Moderators of physical activity and quality of life response to a physical activity intervention for breast cancer survivors

Moderate-to-vigorous physical activity (MVPA) can improve the quality of life (QoL) for breast cancer survivors (BCS), yet, most do not achieve 150 + weekly minutes of MVPA. This study investigated moderators of response to a physical activity (PA) behavior change intervention for BCS. BCS (N = 222) were randomized to the 3-month intervention (BEAT Cancer) or usual care. Measurements occurred at baseline, post-intervention, and 3 months post-intervention. Measures included accelerometry, self-reported MVPA, and Functional Assessment of Cancer Therapy (FACT-General, FACT-Breast, physical well-being (PWB), social well-being (SWB), emotional well-being (EWB), functional well-being (FWB), additional concerns (AC), and Trial Outcome Index (TOI)). Adjusted linear mixed-model analyses indicated individuals ≤ 24 months post-diagnosis and who were single reported smaller increases in weekly self-reported MVPA than those > 24 months (44.07 vs 111.93) and partnered (− 16.24 vs. 49.16 min), all p < 0.05. As for QoL, participants < 12 months post-diagnosis who received chemotherapy experienced smaller improvements than those ≥ 12 months in FACT-General, FACT-Breast, PWB, and SWB scores. Survivors with a history of chemotherapy had smaller improvements in FACT-General, FACT-Breast, PWB, SWB, TOI, and AC scores, all p < 0.05. These findings indicate that being < 2 years post-diagnosis, single, and prior chemotherapy may limit MVPA and QOL responses to a PA intervention. Further studies are needed to determine if and/or what additional PA supports and resources these subgroups of BCS might find beneficial and effective. ClinicalTrials.gov number: NCT00929617.


Background
There is an increasing number of women who are living beyond their breast cancer diagnosis into survivorship, as the relative 5-year survival rate (all breast cancer stages combined) is 90% [1]. While women with breast cancer are living longer, they often have unmet quality of life (QoL) needs and face a range of symptoms and side effects, such as fatigue, decreased physical function, and cognitive impairment [2][3][4][5]. Strong evidence indicates that physical activity such as energy expenditure activities within daily life and exercise, a subset of physical activity which involves planned, structured and repetitive activity [6], is safe for breast cancer survivors and can enhance their physical wellbeing and overall QoL [7][8][9][10][11][12], yet most breast cancer survivors are sedentary and in need of lifestyle intervention [13,14]. A recent meta-analysis identified multiple interventions which successfully promoted physical activity in cancer survivors, yet the sample-weighted effect size was small (i.e., d + = 0.35) [15]. Identifying factors influencing breast cancer survivors' responses to physical activity behavior change interventions will be imperative to enhancing these effects and the expected downstream benefits related to health and well-being. Understanding for whom and/or under what conditions interventions are most effective can be used to better individualize and tailor future interventions for the specific needs of various breast cancer survivor subpopulations. Only eight physical activity and cancer intervention studies have reported factors moderating intervention response [16][17][18][19][20][21][22][23]. Most (six of the eight studies) identified moderators (i.e., age, marital status, education, body mass index (BMI), comorbidities, disease stage, history of chemotherapy, history of radiation, time since treatment), but results have been inconsistent and more research is needed in this area.
Thus, the current study involves secondary analyses for potential demographic, medical, and cancer-related moderators of treatment efficacy in a multicenter physical activity randomized controlled trial for breast cancer survivors, which previously reported significant improvements in physical activity and QoL following a 3-month multicomponent physical activity behavior change intervention (i.e., Better Exercise Adherence after Treatment for Cancer (BEAT Cancer)) [11].
Identifying cancer or treatment-related moderators associated with the observed increases in MVPA physical activity and QoL outcomes from the original BEAT cancer trial was the primary focus for this post hoc analyses of demographic and medical-related moderators. Analyses were exploratory and included moderators based on previous literature findings. Courneya et al. [17,18] were early to examine age, marital status, employment status, BMI, disease stage, and treatment history as moderating variables to their intervention effects among breast [17] and lymphoma [18] patients. Following those trials, other researchers further investigated the same moderators with the addition of education [20], co-morbidities [16,20], time since diagnosis [16], and baseline value of the outcome [21,23]. Additional moderators were selected for our analyses due to a need for further investigation acknowledged by previous studies which included annual income, race, rural residency, prior physician exercise counseling, joint function (WOMAC), and hormone therapy [12,17,18].
Given past studies demonstrating associations between physical activity and social support [24][25][26][27], socioeconomic status [25,28], and geographic region [25,28,29], our hypotheses were older breast cancer survivors who are married, educated, living in more urban areas, more than 12 months postdiagnosis, and did not have a history of chemotherapy would have a greater response to the BEAT Cancer intervention.

Study design
The current study involved secondary, post hoc analyses of potential moderators of treatment efficacy for the BEAT Cancer intervention. Data are taken from the original physical activity randomized controlled efficacy trial (N = 222 breast cancer survivors) between January 2010 and September 2013 [11]. Potential moderator variables (e.g., demographic, medical, and intervention outcome) were measured at baseline. Treatment response was determined by changes in physical activity and QoL from baseline (M0) to month 3 (M3), as measured by accelerometer monitoring, self-reported moderate-to-vigorous physical activity, and QoL questionnaires. The parent multicenter randomized controlled trial has been registered on clinicaltrials.gov (NCT00929617) and described in detail in a previous report [11,12,30].

Participants
Women ages 18-70 years old with a history of ductal carcinoma in situ or stage I-IIIA breast cancer, who self-reported a sedentary lifestyle (engaging in ≤ 30 min of vigorous or ≤ 60 min of moderate-intensity physical activity per week on average over the previous 6 months) were recruited into the trial. Eligible participants must have completed primary cancer treatment, be ≥ 8 weeks post-surgery, English speaking, and received medical clearance from their physician. Exclusion criteria included conditions that contraindicated physical activity or would interfere with assessments and current participation in another physical activity study. Additional details regarding inclusion and exclusion criteria were reported by Rogers et al. [30]. Institutional review board (IRB) approval was obtained, and all participants provided written informed consent.

Better exercise adherence after treatment for cancer (BEAT Cancer) intervention
The 3-month social cognitive theory-based BEAT Cancer intervention included 12 supervised individual exercise sessions tapered over 6 weeks, followed by independent homebased exercise sessions with three face-to-face exercise counseling sessions every 2 weeks with an exercise specialist. Additionally, during the first 8 weeks, six group sessions led by trained facilitators provided behavioral counseling (time management, stress management, behavioral modification strategies, etc.). The goal of the exercise prescription was to gradually progress participants to 150 weekly minutes of moderate-intensity physical activity, although participants were advanced to higher intensity if interested and able. An educational notebook and heart rate monitor were also provided to participants. A detailed description of the BEAT Cancer intervention was provided in prior publications [11,12,30].

Usual care (UC) intervention
The UC group received written materials from the American Cancer Society describing physical activity recommendations for cancer survivors (see Rogers et al. [30] for additional details regarding the UC group intervention).

Measures
Measurements were taken at baseline (M0), immediately post-intervention (M3), and 3 months after intervention completion (M6). Assessment and data entry staff were blinded to study group allocation.

Physical activity
Physical activity was measured by 7-day accelerometer monitoring (MTI/ActiGraph; model numbers GT1M and GT3X). Shortly into the data collection time period, an improved accelerometer model was released, and the study transitioned from using the ActiGraph GT1M to the newer GT3X model. To ensure measurement consistency, accelerometers were numbered and tracked so that participants were given the same accelerometer at each assessment, as the main interest was the difference in physical activity within participants over time. An exercise log and instructions to wear the motion sensor during waking hours for 7 consecutive days apart from bathing and swimming were provided. Participants were asked to perform their usual activities within 7 consecutive days. Activity counts were used to determine activity durations (minutes/day) in specific intensity levels with a focus on moderate (1952-5724 counts/min) and vigorous (5725 + counts/min)) [31,32]. Self-reported physical activity was obtained using a modified version of the Godin Leisure Time Exercise Questionnaire [33,34]. Self-reported activity duration was multiplied by frequency to determine minutes per week spent in each of the three activity levels: light, moderate, and vigorous. To calculate moderate-to-vigorous physical activity (MVPA) as suggested in the physical activity guidelines [35,36], vigorous physical activity minutes were doubled before adding to moderate minutes for accelerometer and self-report physical activity. Since accelerometers do not differentiate between different types of physical activity such as leisure, occupational, and structured exercise, self-reported leisure time physical activity was also obtained and analyzed separately from accelerometer measure to elucidate volitional activity and facilitate comparison with prior research [35,36].

Quality of life
Quality of life was measured using the Functional Assessment of Cancer Therapy-Breast (FACT-Breast) [37]. The 37-item FACT-Breast questionnaire includes the subscales of physical well-being, social well-being, emotional wellbeing, functional well-being, and additional concerns [37]. The sums of the Likert scale items were calculated for each subscale and for FACT-Breast (sum of all subscales combined), FACT-General (sum of all subscales except for additional concerns), and Trial Outcome Index (sum of physical well-being, functional well-being, and breast additional concerns subscales) [38]. Higher scores represent better QoL.

Moderators
All potential moderators were assessed at baseline. Potential moderator variables were identified from the literature and selected based on relevance and plausibility for the breast cancer population and included demographic (age [17,21,22], marital status [18,20], annual income, employment, education [20], race, and rural residence), medical (BMI [18,23], co-morbidities [18], breast cancer stage [17], history of chemotherapy [20,21], history of radiation [20], hormone therapy [39], time since diagnosis [16]), and intervention outcome (baseline value of the outcome [21], joint function (WOMAC) [21]) related factors. Demographic moderators (e.g., age, marital status, annual income, and rurality) were assessed with a demographics questionnaire; rurality was determined by using the United States Department of Agriculture Rural-Urban Continuum Codes (i.e., RUCC, analyzed as a continuous variable) [40]. A medical history questionnaire was used to assess cancer-related factors; time since diagnosis, treatment history, and inquire if any doctor had advised physical activity since their breast cancer diagnosis (yes, no, or I do not know as possible responses). Participants also completed the Comorbidity Index to quantify the number of comorbid conditions [41] and the Western Ontario and McMaster Universities Arthritis Index (WOMAC) to assess lower extremity pain (5 items), stiffness (2 items), and physical dysfunction (17 items) using a 5-point Likert scale (i.e., 1 = none to 5 = extreme) [42]. Finally, BMI was assessed via anthropometric measurements obtained from a scale and stadiometer (Continental Health-O-Meter #400 DML medical and Seca 763 Digital Column Scale). BMI was calculated from the height and weight (weight (kg)/height (m 2 )).

Statistical analysis
We used intent-to-treat and adjusted linear mixed model ANOVA with interaction terms analyses to assess the intervention effect on MVPA, and QoL, by each hypothesized moderator. The mixed models included terms for the group, time, and group-by-time interactions while also adding an additional interaction term for the specific moderator being tested in the model. Moderators were dichotomized and grouped as follows: age (< 55 vs. ≥ 55, cutpoint based on median age), marital status (married or living with significant other vs. other), annual income (< $50,000 vs. ≥ $50,000), employment (unemployed vs. employed), education (≤ high school degree or equivalent vs. > high school degree), race (African American vs. White/other), rurality (RUCC Code 1,3,6), BMI (< 30 vs. ≥ 30), co-morbidities (median Comorbidity Index score used as cutpoint), cancer stage (DCIS or stage I vs. stage II or III), history of chemotherapy (yes vs. no), history of radiation (yes vs. no), time on hormonal therapy (none vs. ≤ 12 months vs. > 12 months), months since diagnosis (≤ 12 vs. > 12; ≤ 24 vs. > 24), and WOMAC (median score used as cutpoint). Dichotomization was based on clinically relevant cutpoints, such as a BMI score of ≥ 30 as a way to categorize for obesity [43] or by using median scores to achieve an equal distribution for comparison, an approach similar to procedures reported previously in the literature [16-18, 20, 21, 23]. The change over time across the intervention-moderator interaction term was modeled. A separate model was run for each moderator for a total of 12 models conducted.
Except when testing moderator effects with interaction terms, all models were adjusted for baseline value of the dependent variable, study site, marital status, breast cancer stage, history of chemotherapy, history of radiation therapy, hormone therapy treatment, and comorbidities (rationale reported previously) [11]. Statistical analyses were performed using SAS® statistical software (Cary, NC). Statistical significance was defined as p < 0.05.

Participant characteristics
The sample consisted of primarily White (84%) women with a mean age of 54 ± 9 years and 16 ± 3 years of education. Cancer stages were as follows: DCIS (11%), stage I (42%), stage II (35%), and stage III (12%). The average time since cancer diagnosis was 54 ± 55 months, with 58% and 68% reporting a history of chemotherapy and radiation therapy, respectively. Approximately half (49%) of all participants reported current hormone therapy. There were no significant baseline group differences with the exception that a greater percentage of participants in the UC group had been on hormonal therapy for less than or equal to one year (p = 0.02) [11]. See prior reports for more details on demographic characteristics [11].

Moderators of physical activity (i.e., primary outcome in the original efficacy study)
Adjusted linear mixed-model analyses revealed that participation in moderate-to-vigorous physical activity (MVPA) in response to the BEAT Cancer intervention was moderated by several factors (Table 1. See below). Greater improvements in MVPA were seen for those who were > 24 months vs. ≤ 24 months post-diagnosis (111.93 vs. 44.07 weekly minutes of self-reported MVPA, p = 0.008) (Fig. 1). As shown in Fig. 2, participants who were married or partnered responded more favorably to the intervention, increasing their accelerometry measured MVPA by 49 min per week compared to single/divorced participants (49.16 vs. − 16.24 weekly minutes of accelerometer MVPA, p = 0.028). The following did not moderate accelerometer or self-reported MVPA response to the intervention: age, race, BMI, employment status, education, RUCC Code, prior physician physical activity counseling, co-morbidities, WOMAC, baseline value of the outcome, cancer stage, history of chemotherapy, radiation, and hormonal therapy.

Moderators of QoL response (i.e., secondary outcomes in the original efficacy study)
Statistically significant results and data are presented in Table 1 with all variables and data from the analyses provided in Supplementary Tables 2, 3, and 4. To summarize, participants > 12 months vs. ≤ 12 months from diagnosis reported larger increases in FACT-G, FACT-Breast, and FACT physical and social well-being sub-scales (Fig. 3. See below). In addition, improvements in the QoL Trial Outcome Index favored participants who were > 12 months postdiagnosis, and who had greater increases in the index score (Fig. 3). Presented in Fig. 4, breast cancer survivors reporting annual household income greater than $50,000/year also reported a more favorable FACT physical well-being sub-scale response to the intervention (Fig. 4. See below). Illustrated in Fig. 5, participants with a history of chemotherapy treatment reported less favorable QoL response to the intervention in several domains: FACT-G, FACT-Breast, physical well-being, social well-being, additional concerns pertaining to the breast and QoL Trial Outcome Index (Fig. 5. See below). Participants who had lower baseline physical well-being scores observed greater physical wellbeing improvements (Fig. 6. See below). Additionally, lower baseline QoL Trial Outcome Index scores moderated greater improvements following the intervention (Fig. 6). The following did not moderate QoL: age, race, BMI, marital status, employment status, education, RUCC Code, prior physician exercise counseling, co-morbidities, WOMAC, cancer stage, radiation, and hormonal therapy (Tables 2, 3, and 4. Supplementary files).

Discussion
This study demonstrates that cancer-related and demographic factors moderate physical activity and QoL outcomes in response to a physical activity behavior change intervention in breast cancer survivors. Participants who were further out from their cancer diagnosis and those who were married/ partnered experienced greater increases in MVPA. Also, improvements in various QoL domains favored participants who were further from diagnosis, reported higher income, had not received chemotherapy, reported lower physical well-being, and had a lower Trial Outcome Index at baseline. Of clinical relevance, the moderating effect of time since diagnosis (e.g., > 12 months post-diagnosis) exceeded the minimally important difference defined by Yost et al. [44] for FACT-General, FACT-Breast, physical well-being, social well-being, and QoL Trial Outcome Index [38]. Minimally important differences also were exceeded for the history of chemotherapy on FACT-General, FACT-Breast, and QoL Trial Outcome Index response. These results indicated breast cancer survivors who are single, have a lower income, are less than 1-year post-diagnosis, and have received chemotherapy may require physical activity interventions that are tailored specifically to the needs and barriers unique to those moderating factors to improve intervention response and ultimately increase physical activity levels and QoL. Moreover, several factors blunted but did not negate the intervention benefits (e.g., < 2 years post-diagnosis, prior chemotherapy) while other factors negated the intervention benefit (e.g., being single). These observations indicated that survivors who do not respond at all might need a completely different intervention, while those who do respond to some extent may need tailoring of the basic BEAT Cancer intervention approach, specific to the moderating factors found in this study.
Using data from the original BEAT Cancer physical activity behavior change intervention trial [11], the current study is able to conclude that while the BEAT Cancer intervention did improve physical activity levels and QoL, several factors blunted or negated the intervention effects (e.g., prior chemotherapy, more recent cancer diagnosis, being single, and lower household income). Notably, time since diagnosis was the only moderator that influenced both MVPA and QoL outcomes, a finding consistent with the known effects of physical activity on QoL and the parallel increase in both MVPA and QoL in the original efficacy trial [11]. We found being 24 months or more from diagnosis yielded greater improvements in weekly minutes of MVPA. Contrary to our results, Buffart et al. [16] found that the further from diagnosis when initiating a physical activity intervention, the less response on global QoL, and physical, emotional, cognitive, and social function subcategories. Similarly, breast cancer survivors who started supervised aerobic plus resistance exercise within 2 years after treatment experienced greater improvements in QoL than the delayed group initiating the exercise 3 months later [45]. Both studies initiated exercise training immediately post-treatment, while our initiation varied with regard to time since treatment completion, which could have also contributed to these inconsistent findings. Additionally, both studies were supervised efficacy trials with health outcomes as the primary outcomes, which differs from the present study in terms of theoretical models adopted, nature of the exercise and behavioral support interventions, and interpretation of trial results [46]. Two factors (i.e., time since diagnosis and marital status) were found to moderate weekly minutes of MVPA. To our knowledge this is the first study to find time since diagnosis as a moderator of MVPA response. While current literature has shown that marital status can influence QoL response in efficacy trials [18,21] to our knowledge no other behavior change studies have shown marital status moderating MVPA outcomes [46]. The BEAT Cancer behavior change intervention encouraged participants to find their own social support and, as such, having a partner/spouse could have made it easier to find someone to exercise with and motivate the participants. This finding could be supported by previous literature which found through social interactions and observations, participants were more likely to exercise because others around them were also exercising [26,47].
A history of chemotherapy moderated QoL response but not MVPA response, suggesting potential explanations for the differential QoL response other than lack of MVPA increase. This is supported by Kalter et al. [20] who reported that a history of chemotherapy did not moderate their exercise training outcomes unless it was combined with cognitive behavioral therapy. Also, Pinto et al. [21] found that those who received chemotherapy had better improvements in self-reported physical function with a physical activity behavior change intervention, possibly due to lower selfreported physical function baseline values than those who did not have a history of chemotherapy. Potential reasoning for there being a moderating effect on self-reported physical and functional well-being yet not on MVPA could be the perception of treatment side effects. Specifically, perception that chemotherapy will cause long term physical limitations may bias breast cancer survivors to perceive their physical function as lower [48]. A few studies have found perception of treatment can affect QoL, but these studies do not assess objectively measured physical function and fitness [48][49][50]. Further research using an objective measure of physical function is needed to better understand these potential relationships with history of chemotherapy. Moreover, Courneya et al. [17] found breast cancer patients receiving nontaxanebased chemotherapies experienced better improvements in muscular strength compared to those receiving taxane-based therapies. While the present study did not collect data on the type of chemotherapy received, specific chemotherapy agents may be a potential moderating factor due to higher toxicities associated with taxane-based chemotherapies that limit muscular strength and blunt the training effect, as well as, taxanes inhibiting the interaction of tublin with myosin  [17] and should be included in future relevant studies.
Annual income greater than $50,000 was associated with greater physical well-being response. Having a higher income may facilitate access to more QoL enhancing experiences, resources, etc. To our knowledge, no other physical activity and cancer study has analyzed income as a moderator, an understudied yet important factor to consider during intervention design. With regard to other moderators, our nonsignificant findings are also notable. By identifying moderators that were not significant, we can better prioritize future intervention targeting and research directions.
Our analyses showed that those who had lower baseline FACT-Breast subscale physical well-being and QoL Trial Outcome Index scores saw greater improvements postintervention. Conversely, Pinto et al. [21] also examined the moderator effect of baseline values and found mixed results, with some significant effects for those with lower baseline values and some with higher baseline values, depending on the measure. While the FACT-Breast survey was used in both studies, Pinto et al. [21] did not find a significant moderator effect of baseline values. Previous literature has proposed that targeting populations with the greatest need through physical activity interventions aimed to improve psychosocial outcomes will see the greatest impact [21,51]. This proposed phenomenon could be possible reasoning for the observed baseline value moderating effect.
Our study strengths include the randomized controlled study design, high retention rate, large sample size, and significant intervention efficacy. Moreover, to our knowledge, this study is the first in cancer survivors to investigate annual income, race, rural residency, prior physician exercise counseling, joint function (WOMAC) and hormone therapy as potential moderators.  corrections for multiple testing were not made due to the exploratory nature of the present study, thus may be limiting. Also, there was limited variability among the rurality of the sample, participants lived in RUCC Code 1, 3 and 6 [40]. Thus, the majority lived in counties in metro areas of fewer than 250,000 population [3] and counties in metro areas of 1 million population or more [1] and may have better access to resources than breast cancer survivors in a more rural or urban location. Over the course of the data collection period, accelerometer devices improved thus, transitioning to using a newer model which may have caused variability between outputs before and after the transition. However, previous studies have reported no significant differences between models at any of the activity levels [52,53] and participants were given the same accelerometer at each assessment, thus offsetting this potential concern. Finally, our findings may not be generalizable to cancer types other than breast or to resistance exercise interventions and the moderating effects of time since treatment and treatment type may change as standard of care for breast cancer treatment evolve over time.

Conclusions
In summary, our moderator analyses revealed QoL and physical activity responses to a physical activity behavior change intervention may be influenced by multiple demographic and cancer-related factors and provides meaningful clinical relevance to better support breast cancer survivors. Our intervention was most effective for helping breast cancer survivors who were married, greater than 12 months postdiagnosis, and did not receive chemotherapy. Additionally, our intervention was effective for breast cancer survivors regardless of age, race, BMI, employment status, education, rurality, prior physician exercise counseling, co-morbidities, WOMAC, cancer stage, history of radiation, and hormonal therapy. Physical activity behavior change interventions that target the needs and barriers of single, low-income breast cancer survivors and/or those who are within 1 to 2 years of diagnosis are warranted. Future research should consider using technology-based delivery [54] and/or tools, such as social media [55,56], to increase social support [57], as well as utilizing low-cost resources and/or home-based strategies to minimize expense, schedule conflicts, child/elderly care, and transportation needs [58]. Additionally, breast cancer survivors should be educated on the benefits of physical activity following their diagnosis [59] and learn strategies to overcome barriers and concerns regarding long-term impact of treatment on QoL (e.g., physical functioning) [8].
Our study also identifies areas for further research, such as understanding why chemotherapy moderated the intervention effects on QoL and if certain types of chemotherapies moderate these effects differently than others. Further, future studies should aim to recruit a large enough sample size to conduct multivariate analyses exploring more complex interrelationships among moderators. The moderating effect on the intervention response provides meaningful insight on subpopulations potentially requiring targeted physical activity behavior change interventions to effectively improve breast cancer survivors' lifestyle behaviors, QoL, and overall health.