A Co-Clinical Trial of Exercise Therapy in Breast Cancer Prevention

Abstract

Purpose.

We conducted a mouse–human co-clinical trial to evaluate the biological efficacy of exercise therapy in breast cancer prevention.

Materials and methods.

In a phase 1 randomized trial, 75 nonexercising women at high-risk of breast cancer were allocated to receive (1:1 ratio): usual care or one of three exercise therapy dose regimens: 75, 150, or 300 minutes/week for 24 consecutive weeks. Biological efficacy was evaluated by changes in breast epithelial cell proliferation (Ki67). Correlative proteomic analysis of paired tissue and plasma samples was also performed. A corresponding preclinical study tested the dose-response of exercise therapy on breast tumor latency.

Results.

Change in Ki67 was not different between groups (global p-value = 0.2). Among participants with paired Ki67 measures, the mean (s.d) change in Ki67 was: −1.26 (4.32) for 75 minutes/week, −1.74 (5.04) for 150 minutes/week, −0.45 (5.16) for 300 minutes/week, and 3.40 (5.53) for usual care (global p-value = 0.04). Only 150 minutes/week associated with significant reductions in Ki67 compared with usual care (Bonferroni-adjusted p-value 0.03). The “response rate” (reduction in Ki67) was 29% for usual care compared with 52% for 150 minutes/week. Proteomics revealed marked reduction in genes involved in epithelial mesenchymal transition in tissues of responding patients. In the preclinical study, only 150 minutes/week significantly delayed tumor latency compared with control (Benjamini- Hochberg-adjusted p-value 0.02).

Conclusion.

Exercise therapy is a promising strategy for early interception of breast cancer in high-risk women.

Trial Registration.

Clinicaltrials.gov identifier NCT02494869

Introduction

Observational data consistently shows self-reported exercise is associated with a significantly lower risk of breast cancer incidence in both pre-and post-menopausal women.^1–5 Nevertheless, despite guidelines from several agencies,^6,7 exercise therapy is not an established aspect of breast cancer prevention.^8,9

Evidence from definitive trials will be required for exercise therapy to be considered a legitimate breast cancer preventive strategy. Since definitive trials require considerable time and resources, it is essential to first optimize dose and design, and to provide preliminary evidence of biological activity via early phase trials.¹⁰ Controlled exercise therapy has been investigated in a limited number of early phase trials among women at modest risk of breast cancer, with all evaluating select plasma (host) biomarkers^11–15 or mammographic density.^16,17 No trial to date has evaluated whether exercise alters breast tissue biology directly using established surrogate biological efficacy end points for breast cancer incidence (e.g., tissue proliferative indices). Consequently, development of exercise therapy as breast cancer preventive strategy has been stymied by the lack of preliminary biological efficacy evidence and mechanistic understanding in humans.

We performed a mouse–human co-clinical trial of exercise therapy in breast cancer prevention. In a “proof-of-concept” human trial we evaluated the biological efficacy of three exercise therapy dose levels on normal breast cell epithelial proliferation (Ki67) via serial collection of tissue at baseline (pre-randomization) and postintervention. We focused on a population of women with features that substantially increase their risk of breast cancer thereby maximizing biomedical relevance. We also conducted a corresponding preclinical study, testing the dose-response of exercise therapy on tumor latency in a mouse model of breast cancer. Our primary objective was to identify the recommended phase 2 dose of exercise therapy. A secondary objective was to perform correlative analyses to shed light on the mechanistic basis for exercise therapy biological activity.

Materials and methods

Human trial

Participants and setting

Full methods are provided in the Supplement and study protocol. Briefly, we conducted a single-center, prospective randomized controlled trial among nonexercising women (i.e., <120 minutes of moderate or vigorous exercise per week) at high-risk of breast cancer under surveillance at Memorial Sloan Kettering Cancer Center. High-risk was defined as any one of the following: (1) family history of breast cancer with a lifetime risk of >20% as estimated by the IBIS calculator,¹⁸ (2) BRCA1/2 mutation carrier, or (3) prior diagnosis of atypical hyperplasia or lobular carcinoma in situ. Additional inclusion criteria were (1) aged 21–80 years old, and (2) if of child-bearing potential, must not be pregnant or plan on becoming pregnant during the study (women <50 years old of child-bearing potential were required to have a negative pregnancy test within 14 days of enrollment). An acceptable baseline (pre-randomization) cardiopulmonary exercise test¹⁹ with 12-lead electrocardiogram clearance was also an inclusion criterion. Women receiving any form of chemopreventive therapy (e.g., tamoxifen, raloxifene) within the previous 6 months, enrolled on an interventional investigational study, with bilateral breast implants, or a history of invasive breast cancer or other invasive cancer diagnosis or ductal carcinoma in situ, metastatic malignancy of any kind, any condition or intercurrent illness that, in the opinion of the investigator, made the individual a poor candidate, and any absolute contraindications to cardiopulmonary exercise testing as per the American Thoracic Society guidelines²⁰ were ineligible. The study was conducted in accordance with the Belmont Report, U.S. Common Rule and approved by the Memorial Sloan Kettering Cancer Center Institutional Review Board. All patients provided written informed consent prior to the initiation of any study procedures.

Randomization and blinding

Participants were randomly allocated in a 1:1:1:1 ratio to receive one of the three exercise therapy dose levels: (i) 75 minutes per week: ~25 minutes per treatment, 3 times per week, (ii) 150 minutes per week: ~50 minutes per treatment, 3 times per week, and (iii) 300 minutes per week: ~60 minutes per treatment, 5 times per week, or (iv) usual care for 24 consecutive weeks. Intra-patient dose escalation or de-escalation was not permitted. The random allocation sequence was generated and implemented using an institutional randomization module with a random permuted block design. Neither participants nor study exercise physiologists were blinded to group allocation.

Exercise therapy

Exercise therapy regimens were standardized between participants based on modality, dose intensity, progression, and schedule. Exercise therapy regimens were also matched in terms of monitoring (one-on-one monitoring/session) and length (24 weeks). Regimens therefore only differed based on total amount of exercise therapy per week (i.e., 75, 150, or 300 minutes per week). Exercise therapy consisted of a total of 72 (75 and 150 minutes per week dose) or 117 (300 minutes per week doses) supervised (i.e., one-on-one) treadmill walking sessions, administered across 3 to 5 sessions per week, depending on dose allocation, for 24 consecutive weeks. The planned exercise dose was scheduled using a nonlinear periodized approach such that physiological stress was continually altered and progressively increased across the entire intervention period.^21,22 Intensity was individualized to each patient based on workload (i.e., treadmill speed / grade) corresponding to specific ventilatory thresholds measured during the pre-randomization cardiopulmonary exercise test.²³ Dose modification of any session was permitted and performed by the exercise physiologist monitoring each session using standardized criteria.²³

Usual care

To optimize accrual and minimize attrition rates, usual care participants received a home-based, general physical activity program comprising an initial, in-person consultation with a study exercise physiologist outlining a structured aerobic walking program with a goal up to 150 minutes/week of moderate exercise consistent with national guidelines for breast cancer prevention.²⁴

Assessments

Normal breast epithelial cell proliferation.

Ki67 was evaluated in normal breast tissue samples acquired from two research ultrasound-guided core biopsies performed at baseline and postintervention (Week 25). Timing of biopsies were not controlled relative to menstrual cycle. The target region of interest was the upper quadrant of the unaffected contralateral breast; a marker was placed to pinpoint the biopsy location. As per international guidelines, we used a monoclonal for IHC staining of the proliferation-associated nuclear protein Ki67 in breast tissue epithelial cells (i.e., Mib-1 monoclonal antibody, 1:200 dilution; Dako, Glostrup, Denmark)^25,26 to determine the percentage of Ki67–positive cells among the total population of breast epithelial cells in tissue sections. Study pathologists (UB, IL) performed all Ki67 staining and quantification and were blinded to group allocation, assessment timepoint, and participant identification number. The same pathologist rescored the slides after at least a 2-week period and the average scores were used.

Participant physiology.

To evaluate whether investigated exercise therapy was delivered at high-fidelity, changes in cardiorespiratory fitness were assessed using a symptom-limited cardiopulmonary exercise test on an electronic motorized treadmill test with 12-lead electrocardiogram monitoring (Mac® 5000, GE Healthcare) according to standard procedures²⁰ and as described previously.²³ Body weight (kilograms) and body composition (percentage of lean and fat mass) was assessed via a dual-energy x-ray absorptiometry (Lunar DPX, General Electric). Both were evaluated at baseline and postintervention.

Compliance.

All modifications to the planned exercise therapy dose were recorded using standardized procedures. “Planned” dose of all sessions was quantified as metabolic equivalent task (MET)-hours per session. The “planned” intensity of each session was multiplied by the corresponding session duration to calculate MET per session; all sessions were summed to derive total “planned” cumulative MET-hours (MET-hrs) per patient.²⁷ “Completed” METs was quantified as the actual intensity and duration of each attended session. All sessions were summed to derive total “completed” cumulative MET-hours (MET-h) per patient. Relative exercise dose intensity was defined as the ratio of total “completed” to total “planned” cumulative dose.²⁷

Safety.

The type and prevalence of serious (e.g., life-threatening, hospitalization) and non-serious (e.g., knee, back pain) adverse events (AEs) during exercise therapy sessions were recorded using standardized procedures. AEs were reviewed by study exercise physiologists and graded and attributed at the end of each session.

Breast tissue and peripheral blood proteomic profiling.

Proteomics was conducted on breast tissue and matched plasma at baseline and postintervention using a Waters M-Class UPLC system connected to a SCIEX ZenoTOF 7600 system, as previously described.²⁸

Statistical analysis

The trial was a “proof-of-concept” study, neither the expected magnitude of change nor the clinically important change in Ki67 in response to exercise therapy in women at high-risk of breast cancer was available. Instead, data analyses were designed to provide distributional summary statistics (e.g., means, variation), as well as preliminary effect sizes to guide future sample size calculations for larger trials. As such, explicit power calculations were not performed for this trial. The per-protocol objective was to enroll 25 participants per dosing level and usual care for an overall accrual goal of 100 participants.

The primary endpoint was change in Ki67 from baseline to post-intervention for all participants with a baseline and/or post-intervention assessment by group allocation as per intention-to-treat. Linear mixed models with a random intercept for each participant and covariates of assessment time, study group, and the interaction between assessment time and study group were applied to evaluate changes in Ki67, cardiorespiratory fitness, and body composition (legs, trunk, and total fat and lean mass). In a secondary, unplanned analysis we evaluated log-fold changes in Ki67 only including participants with paired Ki67 assessment at baseline and post-intervention. Log-fold change in Ki67 was computed as the log (base two) of the ratio of follow-up Ki67 to baseline Ki67. Comparisons across study arms was assessed using a global test, with post-hoc pairwise comparisons of the 150 minutes / week and 300 minutes / week groups vs. usual care with Bonferroni adjustment. The percentage of participants on each study group with any decline in Ki67 (i.e., responder) is also reported. In an unplanned analysis, we visualized the relationship between mean minutes of completed exercise dose / week and change in Ki67 using a scatterplot with a loess curve. Analyses were conducted using R v4.1.2 (Vienna, Austria).

For proteomic analyses, protein fold changes from paired baseline samples were calculated, then log₂ transformed. Unpaired, two-sided t-tests were performed with statistical significance set at p-value < 0.05. For all comparisons, proteins were also rank ordered by log₂ fold change for gene set enrichment analysis (GSEA) using R package cluster Profiler v4.0.5 and the 50 Hallmark gene sets using false discovery rate (FDR) < 0.05 and visualized as networks using R package igraph v1.4.2.

Preclinical trial

Procedures

Female MMTV-PyMT, immunocompetent mice (~3 to 4 weeks of age, fed normal chow ad libitum), a genetically-engineered mouse model of spontaneous development of luminal-like breast cancer,^29–31 were randomly assigned (n=15–19/group) to the same exercise therapy doses (75, 150, 300 minutes per week) or sham control for a maximum of 24 weeks or until the first appearance of a palpable tumor, whichever came first. Exercise therapy comprised progressive treadmill running up to 20 m/min at 0% grade for 45 minutes, 5 days per week. All mice were evaluated (via palpation) for breast tumors three-times weekly. The Institutional Animal Care and Use Committee at Memorial Sloan Kettering Cancer Center approved all study procedures.

Statistical analysis

Kaplan Meier methods evaluated time to tumor appearance from study initiation with mice censored at the time of death if euthanized for reasons prior to tumor development. Mice alive without tumors at the end of the study were censored at Week 24. Median time to tumor development and 95% confidence intervals are reported. Time to tumor development for each exercise therapy group was compared to sham control using log-rank tests, with p-values adjusted using the Benjamini-Hochberg correction for multiple testing.³²

Data availability

The data that support the findings of this study are available upon request from the corresponding author. The data are not publicly available because of privacy restrictions. Code is available at https://codeocean.com/capsule/0444717/tree/v1

Results

Human clinical trial

Participants were assessed for eligibility between October 2016 and November 2019, with final post-treatment assessments completed in June 2020. Due to slow accrual the 75 minutes / week group was closed early; data from the 6 participants allocated to this group are presented for descriptive purposes only. Of note these participants had higher BMI and were all postmenopausal. Trial accrual was closed early due to the Covid-19 pandemic. A total of 75 participants were eligible and randomized, and are included in analyses (Figure 1). Final allocations were: 75 minutes per week (n=6), 150 minutes per week (n=24), 300 minutes per week (n=22), and usual care (n=23). Participant baseline characteristics are in Table 1. Representativeness of study participants is presented in Supplementary Table 1.

Figure 1.

CONSORT diagram

Table 1.

Patient baseline characteristics

Characteristic	All patients (n=75)	75 minutes / week (n=6)	150 minutes / week (n=24)	300 minutes / week (n=22)	Usual care (n=23)
Age (years) at consent
Mean (sd)	56 (10)	62 (8)	54 (10)	53 (11)	57 (10)
Min, max	33, 80	51, 69	38, 72	33, 80	38, 71
Body mass index (kg/m2)
Mean (sd)	28 (5)	32 (5)	27 (5)	28 (5)	28 (6)
Min, max	19, 39	26, 37	19, 37	22, 39	20, 38
Unknown	1	0	0	1	0
Race – no. (%)
Asian	2 (3)	0 (0)	0 (0)	1 (5)	1 (4)
Black	9 (12)	0 (0)	3 (12)	1 (4.5)	5 (22)
White	46 (61)	6 (100)	17 (71)	11 (50)	12 (52)
Unknown1	18 (24)	0 (0)	4 (17)	9 (41)	5 (22)
Ethnicity– no. (%)
Hispanic/Latino	1 (1.3)	1 (17)	0 (0)	0 (0)	0 (0)
Non-Hispanic/Non-Latino	46 (61)	4 (67)	16 (67)	10 (45)	16 (70)
Unknown	28 (37)	1 (17)	8 (33)	12 (55)	7 (30)
Menopausal status – no. (%)
Pre	23 (32)	0 (0)	8 (33)	8 (38)	7 (33)
Peri/post	49 (68)	6 (100)	16 (67)	13 (62)	14 (67)
Unknown	3	0	0	1	2
High-risk diagnosis – no. (%)
Family history	39 (52)	4 (67)	14 (58)	11 (50)	10 (43)
Atypical hyperplasia	22 (29)	2 (33)	6 (25)	5 (23)	9 (39)
Genetic predisposition	8 (11)	0 (0)	3 (12)	2 (9)	3 (13)
BRCA1 carrier	2 (3)	0 (0)	1 (4)	1 (4)	0 (0)
BRCA2 carrier	6 (8)	0 (0)	2 (8)	1 (4)	3 (13)
Lobular carcinoma in situ	1 (1)	0 (0)	0 (0)	1 (5)	0 (0)
Comorbidities – no. (%)
Type 2 diabetes	1 (1)	0 (0)	0 (0)	0 (0)	1 (4.3)
Hyperlipidemia1	13 (18)	3 (50)	2 (8.3)	2 (9)	6 (26)
Hypertension1	14 (19)	2 (33)	5 (21)	2 (9)	5 (22)
Comorbidity status unknown	1	0	0	1	0

¹Not known as participants opted to not provide this information and therefore not available via the electronic health record.

²Hypertension and hyperlipidemia were not defined by the investigative team but rather were pre-existing conditions listed on each participant’s electronic medical record. In all circumstances, participants were receiving medications for these conditions.

Drop-out, compliance (contamination), safety, and changes in patient physiology

A total of 57 (76%) participants completed the postintervention biopsy procedure: 6 (100%) for 75 minutes / week, 21 (88%) for 150 minutes / week, 16 (76%) for 300 minutes / week, and 14 (61%) for usual care. Mean compliance for each dosing level was: 81 ± 13% for 75 minutes / week; 88 ± 15% for 150 minutes / week; and 69 ± 30% for 300 minutes / week (Figure 2a). This corresponded with a difference in the completed exercise therapy dose (mean minutes per week over the treatment period) across group allocations (Figure 2b). The most frequently reasons for missed sessions were scheduling conflicts / lack of time and signs and symptoms (e.g., colds / influenza). At baseline, mean (SD) exercise in the usual care group was 4.1 (6.1) MET-h/week and 2.8 (6.2) MET-h/week at postintervention. Hence, there was minimal contamination (i.e., drop-in) in the usual care group.

Figure 2. Exercise therapy feasibility and biological efficacy in the human trial.

a, feasibility as assessed by mean relative exercise dose intensity (REDI), per participant per dose level. b, completed exercise therapy dose (mean minutes per week over the treatment period) across group allocations. c, log-fold change in breast epithelial cell proliferation (Ki67) from baseline to post-intervention per participant per group. d, log-fold change in breast epithelial cell proliferation (Ki67) from baseline to post-intervention per participant per group by menopausal status. e, percent of participants per group with any reduction in Ki67 from baseline to post-intervention. f, dose – response analysis evaluating completed exercise therapy dose (mean minutes per week) with change in Ki67.

No grade 3 AEs were observed in any participant at any dose level. Approximately two third of participants in exercise groups (33 / 52; 63.5%) experienced at least one non-serious AE, most commonly exercise-induced tachycardia and post-exercise hypertension (Supplementary Table 2). Compared with usual care, improvements in cardiorespiratory fitness were observed in the 150 and 300 minutes / week dose levels (both Bonferroni-adjusted p-values 0.01; Supplementary Figure 1). Cardiorespiratory fitness did not change in the usual care group (Supplementary Figure 1). No significant changes in body composition measures were observed in any group (Supplementary Table 3).

Tissue biological efficacy

In the primary intention-to-treat analysis (n=74), change in Ki67 was not significantly different between study groups (global p-value = 0.2). In the secondary paired Ki67 analysis (n=57), the mean (s.d) log-fold change in Ki67 was: −1.26 (4.32) for 75 minutes per week, −1.74 (5.04) for 150 minutes per week, −0.45 (5.16) for 300 minutes per week, and 3.40 (5.53) for usual care (global p-value = 0.04; Figure 2c). In post hoc pairwise comparisons (i.e., 150 minutes / week vs. usual care, and 300 minutes / week vs. usual care), only the comparison of 150 minutes / week was associated with significant reductions in Ki67 (Bonferroni-adjusted p-value 0.03). In a sensitivity analysis adjusting for menopausal status, point estimates for the change in Ki67 were similar to the primary analysis (Figure 2d), although the p-value was no longer statistically significant (p=0.085), likely a reflection of the small sample sizes and lack of power to formally test for this difference.

The response rate was 33% for 75 minutes per week; 52% for 150 minutes per week; and 38% for 300 minutes per week, compared with 29% for usual care (Figure 2e). Dose – response analysis of completed exercise therapy dose (mean minutes per week) with change in Ki67 revealed a non-linear relationship: exercise dose greater than ~75 minutes / week, in general, maintained, or decreased Ki67 up to a dose of ~120 minutes / week; doses beyond this threshold provided no additional benefit (Figure 2f).

Tissue and plasma proteomic profiling

Comparison of relative changes in protein abundance for all exercise groups combined versus usual care revealed a total of 66 differentially expressed proteins (p < 0.05, Figure. 3a) in breast tissue. Hallmark signatures by gene set enrichment analyses revealed exercise therapy positively enriched several pathways including heme metabolism, epithelial mesenchymal transition, downregulation of ultraviolet response, and coagulation (false discovery rate < 0.05; Figure. 3b). No individual protein consistently changed in response to exercise therapy in plasma.

Figure 3. Effects of exercise therapy on the breast tissue and plasma proteomic landscapes.

a, comparison of relative protein abundance for all exercise groups combined versus usual care revealed 66 DEPs in breast tissue (p < 0.05). Rows are DEPs; columns are patients. Color scale indicates log₂ fold change between post-treatment and baseline protein intensities. b, Network visualization of Hallmark signatures and member proteins enriched by GSEA comparing relative protein abundance for all exercise groups combined versus usual care in breast tissue (FDR < 0.05). c, comparison of relative protein abundance between Ki67 responders and non-responders to exercise therapy revealed 39 DEPs in breast tissue (p < 0.05). Rows are DEPs; columns are patients. Color scale indicates log2 fold change between post-treatment and baseline protein intensities. d, Network visualization of Hallmark signatures and member proteins enriched by GSEA comparing relative protein abundance between Ki67 responders and non-responders to exercise therapy in breast tissue (FDR < 0.05). e, comparison of relative protein abundance between Ki67 responders and non-responders to exercise therapy revealed 18 DEPs in plasma (p < 0.05). Rows are DEPs; columns are patients. Color scale indicates log2 fold change between post-treatment and baseline protein intensities. f, Network visualization of Hallmark signatures and member proteins enriched by GSEA comparing relative protein abundance between Ki67 responders and non-responders to exercise therapy in plasma (FDR < 0.05).

Comparison of protein expression in participants that responded to exercise therapy (any decrease in Ki67) compared with non-responders (no change or increase in Ki67) revealed 39 differentially expressed proteins in breast tissue (p < 0.05; Figure 3c), with no individual protein consistently changing after adjustment for multiple testing. Subsequent Hallmark signatures by gene set enrichment analyses revealed positive enrichment of heme metabolism and negative enrichment of epithelial mesenchymal transition in responders (FDR < 0.05; Figure 3d). In plasma, 18 differentially expressed proteins were identified (p < 0.05; Figure 3e) with gene set enrichment analyses revealing negative enrichment of mammalian target of rapamycin complex 1 signaling, hypoxia, myogenesis, and apical junction in responders (Figure 3f).

Preclinical study

Sham control tumors were detected at a median of 24 days (95% confidence interval: 24–31). In the exercise therapy doses, tumor appearance was 6 to 7 days later (median 30–31 days; log-rank global p-value = 0.03). In post hoc pairwise comparisons, only 150 minutes per week delayed tumor latency compared with sham control (Bonferroni-adjusted p-value 0.02; Figures. 4a–c).

Figure 4. Effects of exercise therapy versus sham control on breast tumor latency in the preclinical trial.

a, time to tumor appearance for 75 minutes per week versus sham control. Median (95% confidence interval) days to first tumor appearance was 24 days (24, 31 days) for sham control (red line) and 31 days (24 days, upper confidence limit not estimable) for 75 minutes per week (blue line) (Benjamini-Hochberg adjusted p=0.14). b, time to tumor appearance for 150 minutes per week versus sham control. Median (95% CI) days to first tumor appearance was 24 days (24, 31 days) for sham control (dark blue line) and 31 days (28, 38 days) for 150 minutes per week (light blue line) (Benjamini-Hochberg adjusted p=0.02). c, time to tumor appearance for 300 minutes per week versus sham control. Median (95% CI) days to first tumor appearance was 24 days (IQR, 24, 31 days) for sham control (dark blue line) and 30 days (24, 35 days) for 300 minutes per week (light blue line)(Benjamini-Hochberg adjusted p=0.09). Shaded area represents the 95% CI.

Discussion

We conducted a mouse-human co-clinical study in which integration of research paired tissue biopsies into the high-risk breast cancer prevention setting permitted testing of the dose-response effect of exercise therapy on biological activity directly in the target organ together with causal experimental testing in a corresponding preclinical study. These data and paired correlative analysis of serially collected normal breast tissue and peripheral blood samples define the biological efficacy and begin to elucidate the mechanistic basis of exercise therapy as a non-pharmacological strategy to suppress or prevent breast cancer in the high-risk setting.

Here, we determined that six months of aerobic exercise therapy showed promising biological efficacy, but in a non-linear (curvilinear) manner. Effects of 150 minutes / week showed promising antitumor efficacy however further doubling to 300 minutes / week, at a cohort level, provided no additional benefit, at least in terms of Ki67 reductions in humans as well as increased tumor latency in the selected preclinical model. Even if efficacy was similar, 150 minutes / week would be indicated since it requires 2.5 fewer hours (per week), therefore enhancing feasibility and testing at scale. In support, dose-response analysis of the human trial data revealed completed doses less than 150 minutes / week, specifically a “therapeutic range” of ~75 to ~120 minutes / week, conferred biological efficacy. An initial prescription of 150 minutes / week will likely be required to account for anticipated compliance rates of ~80–90% as observed in this trial. It is important when interpreting our findings to recognize the variation in Ki67 between participants was high in part because we did not control for menopausal status and on phase of the menstrual cycle. These are important limitations of our study design. Future trials can control these critical issues by only enrolling pre- or postmenopausal women, performing biopsies at the same phase of the menstrual cycle, and using the pre-post difference in each participant as the primary endpoint.

Our study provides an example of the application of mouse-human co-clinical trials in the exercise context. Such trials have been used in oncology drug development,^33,34 but application to translational exercise-oncology research is novel. Co-clinical trials can provide value in predicting the outcome of clinical trials and can rapidly generate new clinically relevant hypotheses that can affect how the concurrent human clinical trial is conducted and analyzed.³⁴ Causal demonstration of the antitumor efficacy of exercise therapy in a clinically-relevant mouse model of breast cancer prevention alongside correlative evidence from the concurrent human trial strengthens study conclusions as well as helps guide design of future studies. Nevertheless, it is important to highlight there were several discrepancies in protocols between our human and mouse trials. These included, but not limited to, differences in age (and therefore menopausal status). Specifically, age range was restricted in the mouse study to primarily study the effects of exercise in the pre-menopausal setting. This also reflected our expectation that most women recruited to the clinical trial would also be pre-menopausal, but we included a wide age-range to enhance generalizability. However, most participants in the human trial were postmenopausal resulting in an important discrepancy between the mouse and human protocols. Other protocol differences included body composition, interspecies scaling of exercise doses and length of exercise exposure between mice to humans as well as different stress responses to exercise modality between species. Finally, the lack of analysis of proliferation in the mouse mammary glands would have permitted direct comparison with those performed in the human trial. Thus, significant caution is needed in comparison of the mouse and human data and findings considered speculative at present; future co-clinical trials that more effectively synchronize protocols across mouse and humans are needed.

How exercise therapy regulates normal breast tissue molecular landscape is not known. Proteomic analysis revealed marked negative enrichment of epithelial mesenchymal transition in responding participants. Epithelial mesenchymal transition is a transient, reversible process of cell de-differentiation established to contribute to breast cancer hallmarks including tumor invasion, metastasis, and chemoresistance.^35,36 We speculate that exercise therapy, via reprogramming of the systemic milieu, changes the signal inputs for distal organs such as the breast altering microenvironmental pressures,³⁷ leading to regulation of numerous biological processes including the epithelial mesenchymal transition repertoire to promote antitumor phenotypes. Additional preclinical and translational efforts to comprehensively interrogate the exercise therapy – epithelial mesenchymal transition link in breast and other distal organs is warranted.

Our study has important limitations. First, our results are limited to inactive women at high-risk of breast cancer and do not generalize to those with higher activity levels or at lower risk of breast cancer in which exercise therapy feasibility (compliance) and biological efficacy may be distinct. Second, we tested highly controlled exercise therapy of one modality with individualized supervision; feasibility, safety, and efficacy could differ under other conditions and with different exercise modalities (e.g., resistance training). Furthermore, initiation of exercise in humans and mice may impact dietary intake as well as dietary composition (humans) but this potential important confounder was not monitored in either species. Our preclinical study only tested one mouse model of breast cancer – effects may be distinct in different breast cancer subtypes. Third, the human trial had slow accrual. Reasons for non-consent were mainly related to lack of interest due to inconvenience (e.g., travel distance, cost) of facility‐based supervised exercise sessions (Manhattan, Upper East Side) required during the first two years of study conduct. The introduction of a “site‐less” telemedicine solution in the third year of trial conduct partially addressed this major issue as it permitted exercise sessions to be conducted remotely in participants home. Use of decentralized solutions may help address major barriers to the rigorous conduct of exercise therapy investigations in clinical populations.³⁸ Finally, due to the small sample size, the statistical power for certain comparisons was limited.

In conclusion, this mouse-human co-clinical study showed relatively short-term aerobic exercise therapy has promising biological efficacy, justifying further clinical development with 150 minutes / week as the recommended dose. These data may help guide further development of exercise therapy as a candidate breast cancer preventive strategy and understanding the mechanism of action to guide rational exercise-drug combination approaches.

Translational relevance.

We conducted a mouse-human co-clinical study in which integration of research paired tissue biopsies into the high-risk breast cancer prevention setting permitted testing of the dose-response effect of controlled exercise therapy on biological activity directly in the target organ together with synchronous causal experimental testing in a preclinical study. These data and paired correlative analysis of serially collected normal breast tissue and peripheral blood samples define the biological efficacy and begin to elucidate the mechanistic basis of a non-pharmacological strategy in women at high-risk of breast cancer.

Data sharing statement.

Patient data is not publicly available due to privacy restrictions.