Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2016 Oct 1.
Published in final edited form as: Health Psychol. 2015 Feb 2;34(10):1022–1032. doi: 10.1037/hea0000188

Antecedents and mediators of physical activity in endometrial cancer survivors: Increasing physical activity through Steps to Health

Matthew Cox 1, Cindy Carmack 1, Daniel Hughes 2, George Baum 1, Jubilee Brown 1, Anuja Jhingran 1, Karen Lu 1, Karen Basen-Engquist 1
PMCID: PMC4522397  NIHMSID: NIHMS659445  PMID: 25642840

Abstract

OBJECTIVE

Research shows that physical activity (PA) has a positive effect on cancer survivors including improving quality of life, improving physical fitness, and decreasing risk for cancer recurrence in some cancer types. Theory-based intervention approaches have identified self-efficacy as a potential mediator of PA interventions. This study examines the temporal relationships at four time points (T1–T4) between several social cognitive theory constructs and PA among a group of endometrial cancer survivors receiving a PA intervention.

METHOD

A sample of 98 sedentary women who were at least six months post treatment for endometrial cancer were given an intervention to increase their PA. The study tested whether modeling, physiological somatic sensations, and social support at previous time points predicted self-efficacy at later time points, which in turn predicted PA at later time points.

RESULTS

Results indicate that as physiological somatic sensations at T2 decrease, self-efficacy at T3 increases, which leads to an increase in PA at T4. This suggests that self-efficacy is a significant mediator between physiological somatic sensations and PA. Exploratory follow up models suggest model fit can be improved with the addition of contemporaneous effects between self-efficacy and PA at T3 and T4, changing the timing of the mediational relationships.

CONCLUSIONS

Physiological somatic sensations appear to be an important construct to target in order to increase PA in this population. While self-efficacy appeared to mediate the relationship between physiological somatic sensations and PA, the timing of this relationship is requires further study.

Keywords: social cognitive theory, cancer survivors, mediation, self-efficacy, endometrial cancer physical activity

Research has demonstrated that physical activity (PA) is not only safe but highly beneficial for cancer survivors. Benefits include improved physical fitness, decreased fatigue, and improved quality of life (Conn, Hafdahl, Porock, McDaniel, & Nielsen, 2006; Speck, Courneya, Mâsse, Duval, & Schmitz, 2010). A recent review of the literature on disease outcomes in cancer survivors suggests PA is associated with decreased risk of breast and colon cancer mortality as well as all-cause mortality (Ballard-Barbash et al., 2012) while another review found that physical activity decreased risk for endometrial cancer by almost 25% (Voskuil et al., 2007).

Despite the positive effects of PA on cancer survivors, PA levels among cancer survivors remain low. The current recommendation for PA in healthy individuals is 150–250 min/week in order to prevent weight gain and receive other health benefits (Haskell et al., 2007). This recommendation has also been extended to cancer survivors including breast, colon, gynecologic, prostate, and hematologic cancers (Rock et al., 2012; Schmitz et al., 2010) as this population has been shown to be less active than those without cancer (Smith, Nolan, Robison, Hudson, & Ness, 2011). Among breast cancer survivors, 32% reported meeting current recommended levels (Irwin et al., 2004) and among endometrial cancer survivors, 30% reported meeting recommended levels (Courneya et al., 2005). Among cancer survivors in general, PA levels tend to decline after diagnosis and do not return to prediagnosis levels (Blanchard et al., 2003).

Social cognitive theory (SCT) has been recommended by the Surgeon General as one of the guiding theories for PA interventions (U.S. Department of Health and Human Services, 1996). Several components make up SCT, including self-efficacy (McAuley & Blissmer, 2000), social support (Courneya & McAuley, 1995), outcome expectancy (Williams, Anderson, & Winett, 2005, and self-regulation (Anderson, Winett, Wojcik, & Williams, 2010).

Evidence from theory-based PA interventions suggests self-efficacy acts as a mediator (as defined by Baron and Kenny (1986)) in the relationship between interventions and PA behaviors. Self-efficacy is defined as one’s confidence in one’s ability to engage in a targeted behavior (Bandura, 1977). Self-efficacy was originally conceived as one of several mediators within SCT (Bandura, 1986) but has also been included as a key component in several other behavioral theories and models, such as the Transtheoretical Model (Prochaska & DiClemente, 1986) and the Theory of Planned Behavior (Ajzen, 1985). Additionally, some researchers have argued that with the abundance of studies utilizing theory-based interventions to improve cancer patients’ and survivors’ lives, more research should be devoted to identifying the mediators of these interventions in order to make them more effective (Stanton, Luecken, MacKinnon, & Thompson, 2013).

Initially, Bandura placed specific emphasis on self-efficacy as a mechanism of change in a variety of behaviors (Bandura, 1977, 1997), and review papers of mediators of PA have shown that even across theories, self-efficacy acts as a significant mediator (Lewis, Marcus, Pate, & Dunn, 2002; Lubans, Foster, & Biddle, 2008); however, one review suggests this relationship is inconsistent (Rhodes & Pfaeffli, 2010). Self efficacy is hypothesized to be gained or lost through the following methods (Bandura, 1997): 1) performance accomplishments, 2) vicarious experience or modeling, 3) verbal and social persuasion or social support, 4) perceptions of emotional and physiological states or somatic sensations. Research by Ewart (1989) suggested that exercise promotion is most effective when patients in a rehabilitation setting are gradually exposed to exercise activities where their self-efficacy is low or virtually zero. Means of enhancing self-efficacy included gradually increasing their “dose” of exercise and arranging the setting so they can observe others engaging in the same exercise, having professional health care providers provide encouragement, positive feedback and verbal praise, and ensuring the activity occurred in a “relaxed but ‘upbeat’ mood”. A number of studies have found that social support can increase the likelihood of engaging in PA through indirect means (Anderson, Wojcik, Winett, & Williams, 2006; Plotnikoff, Lippke, Courneya, Birkett, & Sigal, 2008), often impacting self-efficacy, self-regulation, or outcome expectancy, which in turn affect PA.

Within cancer populations, research over the last decade has shown that self-efficacy is a significant predictor of exercise behavior following an intervention (Mosher et al., 2008; Pinto, Rabin, & Dunsiger, 2009; Vallance, Courneya, Plotnikoff, & Mackey, 2008). Although several studies have examined the role of self-efficacy as a mediator between PA interventions and quality of life, few have examined self-efficacy as a mediator of PA in cancer populations. One study, by Loprinzi and Cardinal (2013), showed that self-efficacy mediated the relationship between behavioral processes of change and PA among breast cancer survivors.

Basen-Engquist et al. (2013) examined the impact of self-efficacy and outcome expectations on endometrial cancer survivors who had undergone an exercise intervention. Endometrial cancer survivors are a relatively under researched population who disproportionately experience factors such as obesity and physical inactivity that put them at risk for cancer reoccurrence and cardiovascular disease (Furberg & Thune, 2003). This study utilized ecological momentary assessment (EMA) methods to examine the same day effects of self-efficacy and outcome expectations on PA as well as whether laboratory assessments of self-efficacy and outcome expectations at baseline, 2-, 4-, and 6-months predicted the following time points’ measure of PA. Results from both analyses indicated that when both self-efficacy and outcome expectations were included in the model, self-efficacy was the only statistically significant predictor of PA for both PA measured that day and PA measured 2-months later. These results suggest that self-efficacy has both a proximal and distal influence on PA.

Despite recent research indicating the potential of self-efficacy as a mediator of interventions and PA, there are several methodological limitations which restrict our ability to evaluate the theory’s utility for intervention development. First, few studies have examined all SCT variables simultaneously, including influences on self-efficacy. The failure of studies to include all relevant theoretical variables in the same model increases the potential for results to be due to unmeasured confounders (Greenland & Robins, 1986), particularly in mediation models (Cox, Kisbu-Sakarya, Miočević, & MacKinnon, 2013; Pearl, 2001). Additionally, cross-sectional mediation limits causal claims of mediation analyses and can result in biased estimates (Cole & Maxwell, 2003; Maxwell & Cole, 2007).

Utilizing data from Basen-Engquist et al. (2013), this study examines the longitudinal relationship between theoretical antecedents of PA as hypothesized by SCT, specifically, somatic sensations, modeling, and social support, and PA with self-efficacy acting as a mediator (see Figure 2 for the hypothesized model). EMA data (data on behaviors, beliefs, symptoms, etc. collected in real time using portable devices) are used in this analysis to avoid the recall bias frequently present in self-report measures of physical activity and its antecedents. SCT provides a theoretical model for testing these constructs with few studies having measured more than one SCT construct and fewer having attempted to model these constructs longitudinally. We hope to fill some of these knowledge gaps by using longitudinal mediation analyses to examine the relationship among several SCT variables.

Figure 2.

Figure 2

Open in a new tab

Model 1 social cognitive theory variables predicting physical activity minutes. Note: Dotted lines represent non-significant paths. Numbers over solid lines represent the respective unstandardized beta coefficients with p < .05 and standard errors in parentheses. Bolded coefficients and standard errors represent hypothesized mediational relationships. Model fit indices: χ2 (144, N = 98) = 406.014, p < .001; RMSEA = .136, TLI = .550; SRMR = .144

Method

Detailed information regarding data collection and intervention methods can be found in Basen-Engquist et al. (2013).

Participants

Participants included 100 women with Stage I, II, or IIIa endometrial cancer who were at least 6 months post-treatment and had no evidence of disease. Participants were excluded from the study if they engaged in moderate activity for at least 30 min or more 5 days weekly or if they engaged in vigorous activity for 20 min or more 3 days weekly.

Procedure

Intervention and assessment procedures are outlined in Figure 1. Baseline EMA data were collected for 7 days on several SCT variables including modeling, somatic sensations, social support, and self-efficacy as well as self-reported number of exercise minutes for each day. Self-efficacy was measured in the morning, modeling and social support were measured in the evening, and somatic sensations were measured after an episode of exercise. Self-efficacy was measured in the morning for two reasons. First is that we wanted to get a measure of self-efficacy that was potentially at the farthest time from when they decided to exercise so that their self-efficacy was not conflated with their decision to exercise. Second, we wanted a measure of self-efficacy even if they did not exercise. Modeling and social support were measured in the evening so that participants could report on social support received or modeling observed during day. After a physical activity bout, participants responded to questions about somatic sensations experienced during the physical activity; this was measured as close as possible to the physical activity bout in order to avoid recall bias.

Figure 1.

Figure 1

Open in a new tab

Graphical display of intervention and measurement procedures.

Following 7 days of baseline EMA data, participants attended a laboratory assessment where participants completed a submaximal cardiorespiratory fitness assessment on a cycle ergometer as well as completing several SCT measures of predictors of PA (laboratory-based assessments were not used in the following analyses and are not reported here). After completion of the laboratory assessment procedures, all participants were given a tailored exercise recommendation. They were then instructed to complete another 5 days of EMA data collection while attempting to exercise at the prescribed levels. At subsequent assessment periods (2-, 4-, and 6-months post-baseline) participants completed 5 days of EMA data collection before and after the laboratory assessment session and were not given feedback about their fitness levels during laboratory assessments.

Measures

For the home-based assessment, participants used a hand-held electronic device (Hewlett-Packard IPAQ RX195) to input self reports of modeling, somatic sensations, social support, self-efficacy, and PA (e.g., self-reported number of minutes of exercise for that day). Physiological somatic sensations were assessed via 10 items from the Pennebaker Inventory of Limbic Languidness (PILL; Pennebaker, 1982), which asked participants to rate their experience of several physiological sensations including “Racing heart”, “Tightness in chest”, and “Stiff or sore muscles”. The questions were rated on a 5 point Likert scale ranging from 0 to 4 and were anchored with “Not at all” at 0 and “Very Much” at 4. The internal consistency for the PILL has been shown to range from .88 to .91 (Pennebaker, 1982) and was .834 at baseline for this study.

Social support was measured using an 11 item measure adapted from a previously validated measure (Sallis, Grossman, Pinski, Patterson, & Nader, 1987). Participants responded “Yes” or “No” to a series of statements about their interactions with other individuals with respect to their exercise behaviors (e.g., “Today, a friend or family member gave me helpful reminders to exercise”). Internal consistency for social support was .831. We changed the responses for this measure from a 5-point Likert to dichotomous in order to assess whether or not the events happened that day instead of rating their frequency the past three months, as was done in the original measure. Items were then averaged together with “Yes” being 1 and “No” being 0 to create a continuous social support score.

Modeling was assessed in a similar method using a 7-item measure where participants responded “Yes” or “No” to statements on whether they had observed someone engaging in exercise that day (e.g., “I was aware that a member of my family exercised today.”). Internal consistency for modeling was .661. Items were then averaged together with “Yes” being 1 and “No” being 0 to create a continuous modeling score.

Due to concerns related to participant burden, self-efficacy was measured using a 1-item measure that asked the participant to rate her confidence in her ability to exercise that day. Participants rated their confidence on a 5-point Likert Scale with 1 being “Not at all confident” and 5 being “Extremely Confident”.

PA data were measured in three different ways. The first method had participants report their number of PA minutes immediately following an exercise activity. They also reported their number of PA minutes in the evening before bed. In addition to the self-reporting, PA was also measured using a GT1M accelerometer (Actigraph L.L.C., Pensacola, FL). When the self-reported PA immediately following exercise was not available (because data were missing), the evening self-reported minutes were used. If both of those measures were not reported, accelerometer data were used to create the combined PA minute per day variable.

Intervention

After the baseline laboratory assessment, participants were given individually tailored PA recommendations based on their current level of PA and the American College of Sports Medicine guidelines, with the ultimate goal of engaging in moderate-intensity exercise for 30 minutes a day, 5 days a week. The primary mode of exercise was walking, but for participants who had difficulty walking, alternative exercises were suggested. Between measurement time points, participants received telephone counseling which reviewed exercise goals and barriers, and provided brief teaching of behavioral and cognitive skills to support their PA behaviors such as goal setting and problem solving. Phone counseling was gradually tapered, such that phone calls were made weekly during months 1 and 2, bi-monthly during months 3 and 4, and once monthly during months 5 and 6. Participants also received a newsletter before each phone session that went over the content of the upcoming counseling session and provided “role model stories” of endometrial cancer survivors who successfully adopted an exercise program.

Analysis

Bandura hypothesizes that SCT constructs occur in an ordered sequence, with modeling, somatic sensations, and social support affecting self-efficacy, and self-efficacy affecting PA (Bandura, 1997). Using the theoretical sequence outlined by SCT, we hypothesized that modeling and social support would have a positive effect on self-efficacy, and physiological somatic sensations (measured by the PILL) would have a negative effect on self-efficacy. In order to incorporate the longitudinal nature of the data into the analyses, data were aggregated by assessment time point (excluding pre-lab assessment baseline data) to create four discrete time points: T1, initial assessment; T2, 2 months after initial assessment; T3, 4 months after initial assessment; T4, 6 months after initial assessment. Baseline pre-lab assessment was excluded because of the fact that an intervention was introduced between pre- and post-lab assessment, thereby making comparisons between coefficients at this time point and other time points impossible and because of issues related to unequal time between assessments. We further hypothesized that for the sequence described above, the measure at each time point would predict the corresponding SCT construct at the following time point. For example, modeling, social support, and somatic sensations at T1 would predict self-efficacy at T2, and self-efficacy at T2 would predict PA at T3. This model is outlined in Figure 2.

To test this hypothesis, we used an autoregressive mediation model (MacKinnon, 2008) with four time points (T1 through T4). The benefit of this model is that it allows testing of the temporal mediational sequence of these variables. An autoregressive model was chosen over a growth curve model for several reasons. First, growth curve models add a level of analytic complexity that would have prevented us from examining the mediating relationships between each antecedent, self-efficacy, and PA. That is, in the parallel process model outlined by Cheong MacKinnon and Khoo (2003), individual growth curve models are fit to assess for change over time for each variable. Once “good” fitting models are developed, and then each growth curve model is added sequentially to a measurement model to ensure adequate fit with each additional model. Finally, mediation paths are specified to test the hypothesized model. Preliminary growth curve models showed that not all antecedent growth curve models were of good fit, which resulted in convergence issues when testing growth curve mediation models. Second, the timing of the effects for SCT variables is not clearly delineated in SCT, and as such, we expected some variables to have effects at certain time points but not others and the autoregession model allows us to assess change between each time point while the growth curve model does not.

While SCT posits the temporal order of the corresponding constructs, few studies have tested such a mediational model. The product of the coefficients (MacKinnon, Lockwood, Hoffman, West, & Sheets, 2002) with the bias corrected bootstrap was used to test for mediation as it has been shown to have better power to detect significant mediation effects in smaller samples than other methods (Fritz & MacKinnon, 2007). In addition to testing the hypothesized model above, we conducted several other exploratory models to determine if another sequence of variables would better fit the data.

Exploratory analyses are warranted because of the lack of research related to the timing of the SCT constructs. While SCT explicates that the passage of time is necessary in order for each construct to affect corresponding constructs, no research has been done to examine the optimal measurement times for these constructs. As such, the addition of contemporaneous mediation paths may be warranted. Statistically significant contemporaneous mediation paths could indicate that two or more constructs occurred in closer temporal proximity than the 2-month time periods measured in this study (MacKinnon, 2008). Fit indices will be used to compare models with and without contemporaneous effects. Model fit will be assessed via several fit indices. The indices and their cutoff are SRMR < .08, TLI ≥ .95, and RMSEA ≤ .05 as well as the χ2 difference test.

All analyses were conducted in Mplus v7.2 (Muthen & Muthen, 2013) and Full Information Maximum Likelihood was used to handle missing data.

Results

The final dataset included 98 women (two were removed because of missing data). Descriptive statistics are presented in Table 1. Overall, the sample was mainly white (70%), educated (72% had some college or more education), and obese (mean BMI = 34.1), with an average age of 57 years. Missing data ranged from 4% to 32.7% missing with the average rate of missingness being 20.7%.

Table 1.

Demographic information for endometrial cancer survivor cohort (N = 98).

Demographic Variable n(%) M(SD) Range
Race
 Black/non-Hispanic   7 (7)
 White/non-Hispanic 70 (71)
 Asian/non-Hispanic   5 (5)
 White/Hispanic 12 (12)
 American Indian/non-Hispanic   1 (1)
Education
 <High school   2 (2)
 High school diploma/GED 13 (13)
 Technical/Vocational degree   8 (8)
 Some college/2-year degree 33 (34)
 4-year degree 22 (22)
 Advanced degree 16 (16)
Disease stage
 I 77 (79)
 II 15 (15)
 IIIa   3 (3)
Treatment
 Surgery only 57 (58)
 Radiotherapy only   0 (0)
 Surgery + Radiotherapy 41 (42)
Age (in years) 57.18 (11.31) 25–78
Time since diagnosis (in months 26.0 (14.3) 3.7–63.8
Body mass index (kg/m2) 34.10 (9.49) 18.71–69.31
Open in a new tab

Model fit indices for each of the two models tested are presented in their respective figures. For Model 1, statistically significant paths are presented in Figure 2 and overall model fit was poor. In this model, self-efficacy was a significant mediator of the relationship between physiological somatic sensations and PA. Physiological somatic sensations at T2 predicted self-efficacy at T3 (B = −0.033, 95% CI [−.061, −0.004]), and self-efficacy at T3 predicted PA at T4 (B = 3.251, 95% CI [0.351, 6.152]). The mediated effect was equal to −0.106 with the 95% CI [−0.320, −0.008], indicating a significant mediated effect. This means that for every one unit decrease in physiological somatic sensations, there is a 0.033 unit increase in self-efficacy and for every one unit increase in self-efficacy, there is a an increase of 3.251 minutes of PA. Model 1 predicted 27% of the variance in PA at T4.

To improve model fit, modification indices from Model 1 were examined to determine if any paths could be added that would improve model fit and were also theoretically relevant. Modification indices suggested that the contemporaneous paths for the effect of modeling on social support at T2, T3, and T4, as well as the effect on self-efficacy on PA at T3 and T4. While the effect of self-efficacy on PA is theoretically hypothesized, the effect of modeling on social support is not. An examination of these items reveals an overlap in terms of the measures, where both measures assess participants’ perceptions and interactions with other people in regard to exercising.

Model 2 (presented in Figure 3) was the same as Model 1 but included the contemporaneous paths as suggested by the modification indices. Statistically significant path coefficients and standard errors from Model 2 are presented in Figure 3. Model fit for this model was significantly improved. The χ2 difference test comparing Model 1 and Model 2 indicated that Model 2 was a significantly better fitting model, χ2 (5, N = 98) = 134.51, p < .001; however, overall model fit did not meet a priori cut off values. Power analysis for the RMSEA (Preacher & Coffman, 2006) revealed power to be 0.989 for this model with alpha equal to .05, the null RMSEA equal to .05, sample size equal to 98, degrees of freedom equal to 139, and the alternative RMSEA equal to .098.

Figure 3.

Figure 3

Open in a new tab

Model 2 social cognitive theory variables predicting physical activity minutes. Note: Dotted lines represent non-significant paths. Numbers over solid lines represent the respective unstandardized beta coefficients with p < .05 and standard errors in parentheses. Bolded coefficients and standard errors represent hypothesized mediational relationships. Model fit indices: χ2 (139, N = 98) = 271.098, p < .001; RMSEA = .098, TLI = .765; SRMR =.111.

In Model 2, self-efficacy at T3 no long mediated the relationship between physiological sensations at T2 and PA at T4. There are four significant mediational relationships worth noting in this model. The first is that physiological sensations at T1 predicted self-efficacy at T2 (B = −0.073, 95% CI [−0.118, −0.017]) and self-efficacy at T2 predicted PA at T3 (B = −7.861, 95% CI [−11.713, −3.714]). The mediated effect was B = 0.572, 95% CI [0.151, 1.133]. The second mediational relationship was physiological sensations at T2 predicted self-efficacy at T3 (B = −0.033, 95% CI [−.063, −0.008]), and self-efficacy at T3 predicted PA at T3 (B = 8.800, 95% CI [5.242, 12.777]). The mediated effect was B = −0.288, 95% CI [−0.607, −0.065]. The third mediational relationship was physiological sensations at T2 predicted self-efficacy at T3 (B = −0.033, 95% CI [−.063, −0.008]), self-efficacy at T3 predicted self-efficacy at T4 (B = 0.813, 95% CI [0.634, 0.996]), and self-efficacy at T4 predicted PA at T4 (B = 7.902, 95% CI3.848, 10.876]). The mediated effect was B = −0.098, 95% CI [−0.241, −0.028]. The fourth mediational relationship is physiological sensations at T2 predicted self-efficacy at T3 (B = −0.04, 95% CI [−.065, −0.015]), self-efficacy at T3 predicted PA at T3 (B = 8.9910, 95% CI [5.2572, 13.030]), and PA at T3 predicted PA at T4 (B = 0.331, 95% CI [0.139, 0.512]). The mediated effect was B = −0.118, 95% CI [−0.261, −0.044]. Lastly, Model 2 predicted 48% of the variance in PA at T4.

Discussion

The current study attempted to examine the mediational relationship of self-efficacy between some of the hypothesized SCT antecedents (modeling, social support, and somatic sensations) and PA. These findings suggest that, at later time points, self-efficacy mediates the relationship between physiological somatic sensations and PA, such that decreases in physiological sensations increase self-efficacy which in turn increases PA. The findings from this study are important for several reasons. This is one of the first studies to longitudinally model the effect of several SCT constructs in the temporal sequence in which they are theoretically hypothesized to occur. Many studies have examined one or two constructs from a specific theory of behavior change, but this study measured four. Including all of these constructs ensures that the estimates of the effects are not due to competing relationships among these other SCT variables. Moreover, because the data were collected over several occasions, we were able to test the temporal sequence of these variables as hypothesized by Bandura (1997).

Comparing the results from the two models reveals several interesting findings related to the order and timing of the relationships among the variables. The first was that in both models, social support and modeling at T1 predicted PA at T2, but not at any other time points. This suggests that these constructs are important components of the behavior change process, but with this sample, they do not operate as hypothesized by Bandura (1997); specifically, that the influence of social support and modeling on PA is not indirect through self-efficacy. Based on these, social support and modeling directly impact PA at the beginning of the intervention, but not at later time points. Future studies should examine the relationship between modeling and social support at various time points to see if these results can be replicated.

Additionally, social support negatively predicted PA, which is the opposite of what SCT hypothesizes. One explanation for this finding is that the type of social support that participants received at the beginning of the intervention was not conducive to facilitating exercise. The social support measure assessed if people received various types of support related to their exercise behaviors from individuals or if anyone exercised with the participant. The measure did not assess the participants’ perceptions of the helpfulness of this “support”. For example, one question asked whether the participants received “encouragement to stick with my exercise program”. Participants beginning an exercise intervention may perceive this type of support as condescending in some contexts, which may have a detrimental effect on their PA behaviors. While no other studies have found this relationship between social support and PA, a study by Hagedoorn et al. (2000) examined social support from the spouses of cancer patients found that certain types of social support such as providing unnecessary help and lavishing excessive praise decreased marital satisfaction. Participants in this study may have experienced something similar which decreased their PA.

The effect of social support on PA did not manifest beyond the effects at T1 indicating this effect was ephemeral. One previous study found that baseline social support predicts adherence to a resistance training program at 3 months, but not at 6 months (Rhodes, Martin, & Taunton, 2001). Literature on social support for PA for cancer survivors is mixed, with there being some evidence that SCT based interventions lead to increases in social support and PA (Barber, 2012), but no studies in that review appeared to examine the indirect role of social support through self-efficacy. In non-clinical populations, the evidence is unclear whether social support has a direct effect on PA through self-efficacy (Courneya & McAuley, 1995), self-efficacy and self-regulation (Anderson et al., 2010; Anderson et al., 2006), or if social support has a direct effect on PA (Anderson-Bill, Winett, & Wojcik, 2011; Becofsky, Baruth, & Wilcox, 2013). Future studies should focus on the type of support participants receive, how they perceive that support in relation to their PA behaviors, and how social support changes over time.

Similarly to social support, modeling at T1 was a significant predictor of PA at T2 but not at any other time point. Recently research using a prospective design for an observational study, found that modeling (also referred to as vicarious experiences) measured at T1 predicted PA at T3 but self-efficacy at T2 did not mediate the relationship (Warner, et al., 2014). A meta-analysis examining interventions that targeted sources of self-efficacy found that interventions that included modeling as a component of the intervention produced greater changes in self-efficacy than those that did not (Ashford, Edmunds, & French, 2010). Neither of these studies involved cancer survivors. The findings from this study do not align with the previous findings and suggests that modeling may differentially impact endometrial cancer survivors as compared to other populations, both in terms of its influence on other variables (i.e., directly or indirectly) as well as the timing of its effect (i.e., proximal vs. distal influence).

Lastly, this study adds to the growing body of literature that suggests that self-efficacy is a significant mediator in the behavior change process; however, when it is a mediator was not clearly delineated by the results. In Model 1, physiological somatic sensations at T2 negatively predicted self-efficacy at T3 and self-efficacy at T3 positively predicted PA at T4, indicating that self-efficacy was an important mediator four months following the beginning of the intervention.

In contrast, Model 2 suggests a more complicated mediational relationship. Self-efficacy was a significant mediator of the relationship between physiological somatic sensations at four different time points with one relationship being in the opposite of the hypothesized direction. The first mediational relationship was with physiological somatic sensations at T1 negatively predicting self-efficacy at T2, and self-efficacy at T2 negatively predicting PA at T3. This would suggest that as physiological somatic sensations decrease, self-efficacy increases and as self-efficacy increases, PA decreases, which is opposite of what SCT hypothesizes. The second mediational relationship was with physiological sensations at T2 predicting self-efficacy at T3, self-efficacy at T3 predicting PA at T3. This relationship is in the correctly hypothesized direction such that as physiological somatic sensations decrease, self-efficacy increases, which in turn increases PA. The third mediational relationship was with physiological sensations at T2 predicting self-efficacy at T3, self-efficacy at T3 predicting self-efficacy at T4, and self-efficacy at T4 predicting PA at T4. Again, this relationship is in the correctly hypothesized direction. The fourth mediational relationship was with physiological sensations at T2 predicting self-efficacy at T3, self-efficacy at T3 predicting PA at T3. This relationship is also in the correctly hypothesize direction.

Research regarding somatic sensations related to PA in cancer survivors is limited. Some research has indicated that somatic sensations (not related to PA) are perceived more negatively by cancer survivors than non-cancer controls and are more readily interpreted as a potential symptom of a health disorder (Benyamini, McClain, Leventhal, & Leventhal, 2003). This in turn may lead cancer survivors to be less amenable to any type of PA that may produce somatic sensation (e.g., racing heart, muscle pain, etc.). The findings from this study suggest that as survivors’ somatic sensations decrease, self-efficacy increases, leading to an increase in PA, which is consistent with previous research. One study looking at girls (Pender, Bar-Or, Wilk, & Mitchell, 2002) and another looking at older adults (Clark & Nothwehr, 1999) also found a negative relationship between self-efficacy and somatic sensations. In a non-randomized control trial, Hughes et al. (2010) showed somatic sensations during exercise negatively predicted post-exercise self-efficacy in both endometrial cancer survivors and non-cancer controls, but that only self-efficacy predicted PA.

With the addition of the contemporaneous effects in Model 2, the temporal sequence of the SCT variables becomes more convoluted. As MacKinnon (2008) notes, significant contemporaneous effects could indicate that the change among variables occurs in closer temporal proximity than 2 month intervals measured in this study. This would mean that the effect of self-efficacy on PA at T3 and T4 may occur more rapidly than 2 months, but that the effect of physiological somatic sensations on self-efficacy is better captured by the 2-month-lag period. Unfortunately, without more measurement time points, exactly when the changes in self-efficacy’s impact on PA cannot be determined.

The other interesting finding from Model 2 was the mediated effect in the opposite direction involving self-efficacy. This may have been due to participants having an inflated sense of self-efficacy at the beginning of the study and being unable to exercise at the expected levels. After three months of engaging in PA, their self-efficacy may have been more realistic and thus accurately predicted their PA behavior. McAuley and Mihalko (1998) suggested that this is in fact how self-efficacy changes over time for older adults.

Both models suggest that physiological sensations and self-efficacy at later time points are better predictors of self-efficacy and PA respectively, than the same variables at earlier time points. This suggests that some passage of time is needed to produce changes in self-efficacy and PA. In this case, a period of 2 months was sufficient to detect the changes in self-efficacy and PA. Future studies using a similar population may benefit from using a similar timing scheme for assessing self-efficacy as a mediator of physiological sensations and PA.

While this study evidenced several strengths, including examining an understudied population and using a prospective longitudinal design, there are a number of limitations that should be noted. First, the study was not a randomized control trial, so none of the results can be considered causal because of the potential for confounding. Also, while the measurement timing used in this study identified a mediational relationship between physiological sensations, self-efficacy and PA, alternative timing schemes may have identified alternative mediational relationships (Cole & Maxwell, 2003). The contemporaneous effects found in Model 2 suggest that the effects of self-efficacy on PA occur in closer temporal proximity than 2 months. More work needs to be done to determine the optimal time for measuring various SCT constructs.

Another limitation is that we were unable to test the full model for SCT because measures of mastery experience for PA and outcome expectations were not included. One of the reasons that mastery experience was not included was that during data collection, a valid measure of mastery experience did not exist. Recently, there have been developments of such a measure (Warner et al., 2014), but items from this measure could be measuring self-regulation self-efficacy as two items pertain to overcoming challenges to PA. Previous measures of PA could be conceptualized as a measure of mastery experience, but modification indices did not indicate model fit would be improved if paths from previous measures of PA to future self-efficacy were added. More work needs to be done to create a valid measure discriminates mastery experience from previous PA and self-regulation self-efficacy.

Outcome expectations were not examined because of the lack of an effect demonstrated previously by Basen-Engquist et al. in this dataset. Self-regulation was measured during laboratory assessments but not during home EMA periods and was excluded from these analyses because it was not assessed via EMA and could potentially cause methodological confounding. It should also be noted that self-efficacy was measured with a one item measure, and although this measure was averaged over several days, a multi-item measure could have provided a more reliable estimate of self-efficacy.

Both models did not provide good fit and results have to be viewed within this context. To improve model fit, modification indices were examined and theoretically relevant paths were added, which included contemporaneous paths for self-efficacy and PA as well as social support and modeling. Although Model 2 fit improved significantly over Model 1, fit indices were all above recommended cutoff values, despite power for the RMSEA being well above .8 for both models. This suggests that either the current model with sources of self-efficacy predicting self-efficacy and self-efficacy predicting PA is the “best” fitting model, or that a model with other unmeasured variables would provide a better fit.

The use of an autoregressive longitudinal model in this study assessed inter-individual change over time, but not intra-individual change. Future studies would benefit from utilizing alternative modeling approaches that capture intra-individual change such as growth curve modeling or time varying effects modeling, which has been advocated for use with EMA data (Tan, Shiyko, Li, Li, & Dierker, 2012). Finally, this study did not include a measure of self-regulation in the EMA measures. As such, the results presented here may be due to self-regulation and not solely somatic sensations and self-efficacy. Future studies should include a valid and reliable measure of self-regulation for PA to parse out the effects of all the hypothesized SCT variables.

There are a number of important implications from these findings. From a theoretical perspective, physiological sensations are linked to self-efficacy; future studies might attempt to target physiological sensations by normalizing the negative sensations and encouraging participants to continue despite these sensations. Through developing a randomized control trial that targets survivors’ interpretations of physiological sensations, researchers could attempt to determine the causal nature of physiological sensations and its effect on self-efficacy.

Table 2.

Model fit indices for longitudinal mediation models.

Model χ2 χ2df RMSEA CFI TLI
Model 1 406.014 144 0.136 0.64 0.55
Model 2 271.098 139 0.098 0.818 0.765
Open in a new tab

Acknowledgments

This research was supported by National Institutes of Health Grants R01CA109919, R25TCA057730, R25ECA056452, and P30 CA016672 (PROSPR Shared Resource)

References

  1. Ajzen I. From Intentions to Actions: A Theory of Planned Behavior. In: Kuhl J, Beckman J, editors. Action Control: From Cognition to Behavior. New York: Springer-Verlag; 1985. pp. 11–39. [Google Scholar]
  2. Anderson-Bill ES, Winett RA, Wojcik JR. Social cognitive determinants of nutrition and physical activity among web-health users enrolling in an online intervention: the influence of social support, self-efficacy, outcome expectations, and self-regulation. Journal of Medical Internet Research. 2011;13(1):e28. doi: 10.2196/jmir.1551. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Anderson ES, Winett RA, Wojcik JR, Williams DM. Social cognitive mediators of change in a group randomized nutrition and physical activity intervention social support, self-efficacy, outcome expectations and self-regulation in the Guide-to-Health Trial. Journal of Health Psychology. 2010;15(1):21–32. doi: 10.1177/1359105309342297. [DOI] [PubMed] [Google Scholar]
  4. Anderson ES, Wojcik JR, Winett RA, Williams DM. Social-cognitive determinants of physical activity: The influence of social support, self-efficacy, outcome expectations, and self-regulation among participants in a church-based health promotion study. Health Psychology. 2006;25(4):510–520. doi: 10.1037/0278-6133.25.4.510. [DOI] [PubMed] [Google Scholar]
  5. Ashford S, Edmunds J, French DP. What is the best way to change self-efficacy to promote lifestyle and recreational physical activity? A systematic review with meta- analysis. British journal of health psychology. 2010;15(2):265–288. doi: 10.1348/135910709X461752. [DOI] [PubMed] [Google Scholar]
  6. 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. Journal of the National Cancer Institute. 2012;104(11):815–840. doi: 10.1093/jnci/djs207. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Bandura A. Self-efficacy: Toward a unifying theory of behavioral change. Psychological Review. 1977;84(2):191–215. doi: 10.1037/0033-295x.84.2.191. [DOI] [PubMed] [Google Scholar]
  8. Bandura A. Social foundations of thought and action: A social cognitive theory. Englewood Cliffs, NJ: Prentice-Hall; 1986. [Google Scholar]
  9. Bandura A. Self-efficacy: The exercise of control. New York, NY: W H Freeman/Times Books/Henry Holt & Co; 1997. [Google Scholar]
  10. Barber FD. Social support and physical activity engagement by cancer survivors. Clinical journal of oncology nursing. 2012;16(3):E84–E98. doi: 10.1188/12.CJON.E84-E98. [DOI] [PubMed] [Google Scholar]
  11. Baron RM, Kenny DA. The moderator–mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology. 1986;51(6):1173–1182. doi: 10.1037//0022-3514.51.6.1173. [DOI] [PubMed] [Google Scholar]
  12. Basen-Engquist K, Carmack C, Li Y, Brown J, Jhingran A, Hughes D, Baum G. Social-Cognitive Theory predictors of exercise behavior in endometrial cancer survivors. Health Psychology. 2013;32(11):1137–1148. doi: 10.1037/a0031712. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Becofsky K, Baruth M, Wilcox S. Psychosocial mediators of two community-based physical activity programs. Annals of Behavioral Medicine. 2013:1–5. doi: 10.1007/s12160-013-9578-3. [DOI] [PubMed] [Google Scholar]
  14. Benyamini Y, McClain CS, Leventhal EA, Leventhal H. Living with the worry of cancer: health perceptions and behaviors of elderly people with self, vicarious, or no history of cancer. Psycho-Oncology. 2003;12(2):161–172. doi: 10.1002/pon.637. [DOI] [PubMed] [Google Scholar]
  15. Blanchard CM, Denniston MM, Baker F, Ainsworth SR, Courneya KS, Hann DM, Kennedy JS. Do adults change their lifestyle behaviors after a cancer diagnosis? American Journal of Health Behavior. 2003;27(3):246–256. doi: 10.5993/ajhb.27.3.6. [DOI] [PubMed] [Google Scholar]
  16. Cheong J, MacKinnon DP, Khoo ST. Investigation of mediational processes using parallel process latent growth curve modeling. Structural Equation Modeling. 2003;10(2):238–262. doi: 10.1207/S15328007SEM1002_5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Clark DO, Nothwehr F. Exercise self-efficacy and its correlates among socioeconomically disadvantaged older adults. Health Education & Behavior. 1999;26(4):535–546. doi: 10.1177/109019819902600410. [DOI] [PubMed] [Google Scholar]
  18. Cole DA, Maxwell SE. Testing mediational models with longitudinal data: questions and tips in the use of structural equation modeling. Journal of Abnormal Psychology. 2003;112(4):558–577. doi: 10.1037/0021-843X.112.4.558. [DOI] [PubMed] [Google Scholar]
  19. Conn VS, Hafdahl AR, Porock DC, McDaniel R, Nielsen PJ. A meta-analysis of exercise interventions among people treated for cancer. Supportive Care in Cancer. 2006;14(7):699–712. doi: 10.1007/s00520-005-0905-5. [DOI] [PubMed] [Google Scholar]
  20. Courneya KS, Karvinen KH, Campbell KL, Pearcey RG, Dundas G, Capstick V, Tonkin KS. Associations among exercise, body weight, and quality of life in a population-based sample of endometrial cancer survivors. Gynecologic Oncology. 2005;97(2):422–430. doi: 10.1016/j.ygyno.2005.01.007. [DOI] [PubMed] [Google Scholar]
  21. Courneya KS, McAuley E. Cognitive mediators of the social influence-exercise adherence relationship: A test of the theory of planned behavior. Journal of Behavioral Medicine. 1995;18(5):499–515. doi: 10.1007/BF01904776. [DOI] [PubMed] [Google Scholar]
  22. Cox MG, Kisbu-Sakarya Y, Miočević M, MacKinnon D. Sensitivity Plots for Confounder Bias in the Single Mediator Model. Evaluation Review. 2013;37(5):405–431. doi: 10.1177/0193841X14524576. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Ewart CK, Taylor CB, Reese LB, DeBusk RF. Effects of early postmyocardial infarction exercise testing on self-perception and subsequent physical activity. The American journal of cardiology. 1983;51(7):1076–1080. doi: 10.1016/0002-9149(83)90348-x. [DOI] [PubMed] [Google Scholar]
  24. Furberg AS, Thune I. Metabolic abnormalities (hypertension, hyperglycemia and overweight), lifestyle (high energy intake and physical inactivity) and endometrial cancer risk in a Norwegian cohort. International Journal of Cancer. 2003;104(6):669–676. doi: 10.1002/ijc.10974. [DOI] [PubMed] [Google Scholar]
  25. Fritz MS, MacKinnon DP. Required sample size to detect the mediated effect. Psychological Science. 2007;18(3):233–239. doi: 10.1111/j.1467-9280.2007.01882.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Greenland S, Robins JM. Identifiability, exchangeability, and epidemiological confounding. International Journal of Epidemiology. 1986;15(3):413–419. doi: 10.1093/ije/15.3.413. [DOI] [PubMed] [Google Scholar]
  27. Hagedoorn M, Kuijer RG, Buunk BP, DeJong GM, Wobbes T, Sanderman R. Marital satisfaction in patients with cancer: Does support from intimate partners benefit those who need it most? Health Psychology. 2000;19(3):274–282. [PubMed] [Google Scholar]
  28. Haskell W, Lee I, Pate R, Powell K, Blair S, Franklin B, Bauman A. Physical activity and public health: updated recommendation for adults from the American College of Sports Medicine and the American Heart Association. Medicine and Science in Sports and Exercise. 2007;39(8):1423–1434. doi: 10.1249/mss.0b013e3180616b27. [DOI] [PubMed] [Google Scholar]
  29. Hughes D, Baum G, Jovanovic J, Carmack C, Greisinger A, Basen-Engquist K. An acute exercise session increases self-efficacy in sedentary endometrial cancer survivors and controls. Journal of Physical Activity & Health. 2010;7(6):784–793. doi: 10.1123/jpah.7.6.784. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Irwin ML, McTiernan A, Bernstein L, Gilliland FD, Baumgartner R, Baumgartner K, Ballard-Barbash R. Physical activity levels among breast cancer survivors. Medicine and Science in Sports and Exercise. 2004;36(9):1484–1491. [PMC free article] [PubMed] [Google Scholar]
  31. Lewis BA, Marcus BH, Pate RR, Dunn AL. Psychosocial mediators of physical activity behavior among adults and children. American Journal of Preventive Medicine. 2002;23(2):26–35. doi: 10.1016/s0749-3797(02)00471-3. [DOI] [PubMed] [Google Scholar]
  32. Loprinzi PD, Cardinal BJ. Self-efficacy mediates the relationship between behavioral processes of change and physical activity in older breast cancer survivors. Breast Cancer. 2013;20(1):47–52. doi: 10.1007/s12282-011-0298-x. [DOI] [PubMed] [Google Scholar]
  33. Lubans DR, Foster C, Biddle SJH. A review of mediators of behavior in interventions to promote physical activity among children and adolescents. Preventive Medicine. 2008;47(5):463–470. doi: 10.1016/j.ypmed.2008.07.011. [DOI] [PubMed] [Google Scholar]
  34. MacKinnon DP. Introduction to statistical mediation analysis. Routledge; 2008. [Google Scholar]
  35. MacKinnon DP, Lockwood CM, Hoffman JM, West SG, Sheets V. A comparison of methods to test mediation and other intervening variable effects. Psychological Methods. 2002;7(1):83–104. doi: 10.1037/1082-989x.7.1.83. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Maxwell SE, Cole DA. Bias in cross-sectional analyses of longitudinal mediation. Psychological Methods. 2007;12(1):23–44. doi: 10.1037/1082-989X.12.1.23. [DOI] [PubMed] [Google Scholar]
  37. McAuley E, Blissmer B. Self-efficacy determinants and consequences of physical activity. Exercise and Sport Science Reviews. 2000;28:85–88. [PubMed] [Google Scholar]
  38. McAuley E, Mihalko SL. Measuring exercise-related self-efficacy. In: Duda JL, editor. Advances in sport and exercise psychology measurement. Morgantown, WV: Fitness Information Technology; 1998. pp. 371–390. [Google Scholar]
  39. Mosher CE, Fuemmeler BF, Sloane R, Kraus WE, Lobach DF, Snyder DC, Demark-Wahnefried W. Change in self-efficacy partially mediates the effects of the FRESH START intervention on cancer survivors’ dietary outcomes. Psycho-Oncology. 2008;17(10):1014–1023. doi: 10.1002/pon.1327. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Muthén LK, Muthén BO. Mplus (Version 7.2) [Computer Software] Los Angeles, CA: Muthén & Muthén; 2014. [Google Scholar]
  41. Pearl J. Direct and indirect effects. Paper presented at the Proceedings of the seventeenth conference on uncertainty in artificial intelligence 2001 [Google Scholar]
  42. Pender NJ, Bar-Or O, Wilk B, Mitchell S. Self-efficacy and perceived exertion of girls during exercise. Nursing Research. 2002;51(2):86–91. doi: 10.1097/00006199-200203000-00004. [DOI] [PubMed] [Google Scholar]
  43. Pennebaker JW. The psychology of physical symptoms. New York, NY: Springer-Verlag; 1982. [Google Scholar]
  44. Pinto BM, Rabin C, Dunsiger S. Home-based exercise among cancer survivors: adherence and its predictors. Psycho-Oncology. 2009;18(4):369–376. doi: 10.1002/pon.1465. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Plotnikoff RC, Lippke S, Courneya KS, Birkett N, Sigal RJ. Physical activity and social cognitive theory: A test in a population of adults with type 1 and 2 diabetes. Applied Psychology: An International Review. 2008;57(4):628–643. [Google Scholar]
  46. Prochaska JO, DiClemente CC. Toward a Comprehensive Model of Change. In: Miller WR, Heather N, editors. Treating Addictive Behaviors: Processes of Change. New York, NY: Plenum Press; 1986. pp. 3–27. [Google Scholar]
  47. Rhodes RE, Martin AD, Taunton JE. Temporal relationships of self-efficacy and social support as predictors of adherence in a 6-month strength-training program for older women. Perceptual and Motor Skills. 2001;93(3):693–703. doi: 10.2466/pms.2001.93.3.693. [DOI] [PubMed] [Google Scholar]
  48. Rhodes RE, Pfaeffli LA. Review Mediators of physical activity behaviour change among adult non-clinical populations: a review update. International Journal of Behavioral Nutrition and Physical Activity. 2010;7:37–48. doi: 10.1186/1479-5868-7-37. [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Rock CL, Doyle C, Demark-Wahnefried W, Meyerhardt J, Courneya KS, Schwartz AL, McCullough M. Nutrition and physical activity guidelines for cancer survivors. CA: A Cancer Journal for Clinicians. 2012;62(4):242–274. doi: 10.3322/caac.21142. [DOI] [PubMed] [Google Scholar]
  50. Preacher KJ, Coffman DL. Computing power and minimum sample size for RMSEA [Computer software] 2006 May; Available from http://quantpsy.org/
  51. Sallis JF, Grossman RM, Pinski RB, Patterson TL, Nader PR. The development of scales to measure social support for diet and exercise behaviors. Preventive Medicine. 1987;16(6):825–836. doi: 10.1016/0091-7435(87)90022-3. [DOI] [PubMed] [Google Scholar]
  52. Schmitz K, Courneya K, Matthews C, Demark-Wahnefried W, Galvão D, Pinto B, Lucia A. American College of Sports Medicine roundtable on exercise guidelines for cancer survivors. Medicine and Science in Sports and Exercise. 2010;42(7):1409–1426. doi: 10.1249/MSS.0b013e3181e0c112. [DOI] [PubMed] [Google Scholar]
  53. Smith WA, Nolan VG, Robison LL, Hudson MM, Ness KK. Physical activity among cancer survivors and those with no history of cancer—a report from the National Health and Nutrition Examination Survey 2003–2006. American journal of translational research. 2011;3(4):342–350. [PMC free article] [PubMed] [Google Scholar]
  54. Speck RM, Courneya KS, Mâsse LC, Duval S, Schmitz KH. An update of controlled physical activity trials in cancer survivors: a systematic review and meta-analysis. Journal of Cancer Survivorship. 2010;4(2):87–100. doi: 10.1007/s11764-009-0110-5. [DOI] [PubMed] [Google Scholar]
  55. Stanton AL, Luecken LJ, MacKinnon DP, Thompson EH. Mechanisms in psychosocial interventions for adults living with cancer: Opportunity for integration of theory, research, and practice. Journal of Consulting and Clinical Psychology. 2013;81(2):318–335. doi: 10.1037/a0028833. [DOI] [PubMed] [Google Scholar]
  56. Tan X, Shiyko MP, Li R, Li Y, Dierker L. A time-varying effect model for intensive longitudinal data. Psychological Methods. 2012;17(1):61–77. doi: 10.1037/a0025814. [DOI] [PMC free article] [PubMed] [Google Scholar]
  57. U.S. Department of Health and Human Services. Physcial activity and health: A report of the Surgeon General. Atlanta, GA: Centers for Disease Control, National Center for Chronic Disease Prevention and Health Promotion; 1996. [Google Scholar]
  58. Vallance JK, Courneya KS, Plotnikoff RC, Mackey JR. Analyzing theoretical mechanisms of physical activity behavior change in breast cancer survivors: results from the activity promotion (ACTION) trial. Annals of Behavioral Medicine. 2008;35(2):150–158. doi: 10.1007/s12160-008-9019-x. [DOI] [PubMed] [Google Scholar]
  59. Warner LM, Schüz B, Wolff JK, Parschau L, Wurm S, Schwarzer R. Sources of Self-Efficacy for Physical Activity. Health Psychology. 2014 doi: 10.1037/hea0000085. Advance online publication. http://dx.doi.org/10.1037/hea0000085. [DOI] [PubMed]
  60. Voskuil DW, Monninkhof EM, Elias SG, Vlems FA, van Leeuwen FE. Physical activity and endometrial cancer risk, a systematic review of current evidence. Cancer Epidemiology Biomarkers & Prevention. 2007;16(4):639–648. doi: 10.1158/1055-9965.EPI-06-0742. [DOI] [PubMed] [Google Scholar]
  61. Williams DM, Anderson ES, Winett RA. A review of the outcome expectancy construct in physical activity research. Annals of Behavioral Medicine. 2005;29(1):70–79. doi: 10.1207/s15324796abm2901_10. [DOI] [PubMed] [Google Scholar]

RESOURCES