Relationships between Obesity, Exercise Preferences, and Related Social Cognitive Theory Variables among Breast Cancer Survivors

Breast cancer survivors with obesity have an increased risk of cancer recurrence, second malignancy, and comorbidities. Though physical activity (PA) interventions are needed, investigation of the relationships between obesity and factors influencing PA program aspects among cancer survivors remain understudied. Thus, we conducted a cross-sectional study examining associations amongst baseline body mass index (BMI), PA program preferences, PA, cardiorespiratory fitness, and related social cognitive theory variables (self-efficacy, exercise barriers interference, social support, positive and negative outcome expectations) from a randomized controlled PA trial with 320 post-treatment breast cancer survivors. BMI was significantly correlated with exercise barriers interference (r = 0.131, p = 0.019). Higher BMI was significantly associated with preference to exercise at a facility (p = 0.038), lower cardiorespiratory fitness (p < 0.001), lower walking self-efficacy (p < 0.001), and higher negative outcome expectations (p = 0.024), independent of covariates (comorbidity score, Western Ontario and McMaster Universities osteoarthritis index score, income, race, education). Those with class I/II obesity reported a higher negative outcome expectations score compared with class III. Location, walking self-efficacy, barriers, negative outcome expectations, and fitness should be considered when designing future PA programs among breast cancer survivors with obesity.


Introduction
The prevalence of breast cancer survivors continues to grow as diagnosis, treatment, and control advance. However, quality of life after diagnosis among this population is affected by modifiable lifestyle behaviors, including physical activity (PA) engagement [1]. American Cancer Society Guidelines for PA recommend at least 150 min of exercise per week and 2 days of strength training per week for cancer survivors to help reduce risk of cancer recurrence, second malignancy, and comorbidities (e.g., obesity) [2]. Notably, a minority (13.9%) of breast cancer survivors meet PA guidelines and only 35% maintain a healthy BMI [3]. Moreover, adult cancer survivors experience a more rapid rise in obesity [4], which has been associated with low cardiorespiratory fitness, a strong predictor criteria: English speaking, between the ages of 18 and 70 years of age with a history of ductal carcinoma in situ (DCIS) or Stage I, II, or IIIA breast cancer and post-primary chemotherapy or radiation therapy, medically cleared for participation by their physician and underactive (participating in no more than 60 min of moderate intensity PA or no more than 30 min of vigorous intensity activity per week, on average, over the past 6 months).

Demographics and BMI
Self-reported participant demographics included age, race, ethnicity, years of education, annual household income, employment status, marital status, cancer stage at diagnosis, history of chemotherapy, history of radiation therapy, hormonal therapy type, functional comorbidity index score [19], and the Western Ontario and McMaster Universities osteoarthritis index (WOMAC) [20]. The functional comorbidity index score, or the number of comorbidities, was assessed by totaling the number of "yes" responses to 18 diagnoses with possible scores of 0-18, 0 indicating no comorbidities and 18 indicating the highest number of comorbidities [19]. The WOMAC, 24-item scale, assessed lower extremity joint pain (5 items), stiffness (2 items), and physical dysfunction (17 items) [20]. Scores from the subscales were summed with possible score ranging from 0 to 68, with higher scores indicating greater pain, stiffness, and physical dysfunction [20]. Weight and height were measured in person by trained research staff using a calibrated scale and stadiometer. Brand and model were study site specific (University of Alabama at Birmingham (Detecto Model 439); Southern Illinois University(Continental Health-O-Meter #400 DML medical scale); University of Illinois (Seca 763 Digital Column Scale)) [21]. BMI was calculated using the measured weight and height (weight (kg)/height (m 2 )) [21].

PA
Weekly minutes of moderate-plus-vigorous intensity PA were assessed with ActiGraph accelerometer (model: GT3X, Pensacola, FL, USA). Participants were instructed (orally and written) to wear the device for at least 10 waking hours for seven (primary trial) or 10 (COM-PARE) consecutive days [24]. The parameters used to validate the minimum wear time of 4 days was comprised of wear time ≥10 waking hours. The cut points that were used to establish moderate-to-vigorous activity intensity were: moderate (1952-5724 counts/min) and vigorous (5725+ counts/min) [21]. Minutes of vigorous intensity activity were doubled prior to adding minutes of moderate intensity activity and calculating weekly minutes of moderate-to-vigorous activity.

Cardiorespiratory Fitness
Following the American College of Sports Medicine guidelines for testing [25], cardiorespiratory fitness (relative VO 2 peak) was estimated with submaximal treadmill testing [26] in which speed and elevation were gradually increased until the participant achieved 85% of age-predicted maximal heart rate. Following the modified Naughton protocol, tests were begun at a slower speed and progressed at lower increments, as in past studies with individuals who are sedentary, older, fatigued, or have balance complications [21,25]. The oxygen cost of walking at the treadmill grade and speed achieved at 85% of predicted heart rate was estimated using published regression equations and is expressed in mL/kg/min [27].

Exercise Barriers Interference
Perceived barriers (or barriers interference), or how often recognized obstacles (i.e., lack of time, fear of injury, fatigue, lack of energy, lack of company, cost of exercising, lack of enjoyment, lack of equipment, family responsibilities, inconvenient exercise schedule, lack of interest, lack of knowledgeable exercise staff, feeling nauseated, no facilities/space, not a priority, procrastination, pain/discomfort, not in routine, lack of self-discipline, lack of skills, weather) interfered with exercise, was assessed using a 21-item, 5-point Likert scale (1 = rare to 5 = very often) measure, Exercise Barriers Interference, that has demonstrated reliability (α = 0.92) among breast cancer survivors [31]. Responses were summed for a total exercise barriers interference score [31][32][33] with a range of possible scores (21-105).

Social Support
Social support, or the perception of encouragement to engage in PA, from other sources (i.e., friends and family) [31] was measured via a 4-item (friends, 2-items, family, 2-items), 5-point Likert scale (0 = none to 4 = very often), Social Support for Physical Activity, with an internal consistency of 0.80 [31]. Responses were summed for a total social support score with possible scores ranging from 0 to 16 [34,35].

Outcome Expectations
Outcome expectations, or the anticipated positive and/or negative consequences of engaging in a behavior (e.g., exercise) was evaluated using a reliable (α = 0.79 and 0.70, respectively) 17-item (14 positive expectations and 3 negative expectations), 5-point Likert scale (1 = strongly disagree to 5 = strongly agree); responses were summed for positive and negative outcomes separately (i.e., higher score indicates greater perceived benefit (positive expectations) or greater perceived risk (negative expectations)) with possible scores ranging from 14 to 70 and 3 to 15, respectively [32].

Statistical Analyses
Descriptive analyses were conducted to summarize participant characteristics. BMI was analyzed as a continuous variable and as a 3-level categorical outcome (i.e., non-obese (BMI ≤ 29.9 kg/m 2 ), obese classes I/II (BMI = 30-39.9 kg/m 2 ), and obese class III (BMI ≥ 40 kg/m 2 )). A 3-level BMI was created to facilitate identifying potentially important differences at a higher BMI that may be missed when analyzing with a continuous BMI.
The associations with continuous BMI were analyzed using Pearson correlation coefficients (continuous correlates) and independent groups t-test (continuous correlates). The associations with 3-level categorical BMI were analyzed using one-way analysis of variance (ANOVA), followed by the Tukey post-hoc test (continuous correlates), and also the chi-square test (categorical or dichotomous correlates). Continuous study variables were examined for normality of distribution using box plots, stem-and-leaf plots, normal probability plots, and the Kolmogorov-Smirnov test. Variables that were determined as being non-normally distributed were also examined using the non-parametric Kruskall-Wallis test. Since the parametric were similar to the non-parametric results, we report the parametric results for ease of interpretation, and previous studies have reported parametric results obtained from these or similar variables. Follow-up multiple linear or logistic regression were performed based on the type of dependent variable.
No imputations were performed for missing data since the amount of missing data were very small (<1%); there were no missing data for most of the study variables. All analyses were performed using IBM SPSS Statistics software (SPSS) Version 28 (Armonk, NY, USA: IBM Corp), and p < 0.05 was deemed as statistically significant.
Exercise program preferences were dichotomized due to the small stratum-specific sample sizes with an emphasis on preferences more likely to alter program design (e.g., facility vs. other). Hence, each preference was reviewed based on the number of participants preferring each option within a specific preference question and then collapsed based on preference with greater potential to alter a program design (e.g., facility vs. other options not requiring a facility).
Accelerometer-measured PA was analyzed as dichotomous, categorical outcome (met PA recommendations: >150 min of moderate-to-vigorous PA vs. did not meet PA recommendations: <150 min of moderate-to-vigorous PA) based on the American Cancer Society guidelines for PA for cancer survivors and to simplify interpretation [2].

Covariates and Adjusted Analyses
The variables investigated as potential covariates included age, race, ethnicity, education, income, employment, marital status, cancer stage at diagnosis, months since diagnosis, history of chemotherapy, history of radiation, hormonal therapy, number of comorbidities, and WOMAC score. A variable was considered a covariate if it was statistically significantly associated with both BMI and one or more of the correlates of interest (PA preferences, PA, cardiorespiratory fitness, and SCT variables). The identified covariates were then used for adjusted multiple variable regression analyses performed as indicated, with linear regression analysis performed for continuous outcomes and logistic regression analysis performed for dichotomous outcomes. All regression coefficients were tested for statistical significance. The R 2 value was examined as goodness of fit measure.

BMI and Program Preferences
A statistically significant relationship between BMI and a dichotomized program preference was found for only one preference, regardless of whether BMI was analyzed as a continuous variable (p = 0.009) or as a 3-level outcome (p = 0.038). Participants who preferred to exercise at a facility had a higher BMI (versus lower) (32.8 ± 8.6 vs. 30.3 ± 6.6 respectively). No other program preferences were associated with the continuous or 3-level BMI outcomes.

BMI, Current PA, and Cardiorespiratory Fitness
There were no statistically significant differences or associations with levels of current PA (moderate-to-vigorous PA or meets recommendations) when BMI was analyzed as a 3-level categorical outcome (Tables 2 and 3). Similarly, no statistically significant association between PA and continuous BMI was noted (Table 4). However, there was a significant inverse correlation between BMI and cardiorespiratory fitness (r = −0.414, p < 0.001) in which higher BMI was related to lower cardiorespiratory fitness (Table 4). Results from a one-way ANOVA yielded statistically significant differences in cardiorespiratory fitness between levels of BMI ( Table 2). A Tukey post-hoc test indicated that there was a significant difference in cardiorespiratory fitness between all levels of BMI (all p < 0.001). Specifically, cardiorespiratory fitness was significantly lower among obese class III vs. non-obese and obese class I/II. Abbreviations: PA, physical activity; SD, standard deviation.

BMI and SCT Variables
Pearson correlation coefficients between continuous BMI and PA-related SCT variables are provided in Table 4. Higher BMI was significantly correlated with higher exercise barriers interference (r = 0.131, p = 0.019) and lower walking self-efficacy (r = −0.364, p < 0.001). No other SCT variables were significantly correlated with BMI.
The one-way ANOVA showed statistically significant differences in walking selfefficacy and negative outcome expectations as reflected in Table 2. A Tukey post-hoc test indicated that there was a significant difference in walking self-efficacy scores between all levels of BMI (all p < 0.001). Walking self-efficacy scores were significantly lower among obese class III vs. non-obese and obese class I/II. As for the negative outcome expectations, obese class I/II reported significantly higher negative outcome expectations than non-obese (p = 0.024). The remaining constructs did not show any statistically significant differences.

Adjusted Associations
Multiple variable linear regression analyses were performed to examine the independent relationships between BMI (dependent variable) and fitness, exercise barriers interference, walking self-efficacy, and identified covariates (comorbidity score, WOMAC, income, race, and education). The model was statistically significant (F(9, 306) = 15.875, p < 0.001, R 2 = 0.318) indicating that walking self-efficacy, cardiorespiratory fitness, comorbidity score, and race are independently associated with BMI (p < 0.05). Table 5 provides the analyses results.

Logistic Regression Models
Because preferring a facility was associated with BMI and walking self-efficacy, a post-hoc binomial logistic regression was performed to determine the relationship between location preference (dependent variable) and BMI independent of walking self-efficacy. The adjusted logistic regression model including BMI, walking self-efficacy, and location was statistically significant, χ 2 (2) = 10.292, p = 0.006. For every 1-unit increase in BMI, there is a 1.04 greater odds of preferring to exercise at a facility independent of walking self-efficacy score with a 95% confidence interval of (1.001-1.073).

Discussion
In this study, we sought to examine the associations between BMI and factors influencing program content and delivery (preferences), current PA (accelerometer-measured), cardiorespiratory fitness, and related SCT constructs (i.e., self-efficacy (barriers and task), exercise barriers interference, social support, outcome expectations (negative outcome expectations and positive outcome expectations)) among breast cancer survivors. We found that BMI was significantly correlated with exercise barriers interference. Further, we found significant associations between higher BMI and preference to exercise at a facility, lower cardiorespiratory fitness, and lower walking self-efficacy, independent of covariates (comorbidity score, Western Ontario and McMaster Universities osteoarthritis index score, income, race, education). These findings suggest that BMI influences PA program preferences, multiple psychosocial factors, and fitness which could be useful for informing future PA interventions for this population.
Overall, our sample of breast cancer survivors preferred to exercise at home or had no preference, which is consistent with the existing literature among diverse cancer survivors [36]. A recent systematic review reported a similar finding that adults with obesity prefer to engage in exercise "close to home", while Hussien et al. found that adults with severe obesity prefer exercising outdoors [37]. However, when assessing preference by BMI category, we found that as obesity increased (i.e., obese class I/II to class III), there was an increased preference for exercising at a facility. This is similar to findings from another study among rural breast cancer survivors with overweight and obesity [38]. Potential hypotheses are that survivors with obesity may appreciate the support/supervision of facility staff (e.g., personal trainers, health coaches, health club exercise specialist), facilityprovided equipment that limits weight bearing, and social context (i.e., seeing others exercise). Further, breast cancer survivors were recruited to participate in the current study, which required in-person visits to a facility.
The current study did not find a statistically significant association between BMI and current PA, similar to a past study among adults with obesity and multiple sclerosis [39]. Our sample consisted of breast cancer survivors who self-reported that they were not currently active (i.e., engaging in no more than 60 min of moderate intensity PA or no more than 30 min of vigorous intensity activity per week), as noted earlier in the inclusion criteria, which may have impacted our ability to test this relationship. However, a majority of the sample met PA recommendations based on free-living PA measured by accelerometer. We did find that overall, breast cancer survivors with and without obesity had low cardiorespiratory fitness levels compared to healthy women without cancer [40]. However, those with higher BMI had significantly lower cardiorespiratory fitness as in past studies among adults with obesity and no history of cancer [41,42]. Engaging cancer survivors, especially those with obesity, in cardiorespiratory fitness enhancing activities (i.e., PA, weight loss) is critical [43].
We found a statistically significant relationship between BMI and several SCT constructs. Specifically, breast cancer survivors with higher BMI had significantly lower walking self-efficacy and higher exercise barriers interference scores. Past studies have found associations between BMI and exercise barriers interference [44], exercise self-efficacy [45], and family social support [46] among adults with obesity, no cancer history [46], and presence of chronic illness (i.e., multiple sclerosis) [39]. The lower levels of self-reported walking self-efficacy among cancer survivors with higher BMI in the current study could possibly be due in part to limited mobility caused by excess body weight and reduced fitness levels, which makes it more difficult to walk at a moderately fast pace without stopping. With regard to exercise barriers interference, participants with increased weight reported more barriers to PA which has been previously described [15]. Similar to walking self-efficacy, bearing more body weight increases the likelihood of barriers (i.e., lack of energy, lack of confidence, lack of enjoyment, lack of interest, pain/discomfort).

Clinical and Research Implications
The findings from the current study have important implications for clinicians, healthcare professionals, and researchers who prescribe exercise and develop PA interventions for breast cancer survivors, especially with obesity. When prescribing exercise/PA for this vulnerable population, clinicians need to individually assess and consider medical (i.e., BMI), preference (i.e., location), SCT (i.e., walking self-efficacy), and fitness-related (i.e., cardiorespiratory fitness) factors as this could influence participation and the likelihood of meeting PA recommendations.
Although breast cancer survivors with obesity were more likely to prefer to engage in exercise at a facility (versus at home or no preference), potential barriers (e.g., financial, lack of accessibility) may arise for those without facility access. Therefore, distance-delivered interventions (e.g., mHealth, telephone, virtual) could serve as alternative options for delivering such programs. As for walking self-efficacy, exercise/PA programs can target and improve this component through incorporating individual behavior change techniques (e.g., goal setting, self-monitoring, and implementing solutions) [47,48]. The low levels of fitness found in the current study are more indicative of a greater need for PA and a potential intervention target. Thus, this suggests that structured exercise progression focused on improving cardiorespiratory fitness may be needed for this group. Lastly, although our study does not assess strength training, future interventions could leverage a preference for facility-based exercise to increase this type of exercise. In turn, individuals with obesity may be better able to do strength training at first, with a progression to greater aerobic exercise as muscle strength improves.
Taken as a whole, future research efforts need to be aimed at gaining a nuanced understanding regarding specific preferences (e.g., Why is facility more preferred by survivors with obesity?) through a qualitative approach. Moreover, future research directions should include assessing whether tailoring interventions and programs to individual health-related factors and preferences increases active living among this population.

Strengths and Limitations
The findings of the current study provide insight on a population that might need tailored or targeted PA programing. Specifically, our data related to location preference and health-related factors (i.e., walking self-efficacy and cardiorespiratory fitness) that can affect participation in programs can be useful for future intervention design. However, there are limitations. First, the entire sample consisted of post-treatment cancer survivors, mainly educated, affluent Caucasian women. Hence, findings may not be generalizable to the larger population of cancer survivors (e.g., those who are not White, have not undergone treatment, or are currently receiving treatment, less educated, less affluent, or male). Furthermore, although we used accelerometers to objectively measure minutes of PA, there are limitations (i.e., estimation of minutes of sedentary behaviors and PA). For example, it is unclear whether the device (waist-worn) accurately distinguishes the difference between sitting and standing idle, and upper body movement [49,50]. Moreover, accelerometers assess free-living PA rather than volitional, which results in a higher prevalence of meeting PA recommendations which are often based on leisure-time exercise alone. Lastly, our sample consisted of fewer participants with class III obesity (than with class I/II), which warrants future research with more representation from this group.

Conclusions
Cancer survivors are living longer due to advances in the diagnosis and control of cancer. Healthy lifestyle habits (physical activity, weight control) can enhance the quality of these years by reducing risk of cancer recurrence and other chronic diseases. Effective lifestyle programs are needed and should take into consideration the physical activity preferences and potential influences specific to this population, particularly among those with obesity who are most in need. Location, walking self-efficacy, and fitness are factors that should be considered for future PA programs among breast cancer survivors who have obesity. Tailoring for such individuals should involve a theoretically driven program which targets walking self-efficacy and involves activities appropriate for various levels of cardiorespiratory fitness. Institutional Review Board Statement: Institutional Review Board (IRB) approval was obtained at all three study sites and participants provided informed consent before initiating data collection.
Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.
Data Availability Statement: Not applicable.