women’s health of different populations while maintaining homogeneity within each study would help to fill this gap, as would greater consistency in the use of analytic approaches to generate results that are readily comparable across studies.

Supplementary Material

Tabls S1

Supplementary Table S1 reports the PRISMA Checklist.

Table S2

Supplementary Table S2 reports the full search strategies for each database.

Table S3

Supplementary Table S3 reports the study inclusion and exclusion criteria for this review/meta-analyses.

Figure S1

Supplementary Figure S1 reports the article identification process using PRISMA research framework

Table S4

Supplementary Table S4 reports the reasons particular studies that meet inclusion criteria were excluded from meta-analyses.

Table S5

Supplementary Table S5 reports the clinical and biological characteristics of participants in each included study.

Table S6

Supplementary Table S6 reports the race and/or ethnicity, education, and income of participants in each included study.

Sup Materials and Methods

Supplementary Materials and Methods reports details on the quality appraisal methods.

Table S7

Supplementary Table S7 reports summary results for the risk of bias assessment conducted using the Newcastle-Ottawa Scale.

Table S8

Supplementary Table S8 reports additional model details, including the outcomes analyzed, model fit statistics, and covariates included in models.

Table S9

Supplementary Table S9 reports additional study characteristics, including retention at exposure (i.e., exercise measurement), time from diagnosis to exposure, follow-up length, time points exposure measured.

Table S10

Supplementary Table S10 reports the aerobic exercise measures and guidelines used, and the proportion of participants meeting guidelines.

Figure S2

Supplementary Figure S2 shows forest plots for meta-analysis of the association between post-diagnosis aerobic exercise and all-cause mortality among breast cancer survivors.

Table S11

Supplementary Table S11 reports subgroup analysis regarding the effects of exercise on risk of all-cause mortality.

Figure S3

Supplementary Figure S3 shows a spline graph of the association between post-diagnosis aerobic exercise and breast cancer-specific mortality among breast cancer survivors.

Table S12

Supplementary Table S12 reports subgroup analysis regarding the effects of exercise on risk of breast cancer-specific mortality.

Figure S4

Supplementary Figure S4 shows forest plots for meta-analyses of the association between post-diagnosis aerobic exercise and breast cancer recurrence among breast cancer survivors.

Figure S5

Supplementary Figure S5 shows a spline graph of the association between post-diagnosis aerobic exercise and breast cancer recurrence among breast cancer survivors.

Table S13

Supplementary Table S13 summarizes findings regarding the effects of exercise on recurrence and/or breast cancer-specific mortality, and other mortality.

Table S14

Supplementary Table S14 reports the heterogeneity and small study effects for each meta-analysis.

Table S15

Supplementary Table S15 reports the sensitivity analyses for each meta-analysis

Authors’ Disclosures

No disclosures were reported.

Disclaimer

The contents reported and views expressed in this review are those of the authors and should not be taken to represent the views of the NIH. Opinions and comments expressed in this review belong to the authors and do not necessarily reflect those of the US government, the Department of Health and Human Services, the NIH, the National Institute Minority Health and Health Disparities, or the NCI. Funders had no role in the review design, the extraction, analyses, interpretation of the data, the writing of the review, and/or the decision to submit the review for publication.