A preliminary trial examining a ‘real world’ approach for increasing physical activity among breast cancer survivors: findings from project MOVE

Project MOVE’s trial rationale and protocol have been described elsewhere [21]. In brief, this study is based on a quasi-experimental pre-post design. Baseline, six-month and one-year follow-up measures were undertaken between May 2015 and March 2017. Minor deviations were made from the original protocol due to the pragmatic nature of this trial. For instance, the analytical framework was modified and some of the findings (e.g., social support) have been reported elsewhere [19, 20]. Informed written consent was obtained prior to baseline assessments from all participants. This study was approved by the Behavioural Research Ethics Board at the University of British Columbia (#H14–02502).

For the purpose of this study, a survivor is defined based on the National Coalition for Cancer Survivorship as someone who has lived with, through and beyond a cancer diagnosis [22]. Groups (8–12 per group) of women (18+ years) who self-defined as BC survivors living in the Okanagan region of British Columbia, Canada were eligible to participate. Based on interests from women to engage in physical activity with both other survivors and healthy women of similar ages, Project MOVE team members adjusted the recruitment eligibility during the initial recruitment phase so that groups comprised of at least 50% BC survivors were eligible. Women living in the Okanagan who wished to participate but were not BC survivors were eligible providing there was space in the groups after all interested BC survivors were accommodated.

A variety of recruitment techniques were employed, including meetings with cancer -related community organizations (e.g., BC Cancer Agency), a project specific website, news items in the local media (e.g., newspapers, online forums) social media announcements (Facebook and Twitter), posters distributed to local businesses, community centers and medical clinics and attendance at the annual Run for the Cure fundraising event.

There was no pre-determined initiative (aka “intervention”) promoted or developed by the researchers. Instead each applicant group was invited to design its own PA intervention and apply for up to $2000 to enable access to facilities, equipment, tools and resources, instruction and/or transportation to implement the intervention. Groups were encouraged to contact members of the research team if they needed support throughout the application process to help conceptualize their project.

For the application process, each group designated a leader who acted as the primary contact and was responsible for submitting an online microgrant application. Each group leader submitted an application with details regarding the PA their group planned to do each week, how this activity would contribute to increasing the group’s overall PA levels, and outline a proposed budget and timeline. All submitted applications were initially screened for eligibility. Those deemed eligible were then processed and further evaluated by a Grant Review Panel consisting of three members from the research team, a representative from the Canadian Cancer Society and a local BC survivor who was not part of a group submitting a microgrant application. Successful applicant groups were notified by email and were informed of program obligations (i.e., participate in data collection, provide final report). Unsuccessful applicants were also notified and provided feedback outlining reasons for the not funded decision. Each group that received a microgrant was also informed that if they increased their group’s mean PA at the 6 month follow-up (assessed via accelerometry), they would receive an additional $500 incentive.

Demographic data were collected at baseline and all other data (i.e., BC information, PA behavior, PA motivation, and QoL) were collected at baseline, 6 months and 12 months, at a location and time convenient for each Project MOVE group (e.g., community center, local fitness center). At these data collection sessions, participants were given accelerometers and accelerometer log sheets, instructed on how to use each item, and asked that the accelerometers and log sheets be returned to their group’s primary contact person after the required wear time. Members of the research team were available to provide instructions and answer participant’s questions during data collection. Participants who were unable to attend the group measurement session were followed up by a research team member to schedule an alternative location (e.g., participants home) and time to complete measurements and receive an accelerometer. All measures were undertaken by two research team members, who collected all data at each time point.

All statistical analyses were conducted using SPSS version 22. Descriptive analyses were completed and presented as means and standard deviations (SD) for continuous variables and as frequencies and proportions for categorical data. Intention-to-treat analyses were conducted with the last known measurement carried forward for those who dropped out. Baseline accelerometer measures for participants who were outliers, or accumulated insufficient wear time at 6-month and 12-month follow-up were carried forward. Those with outlier measurements, as defined by +/− three standard deviations from the mean, at baseline were excluded from the analysis. PA related questions were analyzed using repeated measures ANOVA across the three times points (baseline vs. 6-month vs. 12-month follow-up) with weekly minutes of accelerometer MVPA as the dependent variable. Subsequent repeated measures mixed ANOVA was also conducted with meeting MVPA guidelines (meeting vs. not meeting) as the between-subjects variable, time as the within-subjects variable and weekly minutes of accelerometer MVPA as the dependent variable. Differences in MVPA change between groups were assessed by three one-way ANOVAs with group as the independent variable and MVPA change (baseline to 6-month change, 6-month to 12-month change, baseline to 12-month change) as the dependent variables and a Bonferroni correction applied to the three analysis to adjust for repeated analyses on the same data. For PA motivation and QoL questions repeated measures ANOVA were conducted with time as the within-subject variable. Repeated measures ANOVAs used polynomial contrasts to describe trends over time and one-way ANOVAs used post-hoc t-tests. All analyses assumptions were tested, and in instances of sphericity violations Greenhouse-Geisser or Huynh-Feldt corrections were used depending on epsilon values.

Given the overarching aim of Project MOVE was feasibility and the nature of this paper was to explore an estimate of effectiveness, a power calculation was not conducted, which is common practice amongst feasibility and pilot trails [37‐39].

At baseline, one participant had insufficient accelerometer wear-time, resulting in valid data from 87 participants. At 6-month and 12-month follow-up, four participants had insufficient wear-time and had observations carried forward. At baseline 48 (54.5%) participants were meeting MVPA guidelines and 39 (44.3%) were not, at 6 months follow-up 56 (64.4%) were meeting MVPA guidelines and at 12 month follow-up 40 (46.0%) were meeting MVPA guidelines. Repeated measures ANOVA showed no significant differences between time points on weekly light activity (F(2,174) = 2.37, p = .096). Repeated measures ANOVA showed significant differences and a small to medium effect size between time points on weekly MPVA (F(2,170) = 3.62, p = .03, eta² = .041). Polynomial contrasts showed a significant quadratic trend (F(1,85) = 6.32, p = .01). See Table 3 for means and standard deviations across time points.

Mixed ANOVA further indicated a significant main effect and small effect size between time points on weekly MVPA (F(2,168) = 3.84, p = .02, eta² = .012), but also showed a significant interaction and small effect size between those meeting MVPA guidelines at baseline and those not on weekly MVPA between time points (F(2,168) = 5.61, p = .004, eta² = .02). Polynomial contrasts showed a significant increasing linear trend (F(1,84) = 7.55, p = .007). Descriptive statistics can be seen in Table 3.

One-way ANOVA with group as the independent variable and MVPA change between baseline and 6-month follow up showed no significant group differences (F(9,85) = 1.89, p = .066, eta² = .183). One-way ANOVA with group as the independent variable and MVPA change between 6-month and 12-month follow up showed no significant group differences (F(9,85) = 1.86, p = .07, eta² = .177). One-way ANOVA with group as the independent variable and MVPA change between baseline and 12-month follow up showed no significant group differences (F(9,85) = 1.09, p = .382, eta² = .114). Descriptive statistics by group can be seen in Table 4 and MVPA means by time point by group can be seen in Fig. 1.

Repeated measures mixed ANOVAs, with Huynh-Feldt corrections when applicable, were conducted to assess changes in motivation regulations across time points. There was a significant change and very small effect size for intrinsic regulation (F(2, 168) = 3.95, p = .02, eta² = .0006) and a significant interaction between those meeting and those not meeting MVPA guidelines at baseline (F(2,168) = 3.41, p = .035, eta² = .003). Polynomial contrasts showed a significant linear trend (F(1,84) = .98, p = .02). Descriptive statistics can be seen in Table 3.

Repeated measure mixed ANOVA shows significant differences and small effect size across time points on physical functioning (F(1.32,110.36) = 17.83, p < .001, eta² = .028), with no significant interaction, polynomial contrasts show significant linear (F(1,82) = 14.15, p < .001) and quadratic trends (F(1,82) = 26.16, p < .001). Role functioning-physical shows significant differences and small effect size across time points on role functioning-physical (F(2,164) = 7.82, p = .001, eta² = .037), mediated by a significant interaction between those meeting and those not meeting MVPA guidelines at baseline (F(2,164) = 3.79, p = .024, p = .024, eta² = .014), a significant quadratic trend was found (F(1,82) = 6.78, p = .01). Role-functioning emotional shows significant differences and small effect size across time points (F(2,162) = 3.86, p = .023, eta² = .015) with no significant interaction, and a significant liner trend (F(1,81) = 4.8, p = .03). Social functioning shows significant differences and small effect size across time points (F(1.71, 140.54) = 7.80, p = .001, eta² = .027) with no significant interaction and significant linear (F(1,82) = 5.87, p = .019,) and quadratic trends (F(1,82) = 10.59, p = .002). General health shows significant differences and small effect size across time points (F(1.68, 137.90) = 6.27, p = .004, eta² = .013) with no significant interaction, and significant linear (F(1,82) = 5.42, p = .02) and quadratic trends (F(1,82) =7.84, p = .006). Descriptive statistics are presented in Table 3.

Unique to Project MOVE was the use of the microgrants and a financial incentive. The program was designed to encourage BC survivors to work together to create a strategy that met their specific needs and interests. Contrary to contemporary PA interventions that provide a more structured and pre-determined design, the Project MOVE model represents a highly flexible and adaptable community-based model that requires relatively low follow-up. Additionally, the use of accelerometers as an objective measure of PA as well as a long follow-up period is also a strength of this study. Previous research with this population has often utilized self-report measures of PA and shorter follow-up periods (< 6 months). Moreover, this is the first study to evaluate PA behavior change within in a microgrant model. Previous research has been limited to descriptive research and process evaluations concerning the uptake and implementation of such models [15‐18].

Eligibility was modified to include additional women who wished to participate but were not BC survivors. Although this limits the generalizability of these findings, the inclusion of non-BC female support (i.e., friends and/or relatives) is an interesting strategy to increase participation among BC survivors. As many women may be hesitant to participate alone, allowing friends and/or relatives to join may help to engage more survivors as well as provide an opportunity to strengthen existing support structures. Furthermore, this may also act as a prevention strategy for those who have never been diagnosed with cancer as these friends and/relatives would be exposed to the information and knowledge presented during each session concerning BC and physical and psychological benefits of PA. This study was also limited in terms of diversity, thus further affecting generalizability. The lack in gender, ethnic and social diversity of participants limits representation of the entire BC survivor population. However, it should be noted that the organic nature of the microgrant model (i.e., being able to develop your own intervention) does provide an opportunity for transferability to other cancer, disease, and general populations.

Further, there are also a number of study design and methodological limitations. First, given the exploratory nature of the study and aim of this specific paper (i.e., estimate of effect) a power calculation was not conducted. Second, issues surrounding the of use of accelerometers as a measure of PA are also identified, specifically concerning specificity and sensitivity of accelerometers in differentiating between different modes of physical activity (e.g., habitual activity vs. planned intervention activity) and identifying activity that is undetectable by accelerometers (e.g., flexibility training or some strength training). Third, the use of “last observation carried forward” has limitations [68], however the majority of participants with missing data had only one time point of data collected, making growth curve analysis inappropriate. Fourth, multilevel modeling was not used to analyze the data because we were interested in broad group-level trends rather than how individuals responded to the specific interventions. As such, the clustering of women within groups and over time was not accounted for statistically. This being said, future research could focus on the specific within- and between-person effects of microgrant frameworks on PA, motivation, and QoL. Lastly, we did not collect information concerning physical activity dose, adherence, or intervention variations specific to each group, which would be beneficial with understanding the meaningfulness (interpreted in the context of design, power, and analytical framework) of each separate group intervention.