Searches were performed in the PubMed and Sport Discus databases using the key words “cancer survivor” with “exercise,” “exercise implementation,” “evidence-based exercise,” “exercise translation,” “exercise health promotion,” and “exercise rehabilitation.” The initial time frame for publications was January 1980 to June 2012. The timeframe was selected because the earlier systematic review by White et al. [3] included publications through 2008. In contrast to White, our search was broader in that it was not limited to a specific cancer type or to RCTs. Additional searches of PubMed (US National Library of Medicine, Bethesda, MD), the Cumulative Index to Nursing and Allied Health Literature (CINAHL), and the Library of Congress (Washington, DC) databases were completed by a research librarian to confirm the results of the initial search. The citation collection was archived using End-Note software (Thomson Reuters, 2012; New York, NY).

The abstract of each citation in the collection was reviewed using a priori criteria. The inclusion criteria for abstracts were that the survivor population was described (e.g., race, sex, cancer type) and exercise of any type was the main component of the intervention. Additionally, at least two of the following criteria were required: (1) the delivery and/or setting of the intervention (e.g., individual or group-based delivery, home or health care setting); (2) duration of the intervention (e.g., weeks); and (3) intensity of exercise (e.g., mild, moderate, high, or relative to peak performance) or frequency of exercise (e.g., number of days per week). If an abstract was very brief but the title suggested that it might meet the inclusion criteria, the full manuscript was reviewed. Publications that described study design and methodology, but not the results, of RCTs were retained for further consideration. The rationale for including studies in progress was to survey how investigators were planning to incorporate measures of maintenance in their research that is likely to be reported in the next few years. After this initial screening step, we retained articles that focused on the maintenance of exercise programs for adult cancer survivors in publications from January 2009 to June 2012 by adding the keyword “maintenance” to this subgroup of abstracts.

In keeping with scoping methodology, our initial search strategy included grey literature (i.e., sources other than biomedical databases). The identified citations were either secondary reporting of published studies or did not meet inclusion criteria. Therefore, we discontinued further searches of the grey literature.

Abstracts were screened using a two-step process. In the first screening step, teams from the University of Colorado and Texas A&M Health Science Center independently reviewed the abstracts to determine if inclusion criteria were met. The teams compared their decisions and, when discordant, reached consensus through discussion. In the second screening step, abstracts were judged on “maintenance,” defined as (1) a measure of follow-up of any duration after the exercise program/intervention ended (individual level of maintenance), (2) measurement of continuation/discontinuance (or of intent to measure follow-up in methods papers), or (3) observations about follow-up at the individual or setting level.

The retained publications were coded using a form developed by the research team in an iterative process. The coding form included variables adopted from the RE-AIM Framework Construct Checklist (http://cancercontrol.cancer.gov/IS). The study team tested the form by coding five retained publications in duplicate and commenting on the form. The coding form was revised by three investigators (CMJ, MO, SAB) and the remaining publications were coded one time. Select fields in the completed coding forms were reviewed against the original publication for accuracy and corrections were made prior to data synthesis. If a publication included maintenance data from a previously published study (e.g., secondary outcomes from the seminal publication), the seminal article was retrieved, coded, and included as background information.

Maintenance was addressed on the individual (i.e., participant) and setting (i.e., organizational) levels. On the individual level, maintenance was described by the time interval from the final active intervention visit, the number of data collection time points in the maintenance interval, the primary outcome(s) of the maintenance phase, and the significance of the statistical test(s). An intervention effect was considered “maintained” if the change in the outcome was significantly different from baseline (i.e., baseline versus post-intervention) and not significant during the follow-up period (i.e., post-intervention versus end of follow-up).

Study attrition was defined as the percent of those participants enrolled who were absent in the maintenance phase and reported as the median and range for all studies. Intervention attrition was calculated as the percent of participants enrolled who were absent at the end of the intervention. Maintenance on the setting level was evaluated by the setting, outcomes, and characteristics including whether the program or intervention was ongoing, adapted for the long-term, aligned to an organizational mission, or if the sustainability of a business model (i.e., sources of funding, cost to participants, cost of program) were discussed (http://cancercontrol.cancer.gov/IS).

Of the 377 publications identified in the database searches, 211 abstracts were reviewed for eligibility based on patient and exercise intervention criteria (100 RCTs and 111 other types of studies). This subset was then filtered using the maintenance criteria, leaving 24 publications (19 RCT publications) in the collection. One RCT yielded 2 publications with maintenance data [12, 13], thus the 19 publications in the RCT collection represented 18 unique trials or 8.5 % of studies that reported on maintenance of some type. In addition to the RCTs, we included 5 publications that used qualitative methods, longitudinal surveys, or a secondary regression analysis of results from a RCT (Fig. 2).

Twelve publications represented completed RCTs and seven described planned RCTs. Among the retained RCT publications, mixed cancer types and breast cancer were in the majority (Fig. 2). Of the completed RCTs (Fig. 3a, Table 1), four interventions were home-based [14–17], four were supervised and conducted in an academic health care setting [12, 13, 18, 19], and the others were supervised and home-based exercise [20, 21], group counseling with home-based exercise [22], or supervised exercise with or without group counseling in a clinical setting [23]. The median (range) duration of active intervention intervals (intervention) was 3 months (1.5–10) and duration of the maintenance intervals was 6 months (3–14). Two studies measured maintenance at two time points [16, 23]. The most common primary outcome of the interventions was minutes of exercise, followed by physical function, menopausal symptoms, quality of life, weight loss, and fatigue.

We included five publications that provided contextual information about the maintenance of exercise effects on the individual level [31–35]. One of these publications [35] also included maintenance on the setting level. Due to the diversity of these publications, we present the contextual aspects of each study. Courneya et al. [33] identified several independent predictors of exercise maintenance in breast cancer survivors 6 months after a 4-week, supervised aerobic or resistance exercise training intervention during chemotherapy. Women who were younger (<50 years), exercised more before enrollment, had breast-conserving surgery, greater strength gains from training, a more positive attitude about resistance training, less fatigue, and a lower BMI at the end of training were more likely to maintain exercise [33].

Two longitudinal survey studies [31, 32] provided insight about trends in physical activity up to 3 years after diagnosis. Change in physical activity (MET-hours/week) was assessed for 30 months in women diagnosed with primary invasive breast cancer and increased risk of lymphedema [31]. In the 12 months after diagnosis, physical activity decreased about 50 %. Between 12 and 18 months post-diagnosis, physical activity increased for low, moderate, and vigorous physical activity, and then remained stable from 18 to 30 months post-diagnosis. Women who were younger (<40 years), had a BMI <25, and were more active before cancer diagnosis had the greatest reduction in physical activity and did not rebound to their pre-diagnosis physical activity level.

Chambers et al. [32] surveyed colorectal cancer survivors 5, 12, 24, and 36 months after diagnosis to determine the relationship between psychological distress and physical activity over time. At study entry, higher levels of somatization and anxiety were associated with inactivity. Over time, psychological distress did not motivate colorectal cancer survivors to become more physically active. The authors [32] suggested that cancer survivors with high levels of anxiety after diagnosis and ongoing somatization may benefit from interventions that combine psycho-education, exercise, and other lifestyle changes that mitigate the physical symptoms of stress.

To better understand self-determination and illness resistance in cancer survivors, Midtgaard et al. [34] studied a subset of survivors who were sedentary prior to enrolling in the Copenhagen PACT study and then exercised at least 3 days per week for the next 18 months. Five categories relevant to the maintenance of exercise behavior emerged from interviews with survivors. Survivors who maintained exercise behaviors (1) were able to decide on a new agenda in life, (2) saw physical activity as an act of autonomy, (3) acknowledged the importance of goal setting, (4) prioritized physical activity, and (5) tamed fear.

Evidence for setting-level maintenance was scarce. Fitzpatrick et al. [35] included setting- and individual-level maintenance perspectives in their study of wellness programs for cancer survivors. Their study was driven by underutilization of a wellness center for cancer survivors during its first year of operation (Hope & Cope Wellness Center, Montreal, Canada). One-third of the survivor registrants (148 of 464) never returned to participate and nearly half of registrants participated less than three times in 1 year. For a comparison of utilization trends, 18 other wellness centers for cancer survivors in the USA and Canada were contacted. A consistent finding was that one-third of registrants attended less than three times. Anecdotally, the centers’ staff attributed poor attendance to participants’ health status, medical appointments, and family responsibilities; weather; and suitability of the survivor groups. These were primarily individual-level (i.e. survivor-related) reasons. Setting-level reasons for under-utilization included inadequate recording of attendance, lack of referrals from community organizations, competition from cancer support groups, and insufficient financial resources for services. Insufficient financial resources suggest lack of sustainability of the business model. This is the only evidence in our review of attention to sustainability of a business model supporting an exercise program for cancer survivors.

Fitzpatrick and colleagues [35] then conducted telephone interviews with participants who never (n=16) or seldom (n=17) attended the Hope & Cope Wellness Center. From these interviews, several setting-level strategies to improve attendance were proposed. First, a consistent system of follow-up communication between staff and infrequent users should be in place, including a standardized method for recording survivors’ reasons for not attending. The responses could be used to modify programs and attendance monitored for the effectiveness of the modifications. Second, improve accessibility by linking survivors with appropriate community fitness programs; provide virtual support (online modules, links, chat rooms); and extend hours of operation. Third, provide support and services for family caregivers using center-based and virtual delivery. These findings and strategies addressed the RE-AIM components by assessing the settings of the wellness programs, with the focus on utilization in this case, and adapting programs for the long-term.