Disentangling Efficacy and Expectations: A Prospective, Cross-lagged Panel Study of Cancer Survivors’ Physical Activity

Disentangling Efficacy and Expectations: A Prospective, Cross-lagged Panel Study of Cancer... Abstract Background Despite demonstrated utility of Bandura’s social cognitive theory for increasing physical activity among cancer survivors, the validity of the originally hypothesized relationships among self-efficacy, outcome expectations, and physical activity behavior continues to be debated. Purpose To explore the temporal ordering of outcome expectations and self-efficacy as they relate to moderate-to-vigorous physical activity behavior. Methods Longitudinal data from cancer survivors (N = 1,009) recently completing treatment were used to fit six cross-lagged panel models, including one parent model, one model representing originally hypothesized variable relationships, and four alternative models. All models contained covariates and used full information maximum likelihood and weighted least squares mean and variance adjusted estimation. Tests of equal fit between the parent model and alternative models were conducted. Results The model depicting Bandura’s originally hypothesized relationships showed no statistically significant relationship between outcome expectations and physical activity (p = .18), and was a worse fit to the data, compared with the parent model [Χ2 (1) = 5.92, p = .01]. An alternative model showed evidence of a reciprocal relationship between self-efficacy and outcome expectations, and was statistically equivalent to the parent model [Χ2(1) = 2.01, p = .16]. Conclusions This study provides evidence against Bandura’s theoretical assertions that (a) self-efficacy causes outcome expectations and not vice versa, and (b) outcome expectation has a direct effect on physical activity. Replication within population subgroups and for other health behaviors will determine whether the social cognitive theory needs modification. Future trials should test whether differential construct ordering results in clinically meaningful differences in physical activity behavior change. Physical activity, Neoplasm, Cancer survivor, Social cognitive theory Introduction Physical activity positively benefits the health of cancer survivors after treatment completion by lowering risk of cancer recurrence and cancer-specific mortality [1–3], and improving cancer-related biomarkers, such as insulin-like growth factor 1 [4]. In addition, studies have demonstrated that physical activity reduces fatigue [5–9] and improves quality of life [10–13] among post-treatment survivors. Yet, 66 per cent of cancer survivors fail to regularly meet physical activity recommendations [14] resulting in the need for efficacious, individually tailored, evidence-based behavioral interventions. Foundational to creating highly efficacious behavioral interventions are health behavior theories, which explicitly describe the mechanisms of physical activity behavior change. One such theory, the social cognitive theory [15], has demonstrated utility for explaining, and intervening to increase, physical activity among cancer survivors [16, 17]. Most research focuses on two social cognitive theory constructs: self-efficacy and outcome expectations. Specifically, self-efficacy is defined as perceived ability to perform a behavior [15, 18]. For example, a survivor may be asked how confident they are that they can “walk briskly for 20 min without stopping” or “exercise for 20 min at a level hard enough to cause a large increase in heart rate and breathing” [19]. Outcome expectations are defined as belief that a behavior will lead to certain outcomes [15]. For example, many cancer survivors believe that participating in physical activity will increase energy, maintain a healthy weight, and prevent cancer from coming back [20–22]. Self-efficacy is one of the strongest correlates and longitudinal predictors of successful adoption and maintenance of physical activity in the general population [17], as well as among cancer survivors [16]. Per Bandura’s original hypothesis, self-efficacy influences physical activity behavior directly, as well as indirectly through outcome expectations. Bandura furthermore asserts that the relationship between self-efficacy and outcome expectations is unidirectional such that self-efficacy causes outcome expectations, but outcome expectations do not cause self-efficacy [18, 23–26]. Research studies examining the social cognitive theory for physical activity have found strong empirical support for a positive relationship between self-efficacy and positive outcome expectations [27–31], as well as a positive relationship between self-efficacy and physical activity [17, 27, 29, 30]. The evidence for a direct effect of outcome expectations on physical activity, however, is equivocal [16, 17, 32]. Much of this research has relied on cross-sectional data, randomized trials with low power, or failed to conduct longitudinal mediation analyses. More recently, a properly powered randomized control trial found that a social cognitive theory intervention for breast cancer survivors was successful in significantly raising physical activity and self-efficacy, and significantly lowering negative outcome expectations. The intervention, however, had no effect on positive outcome expectations and provided no evidence that social cognitive theory constructs mediated the intervention’s effect on objectively measured physical activity [33, 34], as hypothesized by Bandura’s original model. Thus, further research examining social cognitive theory construct relationships with one another, and with physical activity over time among cancer survivors, is warranted. One important element of this research involves determining the exact ordering by which self-efficacy and outcome expectations influence physical activity behavior, a topic of continued debate within the theory-testing literature [35]. Contrary to Bandura’s original assertion, evidence exists which suggests that outcome expectations cause self-efficacy [35], and as noted above, evidence of a direct relationship between outcome expectations and physical activity is equivocal [17, 32]. Furthermore, when reviewing the literature examining these construct relationships among cancer survivors, there are methodological limitations prohibiting a firm conclusion for construct ordering for physical activity interventions, such as reliance on cross-sectional designs, low sample sizes, heterogeneity in measurement of physical activity, and social cognitive theory constructs, and relatively few studies focused on cancer survivorship [16, 17, 29, 36, 37]. Salient to this study is a general lack of consideration of alternative models in the literature to challenge the originally hypothesized ordering of constructs. Yet, due to the increasing demand for highly efficacious, theory-driven physical activity interventions delivered via automated technology (e.g., phone, website, and health apps) that increasingly rely on mathematical algorithms for real-time, individually tailored intervention delivery, there is growing need for research to determine the ordering of theoretical constructs, such as outcome expectations and self-efficacy. Clarification in this area can optimize intervention efficacy. Therefore, the purpose of this study is to explore the temporal ordering of outcome expectations and self-efficacy as they relate to moderate-to-vigorous physical activity behavior. Method This was a secondary data analysis of data obtained from the American Cancer Society’s Cancer Survivor Transition Study, a prospective observational study of breast, colorectal, and prostate cancer survivors enrolled within 12 months of completing active treatment. Institutional review board approval (Morehouse School of Medicine #253587) was obtained prior to data collection, and details of this dataset’s survey design, recruitment procedures, and methods were published in an earlier paper [38]. Briefly, participants were randomly selected from the American Cancer Society constituent database, which contains information of individuals who have initiated contact (80% self-referred, 20% provider referral) with the American Cancer Society’s National Cancer Information Center. Inclusion criteria consisted of a diagnosis of breast, colorectal, or prostate cancer; being at least 18 years of age; completing curative treatment < 13 months prior to taking baseline survey; and ability to read and communicate in English. This method yielded a final baseline sample of 1,188 breast, colorectal, and prostate cancer survivors. The follow-up survey was conducted on average 13.3 months after the baseline survey, with 869 individuals completing the follow-up survey (retention rate = 73%). For this study, participants were not eligible if they reported physical disability or a cancer recurrence, metastasis, or multiple cancer diagnoses at baseline or follow-up, as this may interfere with their ability to be physically active (Fig. 1). Final sample size for this study is 1,009 (84.9% of baseline Transition Study sample), with 679 (67.3%) individuals having complete data for all measures at both baseline and follow-up. Fig. 1. View largeDownload slide Flow diagram of participants included in analysis. Fig. 1. View largeDownload slide Flow diagram of participants included in analysis. Measures Sociodemographic factors Participants’ self-reported sociodemographic variables included age in years, gender, education (some college or more vs. high school graduate or less), and marital status (married/marriage like relationship vs. single/separated/divorced/widowed). Cancer-related factors A four-level variable combining cancer type and gender was created to prevent perfect overlap between covariates in multivariable models (female breast, female colorectal, male colorectal, and male prostate). Cancer stage was assessed with two questions asking participants if a doctor had ever told them their cancer spread to their lymph nodes and/or elsewhere in their body (coded local vs. regional or distant). Time since cancer treatment completion was calculated as the number of months since completion of most recent radiation, chemotherapy or surgery. Body mass index Self-reported height and weight were used to compute current body mass index (kg/m2), which was then used to create four body mass index categories using standard cut-off values for underweight (<18.5), normal weight (≥18.5 to <25.0), overweight (≥25.0 to <30), and obese (≥30). Due to a small number of underweight individuals in the analytic sample (n = 12), underweight and normal weight individuals were combined into a three-category variable. Physical comorbid conditions Participants’ number of physical comorbid conditions was assessed using the Older American Resources and Services Comorbidity Scale [39]. The scale assesses the presence of 11 conditions, including arthritis/rheumatism, glaucoma, emphysema/chronic bronchitis, high blood pressure, heart disease, circulation trouble in arms or legs, diabetes, stomach or intestinal problems, osteoporosis, chronic liver or kidney disease, and stroke, and sums them to create a comorbidities score ranging from 0 to 11. Cancer-related symptoms Total number of cancer-related symptoms experienced was calculated using a modified Memorial Symptoms Assessment Scale [38, 40]. A total score, ranging from 0 to 12, was created by summing any occurrence of 12 individual symptoms over the past month: fatigue, difficulty concentrating, difficulty sleeping, feeling nervous or worrying, problems with sexual interest or activity, shortness of breath, feeling sad, physical pain, incontinence, diarrhea/irritable bowel syndrome/problems holding bowels, lymphedema, and numbness or tingling. Self-efficacy Self-efficacy for performing regular physical activity was evaluated using a single item from a nine-item scale developed to assess confidence in performing key behaviors during the transition from active treatment. Participants were asked “How confident are you that you can currently exercise regularly (at least 3 to 4 times per week)?” with response options ranging from 1 (“not at all”) to 5 (“extremely”). The scale was adapted from the Stanford Chronic Disease Self-Efficacy Scale [41] with item modifications and additions informed by results from focus groups with cancer survivors. Outcome expectation Outcome expectations for physical activity were assessed using three items that asked participants how much they currently agree or disagree with the following statements: “exercising regularly (3–4 times per week) can help...” (i) “reduce my chances of getting cancer again”, (ii) “me return to the lifestyle and activities I enjoyed before my cancer treatment,” and (iii) “reduce physical and emotional side effects like fatigue, pain, anxiety, sadness, etc.” Items mirrored those used in other studies assessing this construct in related populations [42, 43]. Response options ranged from strongly disagree [1] to strongly agree [5]. Inter-item reliabilities were good for this study (Cronbach α = 0.83). Confirmatory factor analysis did not meet recommended fit indices at baseline [Χ2 (1) = 290.35, p < .001; CFI = 0.95; TLI = 0.86; RMSEA = 0.54 (95% CI: 0.49, 0.59)] or follow-up [Χ2 (1) = 289.31, p < .001; CFI = 0.94; TLI = 0.81; RMSEA = 0.65 (95% CI: 0.59, 0.72)], but invariance over time was evident [Χ2 (5) = 5.11, p = .40; CFI = 1.00; TLI = 1.00; RMSEA = 0.01 (95% CI: 0.00, 0.04)]. Therefore, a mean score was calculated for those respondents who answered at least two of the three items. Physical activity Moderate-to-vigorous physical activity was assessed using a 15-item scale from the American Cancer Society’s Cancer Prevention Study-3, an epidemiological cohort study surveying cancer risk behaviors of over 300,000 participants [44, 45]. Participants were asked to estimate hours per week spent performing a variety of common physical activities. Response options included “None,” “Less than 1,” “1 to 2,” “3,” “4 to 6,” and “7 or more.” Hour values of 0, 0.5,1.5, 3.0, 5.0, and 7.0, respectively, were multiplied by metabolic equivalent of task values from the 2011 Compendium of Physical Activities [46] and summed to create a metabolic equivalent hours per week score. At baseline, activities included walking, jogging, running, bicycling, tennis/racquetball, lap swimming, aerobics class (e.g., step and kickboxing), aerobic machines (e.g., elliptical and rowing), sports activities, dancing, and other aerobic recreation (e.g., hiking) [45]. At follow-up, some activities were assessed with more detail. Specifically, bicycling was separated into biking for leisure or bicycling vigorously, tennis was separated into singles and doubles, and sports participation was separated into moderate or vigorous sports. Statistical Analysis Descriptive analysis Means, standard deviations, or frequencies were computed, as appropriate, for each study variable and covariate. Descriptive statistics and correlations were examined for evidence of non-normality and multicollinearity. Path analysis All models tested were based on theoretical premise and were created specifically to test the hypothesized relationships among self-efficacy, outcome expectations, and moderate-to-vigorous physical activity. For all path models, non-normality in moderate-to-vigorous physical activity was addressed by using a square root transformation. Covariates included age, comorbidity count, symptom count, body mass index category, marital/cohabitation status, gender by cancer type, cancer stage, and months since treatment. Because self-efficacy was measured with a single-item, it was modeled as an ordinal variable and, accordingly, the weighted least squares mean and variance adjusted estimator was used. Full information maximum likelihood was used to address missing data. Commonly accepted cutoffs for the model chi-square (p > 0.05), root mean square error of approximation (≤.06), comparative fit index (≥0.95), and Tucker–Lewis index (≥0.95) indices were used to evaluate model fit [47]. In the first path analysis (Fig. 2A; Mplus 7.4), all autocorrelations were included because measures of the same phenomena within subjects are expected to be highly correlated over time. This is especially true for physical activity behavior, where past behavior is the best predictor of future behavior, and failure to include the autocorrelation leads to overestimation of construct-behavior relationships [17]. Paths leading from baseline moderate-to-vigorous physical activity to follow-up self-efficacy and from baseline moderate-to-vigorous physical activity to follow-up outcome expectations were set to zero because, according to behavioral theory, behavior change is considered an outcome (not a predictor) of psychosocial processes. This model serves as a parent model for both the Bandura (Fig. 2B) and alternative models (Fig. 2C), allowing chi-square difference tests of the equal fit hypothesis. Fig. 2. View largeDownload slide Three path models conducted to test hypothesized relationships among Time 1 (T1) and Time 2 (T2) outcome expectations (OE), self-efficacy (SE), and moderate-to-vigorous physical activity (MVPA). Fig. 2. View largeDownload slide Three path models conducted to test hypothesized relationships among Time 1 (T1) and Time 2 (T2) outcome expectations (OE), self-efficacy (SE), and moderate-to-vigorous physical activity (MVPA). Second, a path model representing Bandura’s original hypothesis that self-efficacy directly leads to outcome expectations, and outcome expectations directly lead to physical activity was conducted (Fig. 2B). The path from baseline outcome expectations to follow-up self-efficacy was set to zero to represent Bandura’s assertion that outcome expectations cannot influence self-efficacy. Subsequently, three alternative path models were tested, each representing theoretical alternatives to Bandura’s original hypotheses regarding the relationships among self-efficacy, outcome expectations, and behavior. Only one alternative model demonstrated acceptable fit statistics (Fig. 2C) and is presented in the published manuscript. Data for the two path models without acceptable fit statistics are presented in Supplementary Material on the journal website. In the alternative model with acceptable fit, Bandura’s assertion that baseline self-efficacy will influence follow-up outcome expectations, and not vice versa, is directly tested by hypothesizing a reciprocal relationship between the two constructs over time. Accordingly, the path from baseline outcome expectations to follow-up self-efficacy was freely estimated. Furthermore, the hypothesized relationship between baseline outcome expectations and follow-up moderate-to-vigorous physical activity was set to zero to directly test the assertion that outcome expectations has a direct effect on physical activity—a pathway necessary for Bandura’s assertion that outcome expectations mediate the relationship between self-efficacy and behavior. Setting this path to zero also reflects previous literature that outcome expectations do not have a significant, direct effect on physical activity [17, 32, 35]. Sensitivity analyses To explore possible bias due to missing data, chi-square, t-tests, and one-way analysis of variance were used to compare individuals with and without missing data for all covariates, social cognitive theory constructs, and physical activity behavior. Additionally, each path analysis was conducted using full information maximum likelihood (n = 1,009) and complete cases only (n = 679). Finally, additional structural equation analyses were conducted to model the measurement error. Like the path analyses, these models were conducted using the weighted least squares mean and variance adjusted estimator, full information maximum likelihood (n = 1,009) and using complete cases only (n = 679). Results Descriptive statistics for the study sample (N = 1,009) are presented in Table 1. Comparisons of individuals with (n = 330) and without (n = 679) missing data at follow-up showed that individuals with complete data more likely to be married/cohabitating (Χ2 = 7.71; df = 1; p < .01), had higher baseline self-efficacy (Χ2 =20.12; df = 4; p < .01), and higher baseline outcome expectations (t = 3.26; df = 1001; p < .01). Although these differences were statistically significant, size of the difference was small for marital status and self-efficacy (Cramer’s V < 0.15) and were less than half a standard deviation for outcome expectations. Additionally, correlations were not indicative of multicollinearity among study variables (Table 2). The unstandardized and standardized path coefficients for the parent model, Bandura model, and the alternative model with acceptable fit statistics are presented (Tables 3 and 4). Table 1 Descriptive statistics for study sample (N = 1,009) Characteristic N (%) Missing n (%) Cancer type/gender 0  Female—breast 359 (35.6)  Male—colorectal 228 (22.6)  Female—colorectal 130 (12.9)  Male—prostate 292 (28.9) Stage 0  Local 630 (62.4)  Regional/distant 379 (37.6) Education 0  ≤High school grad 342 (33.9)  Some college or more 667 (66.1) Marital status 0  Married/cohabitating 712 (70.6)  Not married/cohabitating 297 (29.4) Body mass index category 0  Normal/underweight 288 (28.5)  Overweight 345 (34.2)  Obese 376 (37.3) Time 1 self-efficacy 6  Not at all 116 (11.5)  A little bit 216 (21.4)  Moderately 270 (26.8)  Quite a bit 214 (21.2)  Extremely 187 (18.5) Time 2 self-efficacy 294 (29.1)  Not at all 81 (8.0)  A little bit 152 (15.1)  Moderately 182 (18.0)  Quite a bit 155 (15.4)  Extremely 145 (14.4) Mean (SD) Missing n (%) Age (years) 61.94 (11.00) 0 Months since treatment 7.84 (3.28) 0 Number of symptoms 7.06 (3.25) 0 Number of comorbidities 1.82 (1.56) 0 Time 1 moderate-to-vigorous physical activity 18.43 (21.02) 11 (1.1) Time 2 moderate-to-vigorous physical activity 18.76 (22.00) 301 (29.8) Time 1 outcome expectations 3.97 (0.74) 6 (0.6) Time 2 outcome expectations 4.00 (0.73) 295 (29.2) Characteristic N (%) Missing n (%) Cancer type/gender 0  Female—breast 359 (35.6)  Male—colorectal 228 (22.6)  Female—colorectal 130 (12.9)  Male—prostate 292 (28.9) Stage 0  Local 630 (62.4)  Regional/distant 379 (37.6) Education 0  ≤High school grad 342 (33.9)  Some college or more 667 (66.1) Marital status 0  Married/cohabitating 712 (70.6)  Not married/cohabitating 297 (29.4) Body mass index category 0  Normal/underweight 288 (28.5)  Overweight 345 (34.2)  Obese 376 (37.3) Time 1 self-efficacy 6  Not at all 116 (11.5)  A little bit 216 (21.4)  Moderately 270 (26.8)  Quite a bit 214 (21.2)  Extremely 187 (18.5) Time 2 self-efficacy 294 (29.1)  Not at all 81 (8.0)  A little bit 152 (15.1)  Moderately 182 (18.0)  Quite a bit 155 (15.4)  Extremely 145 (14.4) Mean (SD) Missing n (%) Age (years) 61.94 (11.00) 0 Months since treatment 7.84 (3.28) 0 Number of symptoms 7.06 (3.25) 0 Number of comorbidities 1.82 (1.56) 0 Time 1 moderate-to-vigorous physical activity 18.43 (21.02) 11 (1.1) Time 2 moderate-to-vigorous physical activity 18.76 (22.00) 301 (29.8) Time 1 outcome expectations 3.97 (0.74) 6 (0.6) Time 2 outcome expectations 4.00 (0.73) 295 (29.2) View Large Table 1 Descriptive statistics for study sample (N = 1,009) Characteristic N (%) Missing n (%) Cancer type/gender 0  Female—breast 359 (35.6)  Male—colorectal 228 (22.6)  Female—colorectal 130 (12.9)  Male—prostate 292 (28.9) Stage 0  Local 630 (62.4)  Regional/distant 379 (37.6) Education 0  ≤High school grad 342 (33.9)  Some college or more 667 (66.1) Marital status 0  Married/cohabitating 712 (70.6)  Not married/cohabitating 297 (29.4) Body mass index category 0  Normal/underweight 288 (28.5)  Overweight 345 (34.2)  Obese 376 (37.3) Time 1 self-efficacy 6  Not at all 116 (11.5)  A little bit 216 (21.4)  Moderately 270 (26.8)  Quite a bit 214 (21.2)  Extremely 187 (18.5) Time 2 self-efficacy 294 (29.1)  Not at all 81 (8.0)  A little bit 152 (15.1)  Moderately 182 (18.0)  Quite a bit 155 (15.4)  Extremely 145 (14.4) Mean (SD) Missing n (%) Age (years) 61.94 (11.00) 0 Months since treatment 7.84 (3.28) 0 Number of symptoms 7.06 (3.25) 0 Number of comorbidities 1.82 (1.56) 0 Time 1 moderate-to-vigorous physical activity 18.43 (21.02) 11 (1.1) Time 2 moderate-to-vigorous physical activity 18.76 (22.00) 301 (29.8) Time 1 outcome expectations 3.97 (0.74) 6 (0.6) Time 2 outcome expectations 4.00 (0.73) 295 (29.2) Characteristic N (%) Missing n (%) Cancer type/gender 0  Female—breast 359 (35.6)  Male—colorectal 228 (22.6)  Female—colorectal 130 (12.9)  Male—prostate 292 (28.9) Stage 0  Local 630 (62.4)  Regional/distant 379 (37.6) Education 0  ≤High school grad 342 (33.9)  Some college or more 667 (66.1) Marital status 0  Married/cohabitating 712 (70.6)  Not married/cohabitating 297 (29.4) Body mass index category 0  Normal/underweight 288 (28.5)  Overweight 345 (34.2)  Obese 376 (37.3) Time 1 self-efficacy 6  Not at all 116 (11.5)  A little bit 216 (21.4)  Moderately 270 (26.8)  Quite a bit 214 (21.2)  Extremely 187 (18.5) Time 2 self-efficacy 294 (29.1)  Not at all 81 (8.0)  A little bit 152 (15.1)  Moderately 182 (18.0)  Quite a bit 155 (15.4)  Extremely 145 (14.4) Mean (SD) Missing n (%) Age (years) 61.94 (11.00) 0 Months since treatment 7.84 (3.28) 0 Number of symptoms 7.06 (3.25) 0 Number of comorbidities 1.82 (1.56) 0 Time 1 moderate-to-vigorous physical activity 18.43 (21.02) 11 (1.1) Time 2 moderate-to-vigorous physical activity 18.76 (22.00) 301 (29.8) Time 1 outcome expectations 3.97 (0.74) 6 (0.6) Time 2 outcome expectations 4.00 (0.73) 295 (29.2) View Large Table 2 Correlation matrix among self-efficacy, outcome expectations, and moderate-to-vigorous physical activity over time 1 2 3 4 5 1 Time 1 moderate-to-vigorous physical activity 2 Time 2 moderate-to-vigorous physical activity 0.563 3 Time 1 outcome expectations 0.247 0.225 4 Time 2 outcome expectations 0.218 0.213 0.556 5 Time 1 self-efficacy 0.504 0.404 0.395 0.302 6 Time 2 self-efficacy 0.374 0.517 0.322 0.372 0.631 1 2 3 4 5 1 Time 1 moderate-to-vigorous physical activity 2 Time 2 moderate-to-vigorous physical activity 0.563 3 Time 1 outcome expectations 0.247 0.225 4 Time 2 outcome expectations 0.218 0.213 0.556 5 Time 1 self-efficacy 0.504 0.404 0.395 0.302 6 Time 2 self-efficacy 0.374 0.517 0.322 0.372 0.631 Correlations reported in this table are from the saturated model, with covariates (N = 1,009, weighted least squares mean and variance adjusted estimation, and full information maximum likelihood). View Large Table 2 Correlation matrix among self-efficacy, outcome expectations, and moderate-to-vigorous physical activity over time 1 2 3 4 5 1 Time 1 moderate-to-vigorous physical activity 2 Time 2 moderate-to-vigorous physical activity 0.563 3 Time 1 outcome expectations 0.247 0.225 4 Time 2 outcome expectations 0.218 0.213 0.556 5 Time 1 self-efficacy 0.504 0.404 0.395 0.302 6 Time 2 self-efficacy 0.374 0.517 0.322 0.372 0.631 1 2 3 4 5 1 Time 1 moderate-to-vigorous physical activity 2 Time 2 moderate-to-vigorous physical activity 0.563 3 Time 1 outcome expectations 0.247 0.225 4 Time 2 outcome expectations 0.218 0.213 0.556 5 Time 1 self-efficacy 0.504 0.404 0.395 0.302 6 Time 2 self-efficacy 0.374 0.517 0.322 0.372 0.631 Correlations reported in this table are from the saturated model, with covariates (N = 1,009, weighted least squares mean and variance adjusted estimation, and full information maximum likelihood). View Large Table 3 Results of parent path model Parent model Unstandardized Std est Est (s.e.) 95% CI p-value Paths of interest  T2OE on T1OE 0.51 (0.03) [0.45, 0.58] <.001 0.52  T2OE on T1MVPA @0 @0 @0 @0  T2OE on T1SE 0.08 (0.03) [0.03, 0.13] <.01 0.12  T2SE on T1MVPA @0 @0 @0 @0  T2SE on T1SE 0.61 (0.03) [0.55, 0.66] <.001 0.61  T2SE on T1OE 0.13 (0.05) [0.03, 0.23] .01 0.09  T2MVPA on T1OE 0.15 (0.10) [−0.06, 0.35] .15 0.05  T2MVPA on T1MVPA 0.50 (0.03) [0.43, 0.56] <.001 0.48  T2MVPA on T1SE 0.29 (0.09) [0.11, 0.47] <.01 0.14 Covariances  T1OE with T1MVPA 0.38 (0.05) [0.29, 0.47] <.001 0.26  T1OE with T1SE 0.27 (0.02) [0.23, 0.31] <.001 0.39  T1SE with T1MVPA 1.09 (0.06) [0.97, 1.21] <.001 0.52  T2OE with T2MVPA 0.08 (0.04) [0.00, 0.16] .05 0.08  T2OE with T2SE 0.09 (0.02) [0.06, 0.13] <.001 0.21  T2SE with T2MVPA 0.55 (0.06) [0.43, 0.66] <.001 0.41 Parent model Unstandardized Std est Est (s.e.) 95% CI p-value Paths of interest  T2OE on T1OE 0.51 (0.03) [0.45, 0.58] <.001 0.52  T2OE on T1MVPA @0 @0 @0 @0  T2OE on T1SE 0.08 (0.03) [0.03, 0.13] <.01 0.12  T2SE on T1MVPA @0 @0 @0 @0  T2SE on T1SE 0.61 (0.03) [0.55, 0.66] <.001 0.61  T2SE on T1OE 0.13 (0.05) [0.03, 0.23] .01 0.09  T2MVPA on T1OE 0.15 (0.10) [−0.06, 0.35] .15 0.05  T2MVPA on T1MVPA 0.50 (0.03) [0.43, 0.56] <.001 0.48  T2MVPA on T1SE 0.29 (0.09) [0.11, 0.47] <.01 0.14 Covariances  T1OE with T1MVPA 0.38 (0.05) [0.29, 0.47] <.001 0.26  T1OE with T1SE 0.27 (0.02) [0.23, 0.31] <.001 0.39  T1SE with T1MVPA 1.09 (0.06) [0.97, 1.21] <.001 0.52  T2OE with T2MVPA 0.08 (0.04) [0.00, 0.16] .05 0.08  T2OE with T2SE 0.09 (0.02) [0.06, 0.13] <.001 0.21  T2SE with T2MVPA 0.55 (0.06) [0.43, 0.66] <.001 0.41 Relationships between covariates and self-efficacy (SE), outcome expectations (OE), and moderate-to-vigorous physical activity (MVPA) were freely estimated for both time points (Supplementary Material). MVPA was transformed using a square root transformation to adjust for non-normality, and weighted least squares mean and variance adjusted estimation, and full information maximum likelihood were used. Est estimate; s.e. standard error; 95% CI 95% confidence interval; Std est standardized estimate; T1 time point 1; T2 timepoint 2. View Large Table 3 Results of parent path model Parent model Unstandardized Std est Est (s.e.) 95% CI p-value Paths of interest  T2OE on T1OE 0.51 (0.03) [0.45, 0.58] <.001 0.52  T2OE on T1MVPA @0 @0 @0 @0  T2OE on T1SE 0.08 (0.03) [0.03, 0.13] <.01 0.12  T2SE on T1MVPA @0 @0 @0 @0  T2SE on T1SE 0.61 (0.03) [0.55, 0.66] <.001 0.61  T2SE on T1OE 0.13 (0.05) [0.03, 0.23] .01 0.09  T2MVPA on T1OE 0.15 (0.10) [−0.06, 0.35] .15 0.05  T2MVPA on T1MVPA 0.50 (0.03) [0.43, 0.56] <.001 0.48  T2MVPA on T1SE 0.29 (0.09) [0.11, 0.47] <.01 0.14 Covariances  T1OE with T1MVPA 0.38 (0.05) [0.29, 0.47] <.001 0.26  T1OE with T1SE 0.27 (0.02) [0.23, 0.31] <.001 0.39  T1SE with T1MVPA 1.09 (0.06) [0.97, 1.21] <.001 0.52  T2OE with T2MVPA 0.08 (0.04) [0.00, 0.16] .05 0.08  T2OE with T2SE 0.09 (0.02) [0.06, 0.13] <.001 0.21  T2SE with T2MVPA 0.55 (0.06) [0.43, 0.66] <.001 0.41 Parent model Unstandardized Std est Est (s.e.) 95% CI p-value Paths of interest  T2OE on T1OE 0.51 (0.03) [0.45, 0.58] <.001 0.52  T2OE on T1MVPA @0 @0 @0 @0  T2OE on T1SE 0.08 (0.03) [0.03, 0.13] <.01 0.12  T2SE on T1MVPA @0 @0 @0 @0  T2SE on T1SE 0.61 (0.03) [0.55, 0.66] <.001 0.61  T2SE on T1OE 0.13 (0.05) [0.03, 0.23] .01 0.09  T2MVPA on T1OE 0.15 (0.10) [−0.06, 0.35] .15 0.05  T2MVPA on T1MVPA 0.50 (0.03) [0.43, 0.56] <.001 0.48  T2MVPA on T1SE 0.29 (0.09) [0.11, 0.47] <.01 0.14 Covariances  T1OE with T1MVPA 0.38 (0.05) [0.29, 0.47] <.001 0.26  T1OE with T1SE 0.27 (0.02) [0.23, 0.31] <.001 0.39  T1SE with T1MVPA 1.09 (0.06) [0.97, 1.21] <.001 0.52  T2OE with T2MVPA 0.08 (0.04) [0.00, 0.16] .05 0.08  T2OE with T2SE 0.09 (0.02) [0.06, 0.13] <.001 0.21  T2SE with T2MVPA 0.55 (0.06) [0.43, 0.66] <.001 0.41 Relationships between covariates and self-efficacy (SE), outcome expectations (OE), and moderate-to-vigorous physical activity (MVPA) were freely estimated for both time points (Supplementary Material). MVPA was transformed using a square root transformation to adjust for non-normality, and weighted least squares mean and variance adjusted estimation, and full information maximum likelihood were used. Est estimate; s.e. standard error; 95% CI 95% confidence interval; Std est standardized estimate; T1 time point 1; T2 timepoint 2. View Large Table 4 Results of the Bandura and alternative path model Bandura model Alternative model Unstandardized Std est Unstandardized Std est Est (s.e.) 95% CI p-value Est (s.e.) 95% CI p-value Paths of interest  T2OE on T1OE 0.52 (0.03) [0.45, 0.58] <.001 0.52 0.52 (0.03) [0.45, 0.58] <.001 0.52  T2OE on T1MVPA @0 @0 @0 @0 @0 @0 @0 @0  T2OE on T1SE 0.07 (0.03) [0.02, 0.12] <.01 0.11 0.08 (0.03) [0.03, 0.12] <.01 0.11  T2SE on T1MVPA @0 @0 @0 @0 @0 @0 @0 @0  T2SE on T1SE 0.65 (0.02) [0.60, 0.70] <.001 0.65 0.61 (0.03) [0.55, 0.66] <.001 0.61  T2SE on T1OE @0 @0 @0 @0 0.13 (0.05) [0.03, 0.23] .01 0.09  T2MVPA on T1OE 0.14 (0.10) [−0.06, 0.34] .18 0.05 @0 @0 @0 @0  T2MVPA on T1MVPA 0.50 (0.03) [0.43, 0.56] <.001 0.48 0.50 (0.03) [0.43, 0.56] <.001 0.48  T2MVPA on T1SE 0.29 (0.09) [0.11, 0.48] <.01 0.14 0.36 (0.09) [0.19, 0.53] <.001 0.17 Covariances  T1OE with T1MVPA 0.38 (0.05) [0.29, 0.47] <.001 0.26 0.40 (0.05) [0.30, 0.49] <.001 0.27  T1OE with T1SE 0.29 (0.02) [0.25, 0.33] <.001 0.41 0.27 (0.02) [0.23, 0.31] <.001 0.39  T1SE with T1MVPA 1.09 (0.06) [0.98, 1.21] <.001 0.52 1.09 (0.06) [0.97, 1.20] <.001 0.52  T2OE with T2MVPA 0.08 (0.04) [0.01, 0.16] .03 0.08 0.10 (0.04) [0.02, 0.18] .01 0.10  T2OE with T2SE 0.12 (0.02) [0.08, 0.16] <.001 0.27 0.09 (0.02) [0.06, 0.13] <.001 0.22  T2SE with T2MVPA 0.55 (0.06) [0.44, 0.67] <.001 0.41 0.54 (0.06) [0.43, 0.66] <.001 0.40 Bandura model Alternative model Unstandardized Std est Unstandardized Std est Est (s.e.) 95% CI p-value Est (s.e.) 95% CI p-value Paths of interest  T2OE on T1OE 0.52 (0.03) [0.45, 0.58] <.001 0.52 0.52 (0.03) [0.45, 0.58] <.001 0.52  T2OE on T1MVPA @0 @0 @0 @0 @0 @0 @0 @0  T2OE on T1SE 0.07 (0.03) [0.02, 0.12] <.01 0.11 0.08 (0.03) [0.03, 0.12] <.01 0.11  T2SE on T1MVPA @0 @0 @0 @0 @0 @0 @0 @0  T2SE on T1SE 0.65 (0.02) [0.60, 0.70] <.001 0.65 0.61 (0.03) [0.55, 0.66] <.001 0.61  T2SE on T1OE @0 @0 @0 @0 0.13 (0.05) [0.03, 0.23] .01 0.09  T2MVPA on T1OE 0.14 (0.10) [−0.06, 0.34] .18 0.05 @0 @0 @0 @0  T2MVPA on T1MVPA 0.50 (0.03) [0.43, 0.56] <.001 0.48 0.50 (0.03) [0.43, 0.56] <.001 0.48  T2MVPA on T1SE 0.29 (0.09) [0.11, 0.48] <.01 0.14 0.36 (0.09) [0.19, 0.53] <.001 0.17 Covariances  T1OE with T1MVPA 0.38 (0.05) [0.29, 0.47] <.001 0.26 0.40 (0.05) [0.30, 0.49] <.001 0.27  T1OE with T1SE 0.29 (0.02) [0.25, 0.33] <.001 0.41 0.27 (0.02) [0.23, 0.31] <.001 0.39  T1SE with T1MVPA 1.09 (0.06) [0.98, 1.21] <.001 0.52 1.09 (0.06) [0.97, 1.20] <.001 0.52  T2OE with T2MVPA 0.08 (0.04) [0.01, 0.16] .03 0.08 0.10 (0.04) [0.02, 0.18] .01 0.10  T2OE with T2SE 0.12 (0.02) [0.08, 0.16] <.001 0.27 0.09 (0.02) [0.06, 0.13] <.001 0.22  T2SE with T2MVPA 0.55 (0.06) [0.44, 0.67] <.001 0.41 0.54 (0.06) [0.43, 0.66] <.001 0.40 Relationships between covariates and self-efficacy (SE), outcome expectations (OE), and moderate-to-vigorous physical activity (MVPA) were freely estimated for both time points (Supplementary Material). MVPA was transformed using a square root transformation to adjust for non-normality, and weighted least squares mean and variance adjusted estimation, and full information maximum likelihood were used. Est estimate; s.e. standard error; 95% CI 95% confidence interval; Std est standardized estimate; T1 time point 1; T2 timepoint 2. View Large Table 4 Results of the Bandura and alternative path model Bandura model Alternative model Unstandardized Std est Unstandardized Std est Est (s.e.) 95% CI p-value Est (s.e.) 95% CI p-value Paths of interest  T2OE on T1OE 0.52 (0.03) [0.45, 0.58] <.001 0.52 0.52 (0.03) [0.45, 0.58] <.001 0.52  T2OE on T1MVPA @0 @0 @0 @0 @0 @0 @0 @0  T2OE on T1SE 0.07 (0.03) [0.02, 0.12] <.01 0.11 0.08 (0.03) [0.03, 0.12] <.01 0.11  T2SE on T1MVPA @0 @0 @0 @0 @0 @0 @0 @0  T2SE on T1SE 0.65 (0.02) [0.60, 0.70] <.001 0.65 0.61 (0.03) [0.55, 0.66] <.001 0.61  T2SE on T1OE @0 @0 @0 @0 0.13 (0.05) [0.03, 0.23] .01 0.09  T2MVPA on T1OE 0.14 (0.10) [−0.06, 0.34] .18 0.05 @0 @0 @0 @0  T2MVPA on T1MVPA 0.50 (0.03) [0.43, 0.56] <.001 0.48 0.50 (0.03) [0.43, 0.56] <.001 0.48  T2MVPA on T1SE 0.29 (0.09) [0.11, 0.48] <.01 0.14 0.36 (0.09) [0.19, 0.53] <.001 0.17 Covariances  T1OE with T1MVPA 0.38 (0.05) [0.29, 0.47] <.001 0.26 0.40 (0.05) [0.30, 0.49] <.001 0.27  T1OE with T1SE 0.29 (0.02) [0.25, 0.33] <.001 0.41 0.27 (0.02) [0.23, 0.31] <.001 0.39  T1SE with T1MVPA 1.09 (0.06) [0.98, 1.21] <.001 0.52 1.09 (0.06) [0.97, 1.20] <.001 0.52  T2OE with T2MVPA 0.08 (0.04) [0.01, 0.16] .03 0.08 0.10 (0.04) [0.02, 0.18] .01 0.10  T2OE with T2SE 0.12 (0.02) [0.08, 0.16] <.001 0.27 0.09 (0.02) [0.06, 0.13] <.001 0.22  T2SE with T2MVPA 0.55 (0.06) [0.44, 0.67] <.001 0.41 0.54 (0.06) [0.43, 0.66] <.001 0.40 Bandura model Alternative model Unstandardized Std est Unstandardized Std est Est (s.e.) 95% CI p-value Est (s.e.) 95% CI p-value Paths of interest  T2OE on T1OE 0.52 (0.03) [0.45, 0.58] <.001 0.52 0.52 (0.03) [0.45, 0.58] <.001 0.52  T2OE on T1MVPA @0 @0 @0 @0 @0 @0 @0 @0  T2OE on T1SE 0.07 (0.03) [0.02, 0.12] <.01 0.11 0.08 (0.03) [0.03, 0.12] <.01 0.11  T2SE on T1MVPA @0 @0 @0 @0 @0 @0 @0 @0  T2SE on T1SE 0.65 (0.02) [0.60, 0.70] <.001 0.65 0.61 (0.03) [0.55, 0.66] <.001 0.61  T2SE on T1OE @0 @0 @0 @0 0.13 (0.05) [0.03, 0.23] .01 0.09  T2MVPA on T1OE 0.14 (0.10) [−0.06, 0.34] .18 0.05 @0 @0 @0 @0  T2MVPA on T1MVPA 0.50 (0.03) [0.43, 0.56] <.001 0.48 0.50 (0.03) [0.43, 0.56] <.001 0.48  T2MVPA on T1SE 0.29 (0.09) [0.11, 0.48] <.01 0.14 0.36 (0.09) [0.19, 0.53] <.001 0.17 Covariances  T1OE with T1MVPA 0.38 (0.05) [0.29, 0.47] <.001 0.26 0.40 (0.05) [0.30, 0.49] <.001 0.27  T1OE with T1SE 0.29 (0.02) [0.25, 0.33] <.001 0.41 0.27 (0.02) [0.23, 0.31] <.001 0.39  T1SE with T1MVPA 1.09 (0.06) [0.98, 1.21] <.001 0.52 1.09 (0.06) [0.97, 1.20] <.001 0.52  T2OE with T2MVPA 0.08 (0.04) [0.01, 0.16] .03 0.08 0.10 (0.04) [0.02, 0.18] .01 0.10  T2OE with T2SE 0.12 (0.02) [0.08, 0.16] <.001 0.27 0.09 (0.02) [0.06, 0.13] <.001 0.22  T2SE with T2MVPA 0.55 (0.06) [0.44, 0.67] <.001 0.41 0.54 (0.06) [0.43, 0.66] <.001 0.40 Relationships between covariates and self-efficacy (SE), outcome expectations (OE), and moderate-to-vigorous physical activity (MVPA) were freely estimated for both time points (Supplementary Material). MVPA was transformed using a square root transformation to adjust for non-normality, and weighted least squares mean and variance adjusted estimation, and full information maximum likelihood were used. Est estimate; s.e. standard error; 95% CI 95% confidence interval; Std est standardized estimate; T1 time point 1; T2 timepoint 2. View Large The parent model (Fig. 2A; Table 3) was a good fit to the data [Χ2 (2) = 5.22, p = .07; CFI = 0.99; TLI = 0.93; RMSEA = 0.04 (95% CI: 0.00, 0.08)] and accounted for 41, 39, and 52 per cent of variance in follow-up moderate-to-vigorous physical activity, outcome expectations, and self-efficacy, respectively. Standardized and unstandardized path coefficients, confidence intervals, and p-values are presented in Table 3. The path model representing Bandura’s hypothesis (Fig. 2B; Table 4) that self-efficacy causes outcome expectations, and not vice versa, did not meet recommended cutoffs for fit indices [Χ2 (3) = 11.49, p = .009; CFI = 0.99; TLI = 0.88; RMSEA = 0.05 (95% CI:0.02, 0.09)], indicating that this model is not a good fit to the data (Table 3). When directly comparing the parent model and the Bandura model, a Χ2 difference test resulted in the rejection of the equal fit hypothesis [Χ2 (1) = 5.92, p = .01], indicating a significant difference in model fit between the parent model and the model proposed by Bandura (i.e., the parent model is a better fitting model). Path coefficients show that the path from baseline outcome expectations to follow-up moderate-to-vigorous physical activity is not statistically significant (p = .18; Table 4). Finally, the path model examining the alternative hypothesis (Fig. 2C; Table 4) met recommended cutoffs for fit indices [Χ2 (3) = 7.22, p = .07; CFI = 0.99; TLI = 0.94; RMSEA = 0.04 (95% CI: 0.00, 0.07)], indicating that it is a good fit to the data. When directly comparing the parent model and the alternative model, the Χ2 difference test failed to reject the equal fit hypothesis [Χ2 (1) = 2.01, p = .16], indicating no significant differences in model fit between the parent model and the alternative model. Path coefficients show that the path from baseline outcome expectations to follow-up self-efficacy and the path from baseline self-efficacy to follow-up outcome expectations are both statistically significant (Table 4). Detailed results of sensitivity analyses are presented in Supplementary Material. Briefly, results of the structural equation models replicated the findings from the path analyses such that the parent model was a good fit to the data [Χ2 (71) = 88.61, p = .07; CFI = 0.99; TLI = 0.99; RMSEA = 0.02 (95% CI: 0.00, 0.02)]. Additionally, the Bandura model was not a good fit to the data [Χ2 (72) = 101.01, p = .01; CFI = 0.99; TLI = 0.99; RMSEA = 0.02 (95% CI: 0.01, 0.03)] and failed the equal fit hypothesis [Χ2 (1) = 6.06, p = .02]. Meanwhile, the alternative model was a good fit to the data [Χ2 (72) = 92.06, p = .06; CFI = 0.99; TLI = 0.99; RMSEA = 0.02 (95% CI: 0.00, 0.03)], and the chi-square test of equal fit showed that the alternative model and the parent model were statistically equivalent [Χ2 (1) = 2.49, p = .11]. Models without covariates were not a sufficient fit to the data [parent path model without covariates: [Χ2 (2) = 15.17, p < .001; CFI = 0.99; TLI = 0.92; RMSEA = 0.08 (95% CI: 0.05, 0.12)], indicating the need to incorporate covariates in the model to provide more accurate parameter estimates for social cognitive theory constructs. Finally, complete case (n = 679) path analyses yielded similar results to full information maximum likelihood path models (n = 1,009; results not presented). Discussion The current study examined, at two time points, the relationship between self-efficacy, outcome expectations, and physical activity among a cohort of breast, prostate, and colorectal cancer survivors who recently completed treatment. The parent model and the alterative model met recommended fit indices and like previous research [17, 29], accounted for 41, 39, and 52 per cent of variance in follow-up moderate-to-vigorous physical activity, outcome expectations, and self-efficacy, respectively. Upon examining the path coefficient of the alternative model, there is evidence of a reciprocal relationship between self-efficacy and outcome expectations over time, which is inconsistent with Bandura’s original assertion that outcome expectations do not predict self-efficacy. Furthermore, fixing the outcome expectations to physical activity path to zero is statistically defensible, which negates one of the pathways needed to complete Bandura’s proposed mediational pathway. The Bandura model explained similar levels of variance in follow-up moderate-to-vigorous physical activity (41%), outcome expectations (39%), and self-efficacy (52%), but did not meet fit indices. When examining the path coefficients of the Bandura model, baseline self-efficacy significantly influences follow-up outcome expectations, which is consistent with Bandura’s assertion. Inconsistent with Bandura’s assertion, however, is the lack of a significant effect of baseline outcome expectations on follow-up moderate-to-vigorous physical activity, again suggesting that Bandura’s proposed mediational pathway is uncertain. Taken together, the path models presented in this study fail to support Bandura’s original assertions within the social cognitive theory that (a) self-efficacy precedes outcome expectations and not vice versa, and (b) outcome expectations directly effects moderate-to-vigorous physical activity, a necessary relationship for Bandura’s proposed mediational pathway (self-efficacy → outcome expectations → moderate-to-vigorous physical activity). Instead, our data suggest that outcome expectations precede self-efficacy, or that there may be a reciprocal relationship between self-efficacy and outcome expectations (self-efficacy ↔ outcome expectations). Additionally, our data support previous reviews and commentaries [32, 35], suggesting that outcome expectations do not directly affect moderate-to-vigorous physical activity. If these findings were extrapolated to longitudinal data with three or more time points, we would hypothesize that self-efficacy mediates the relationship between outcome expectations and moderate-to-vigorous physical activity. Some limitations of this study should be noted. First, this was a secondary data analysis relying on self-reported data from a convenience sample of breast, colorectal, and prostate cancer survivors transitioning out of primary treatment. Most participants reported their race/ethnicity as non-Hispanic white/Caucasian, and most had received at least some college education. Therefore, results may be subject to self-report biases, and we cannot confidently generalize our results to individuals with other cancers, survivors undergoing active treatment, cancer survivors who are in advanced stage cancer or end-of-life care, minority groups, or less educated groups. Second, random selection from an American Cancer Society national call center constituent database may have produced a sample more likely to seek cancer-related and health-related information, which may be an indicator of overall motivation and activation for healthier lifestyle. Specifically, the call center reports that 80 per cent of the callers are self-referred and 20 per cent are referred by their physician [38], and therefore, our sample may not be representative of all cancer survivors. Third, the current study assessed theoretical constructs and behavior at two time points, an average of 13.3 months apart. Future research is needed with assessments at three or more time points to allow for testing of full mediation models. Additionally, although our timeframe is not inconsistent with prospective and intervention studies using 12 or 18 month time intervals [16, 17], or the mean changes in constructs in previous research [29, 33], it is important to think about how the length of time between assessments limits our ability to understand when changes in constructs occurred during this time, how many changes occurred, the magnitude of the changes that occurred, or the causal mechanisms of these changes. The common use of 6, 12, or 18 month assessment periods is a normative methodology, not based on the social cognitive theory, nor empirical research examining the optimal timeframe for studying these relationships. Thus, future research is needed to examine whether ordering of theoretical constructs is consistent or varies depending on length of time between assessments. This research would have implications for theory, but also implications for intervention delivery and expected intervention outcomes across different contexts (e.g., phone, face-to-face, and adaptive technology). Finally, regarding construct measurement, we did not include other constructs within the social cognitive theory, and therefore, this study is subject to missing variable bias. Furthermore, self-efficacy was assessed with a single item which limits the reliability and validity of the measure, although we did account for the ordinal nature of the item by using the weighted least squares mean and variance adjusted estimation. Outcome expectations was assessed using three items often used within the cancer population that, despite good inter-item correlations, did not meet confirmatory factor analysis cut-off criteria, and therefore, outcome expectations were modeled as an observed variable. Additional analyses incorporating measurement error using structural equation modeling produced similar overall results, with the alternative model meeting model fit indices and proving to be a superior model to Bandura’s original model (Supplementary Tables A–E). While a better measure of outcome expectations would be preferable, the additional structural equation models, in combination with the explicit testing of alternative models using the same data, suggest that our findings cannot be dismissed due to measurement error. Despite these limitations, this study has several advantages. First, because cancer survivors would gain positive health benefits from physical activity, but are less likely than the general adult population to meet physical activity recommendations [48], there is greater need to focus on this population. Second, we improved upon previous cross-sectional research by using a longitudinal, prospective survey design. Bias introduced by limiting analysis to only complete cases was avoided by using full information maximum likelihood methods. Finally, we have avoided confirmation bias by explicitly testing several alternatives to the original construct ordering suggested within the social cognitive theory. This study has implications for future behavioral research as well as applied public health practice regarding the use of the social cognitive theory as a framework for understanding and promoting physical activity behavior. Specifically, future research should seek to replicate these findings in different population subgroups (e.g., race/ethnicity, age, body mass index categories, by cancer type, and noncancer populations) and for other health behaviors (e.g., diet, stress management, and smoking cessation) with the goal of determining whether the social cognitive theory requires modification. These results should also be replicated across different time points in the cancer survivorship trajectory, specifically during treatment, where physical activity initiation and maintenance are associated with cancer treatment adherence [49], reduced cancer-related fatigue [4–9], and improvements in quality of life [10–13]. Replication studies are needed that use reliable and valid measures of self-efficacy and outcome expectations and assess additional social cognitive theory constructs to reduce measurement error and avoid missing variable bias. It seems particularly important to include negative outcome expectations [33], perceived barriers [33], and goal-setting [29, 33]. Mastery experience is a less commonly studied construct proposed to predict self-efficacy and promote its sustainability [15], and may account for variation in self-efficacy due to past physical activity experiences. Future studies may also benefit from emerging statistical techniques that seek to reduce model complexity [50]. Results of this research inform development of traditional in-person interventions, as well as interventions delivered via automated technology (e.g., phone, website, and health apps), to ensure that intervention components adhere to processes supported by effective theory. Specifically, an intervention approach that would strengthen outcome expectations over self-efficacy beliefs would not be supported. But acknowledging and supporting an interactional approach between these two variables or an ordered approach (outcome expectations → self-efficacy) would be supported. Furthermore, randomized control trials or comparative effectiveness studies should be designed to determine whether changing the ordering of outcome expectations and self-efficacy in intervention modules will result in clinically meaningful differences in physical activity outcomes. If successful, this would be a simple change in existing research and practice with little cost for implementation. In conclusion, most cancer survivors do not engage in sufficient moderate-to-vigorous physical activity to attain health benefits, and therefore, efficacious interventions are needed to increase the prevalence and long-term maintenance of moderate-to-vigorous physical activity in this population. This study advances our understanding of how the social cognitive theory constructs might be optimally applied to better explain, and ultimately, change moderate-to-vigorous physical activity behavior among cancer survivors. Studies using similar methodologies to replicate these findings, as well as randomized controlled trials and/or comparative effectiveness studies, are needed to systematically improve existing social cognitive theory interventions for physical activity among cancer survivors. Supplementary Material Supplementary material is available at Annals of Behavioral Medicine online. Compliance with Ethical Standards Authors’ Statement of Conflict of Interest and Adherence to Ethical Standards Authors Elizabeth Fallon, Robert Stephens, Bennett McDonald, and Corinne Leach acknowledge their employment at American Cancer Society, and declare no other conflicts of interest. Author Michael Diefenbach declares no conflict of interest. All procedures, including the informed consent process, were conducted in accordance with the ethical standards of the responsible committee on human experimentation (institutional and national) and with the Helsinki Declaration of 1975, as revised in 2000. Authors’ Contributions E.F. generated the paper concept. C.L. is the PI of the Transition Study, and was responsible for designing the survey and supervising the data collection and management process. M.D. served as an expert consultant for the Transition Study. B.M. led the data management process. E.F. led the data analysis, supported by R.S., and B.M. E.F. led the manuscript writing, supported by B.M. and R.S. All authors edited the manuscript for substantive content, and approved the manuscript for publication. Ethical Approval All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the Helsinki declaration or comparable ethical standards. Informed Consent Informed consent was obtained from all individual participants included in the study. Acknowledgments We thank the study participants, who demonstrated their leadership in the fight against cancer by giving their time and insight into the cancer survivorship experience. Without their support, this research and future application of what we have learned would not have been possible. Data for this manuscript came from the Transition Study, funded by the American Cancer Society. We acknowledge Alyssa Troeschel and Rhyan Vereen for their role in survey design, data collection, and cleaning and Dawn Wiatrek who served as co-PI in the Transition study design. References 1. Friedenreich CM , Wang Q , Neilson HK , Kopciuk KA , McGregor SE , Courneya KS . Physical activity and survival after prostate cancer . Eur Urol . 2016 ; 70 : 576 – 585 . Google Scholar CrossRef Search ADS PubMed 2. Lee IM , Wolin KY , Freeman SE , Sattlemair J , Sesso HD . Physical activity and survival after cancer diagnosis in men . J Phys Act Health . 2014 ; 11 : 85 – 90 . Google Scholar CrossRef Search ADS PubMed 3. Lahart IM , Metsios GS , Nevill AM , Carmichael AR . Physical activity, risk of death and recurrence in breast cancer survivors: A systematic review and meta-analysis of epidemiological studies . Acta Oncol . 2015 ; 54 : 635 – 654 . Google Scholar CrossRef Search ADS PubMed 4. Ballard-Barbash R , Friedenreich CM , Courneya KS , Siddiqi SM , McTiernan A , Alfano CM . Physical activity, biomarkers, and disease outcomes in cancer survivors: A systematic review . J Natl Cancer Inst . 2012 ; 104 : 815 – 840 . Google Scholar CrossRef Search ADS PubMed 5. Mustian KM , Alfano CM , Heckler C , et al. Comparison of pharmaceutical, psychological, and exercise treatments for cancer-related fatigue: A meta-analysis . JAMA Oncol . 2017 ; 3 : 961 – 968 . Google Scholar CrossRef Search ADS PubMed 6. Lipsett A , Barrett S , Haruna F , Mustian K , O’Donovan A . The impact of exercise during adjuvant radiotherapy for breast cancer on fatigue and quality of life: A systematic review and meta-analysis . Breast . 2017 ; 32 : 144 – 155 . Google Scholar CrossRef Search ADS PubMed 7. Dennett AM , Peiris CL , Shields N , Prendergast LA , Taylor NF . Moderate-intensity exercise reduces fatigue and improves mobility in cancer survivors: A systematic review and meta-regression . J Physiother . 2016 ; 62 : 68 – 82 . Google Scholar CrossRef Search ADS PubMed 8. Meneses-Echávez JF , González-Jiménez E , Ramírez-Vélez R . Supervised exercise reduces cancer-related fatigue: A systematic review . J Physiother . 2015 ; 61 : 3 – 9 . Google Scholar CrossRef Search ADS PubMed 9. Meneses-Echávez JF , González-Jiménez E , Ramírez-Vélez R . Effects of supervised multimodal exercise interventions on cancer-related fatigue: Systematic review and meta-analysis of randomized controlled trials . Biomed Res Int . 2015 ; 2015 : 328636 . Google Scholar CrossRef Search ADS PubMed 10. Menichetti J , Villa S , Magnani T , et al. Lifestyle interventions to improve the quality of life of men with prostate cancer: A systematic review of randomized controlled trials . Crit Rev Oncol Hematol . 2016 ; 108 : 13 – 22 . Google Scholar CrossRef Search ADS PubMed 11. Bourke L , Boorjian SA , Briganti A , et al. Survivorship and improving quality of life in men with prostate cancer . Eur Urol . 2015 ; 68 : 374 – 383 . Google Scholar CrossRef Search ADS PubMed 12. Mishra SI , Scherer RW , Geigle PM , et al. Exercise interventions on health-related quality of life for cancer survivors . Cochrane DB Syst Rev . 2012 ; 8 : Cd007566 . 13. Mishra SI , Scherer RW , Snyder C , Geigle PM , Berlanstein DR , Topaloglu O . Exercise interventions on health-related quality of life for people with cancer during active treatment . Cochrane DB Syst Rev . 2012 ( 8 ): Cd008465 . 14. Tannenbaum SL , McClure LA , Asfar T , Sherman RL , LeBlanc WG , Lee DJ . Are cancer survivors physically active? a comparison by US States . J Phys Act Health . 2016 ; 13 : 159 – 167 . Google Scholar CrossRef Search ADS PubMed 15. Bandura A . Self-efficacy: Toward a unifying theory of behavioral change . Psychol Rev . 1977 ; 84 : 191 – 215 . Google Scholar CrossRef Search ADS PubMed 16. Stacey FG , James EL , Chapman K , Courneya KS , Lubans DR . A systematic review and meta-analysis of social cognitive theory-based physical activity and/or nutrition behavior change interventions for cancer survivors . J Cancer Surviv . 2015 ; 9 : 305 – 338 . Google Scholar CrossRef Search ADS PubMed 17. Young MD , Plotnikoff RC , Collins CE , Callister R , Morgan PJ . Social cognitive theory and physical activity: A systematic review and meta-analysis . Obes Rev . 2014 ; 15 : 983 – 995 . Google Scholar CrossRef Search ADS PubMed 18. Bandura A. Self-Efficacy: The Exercise of Control . New York : Freeman ; 1997 . 19. Rogers LQ , Courneya KS , Verhulst S , Markwell S , Lanzotti V , Shah P . Exercise barrier and task self-efficacy in breast cancer patients during treatment . Support Care Cancer . 2006 ; 14 : 84 – 90 . Google Scholar CrossRef Search ADS PubMed 20. Short CE , James EL , Plotnikoff RC . How social cognitive theory can help oncology-based health professionals promote physical activity among breast cancer survivors . Eur J Oncol Nurs . 2013 ; 17 : 482 – 489 . Google Scholar CrossRef Search ADS PubMed 21. Hirschey R , Docherty SL , Pan W , Lipkus I . Exploration of exercise outcome expectations among breast cancer survivors . Cancer Nurs . 2017 ; 40 : E39 – E46 . Google Scholar CrossRef Search ADS PubMed 22. Brunet J , Taran S , Burke S , Sabiston CM . A qualitative exploration of barriers and motivators to physical activity participation in women treated for breast cancer . Disabil Rehabil . 2013 ; 35 : 2038 – 2045 . Google Scholar CrossRef Search ADS PubMed 23. Bandura A . Reflections on self-efficacy . In: Rachman S , ed. Advances in Behavior Research and Therapy . Oxford, UK: Pergamon Press Ltd.; 1978; 1 : 237 – 269 . Google Scholar CrossRef Search ADS 24. Bandura A . Recycling misconceptions of perceived self-efficacy . Cognitive Ther Res . 1984 ; 8 ( 3 ): 231 – 255 . Google Scholar CrossRef Search ADS 25. Bandura A . Health promotion from the perspective of social cognitive theory . Psychol Health . 1998 ; 13 ( 4 ): 623 – 649 . Google Scholar CrossRef Search ADS 26. Bandura A . Toward a psychology of human agency . Perspect Psychol Sci . 2006 ; 1 : 164 – 180 . Google Scholar CrossRef Search ADS PubMed 27. Esmaeily H , Peyman N , Taghipour A , KHorashadizadeh F , Mahdizadeh M . A structural equation model to predict the social-cognitive determinants related to physical activity in Iranian women with diabetes mellitus . J Res Health Sci . 2014 ; 14 : 296 – 302 . Google Scholar PubMed 28. Rovniak LS , Anderson ES , Winett RA , Stephens RS . Social cognitive determinants of physical activity in young adults: A prospective structural equation analysis . Ann Behav Med . 2002 ; 24 : 149 – 156 . Google Scholar CrossRef Search ADS PubMed 29. Phillips SM , McAuley E . Social cognitive influences on physical activity participation in long-term breast cancer survivors . Psychooncology . 2013 ; 22 : 783 – 791 . Google Scholar CrossRef Search ADS PubMed 30. Plotnikoff RC , Lubans DR , Penfold CM , Courneya KS . Testing the utility of three social-cognitive models for predicting objective and self-report physical activity in adults with type 2 diabetes . Br J Health Psychol . 2014 ; 19 : 329 – 346 . Google Scholar CrossRef Search ADS PubMed 31. Dewar DL , Plotnikoff RC , Morgan PJ , Okely AD , Costigan SA , Lubans DR . Testing social-cognitive theory to explain physical activity change in adolescent girls from low-income communities . Res Q Exerc Sport . 2013 ; 84 : 483 – 491 . Google Scholar CrossRef Search ADS PubMed 32. Williams DM , Anderson ES , Winett RA . A review of the outcome expectancy construct in physical activity research . Ann Behav Med . 2005 ; 29 : 70 – 79 . Google Scholar CrossRef Search ADS PubMed 33. Rogers LQ , Courneya KS , Anton PM , et al. Social cognitive constructs did not mediate the BEAT cancer intervention effects on objective physical activity behavior based on multivariable path analysis . Ann Behav Med . 2017 ; 51 : 321 – 326 . Google Scholar CrossRef Search ADS PubMed 34. Rogers LQ , Courneya KS , Anton PM , et al. Effects of the BEAT Cancer physical activity behavior change intervention on physical activity, aerobic fitness, and quality of life in breast cancer survivors: A multicenter randomized controlled trial . Breast Cancer Res Treat . 2015 ; 149 : 109 – 119 . Google Scholar CrossRef Search ADS PubMed 35. Williams DM . Outcome expectancy and self-efficacy: Theoretical implications of an unresolved contradiction . Pers Soc Psychol Rev . 2010 ; 14 : 417 – 425 . Google Scholar CrossRef Search ADS PubMed 36. Coups EJ , Park BJ , Feinstein MB , et al. Correlates of physical activity among lung cancer survivors . Psychooncology . 2009 ; 18 : 395 – 404 . Google Scholar CrossRef Search ADS PubMed 37. Hsu HT , Dodd MJ , Guo SE , Lee KA , Hwang SL , Lai YH . Predictors of exercise frequency in breast cancer survivors in Taiwan . J Clin Nurs . 2011 ; 20 : 1923 – 1935 . Google Scholar CrossRef Search ADS PubMed 38. Leach CR , Troeschel AN , Wiatrek D , et al. Preparedness and cancer-related symptom management among cancer survivors in the first year post-treatment . Ann Behav Med . 2017 ; 51 : 587 – 598 . Google Scholar CrossRef Search ADS PubMed 39. Fillenbaum GG , Smyer MA . The development, validity, and reliability of the OARS multidimensional functional assessment questionnaire . J Gerontol . 1981 ; 36 : 428 – 434 . Google Scholar CrossRef Search ADS PubMed 40. Portenoy RK , Thaler HT , Kornblith AB , et al. The memorial symptom assessment scale: An instrument for the evaluation of symptom prevalence, characteristics and distress . Eur J Cancer . 1994 ; 30A : 1326 – 1336 . Google Scholar CrossRef Search ADS PubMed 41. Lorig KR , Sobel DS , Ritter PL , Laurent D , Hobbs M . Effect of a self-management program on patients with chronic disease . Eff Clin Pract . 2001 ; 4 : 256 – 262 . Google Scholar PubMed 42. Basen-Engquist K , Carmack CL , Perkins H , et al. Design of the steps to health study of physical activity in survivors of endometrial cancer: Testing a social cognitive theory model . Psychol Sport Exerc . 2011 ; 12 : 27 – 35 . Google Scholar CrossRef Search ADS PubMed 43. Wójcicki TR , White SM , McAuley E . Assessing outcome expectations in older adults: The multidimensional outcome expectations for exercise scale . J Gerontol B Psychol Sci Soc Sci . 2009 ; 64 : 33 – 40 . Google Scholar CrossRef Search ADS PubMed 44. Patel AV , Jacobs EJ , Dudas DM , et al. The American cancer society’s cancer prevention study 3 (CPS-3): Recruitment, study design, and baseline characteristics . Cancer . 2017 ; 123 : 2014 – 2024 . Google Scholar CrossRef Search ADS PubMed 45. Troeschel AN , Leach CR , Shuval K , Stein KD , Patel AV . Prevalence and medico-demographic correlates of physical activity in cancer survivors during the “re-entry” phase . Prev Chronic Dis . (in press). 46. Ainsworth BE , Haskell WL , Herrmann SD , et al. 2011 Compendium of physical activities: A second update of codes and MET values . Med Sci Sports Exerc . 2011 ; 43 : 1575 – 1581 . Google Scholar CrossRef Search ADS PubMed 47. Hu L-t , Bentler PM . Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives . Struct Equ Modeling . 1999 ; 6 ( 1 ): 1 – 55 . Google Scholar CrossRef Search ADS 48. Brawner CA , Churilla JR , Keteyian SJ . Prevalence of physical activity is lower among individuals with chronic disease . Med Sci Sports Exerc . 2016 ; 48 : 1062 – 1067 . Google Scholar CrossRef Search ADS PubMed 49. Courneya KS , Segal RJ , Mackey JR , et al. Effects of aerobic and resistance exercise in breast cancer patients receiving adjuvant chemotherapy: A multicenter randomized controlled trial . J Clin Oncol . 2007 ; 25 : 4396 – 4404 . Google Scholar CrossRef Search ADS PubMed 50. Jacobucci R , Grimm KJ , McArdle JJ . Regularized structural equation modeling . Struct Equ Modeling . 2016 ; 23 : 555 – 566 . Google Scholar CrossRef Search ADS PubMed © Society of Behavioral Medicine 2018. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Annals of Behavioral Medicine Oxford University Press

Disentangling Efficacy and Expectations: A Prospective, Cross-lagged Panel Study of Cancer Survivors’ Physical Activity

Loading next page...
 
/lp/ou_press/disentangling-efficacy-and-expectations-a-prospective-cross-lagged-E0OpizXha9
Copyright
© Society of Behavioral Medicine 2018. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.
ISSN
0883-6612
eISSN
1532-4796
D.O.I.
10.1093/abm/kay022
Publisher site
See Article on Publisher Site

Abstract

Abstract Background Despite demonstrated utility of Bandura’s social cognitive theory for increasing physical activity among cancer survivors, the validity of the originally hypothesized relationships among self-efficacy, outcome expectations, and physical activity behavior continues to be debated. Purpose To explore the temporal ordering of outcome expectations and self-efficacy as they relate to moderate-to-vigorous physical activity behavior. Methods Longitudinal data from cancer survivors (N = 1,009) recently completing treatment were used to fit six cross-lagged panel models, including one parent model, one model representing originally hypothesized variable relationships, and four alternative models. All models contained covariates and used full information maximum likelihood and weighted least squares mean and variance adjusted estimation. Tests of equal fit between the parent model and alternative models were conducted. Results The model depicting Bandura’s originally hypothesized relationships showed no statistically significant relationship between outcome expectations and physical activity (p = .18), and was a worse fit to the data, compared with the parent model [Χ2 (1) = 5.92, p = .01]. An alternative model showed evidence of a reciprocal relationship between self-efficacy and outcome expectations, and was statistically equivalent to the parent model [Χ2(1) = 2.01, p = .16]. Conclusions This study provides evidence against Bandura’s theoretical assertions that (a) self-efficacy causes outcome expectations and not vice versa, and (b) outcome expectation has a direct effect on physical activity. Replication within population subgroups and for other health behaviors will determine whether the social cognitive theory needs modification. Future trials should test whether differential construct ordering results in clinically meaningful differences in physical activity behavior change. Physical activity, Neoplasm, Cancer survivor, Social cognitive theory Introduction Physical activity positively benefits the health of cancer survivors after treatment completion by lowering risk of cancer recurrence and cancer-specific mortality [1–3], and improving cancer-related biomarkers, such as insulin-like growth factor 1 [4]. In addition, studies have demonstrated that physical activity reduces fatigue [5–9] and improves quality of life [10–13] among post-treatment survivors. Yet, 66 per cent of cancer survivors fail to regularly meet physical activity recommendations [14] resulting in the need for efficacious, individually tailored, evidence-based behavioral interventions. Foundational to creating highly efficacious behavioral interventions are health behavior theories, which explicitly describe the mechanisms of physical activity behavior change. One such theory, the social cognitive theory [15], has demonstrated utility for explaining, and intervening to increase, physical activity among cancer survivors [16, 17]. Most research focuses on two social cognitive theory constructs: self-efficacy and outcome expectations. Specifically, self-efficacy is defined as perceived ability to perform a behavior [15, 18]. For example, a survivor may be asked how confident they are that they can “walk briskly for 20 min without stopping” or “exercise for 20 min at a level hard enough to cause a large increase in heart rate and breathing” [19]. Outcome expectations are defined as belief that a behavior will lead to certain outcomes [15]. For example, many cancer survivors believe that participating in physical activity will increase energy, maintain a healthy weight, and prevent cancer from coming back [20–22]. Self-efficacy is one of the strongest correlates and longitudinal predictors of successful adoption and maintenance of physical activity in the general population [17], as well as among cancer survivors [16]. Per Bandura’s original hypothesis, self-efficacy influences physical activity behavior directly, as well as indirectly through outcome expectations. Bandura furthermore asserts that the relationship between self-efficacy and outcome expectations is unidirectional such that self-efficacy causes outcome expectations, but outcome expectations do not cause self-efficacy [18, 23–26]. Research studies examining the social cognitive theory for physical activity have found strong empirical support for a positive relationship between self-efficacy and positive outcome expectations [27–31], as well as a positive relationship between self-efficacy and physical activity [17, 27, 29, 30]. The evidence for a direct effect of outcome expectations on physical activity, however, is equivocal [16, 17, 32]. Much of this research has relied on cross-sectional data, randomized trials with low power, or failed to conduct longitudinal mediation analyses. More recently, a properly powered randomized control trial found that a social cognitive theory intervention for breast cancer survivors was successful in significantly raising physical activity and self-efficacy, and significantly lowering negative outcome expectations. The intervention, however, had no effect on positive outcome expectations and provided no evidence that social cognitive theory constructs mediated the intervention’s effect on objectively measured physical activity [33, 34], as hypothesized by Bandura’s original model. Thus, further research examining social cognitive theory construct relationships with one another, and with physical activity over time among cancer survivors, is warranted. One important element of this research involves determining the exact ordering by which self-efficacy and outcome expectations influence physical activity behavior, a topic of continued debate within the theory-testing literature [35]. Contrary to Bandura’s original assertion, evidence exists which suggests that outcome expectations cause self-efficacy [35], and as noted above, evidence of a direct relationship between outcome expectations and physical activity is equivocal [17, 32]. Furthermore, when reviewing the literature examining these construct relationships among cancer survivors, there are methodological limitations prohibiting a firm conclusion for construct ordering for physical activity interventions, such as reliance on cross-sectional designs, low sample sizes, heterogeneity in measurement of physical activity, and social cognitive theory constructs, and relatively few studies focused on cancer survivorship [16, 17, 29, 36, 37]. Salient to this study is a general lack of consideration of alternative models in the literature to challenge the originally hypothesized ordering of constructs. Yet, due to the increasing demand for highly efficacious, theory-driven physical activity interventions delivered via automated technology (e.g., phone, website, and health apps) that increasingly rely on mathematical algorithms for real-time, individually tailored intervention delivery, there is growing need for research to determine the ordering of theoretical constructs, such as outcome expectations and self-efficacy. Clarification in this area can optimize intervention efficacy. Therefore, the purpose of this study is to explore the temporal ordering of outcome expectations and self-efficacy as they relate to moderate-to-vigorous physical activity behavior. Method This was a secondary data analysis of data obtained from the American Cancer Society’s Cancer Survivor Transition Study, a prospective observational study of breast, colorectal, and prostate cancer survivors enrolled within 12 months of completing active treatment. Institutional review board approval (Morehouse School of Medicine #253587) was obtained prior to data collection, and details of this dataset’s survey design, recruitment procedures, and methods were published in an earlier paper [38]. Briefly, participants were randomly selected from the American Cancer Society constituent database, which contains information of individuals who have initiated contact (80% self-referred, 20% provider referral) with the American Cancer Society’s National Cancer Information Center. Inclusion criteria consisted of a diagnosis of breast, colorectal, or prostate cancer; being at least 18 years of age; completing curative treatment < 13 months prior to taking baseline survey; and ability to read and communicate in English. This method yielded a final baseline sample of 1,188 breast, colorectal, and prostate cancer survivors. The follow-up survey was conducted on average 13.3 months after the baseline survey, with 869 individuals completing the follow-up survey (retention rate = 73%). For this study, participants were not eligible if they reported physical disability or a cancer recurrence, metastasis, or multiple cancer diagnoses at baseline or follow-up, as this may interfere with their ability to be physically active (Fig. 1). Final sample size for this study is 1,009 (84.9% of baseline Transition Study sample), with 679 (67.3%) individuals having complete data for all measures at both baseline and follow-up. Fig. 1. View largeDownload slide Flow diagram of participants included in analysis. Fig. 1. View largeDownload slide Flow diagram of participants included in analysis. Measures Sociodemographic factors Participants’ self-reported sociodemographic variables included age in years, gender, education (some college or more vs. high school graduate or less), and marital status (married/marriage like relationship vs. single/separated/divorced/widowed). Cancer-related factors A four-level variable combining cancer type and gender was created to prevent perfect overlap between covariates in multivariable models (female breast, female colorectal, male colorectal, and male prostate). Cancer stage was assessed with two questions asking participants if a doctor had ever told them their cancer spread to their lymph nodes and/or elsewhere in their body (coded local vs. regional or distant). Time since cancer treatment completion was calculated as the number of months since completion of most recent radiation, chemotherapy or surgery. Body mass index Self-reported height and weight were used to compute current body mass index (kg/m2), which was then used to create four body mass index categories using standard cut-off values for underweight (<18.5), normal weight (≥18.5 to <25.0), overweight (≥25.0 to <30), and obese (≥30). Due to a small number of underweight individuals in the analytic sample (n = 12), underweight and normal weight individuals were combined into a three-category variable. Physical comorbid conditions Participants’ number of physical comorbid conditions was assessed using the Older American Resources and Services Comorbidity Scale [39]. The scale assesses the presence of 11 conditions, including arthritis/rheumatism, glaucoma, emphysema/chronic bronchitis, high blood pressure, heart disease, circulation trouble in arms or legs, diabetes, stomach or intestinal problems, osteoporosis, chronic liver or kidney disease, and stroke, and sums them to create a comorbidities score ranging from 0 to 11. Cancer-related symptoms Total number of cancer-related symptoms experienced was calculated using a modified Memorial Symptoms Assessment Scale [38, 40]. A total score, ranging from 0 to 12, was created by summing any occurrence of 12 individual symptoms over the past month: fatigue, difficulty concentrating, difficulty sleeping, feeling nervous or worrying, problems with sexual interest or activity, shortness of breath, feeling sad, physical pain, incontinence, diarrhea/irritable bowel syndrome/problems holding bowels, lymphedema, and numbness or tingling. Self-efficacy Self-efficacy for performing regular physical activity was evaluated using a single item from a nine-item scale developed to assess confidence in performing key behaviors during the transition from active treatment. Participants were asked “How confident are you that you can currently exercise regularly (at least 3 to 4 times per week)?” with response options ranging from 1 (“not at all”) to 5 (“extremely”). The scale was adapted from the Stanford Chronic Disease Self-Efficacy Scale [41] with item modifications and additions informed by results from focus groups with cancer survivors. Outcome expectation Outcome expectations for physical activity were assessed using three items that asked participants how much they currently agree or disagree with the following statements: “exercising regularly (3–4 times per week) can help...” (i) “reduce my chances of getting cancer again”, (ii) “me return to the lifestyle and activities I enjoyed before my cancer treatment,” and (iii) “reduce physical and emotional side effects like fatigue, pain, anxiety, sadness, etc.” Items mirrored those used in other studies assessing this construct in related populations [42, 43]. Response options ranged from strongly disagree [1] to strongly agree [5]. Inter-item reliabilities were good for this study (Cronbach α = 0.83). Confirmatory factor analysis did not meet recommended fit indices at baseline [Χ2 (1) = 290.35, p < .001; CFI = 0.95; TLI = 0.86; RMSEA = 0.54 (95% CI: 0.49, 0.59)] or follow-up [Χ2 (1) = 289.31, p < .001; CFI = 0.94; TLI = 0.81; RMSEA = 0.65 (95% CI: 0.59, 0.72)], but invariance over time was evident [Χ2 (5) = 5.11, p = .40; CFI = 1.00; TLI = 1.00; RMSEA = 0.01 (95% CI: 0.00, 0.04)]. Therefore, a mean score was calculated for those respondents who answered at least two of the three items. Physical activity Moderate-to-vigorous physical activity was assessed using a 15-item scale from the American Cancer Society’s Cancer Prevention Study-3, an epidemiological cohort study surveying cancer risk behaviors of over 300,000 participants [44, 45]. Participants were asked to estimate hours per week spent performing a variety of common physical activities. Response options included “None,” “Less than 1,” “1 to 2,” “3,” “4 to 6,” and “7 or more.” Hour values of 0, 0.5,1.5, 3.0, 5.0, and 7.0, respectively, were multiplied by metabolic equivalent of task values from the 2011 Compendium of Physical Activities [46] and summed to create a metabolic equivalent hours per week score. At baseline, activities included walking, jogging, running, bicycling, tennis/racquetball, lap swimming, aerobics class (e.g., step and kickboxing), aerobic machines (e.g., elliptical and rowing), sports activities, dancing, and other aerobic recreation (e.g., hiking) [45]. At follow-up, some activities were assessed with more detail. Specifically, bicycling was separated into biking for leisure or bicycling vigorously, tennis was separated into singles and doubles, and sports participation was separated into moderate or vigorous sports. Statistical Analysis Descriptive analysis Means, standard deviations, or frequencies were computed, as appropriate, for each study variable and covariate. Descriptive statistics and correlations were examined for evidence of non-normality and multicollinearity. Path analysis All models tested were based on theoretical premise and were created specifically to test the hypothesized relationships among self-efficacy, outcome expectations, and moderate-to-vigorous physical activity. For all path models, non-normality in moderate-to-vigorous physical activity was addressed by using a square root transformation. Covariates included age, comorbidity count, symptom count, body mass index category, marital/cohabitation status, gender by cancer type, cancer stage, and months since treatment. Because self-efficacy was measured with a single-item, it was modeled as an ordinal variable and, accordingly, the weighted least squares mean and variance adjusted estimator was used. Full information maximum likelihood was used to address missing data. Commonly accepted cutoffs for the model chi-square (p > 0.05), root mean square error of approximation (≤.06), comparative fit index (≥0.95), and Tucker–Lewis index (≥0.95) indices were used to evaluate model fit [47]. In the first path analysis (Fig. 2A; Mplus 7.4), all autocorrelations were included because measures of the same phenomena within subjects are expected to be highly correlated over time. This is especially true for physical activity behavior, where past behavior is the best predictor of future behavior, and failure to include the autocorrelation leads to overestimation of construct-behavior relationships [17]. Paths leading from baseline moderate-to-vigorous physical activity to follow-up self-efficacy and from baseline moderate-to-vigorous physical activity to follow-up outcome expectations were set to zero because, according to behavioral theory, behavior change is considered an outcome (not a predictor) of psychosocial processes. This model serves as a parent model for both the Bandura (Fig. 2B) and alternative models (Fig. 2C), allowing chi-square difference tests of the equal fit hypothesis. Fig. 2. View largeDownload slide Three path models conducted to test hypothesized relationships among Time 1 (T1) and Time 2 (T2) outcome expectations (OE), self-efficacy (SE), and moderate-to-vigorous physical activity (MVPA). Fig. 2. View largeDownload slide Three path models conducted to test hypothesized relationships among Time 1 (T1) and Time 2 (T2) outcome expectations (OE), self-efficacy (SE), and moderate-to-vigorous physical activity (MVPA). Second, a path model representing Bandura’s original hypothesis that self-efficacy directly leads to outcome expectations, and outcome expectations directly lead to physical activity was conducted (Fig. 2B). The path from baseline outcome expectations to follow-up self-efficacy was set to zero to represent Bandura’s assertion that outcome expectations cannot influence self-efficacy. Subsequently, three alternative path models were tested, each representing theoretical alternatives to Bandura’s original hypotheses regarding the relationships among self-efficacy, outcome expectations, and behavior. Only one alternative model demonstrated acceptable fit statistics (Fig. 2C) and is presented in the published manuscript. Data for the two path models without acceptable fit statistics are presented in Supplementary Material on the journal website. In the alternative model with acceptable fit, Bandura’s assertion that baseline self-efficacy will influence follow-up outcome expectations, and not vice versa, is directly tested by hypothesizing a reciprocal relationship between the two constructs over time. Accordingly, the path from baseline outcome expectations to follow-up self-efficacy was freely estimated. Furthermore, the hypothesized relationship between baseline outcome expectations and follow-up moderate-to-vigorous physical activity was set to zero to directly test the assertion that outcome expectations has a direct effect on physical activity—a pathway necessary for Bandura’s assertion that outcome expectations mediate the relationship between self-efficacy and behavior. Setting this path to zero also reflects previous literature that outcome expectations do not have a significant, direct effect on physical activity [17, 32, 35]. Sensitivity analyses To explore possible bias due to missing data, chi-square, t-tests, and one-way analysis of variance were used to compare individuals with and without missing data for all covariates, social cognitive theory constructs, and physical activity behavior. Additionally, each path analysis was conducted using full information maximum likelihood (n = 1,009) and complete cases only (n = 679). Finally, additional structural equation analyses were conducted to model the measurement error. Like the path analyses, these models were conducted using the weighted least squares mean and variance adjusted estimator, full information maximum likelihood (n = 1,009) and using complete cases only (n = 679). Results Descriptive statistics for the study sample (N = 1,009) are presented in Table 1. Comparisons of individuals with (n = 330) and without (n = 679) missing data at follow-up showed that individuals with complete data more likely to be married/cohabitating (Χ2 = 7.71; df = 1; p < .01), had higher baseline self-efficacy (Χ2 =20.12; df = 4; p < .01), and higher baseline outcome expectations (t = 3.26; df = 1001; p < .01). Although these differences were statistically significant, size of the difference was small for marital status and self-efficacy (Cramer’s V < 0.15) and were less than half a standard deviation for outcome expectations. Additionally, correlations were not indicative of multicollinearity among study variables (Table 2). The unstandardized and standardized path coefficients for the parent model, Bandura model, and the alternative model with acceptable fit statistics are presented (Tables 3 and 4). Table 1 Descriptive statistics for study sample (N = 1,009) Characteristic N (%) Missing n (%) Cancer type/gender 0  Female—breast 359 (35.6)  Male—colorectal 228 (22.6)  Female—colorectal 130 (12.9)  Male—prostate 292 (28.9) Stage 0  Local 630 (62.4)  Regional/distant 379 (37.6) Education 0  ≤High school grad 342 (33.9)  Some college or more 667 (66.1) Marital status 0  Married/cohabitating 712 (70.6)  Not married/cohabitating 297 (29.4) Body mass index category 0  Normal/underweight 288 (28.5)  Overweight 345 (34.2)  Obese 376 (37.3) Time 1 self-efficacy 6  Not at all 116 (11.5)  A little bit 216 (21.4)  Moderately 270 (26.8)  Quite a bit 214 (21.2)  Extremely 187 (18.5) Time 2 self-efficacy 294 (29.1)  Not at all 81 (8.0)  A little bit 152 (15.1)  Moderately 182 (18.0)  Quite a bit 155 (15.4)  Extremely 145 (14.4) Mean (SD) Missing n (%) Age (years) 61.94 (11.00) 0 Months since treatment 7.84 (3.28) 0 Number of symptoms 7.06 (3.25) 0 Number of comorbidities 1.82 (1.56) 0 Time 1 moderate-to-vigorous physical activity 18.43 (21.02) 11 (1.1) Time 2 moderate-to-vigorous physical activity 18.76 (22.00) 301 (29.8) Time 1 outcome expectations 3.97 (0.74) 6 (0.6) Time 2 outcome expectations 4.00 (0.73) 295 (29.2) Characteristic N (%) Missing n (%) Cancer type/gender 0  Female—breast 359 (35.6)  Male—colorectal 228 (22.6)  Female—colorectal 130 (12.9)  Male—prostate 292 (28.9) Stage 0  Local 630 (62.4)  Regional/distant 379 (37.6) Education 0  ≤High school grad 342 (33.9)  Some college or more 667 (66.1) Marital status 0  Married/cohabitating 712 (70.6)  Not married/cohabitating 297 (29.4) Body mass index category 0  Normal/underweight 288 (28.5)  Overweight 345 (34.2)  Obese 376 (37.3) Time 1 self-efficacy 6  Not at all 116 (11.5)  A little bit 216 (21.4)  Moderately 270 (26.8)  Quite a bit 214 (21.2)  Extremely 187 (18.5) Time 2 self-efficacy 294 (29.1)  Not at all 81 (8.0)  A little bit 152 (15.1)  Moderately 182 (18.0)  Quite a bit 155 (15.4)  Extremely 145 (14.4) Mean (SD) Missing n (%) Age (years) 61.94 (11.00) 0 Months since treatment 7.84 (3.28) 0 Number of symptoms 7.06 (3.25) 0 Number of comorbidities 1.82 (1.56) 0 Time 1 moderate-to-vigorous physical activity 18.43 (21.02) 11 (1.1) Time 2 moderate-to-vigorous physical activity 18.76 (22.00) 301 (29.8) Time 1 outcome expectations 3.97 (0.74) 6 (0.6) Time 2 outcome expectations 4.00 (0.73) 295 (29.2) View Large Table 1 Descriptive statistics for study sample (N = 1,009) Characteristic N (%) Missing n (%) Cancer type/gender 0  Female—breast 359 (35.6)  Male—colorectal 228 (22.6)  Female—colorectal 130 (12.9)  Male—prostate 292 (28.9) Stage 0  Local 630 (62.4)  Regional/distant 379 (37.6) Education 0  ≤High school grad 342 (33.9)  Some college or more 667 (66.1) Marital status 0  Married/cohabitating 712 (70.6)  Not married/cohabitating 297 (29.4) Body mass index category 0  Normal/underweight 288 (28.5)  Overweight 345 (34.2)  Obese 376 (37.3) Time 1 self-efficacy 6  Not at all 116 (11.5)  A little bit 216 (21.4)  Moderately 270 (26.8)  Quite a bit 214 (21.2)  Extremely 187 (18.5) Time 2 self-efficacy 294 (29.1)  Not at all 81 (8.0)  A little bit 152 (15.1)  Moderately 182 (18.0)  Quite a bit 155 (15.4)  Extremely 145 (14.4) Mean (SD) Missing n (%) Age (years) 61.94 (11.00) 0 Months since treatment 7.84 (3.28) 0 Number of symptoms 7.06 (3.25) 0 Number of comorbidities 1.82 (1.56) 0 Time 1 moderate-to-vigorous physical activity 18.43 (21.02) 11 (1.1) Time 2 moderate-to-vigorous physical activity 18.76 (22.00) 301 (29.8) Time 1 outcome expectations 3.97 (0.74) 6 (0.6) Time 2 outcome expectations 4.00 (0.73) 295 (29.2) Characteristic N (%) Missing n (%) Cancer type/gender 0  Female—breast 359 (35.6)  Male—colorectal 228 (22.6)  Female—colorectal 130 (12.9)  Male—prostate 292 (28.9) Stage 0  Local 630 (62.4)  Regional/distant 379 (37.6) Education 0  ≤High school grad 342 (33.9)  Some college or more 667 (66.1) Marital status 0  Married/cohabitating 712 (70.6)  Not married/cohabitating 297 (29.4) Body mass index category 0  Normal/underweight 288 (28.5)  Overweight 345 (34.2)  Obese 376 (37.3) Time 1 self-efficacy 6  Not at all 116 (11.5)  A little bit 216 (21.4)  Moderately 270 (26.8)  Quite a bit 214 (21.2)  Extremely 187 (18.5) Time 2 self-efficacy 294 (29.1)  Not at all 81 (8.0)  A little bit 152 (15.1)  Moderately 182 (18.0)  Quite a bit 155 (15.4)  Extremely 145 (14.4) Mean (SD) Missing n (%) Age (years) 61.94 (11.00) 0 Months since treatment 7.84 (3.28) 0 Number of symptoms 7.06 (3.25) 0 Number of comorbidities 1.82 (1.56) 0 Time 1 moderate-to-vigorous physical activity 18.43 (21.02) 11 (1.1) Time 2 moderate-to-vigorous physical activity 18.76 (22.00) 301 (29.8) Time 1 outcome expectations 3.97 (0.74) 6 (0.6) Time 2 outcome expectations 4.00 (0.73) 295 (29.2) View Large Table 2 Correlation matrix among self-efficacy, outcome expectations, and moderate-to-vigorous physical activity over time 1 2 3 4 5 1 Time 1 moderate-to-vigorous physical activity 2 Time 2 moderate-to-vigorous physical activity 0.563 3 Time 1 outcome expectations 0.247 0.225 4 Time 2 outcome expectations 0.218 0.213 0.556 5 Time 1 self-efficacy 0.504 0.404 0.395 0.302 6 Time 2 self-efficacy 0.374 0.517 0.322 0.372 0.631 1 2 3 4 5 1 Time 1 moderate-to-vigorous physical activity 2 Time 2 moderate-to-vigorous physical activity 0.563 3 Time 1 outcome expectations 0.247 0.225 4 Time 2 outcome expectations 0.218 0.213 0.556 5 Time 1 self-efficacy 0.504 0.404 0.395 0.302 6 Time 2 self-efficacy 0.374 0.517 0.322 0.372 0.631 Correlations reported in this table are from the saturated model, with covariates (N = 1,009, weighted least squares mean and variance adjusted estimation, and full information maximum likelihood). View Large Table 2 Correlation matrix among self-efficacy, outcome expectations, and moderate-to-vigorous physical activity over time 1 2 3 4 5 1 Time 1 moderate-to-vigorous physical activity 2 Time 2 moderate-to-vigorous physical activity 0.563 3 Time 1 outcome expectations 0.247 0.225 4 Time 2 outcome expectations 0.218 0.213 0.556 5 Time 1 self-efficacy 0.504 0.404 0.395 0.302 6 Time 2 self-efficacy 0.374 0.517 0.322 0.372 0.631 1 2 3 4 5 1 Time 1 moderate-to-vigorous physical activity 2 Time 2 moderate-to-vigorous physical activity 0.563 3 Time 1 outcome expectations 0.247 0.225 4 Time 2 outcome expectations 0.218 0.213 0.556 5 Time 1 self-efficacy 0.504 0.404 0.395 0.302 6 Time 2 self-efficacy 0.374 0.517 0.322 0.372 0.631 Correlations reported in this table are from the saturated model, with covariates (N = 1,009, weighted least squares mean and variance adjusted estimation, and full information maximum likelihood). View Large Table 3 Results of parent path model Parent model Unstandardized Std est Est (s.e.) 95% CI p-value Paths of interest  T2OE on T1OE 0.51 (0.03) [0.45, 0.58] <.001 0.52  T2OE on T1MVPA @0 @0 @0 @0  T2OE on T1SE 0.08 (0.03) [0.03, 0.13] <.01 0.12  T2SE on T1MVPA @0 @0 @0 @0  T2SE on T1SE 0.61 (0.03) [0.55, 0.66] <.001 0.61  T2SE on T1OE 0.13 (0.05) [0.03, 0.23] .01 0.09  T2MVPA on T1OE 0.15 (0.10) [−0.06, 0.35] .15 0.05  T2MVPA on T1MVPA 0.50 (0.03) [0.43, 0.56] <.001 0.48  T2MVPA on T1SE 0.29 (0.09) [0.11, 0.47] <.01 0.14 Covariances  T1OE with T1MVPA 0.38 (0.05) [0.29, 0.47] <.001 0.26  T1OE with T1SE 0.27 (0.02) [0.23, 0.31] <.001 0.39  T1SE with T1MVPA 1.09 (0.06) [0.97, 1.21] <.001 0.52  T2OE with T2MVPA 0.08 (0.04) [0.00, 0.16] .05 0.08  T2OE with T2SE 0.09 (0.02) [0.06, 0.13] <.001 0.21  T2SE with T2MVPA 0.55 (0.06) [0.43, 0.66] <.001 0.41 Parent model Unstandardized Std est Est (s.e.) 95% CI p-value Paths of interest  T2OE on T1OE 0.51 (0.03) [0.45, 0.58] <.001 0.52  T2OE on T1MVPA @0 @0 @0 @0  T2OE on T1SE 0.08 (0.03) [0.03, 0.13] <.01 0.12  T2SE on T1MVPA @0 @0 @0 @0  T2SE on T1SE 0.61 (0.03) [0.55, 0.66] <.001 0.61  T2SE on T1OE 0.13 (0.05) [0.03, 0.23] .01 0.09  T2MVPA on T1OE 0.15 (0.10) [−0.06, 0.35] .15 0.05  T2MVPA on T1MVPA 0.50 (0.03) [0.43, 0.56] <.001 0.48  T2MVPA on T1SE 0.29 (0.09) [0.11, 0.47] <.01 0.14 Covariances  T1OE with T1MVPA 0.38 (0.05) [0.29, 0.47] <.001 0.26  T1OE with T1SE 0.27 (0.02) [0.23, 0.31] <.001 0.39  T1SE with T1MVPA 1.09 (0.06) [0.97, 1.21] <.001 0.52  T2OE with T2MVPA 0.08 (0.04) [0.00, 0.16] .05 0.08  T2OE with T2SE 0.09 (0.02) [0.06, 0.13] <.001 0.21  T2SE with T2MVPA 0.55 (0.06) [0.43, 0.66] <.001 0.41 Relationships between covariates and self-efficacy (SE), outcome expectations (OE), and moderate-to-vigorous physical activity (MVPA) were freely estimated for both time points (Supplementary Material). MVPA was transformed using a square root transformation to adjust for non-normality, and weighted least squares mean and variance adjusted estimation, and full information maximum likelihood were used. Est estimate; s.e. standard error; 95% CI 95% confidence interval; Std est standardized estimate; T1 time point 1; T2 timepoint 2. View Large Table 3 Results of parent path model Parent model Unstandardized Std est Est (s.e.) 95% CI p-value Paths of interest  T2OE on T1OE 0.51 (0.03) [0.45, 0.58] <.001 0.52  T2OE on T1MVPA @0 @0 @0 @0  T2OE on T1SE 0.08 (0.03) [0.03, 0.13] <.01 0.12  T2SE on T1MVPA @0 @0 @0 @0  T2SE on T1SE 0.61 (0.03) [0.55, 0.66] <.001 0.61  T2SE on T1OE 0.13 (0.05) [0.03, 0.23] .01 0.09  T2MVPA on T1OE 0.15 (0.10) [−0.06, 0.35] .15 0.05  T2MVPA on T1MVPA 0.50 (0.03) [0.43, 0.56] <.001 0.48  T2MVPA on T1SE 0.29 (0.09) [0.11, 0.47] <.01 0.14 Covariances  T1OE with T1MVPA 0.38 (0.05) [0.29, 0.47] <.001 0.26  T1OE with T1SE 0.27 (0.02) [0.23, 0.31] <.001 0.39  T1SE with T1MVPA 1.09 (0.06) [0.97, 1.21] <.001 0.52  T2OE with T2MVPA 0.08 (0.04) [0.00, 0.16] .05 0.08  T2OE with T2SE 0.09 (0.02) [0.06, 0.13] <.001 0.21  T2SE with T2MVPA 0.55 (0.06) [0.43, 0.66] <.001 0.41 Parent model Unstandardized Std est Est (s.e.) 95% CI p-value Paths of interest  T2OE on T1OE 0.51 (0.03) [0.45, 0.58] <.001 0.52  T2OE on T1MVPA @0 @0 @0 @0  T2OE on T1SE 0.08 (0.03) [0.03, 0.13] <.01 0.12  T2SE on T1MVPA @0 @0 @0 @0  T2SE on T1SE 0.61 (0.03) [0.55, 0.66] <.001 0.61  T2SE on T1OE 0.13 (0.05) [0.03, 0.23] .01 0.09  T2MVPA on T1OE 0.15 (0.10) [−0.06, 0.35] .15 0.05  T2MVPA on T1MVPA 0.50 (0.03) [0.43, 0.56] <.001 0.48  T2MVPA on T1SE 0.29 (0.09) [0.11, 0.47] <.01 0.14 Covariances  T1OE with T1MVPA 0.38 (0.05) [0.29, 0.47] <.001 0.26  T1OE with T1SE 0.27 (0.02) [0.23, 0.31] <.001 0.39  T1SE with T1MVPA 1.09 (0.06) [0.97, 1.21] <.001 0.52  T2OE with T2MVPA 0.08 (0.04) [0.00, 0.16] .05 0.08  T2OE with T2SE 0.09 (0.02) [0.06, 0.13] <.001 0.21  T2SE with T2MVPA 0.55 (0.06) [0.43, 0.66] <.001 0.41 Relationships between covariates and self-efficacy (SE), outcome expectations (OE), and moderate-to-vigorous physical activity (MVPA) were freely estimated for both time points (Supplementary Material). MVPA was transformed using a square root transformation to adjust for non-normality, and weighted least squares mean and variance adjusted estimation, and full information maximum likelihood were used. Est estimate; s.e. standard error; 95% CI 95% confidence interval; Std est standardized estimate; T1 time point 1; T2 timepoint 2. View Large Table 4 Results of the Bandura and alternative path model Bandura model Alternative model Unstandardized Std est Unstandardized Std est Est (s.e.) 95% CI p-value Est (s.e.) 95% CI p-value Paths of interest  T2OE on T1OE 0.52 (0.03) [0.45, 0.58] <.001 0.52 0.52 (0.03) [0.45, 0.58] <.001 0.52  T2OE on T1MVPA @0 @0 @0 @0 @0 @0 @0 @0  T2OE on T1SE 0.07 (0.03) [0.02, 0.12] <.01 0.11 0.08 (0.03) [0.03, 0.12] <.01 0.11  T2SE on T1MVPA @0 @0 @0 @0 @0 @0 @0 @0  T2SE on T1SE 0.65 (0.02) [0.60, 0.70] <.001 0.65 0.61 (0.03) [0.55, 0.66] <.001 0.61  T2SE on T1OE @0 @0 @0 @0 0.13 (0.05) [0.03, 0.23] .01 0.09  T2MVPA on T1OE 0.14 (0.10) [−0.06, 0.34] .18 0.05 @0 @0 @0 @0  T2MVPA on T1MVPA 0.50 (0.03) [0.43, 0.56] <.001 0.48 0.50 (0.03) [0.43, 0.56] <.001 0.48  T2MVPA on T1SE 0.29 (0.09) [0.11, 0.48] <.01 0.14 0.36 (0.09) [0.19, 0.53] <.001 0.17 Covariances  T1OE with T1MVPA 0.38 (0.05) [0.29, 0.47] <.001 0.26 0.40 (0.05) [0.30, 0.49] <.001 0.27  T1OE with T1SE 0.29 (0.02) [0.25, 0.33] <.001 0.41 0.27 (0.02) [0.23, 0.31] <.001 0.39  T1SE with T1MVPA 1.09 (0.06) [0.98, 1.21] <.001 0.52 1.09 (0.06) [0.97, 1.20] <.001 0.52  T2OE with T2MVPA 0.08 (0.04) [0.01, 0.16] .03 0.08 0.10 (0.04) [0.02, 0.18] .01 0.10  T2OE with T2SE 0.12 (0.02) [0.08, 0.16] <.001 0.27 0.09 (0.02) [0.06, 0.13] <.001 0.22  T2SE with T2MVPA 0.55 (0.06) [0.44, 0.67] <.001 0.41 0.54 (0.06) [0.43, 0.66] <.001 0.40 Bandura model Alternative model Unstandardized Std est Unstandardized Std est Est (s.e.) 95% CI p-value Est (s.e.) 95% CI p-value Paths of interest  T2OE on T1OE 0.52 (0.03) [0.45, 0.58] <.001 0.52 0.52 (0.03) [0.45, 0.58] <.001 0.52  T2OE on T1MVPA @0 @0 @0 @0 @0 @0 @0 @0  T2OE on T1SE 0.07 (0.03) [0.02, 0.12] <.01 0.11 0.08 (0.03) [0.03, 0.12] <.01 0.11  T2SE on T1MVPA @0 @0 @0 @0 @0 @0 @0 @0  T2SE on T1SE 0.65 (0.02) [0.60, 0.70] <.001 0.65 0.61 (0.03) [0.55, 0.66] <.001 0.61  T2SE on T1OE @0 @0 @0 @0 0.13 (0.05) [0.03, 0.23] .01 0.09  T2MVPA on T1OE 0.14 (0.10) [−0.06, 0.34] .18 0.05 @0 @0 @0 @0  T2MVPA on T1MVPA 0.50 (0.03) [0.43, 0.56] <.001 0.48 0.50 (0.03) [0.43, 0.56] <.001 0.48  T2MVPA on T1SE 0.29 (0.09) [0.11, 0.48] <.01 0.14 0.36 (0.09) [0.19, 0.53] <.001 0.17 Covariances  T1OE with T1MVPA 0.38 (0.05) [0.29, 0.47] <.001 0.26 0.40 (0.05) [0.30, 0.49] <.001 0.27  T1OE with T1SE 0.29 (0.02) [0.25, 0.33] <.001 0.41 0.27 (0.02) [0.23, 0.31] <.001 0.39  T1SE with T1MVPA 1.09 (0.06) [0.98, 1.21] <.001 0.52 1.09 (0.06) [0.97, 1.20] <.001 0.52  T2OE with T2MVPA 0.08 (0.04) [0.01, 0.16] .03 0.08 0.10 (0.04) [0.02, 0.18] .01 0.10  T2OE with T2SE 0.12 (0.02) [0.08, 0.16] <.001 0.27 0.09 (0.02) [0.06, 0.13] <.001 0.22  T2SE with T2MVPA 0.55 (0.06) [0.44, 0.67] <.001 0.41 0.54 (0.06) [0.43, 0.66] <.001 0.40 Relationships between covariates and self-efficacy (SE), outcome expectations (OE), and moderate-to-vigorous physical activity (MVPA) were freely estimated for both time points (Supplementary Material). MVPA was transformed using a square root transformation to adjust for non-normality, and weighted least squares mean and variance adjusted estimation, and full information maximum likelihood were used. Est estimate; s.e. standard error; 95% CI 95% confidence interval; Std est standardized estimate; T1 time point 1; T2 timepoint 2. View Large Table 4 Results of the Bandura and alternative path model Bandura model Alternative model Unstandardized Std est Unstandardized Std est Est (s.e.) 95% CI p-value Est (s.e.) 95% CI p-value Paths of interest  T2OE on T1OE 0.52 (0.03) [0.45, 0.58] <.001 0.52 0.52 (0.03) [0.45, 0.58] <.001 0.52  T2OE on T1MVPA @0 @0 @0 @0 @0 @0 @0 @0  T2OE on T1SE 0.07 (0.03) [0.02, 0.12] <.01 0.11 0.08 (0.03) [0.03, 0.12] <.01 0.11  T2SE on T1MVPA @0 @0 @0 @0 @0 @0 @0 @0  T2SE on T1SE 0.65 (0.02) [0.60, 0.70] <.001 0.65 0.61 (0.03) [0.55, 0.66] <.001 0.61  T2SE on T1OE @0 @0 @0 @0 0.13 (0.05) [0.03, 0.23] .01 0.09  T2MVPA on T1OE 0.14 (0.10) [−0.06, 0.34] .18 0.05 @0 @0 @0 @0  T2MVPA on T1MVPA 0.50 (0.03) [0.43, 0.56] <.001 0.48 0.50 (0.03) [0.43, 0.56] <.001 0.48  T2MVPA on T1SE 0.29 (0.09) [0.11, 0.48] <.01 0.14 0.36 (0.09) [0.19, 0.53] <.001 0.17 Covariances  T1OE with T1MVPA 0.38 (0.05) [0.29, 0.47] <.001 0.26 0.40 (0.05) [0.30, 0.49] <.001 0.27  T1OE with T1SE 0.29 (0.02) [0.25, 0.33] <.001 0.41 0.27 (0.02) [0.23, 0.31] <.001 0.39  T1SE with T1MVPA 1.09 (0.06) [0.98, 1.21] <.001 0.52 1.09 (0.06) [0.97, 1.20] <.001 0.52  T2OE with T2MVPA 0.08 (0.04) [0.01, 0.16] .03 0.08 0.10 (0.04) [0.02, 0.18] .01 0.10  T2OE with T2SE 0.12 (0.02) [0.08, 0.16] <.001 0.27 0.09 (0.02) [0.06, 0.13] <.001 0.22  T2SE with T2MVPA 0.55 (0.06) [0.44, 0.67] <.001 0.41 0.54 (0.06) [0.43, 0.66] <.001 0.40 Bandura model Alternative model Unstandardized Std est Unstandardized Std est Est (s.e.) 95% CI p-value Est (s.e.) 95% CI p-value Paths of interest  T2OE on T1OE 0.52 (0.03) [0.45, 0.58] <.001 0.52 0.52 (0.03) [0.45, 0.58] <.001 0.52  T2OE on T1MVPA @0 @0 @0 @0 @0 @0 @0 @0  T2OE on T1SE 0.07 (0.03) [0.02, 0.12] <.01 0.11 0.08 (0.03) [0.03, 0.12] <.01 0.11  T2SE on T1MVPA @0 @0 @0 @0 @0 @0 @0 @0  T2SE on T1SE 0.65 (0.02) [0.60, 0.70] <.001 0.65 0.61 (0.03) [0.55, 0.66] <.001 0.61  T2SE on T1OE @0 @0 @0 @0 0.13 (0.05) [0.03, 0.23] .01 0.09  T2MVPA on T1OE 0.14 (0.10) [−0.06, 0.34] .18 0.05 @0 @0 @0 @0  T2MVPA on T1MVPA 0.50 (0.03) [0.43, 0.56] <.001 0.48 0.50 (0.03) [0.43, 0.56] <.001 0.48  T2MVPA on T1SE 0.29 (0.09) [0.11, 0.48] <.01 0.14 0.36 (0.09) [0.19, 0.53] <.001 0.17 Covariances  T1OE with T1MVPA 0.38 (0.05) [0.29, 0.47] <.001 0.26 0.40 (0.05) [0.30, 0.49] <.001 0.27  T1OE with T1SE 0.29 (0.02) [0.25, 0.33] <.001 0.41 0.27 (0.02) [0.23, 0.31] <.001 0.39  T1SE with T1MVPA 1.09 (0.06) [0.98, 1.21] <.001 0.52 1.09 (0.06) [0.97, 1.20] <.001 0.52  T2OE with T2MVPA 0.08 (0.04) [0.01, 0.16] .03 0.08 0.10 (0.04) [0.02, 0.18] .01 0.10  T2OE with T2SE 0.12 (0.02) [0.08, 0.16] <.001 0.27 0.09 (0.02) [0.06, 0.13] <.001 0.22  T2SE with T2MVPA 0.55 (0.06) [0.44, 0.67] <.001 0.41 0.54 (0.06) [0.43, 0.66] <.001 0.40 Relationships between covariates and self-efficacy (SE), outcome expectations (OE), and moderate-to-vigorous physical activity (MVPA) were freely estimated for both time points (Supplementary Material). MVPA was transformed using a square root transformation to adjust for non-normality, and weighted least squares mean and variance adjusted estimation, and full information maximum likelihood were used. Est estimate; s.e. standard error; 95% CI 95% confidence interval; Std est standardized estimate; T1 time point 1; T2 timepoint 2. View Large The parent model (Fig. 2A; Table 3) was a good fit to the data [Χ2 (2) = 5.22, p = .07; CFI = 0.99; TLI = 0.93; RMSEA = 0.04 (95% CI: 0.00, 0.08)] and accounted for 41, 39, and 52 per cent of variance in follow-up moderate-to-vigorous physical activity, outcome expectations, and self-efficacy, respectively. Standardized and unstandardized path coefficients, confidence intervals, and p-values are presented in Table 3. The path model representing Bandura’s hypothesis (Fig. 2B; Table 4) that self-efficacy causes outcome expectations, and not vice versa, did not meet recommended cutoffs for fit indices [Χ2 (3) = 11.49, p = .009; CFI = 0.99; TLI = 0.88; RMSEA = 0.05 (95% CI:0.02, 0.09)], indicating that this model is not a good fit to the data (Table 3). When directly comparing the parent model and the Bandura model, a Χ2 difference test resulted in the rejection of the equal fit hypothesis [Χ2 (1) = 5.92, p = .01], indicating a significant difference in model fit between the parent model and the model proposed by Bandura (i.e., the parent model is a better fitting model). Path coefficients show that the path from baseline outcome expectations to follow-up moderate-to-vigorous physical activity is not statistically significant (p = .18; Table 4). Finally, the path model examining the alternative hypothesis (Fig. 2C; Table 4) met recommended cutoffs for fit indices [Χ2 (3) = 7.22, p = .07; CFI = 0.99; TLI = 0.94; RMSEA = 0.04 (95% CI: 0.00, 0.07)], indicating that it is a good fit to the data. When directly comparing the parent model and the alternative model, the Χ2 difference test failed to reject the equal fit hypothesis [Χ2 (1) = 2.01, p = .16], indicating no significant differences in model fit between the parent model and the alternative model. Path coefficients show that the path from baseline outcome expectations to follow-up self-efficacy and the path from baseline self-efficacy to follow-up outcome expectations are both statistically significant (Table 4). Detailed results of sensitivity analyses are presented in Supplementary Material. Briefly, results of the structural equation models replicated the findings from the path analyses such that the parent model was a good fit to the data [Χ2 (71) = 88.61, p = .07; CFI = 0.99; TLI = 0.99; RMSEA = 0.02 (95% CI: 0.00, 0.02)]. Additionally, the Bandura model was not a good fit to the data [Χ2 (72) = 101.01, p = .01; CFI = 0.99; TLI = 0.99; RMSEA = 0.02 (95% CI: 0.01, 0.03)] and failed the equal fit hypothesis [Χ2 (1) = 6.06, p = .02]. Meanwhile, the alternative model was a good fit to the data [Χ2 (72) = 92.06, p = .06; CFI = 0.99; TLI = 0.99; RMSEA = 0.02 (95% CI: 0.00, 0.03)], and the chi-square test of equal fit showed that the alternative model and the parent model were statistically equivalent [Χ2 (1) = 2.49, p = .11]. Models without covariates were not a sufficient fit to the data [parent path model without covariates: [Χ2 (2) = 15.17, p < .001; CFI = 0.99; TLI = 0.92; RMSEA = 0.08 (95% CI: 0.05, 0.12)], indicating the need to incorporate covariates in the model to provide more accurate parameter estimates for social cognitive theory constructs. Finally, complete case (n = 679) path analyses yielded similar results to full information maximum likelihood path models (n = 1,009; results not presented). Discussion The current study examined, at two time points, the relationship between self-efficacy, outcome expectations, and physical activity among a cohort of breast, prostate, and colorectal cancer survivors who recently completed treatment. The parent model and the alterative model met recommended fit indices and like previous research [17, 29], accounted for 41, 39, and 52 per cent of variance in follow-up moderate-to-vigorous physical activity, outcome expectations, and self-efficacy, respectively. Upon examining the path coefficient of the alternative model, there is evidence of a reciprocal relationship between self-efficacy and outcome expectations over time, which is inconsistent with Bandura’s original assertion that outcome expectations do not predict self-efficacy. Furthermore, fixing the outcome expectations to physical activity path to zero is statistically defensible, which negates one of the pathways needed to complete Bandura’s proposed mediational pathway. The Bandura model explained similar levels of variance in follow-up moderate-to-vigorous physical activity (41%), outcome expectations (39%), and self-efficacy (52%), but did not meet fit indices. When examining the path coefficients of the Bandura model, baseline self-efficacy significantly influences follow-up outcome expectations, which is consistent with Bandura’s assertion. Inconsistent with Bandura’s assertion, however, is the lack of a significant effect of baseline outcome expectations on follow-up moderate-to-vigorous physical activity, again suggesting that Bandura’s proposed mediational pathway is uncertain. Taken together, the path models presented in this study fail to support Bandura’s original assertions within the social cognitive theory that (a) self-efficacy precedes outcome expectations and not vice versa, and (b) outcome expectations directly effects moderate-to-vigorous physical activity, a necessary relationship for Bandura’s proposed mediational pathway (self-efficacy → outcome expectations → moderate-to-vigorous physical activity). Instead, our data suggest that outcome expectations precede self-efficacy, or that there may be a reciprocal relationship between self-efficacy and outcome expectations (self-efficacy ↔ outcome expectations). Additionally, our data support previous reviews and commentaries [32, 35], suggesting that outcome expectations do not directly affect moderate-to-vigorous physical activity. If these findings were extrapolated to longitudinal data with three or more time points, we would hypothesize that self-efficacy mediates the relationship between outcome expectations and moderate-to-vigorous physical activity. Some limitations of this study should be noted. First, this was a secondary data analysis relying on self-reported data from a convenience sample of breast, colorectal, and prostate cancer survivors transitioning out of primary treatment. Most participants reported their race/ethnicity as non-Hispanic white/Caucasian, and most had received at least some college education. Therefore, results may be subject to self-report biases, and we cannot confidently generalize our results to individuals with other cancers, survivors undergoing active treatment, cancer survivors who are in advanced stage cancer or end-of-life care, minority groups, or less educated groups. Second, random selection from an American Cancer Society national call center constituent database may have produced a sample more likely to seek cancer-related and health-related information, which may be an indicator of overall motivation and activation for healthier lifestyle. Specifically, the call center reports that 80 per cent of the callers are self-referred and 20 per cent are referred by their physician [38], and therefore, our sample may not be representative of all cancer survivors. Third, the current study assessed theoretical constructs and behavior at two time points, an average of 13.3 months apart. Future research is needed with assessments at three or more time points to allow for testing of full mediation models. Additionally, although our timeframe is not inconsistent with prospective and intervention studies using 12 or 18 month time intervals [16, 17], or the mean changes in constructs in previous research [29, 33], it is important to think about how the length of time between assessments limits our ability to understand when changes in constructs occurred during this time, how many changes occurred, the magnitude of the changes that occurred, or the causal mechanisms of these changes. The common use of 6, 12, or 18 month assessment periods is a normative methodology, not based on the social cognitive theory, nor empirical research examining the optimal timeframe for studying these relationships. Thus, future research is needed to examine whether ordering of theoretical constructs is consistent or varies depending on length of time between assessments. This research would have implications for theory, but also implications for intervention delivery and expected intervention outcomes across different contexts (e.g., phone, face-to-face, and adaptive technology). Finally, regarding construct measurement, we did not include other constructs within the social cognitive theory, and therefore, this study is subject to missing variable bias. Furthermore, self-efficacy was assessed with a single item which limits the reliability and validity of the measure, although we did account for the ordinal nature of the item by using the weighted least squares mean and variance adjusted estimation. Outcome expectations was assessed using three items often used within the cancer population that, despite good inter-item correlations, did not meet confirmatory factor analysis cut-off criteria, and therefore, outcome expectations were modeled as an observed variable. Additional analyses incorporating measurement error using structural equation modeling produced similar overall results, with the alternative model meeting model fit indices and proving to be a superior model to Bandura’s original model (Supplementary Tables A–E). While a better measure of outcome expectations would be preferable, the additional structural equation models, in combination with the explicit testing of alternative models using the same data, suggest that our findings cannot be dismissed due to measurement error. Despite these limitations, this study has several advantages. First, because cancer survivors would gain positive health benefits from physical activity, but are less likely than the general adult population to meet physical activity recommendations [48], there is greater need to focus on this population. Second, we improved upon previous cross-sectional research by using a longitudinal, prospective survey design. Bias introduced by limiting analysis to only complete cases was avoided by using full information maximum likelihood methods. Finally, we have avoided confirmation bias by explicitly testing several alternatives to the original construct ordering suggested within the social cognitive theory. This study has implications for future behavioral research as well as applied public health practice regarding the use of the social cognitive theory as a framework for understanding and promoting physical activity behavior. Specifically, future research should seek to replicate these findings in different population subgroups (e.g., race/ethnicity, age, body mass index categories, by cancer type, and noncancer populations) and for other health behaviors (e.g., diet, stress management, and smoking cessation) with the goal of determining whether the social cognitive theory requires modification. These results should also be replicated across different time points in the cancer survivorship trajectory, specifically during treatment, where physical activity initiation and maintenance are associated with cancer treatment adherence [49], reduced cancer-related fatigue [4–9], and improvements in quality of life [10–13]. Replication studies are needed that use reliable and valid measures of self-efficacy and outcome expectations and assess additional social cognitive theory constructs to reduce measurement error and avoid missing variable bias. It seems particularly important to include negative outcome expectations [33], perceived barriers [33], and goal-setting [29, 33]. Mastery experience is a less commonly studied construct proposed to predict self-efficacy and promote its sustainability [15], and may account for variation in self-efficacy due to past physical activity experiences. Future studies may also benefit from emerging statistical techniques that seek to reduce model complexity [50]. Results of this research inform development of traditional in-person interventions, as well as interventions delivered via automated technology (e.g., phone, website, and health apps), to ensure that intervention components adhere to processes supported by effective theory. Specifically, an intervention approach that would strengthen outcome expectations over self-efficacy beliefs would not be supported. But acknowledging and supporting an interactional approach between these two variables or an ordered approach (outcome expectations → self-efficacy) would be supported. Furthermore, randomized control trials or comparative effectiveness studies should be designed to determine whether changing the ordering of outcome expectations and self-efficacy in intervention modules will result in clinically meaningful differences in physical activity outcomes. If successful, this would be a simple change in existing research and practice with little cost for implementation. In conclusion, most cancer survivors do not engage in sufficient moderate-to-vigorous physical activity to attain health benefits, and therefore, efficacious interventions are needed to increase the prevalence and long-term maintenance of moderate-to-vigorous physical activity in this population. This study advances our understanding of how the social cognitive theory constructs might be optimally applied to better explain, and ultimately, change moderate-to-vigorous physical activity behavior among cancer survivors. Studies using similar methodologies to replicate these findings, as well as randomized controlled trials and/or comparative effectiveness studies, are needed to systematically improve existing social cognitive theory interventions for physical activity among cancer survivors. Supplementary Material Supplementary material is available at Annals of Behavioral Medicine online. Compliance with Ethical Standards Authors’ Statement of Conflict of Interest and Adherence to Ethical Standards Authors Elizabeth Fallon, Robert Stephens, Bennett McDonald, and Corinne Leach acknowledge their employment at American Cancer Society, and declare no other conflicts of interest. Author Michael Diefenbach declares no conflict of interest. All procedures, including the informed consent process, were conducted in accordance with the ethical standards of the responsible committee on human experimentation (institutional and national) and with the Helsinki Declaration of 1975, as revised in 2000. Authors’ Contributions E.F. generated the paper concept. C.L. is the PI of the Transition Study, and was responsible for designing the survey and supervising the data collection and management process. M.D. served as an expert consultant for the Transition Study. B.M. led the data management process. E.F. led the data analysis, supported by R.S., and B.M. E.F. led the manuscript writing, supported by B.M. and R.S. All authors edited the manuscript for substantive content, and approved the manuscript for publication. Ethical Approval All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the Helsinki declaration or comparable ethical standards. Informed Consent Informed consent was obtained from all individual participants included in the study. Acknowledgments We thank the study participants, who demonstrated their leadership in the fight against cancer by giving their time and insight into the cancer survivorship experience. Without their support, this research and future application of what we have learned would not have been possible. Data for this manuscript came from the Transition Study, funded by the American Cancer Society. We acknowledge Alyssa Troeschel and Rhyan Vereen for their role in survey design, data collection, and cleaning and Dawn Wiatrek who served as co-PI in the Transition study design. References 1. Friedenreich CM , Wang Q , Neilson HK , Kopciuk KA , McGregor SE , Courneya KS . Physical activity and survival after prostate cancer . Eur Urol . 2016 ; 70 : 576 – 585 . Google Scholar CrossRef Search ADS PubMed 2. Lee IM , Wolin KY , Freeman SE , Sattlemair J , Sesso HD . Physical activity and survival after cancer diagnosis in men . J Phys Act Health . 2014 ; 11 : 85 – 90 . Google Scholar CrossRef Search ADS PubMed 3. Lahart IM , Metsios GS , Nevill AM , Carmichael AR . Physical activity, risk of death and recurrence in breast cancer survivors: A systematic review and meta-analysis of epidemiological studies . Acta Oncol . 2015 ; 54 : 635 – 654 . Google Scholar CrossRef Search ADS PubMed 4. Ballard-Barbash R , Friedenreich CM , Courneya KS , Siddiqi SM , McTiernan A , Alfano CM . Physical activity, biomarkers, and disease outcomes in cancer survivors: A systematic review . J Natl Cancer Inst . 2012 ; 104 : 815 – 840 . Google Scholar CrossRef Search ADS PubMed 5. Mustian KM , Alfano CM , Heckler C , et al. Comparison of pharmaceutical, psychological, and exercise treatments for cancer-related fatigue: A meta-analysis . JAMA Oncol . 2017 ; 3 : 961 – 968 . Google Scholar CrossRef Search ADS PubMed 6. Lipsett A , Barrett S , Haruna F , Mustian K , O’Donovan A . The impact of exercise during adjuvant radiotherapy for breast cancer on fatigue and quality of life: A systematic review and meta-analysis . Breast . 2017 ; 32 : 144 – 155 . Google Scholar CrossRef Search ADS PubMed 7. Dennett AM , Peiris CL , Shields N , Prendergast LA , Taylor NF . Moderate-intensity exercise reduces fatigue and improves mobility in cancer survivors: A systematic review and meta-regression . J Physiother . 2016 ; 62 : 68 – 82 . Google Scholar CrossRef Search ADS PubMed 8. Meneses-Echávez JF , González-Jiménez E , Ramírez-Vélez R . Supervised exercise reduces cancer-related fatigue: A systematic review . J Physiother . 2015 ; 61 : 3 – 9 . Google Scholar CrossRef Search ADS PubMed 9. Meneses-Echávez JF , González-Jiménez E , Ramírez-Vélez R . Effects of supervised multimodal exercise interventions on cancer-related fatigue: Systematic review and meta-analysis of randomized controlled trials . Biomed Res Int . 2015 ; 2015 : 328636 . Google Scholar CrossRef Search ADS PubMed 10. Menichetti J , Villa S , Magnani T , et al. Lifestyle interventions to improve the quality of life of men with prostate cancer: A systematic review of randomized controlled trials . Crit Rev Oncol Hematol . 2016 ; 108 : 13 – 22 . Google Scholar CrossRef Search ADS PubMed 11. Bourke L , Boorjian SA , Briganti A , et al. Survivorship and improving quality of life in men with prostate cancer . Eur Urol . 2015 ; 68 : 374 – 383 . Google Scholar CrossRef Search ADS PubMed 12. Mishra SI , Scherer RW , Geigle PM , et al. Exercise interventions on health-related quality of life for cancer survivors . Cochrane DB Syst Rev . 2012 ; 8 : Cd007566 . 13. Mishra SI , Scherer RW , Snyder C , Geigle PM , Berlanstein DR , Topaloglu O . Exercise interventions on health-related quality of life for people with cancer during active treatment . Cochrane DB Syst Rev . 2012 ( 8 ): Cd008465 . 14. Tannenbaum SL , McClure LA , Asfar T , Sherman RL , LeBlanc WG , Lee DJ . Are cancer survivors physically active? a comparison by US States . J Phys Act Health . 2016 ; 13 : 159 – 167 . Google Scholar CrossRef Search ADS PubMed 15. Bandura A . Self-efficacy: Toward a unifying theory of behavioral change . Psychol Rev . 1977 ; 84 : 191 – 215 . Google Scholar CrossRef Search ADS PubMed 16. Stacey FG , James EL , Chapman K , Courneya KS , Lubans DR . A systematic review and meta-analysis of social cognitive theory-based physical activity and/or nutrition behavior change interventions for cancer survivors . J Cancer Surviv . 2015 ; 9 : 305 – 338 . Google Scholar CrossRef Search ADS PubMed 17. Young MD , Plotnikoff RC , Collins CE , Callister R , Morgan PJ . Social cognitive theory and physical activity: A systematic review and meta-analysis . Obes Rev . 2014 ; 15 : 983 – 995 . Google Scholar CrossRef Search ADS PubMed 18. Bandura A. Self-Efficacy: The Exercise of Control . New York : Freeman ; 1997 . 19. Rogers LQ , Courneya KS , Verhulst S , Markwell S , Lanzotti V , Shah P . Exercise barrier and task self-efficacy in breast cancer patients during treatment . Support Care Cancer . 2006 ; 14 : 84 – 90 . Google Scholar CrossRef Search ADS PubMed 20. Short CE , James EL , Plotnikoff RC . How social cognitive theory can help oncology-based health professionals promote physical activity among breast cancer survivors . Eur J Oncol Nurs . 2013 ; 17 : 482 – 489 . Google Scholar CrossRef Search ADS PubMed 21. Hirschey R , Docherty SL , Pan W , Lipkus I . Exploration of exercise outcome expectations among breast cancer survivors . Cancer Nurs . 2017 ; 40 : E39 – E46 . Google Scholar CrossRef Search ADS PubMed 22. Brunet J , Taran S , Burke S , Sabiston CM . A qualitative exploration of barriers and motivators to physical activity participation in women treated for breast cancer . Disabil Rehabil . 2013 ; 35 : 2038 – 2045 . Google Scholar CrossRef Search ADS PubMed 23. Bandura A . Reflections on self-efficacy . In: Rachman S , ed. Advances in Behavior Research and Therapy . Oxford, UK: Pergamon Press Ltd.; 1978; 1 : 237 – 269 . Google Scholar CrossRef Search ADS 24. Bandura A . Recycling misconceptions of perceived self-efficacy . Cognitive Ther Res . 1984 ; 8 ( 3 ): 231 – 255 . Google Scholar CrossRef Search ADS 25. Bandura A . Health promotion from the perspective of social cognitive theory . Psychol Health . 1998 ; 13 ( 4 ): 623 – 649 . Google Scholar CrossRef Search ADS 26. Bandura A . Toward a psychology of human agency . Perspect Psychol Sci . 2006 ; 1 : 164 – 180 . Google Scholar CrossRef Search ADS PubMed 27. Esmaeily H , Peyman N , Taghipour A , KHorashadizadeh F , Mahdizadeh M . A structural equation model to predict the social-cognitive determinants related to physical activity in Iranian women with diabetes mellitus . J Res Health Sci . 2014 ; 14 : 296 – 302 . Google Scholar PubMed 28. Rovniak LS , Anderson ES , Winett RA , Stephens RS . Social cognitive determinants of physical activity in young adults: A prospective structural equation analysis . Ann Behav Med . 2002 ; 24 : 149 – 156 . Google Scholar CrossRef Search ADS PubMed 29. Phillips SM , McAuley E . Social cognitive influences on physical activity participation in long-term breast cancer survivors . Psychooncology . 2013 ; 22 : 783 – 791 . Google Scholar CrossRef Search ADS PubMed 30. Plotnikoff RC , Lubans DR , Penfold CM , Courneya KS . Testing the utility of three social-cognitive models for predicting objective and self-report physical activity in adults with type 2 diabetes . Br J Health Psychol . 2014 ; 19 : 329 – 346 . Google Scholar CrossRef Search ADS PubMed 31. Dewar DL , Plotnikoff RC , Morgan PJ , Okely AD , Costigan SA , Lubans DR . Testing social-cognitive theory to explain physical activity change in adolescent girls from low-income communities . Res Q Exerc Sport . 2013 ; 84 : 483 – 491 . Google Scholar CrossRef Search ADS PubMed 32. Williams DM , Anderson ES , Winett RA . A review of the outcome expectancy construct in physical activity research . Ann Behav Med . 2005 ; 29 : 70 – 79 . Google Scholar CrossRef Search ADS PubMed 33. Rogers LQ , Courneya KS , Anton PM , et al. Social cognitive constructs did not mediate the BEAT cancer intervention effects on objective physical activity behavior based on multivariable path analysis . Ann Behav Med . 2017 ; 51 : 321 – 326 . Google Scholar CrossRef Search ADS PubMed 34. Rogers LQ , Courneya KS , Anton PM , et al. Effects of the BEAT Cancer physical activity behavior change intervention on physical activity, aerobic fitness, and quality of life in breast cancer survivors: A multicenter randomized controlled trial . Breast Cancer Res Treat . 2015 ; 149 : 109 – 119 . Google Scholar CrossRef Search ADS PubMed 35. Williams DM . Outcome expectancy and self-efficacy: Theoretical implications of an unresolved contradiction . Pers Soc Psychol Rev . 2010 ; 14 : 417 – 425 . Google Scholar CrossRef Search ADS PubMed 36. Coups EJ , Park BJ , Feinstein MB , et al. Correlates of physical activity among lung cancer survivors . Psychooncology . 2009 ; 18 : 395 – 404 . Google Scholar CrossRef Search ADS PubMed 37. Hsu HT , Dodd MJ , Guo SE , Lee KA , Hwang SL , Lai YH . Predictors of exercise frequency in breast cancer survivors in Taiwan . J Clin Nurs . 2011 ; 20 : 1923 – 1935 . Google Scholar CrossRef Search ADS PubMed 38. Leach CR , Troeschel AN , Wiatrek D , et al. Preparedness and cancer-related symptom management among cancer survivors in the first year post-treatment . Ann Behav Med . 2017 ; 51 : 587 – 598 . Google Scholar CrossRef Search ADS PubMed 39. Fillenbaum GG , Smyer MA . The development, validity, and reliability of the OARS multidimensional functional assessment questionnaire . J Gerontol . 1981 ; 36 : 428 – 434 . Google Scholar CrossRef Search ADS PubMed 40. Portenoy RK , Thaler HT , Kornblith AB , et al. The memorial symptom assessment scale: An instrument for the evaluation of symptom prevalence, characteristics and distress . Eur J Cancer . 1994 ; 30A : 1326 – 1336 . Google Scholar CrossRef Search ADS PubMed 41. Lorig KR , Sobel DS , Ritter PL , Laurent D , Hobbs M . Effect of a self-management program on patients with chronic disease . Eff Clin Pract . 2001 ; 4 : 256 – 262 . Google Scholar PubMed 42. Basen-Engquist K , Carmack CL , Perkins H , et al. Design of the steps to health study of physical activity in survivors of endometrial cancer: Testing a social cognitive theory model . Psychol Sport Exerc . 2011 ; 12 : 27 – 35 . Google Scholar CrossRef Search ADS PubMed 43. Wójcicki TR , White SM , McAuley E . Assessing outcome expectations in older adults: The multidimensional outcome expectations for exercise scale . J Gerontol B Psychol Sci Soc Sci . 2009 ; 64 : 33 – 40 . Google Scholar CrossRef Search ADS PubMed 44. Patel AV , Jacobs EJ , Dudas DM , et al. The American cancer society’s cancer prevention study 3 (CPS-3): Recruitment, study design, and baseline characteristics . Cancer . 2017 ; 123 : 2014 – 2024 . Google Scholar CrossRef Search ADS PubMed 45. Troeschel AN , Leach CR , Shuval K , Stein KD , Patel AV . Prevalence and medico-demographic correlates of physical activity in cancer survivors during the “re-entry” phase . Prev Chronic Dis . (in press). 46. Ainsworth BE , Haskell WL , Herrmann SD , et al. 2011 Compendium of physical activities: A second update of codes and MET values . Med Sci Sports Exerc . 2011 ; 43 : 1575 – 1581 . Google Scholar CrossRef Search ADS PubMed 47. Hu L-t , Bentler PM . Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives . Struct Equ Modeling . 1999 ; 6 ( 1 ): 1 – 55 . Google Scholar CrossRef Search ADS 48. Brawner CA , Churilla JR , Keteyian SJ . Prevalence of physical activity is lower among individuals with chronic disease . Med Sci Sports Exerc . 2016 ; 48 : 1062 – 1067 . Google Scholar CrossRef Search ADS PubMed 49. Courneya KS , Segal RJ , Mackey JR , et al. Effects of aerobic and resistance exercise in breast cancer patients receiving adjuvant chemotherapy: A multicenter randomized controlled trial . J Clin Oncol . 2007 ; 25 : 4396 – 4404 . Google Scholar CrossRef Search ADS PubMed 50. Jacobucci R , Grimm KJ , McArdle JJ . Regularized structural equation modeling . Struct Equ Modeling . 2016 ; 23 : 555 – 566 . Google Scholar CrossRef Search ADS PubMed © Society of Behavioral Medicine 2018. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)

Journal

Annals of Behavioral MedicineOxford University Press

Published: Apr 23, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off