Predictors of adherence to exercise interventions during and after cancer treatment: A systematic review

Abstract Objective Exercise interventions benefit cancer patients. However, only low numbers of patients adhere to these interventions. This review aimed to identify predictors of exercise intervention adherence in patients with cancer, during and after multimodality cancer treatment. Methods A literature search was performed using electronic databases (PubMed, Embase, and Cochrane) to identify relevant papers published before February 1, 2017. Papers reporting randomized controlled trials, conducted in adult cancer patients who participated in an exercise intervention during and/or after multimodality cancer treatment, and providing outcome of factors predicting exercise adherence were included. Papers were assessed for methodological quality by using the Physiotherapy Evidence Database scale. Results The search identified 720 potentially relevant papers, of which 15 fulfilled the eligibility criteria. In these 15 studies, 2279 patients were included and 1383 of these patients were randomized to an exercise intervention. During cancer treatment, the factors predicting exercise adherence were as follows: location of the rehabilitation center, extensive exercise history, high motivation for exercise, and fewer exercise limitations. After cancer treatment, factors that predicted adherence were as follows: less extensive surgery, low alcohol consumption, high previous exercise adherence, family support, feedback by trainers, and knowledge and skills of exercise. Methodological quality of the included papers was rated “high”. Conclusions The most prominent predictors of adherence to exercise interventions were location of the rehabilitation center, extensive exercise history, high motivation for exercise, and fewer exercise limitations. To increase the number of cancer patients who will benefit, these results should be considered into the development and implementation of future exercise interventions.

treatment can result in deterioration of physical fitness, decreased muscle strength, fatigue, and a reduced quality of life. [6][7][8] Cancer treatment can also result in inactivity and weight gain, as previously described in patients diagnosed with breast cancer, prostate cancer, testicular cancer, and leukemia. [9][10][11] Moreover, cancer survivors frequently experience long-term adverse events related to the cancer treatment such as the development of metabolic syndrome and subsequent cardiovascular disease. [12][13][14] Evidence is accumulating that physical exercise complementary to cancer treatment is safe and feasible. 15,16 Encouraging effects of exercise interventions to improve lifestyle in patients with various cancer diagnoses have been reported. 15,17,18 In general, exercise interventions can alleviate common side effects of cancer treatment, for example, by increasing patients' physical fitness, improving quality of life, and reducing cancer-related fatigue. 6,7,17,19 In patients diagnosed with lymphoma, breast cancer, colorectal cancer, or prostate cancer, physical exercise may be associated with improved progression-free survival. [20][21][22] Importantly, an increase in physical exercise behavior and maintenance of this behavior after completion of cancer treatment may lower the risk of cancer recurrence, as reported in patients diagnosed with breast or prostate cancer. 23,24 In various cancer types, physical exercise appears to decrease disease-related morbidity and mortality. [24][25][26][27] A meta-analysis of 23 prospective studies in breast and colorectal cancer survivors found that engaging in at least 150 minutes of moderate to vigorous intensity physical exercise was associated with a reduction in the risk of overall mortality of approximately 24% compared to being less physically active. 24 These benefits are comparable to the effect of smoking cessation on reducing the risk of cancer mortality. 28 Behavioral change, focused on adaption of a healthier lifestyle, is complicated. A cancer diagnosis and subsequent treatment may potentially motivate patients to change their lifestyle (eg, to become more active, follow a healthier diet, or quit smoking). [29][30][31] In observational studies, however, a decrease in patients' physical exercise frequency was found after being diagnosed with breast cancer and this effect was more distinct in obese, sedentary, and elderly patients. 32,33 Accumulating data on the negative effects of being overweight on the development of cancer and cancer survival fuel the sense of urgency for successful interventions to enhance a healthy lifestyle. 8,21,34 Unfortunately, low adherence to the interventions and limited recruitment rates are frequently reported in studies investigating exercise interventions in cancer patients, both during and after cancer treatment. 35,36 Several barriers to physical exercise (eg, fatigue, time restraints, and discomfort) have been reported. 35,37,38 Understanding which factors predict adherence to exercise interventions is essential to identify patients that are intending to increase their physical exercise intensity but who are at risk of nonadherence. Identifying predictors of exercise adherence can contribute to an increased number of cancer patients participating in exercise interventions, with potential benefits in cancer outcome. 36,39 The aim of this review is to identify predictors of adherence to exercise interventions in patients with cancer, during and after multimodality cancer treatment. This knowledge will help optimize implementation strategies and eventually help in improving cancer treatment outcome.

| Design
A systematic review was performed to identify predictors of adherence to exercise interventions from randomized controlled trials (RCTs) and to discuss the methodological quality and results of included papers. This systematic review was conducted in accordance with Preferred Reporting Items of Systematic Reviews and Meta-Analyses (PRISMA) guidelines. 40

| Literature search
A literature search was performed using electronic databases (PubMed, Embase, and Cochrane) to identify relevant paper published before February 1, 2017. The complete search including Medical Subject Headings (MeSH) terms and keywords is described in Tables S1 and S2. In addition, reference tracking of all papers was performed. Full papers were eligible for inclusion when they reported an RCT design, were conducted in adult cancer patients who participated in a physical exercise intervention during or after systemic (neo-) adjuvant cancer treatment, provided outcome of factors predicting exercise intervention adherence, and were written in English. An exercise intervention was defined as exercise interventions involving any physical movement produced by skeletal muscles that require energy expenditure 41 ; that were planned, structured, and repetitive; that were of at least moderate to vigorous intensity; and that were aimed to improve or maintain physical fitness over a predetermined time period. 42 Pilot studies, case studies, and papers of low methodological quality were not included.

| Selection of studies
Selected papers were screened based on title and abstract. In cases when titles and abstracts implied that a paper was potentially eligible for inclusion, a full paper copy of the report was obtained and evaluated for inclusion.

| Data extraction and assessment of methodological quality
Data were extracted using a predetermined extraction form and in accordance with PRISMA guidelines. 40 Data extracted were as follows: (1) first author's last name, year of publication, country, and trial name; (2) design; (3) population (number of participants, gender, age, cancer type(s), stage, and treatment modalities); (4) exercise intervention (extent, duration, type, frequency, treatment phase, intensity, adherence facilitation, and control group program);  43 The scale is composed of 11 items, of which the first item is only applicable for specification of eligibility criteria and is not considered as part of calculating the overall PEDro score. Studies scored one point for each item present and could score between 0 to 10 points. Studies that scored ≥4 points were classified as "high" quality and studies that scored <4 points were considered to be of "low" methodological quality. 44 Disagreement between the 2 investigators regarding a papers' quality score was resolved by discussion with a third investigator (A.W.) until consensus was reached. Cohen's Kappa and percentage of agreement on methodological quality were calculated.

| Characteristics of included studies
In total, 2279 cancer patients were included in the 15 studies analyzed. [45][46][47][48][49][50][51][52][53][54][55][56][57][58][59] Of these patients, 1383 were assigned to an exercise intervention and these patients had a mean age of 55.5 years. All studies used an RCT design and were conducted in the United States, Canada, Australia, the Netherlands, Germany, or Taiwan. Eligibility criteria were heterogeneous among studies, with differences in cancer type(s), cancer treatment phase, exercise interventions, and patient characteristics. A full description of the different study characteristics is depicted in Table 1.

Identification
Articles identified through database searching (n = 720) Articles screened (n = 502) Full-text articles assessed for eligibility (n = 30) Studies included in qualitative synthesis (n = 15) FIGURE 1 Flowchart of the literature search strategy and study selection in the systematic review

| Assessment of methodological quality and quantitative analysis
The 15 included studies were scored using the PEDro scale.  48,50,52,53,58,59 The methodological quality assessment is summarized in Table 2. Unfortunately, a quantitative analysis by pooling outcome data (meta-analysis) or a best-evidence synthesis was inappropriate. This is due to incomparability of outcome data caused by heterogeneity of study sample characteristics (eg, divergent exercise interventions, patient characteristics, and outcome as summarized in Table 1).

| Univariable and multivariable analyses of selected studies
A wide range of predictive factors were investigated, which were classified as socio-demographic (eg, gender, marital status, education, employment, location of the rehabilitation center in relation to the residential area, family support, and feedback by trainers), medical (eg, cancer type, treatment regimen, pretreatment fatigue, and disease stage), physical and physiological (eg, physical fitness and body mass index) and behavioral factors (eg, exercise history, baseline self-efficacy, exercise motivation, smoking behavior, and alcohol consumption). Study results are depicted in Tables 3 and S3.
Highly significant (P ≤ .01) and significant (P < .05) or borderline significant (P < .10) associations between exercise intervention adherence and various factors were identified in univariable [45][46][47][48]50,52,53,[55][56][57][58][59] or bivariable analysis. 49,51 Thereafter, these factors were included in a multivariable analysis to finally derive predictors of adherence to exercise intervention. One study did not describe a univariable or bivariable analysis. 54 An overview of the significant predictors of adherence to exercise interventions during and after cancer treatment is TABLE 3 Overview of significant predictors of exercise intervention adherence found in multivariable analysis

During Treatment After Treatment
Exercise Intervention Adherence High Low High Low

Socio-demographic factors
Being married 49 51 Gender (male) 59a 59a Close location/center 46,47 Having children at home 55a More knowledge and skills of exercises 51 High intensity exercise group assignment 46 More family support 51 More feedback by trainers 51 Low employment status 58a 58a Medical factors Extensive treatment protocol 46; 58a 50; 58a Pretreatment fatigue 49; 55a Advanced disease stage 47 Cancer types other than breast cancer 52 Low psychological distress 52 Exercise limitations due to cancer treatment 46 Endocrine symptoms 46 High depression 46,47 Physiological and physical factors High physical fitness 49 High age 45; 57a 48 57a High VO 2 peak [45][46][47] High submaximal endurance capacity 55a Low BMI 53 Behavioral factors High exercise stage of change 48 High exercise history 57a 54; 55a; 57a; 59a High self-efficacy 52; 54 Being a nonsmoker 52 High previous exercise adherence 51; 55a High alcohol consumption 50 High exercise motivation 58a 53; 58a High role functioning 56a 56a High mid-treatment mood disturbance 49 a Exercise intervention covered both time periods, during and after treatment.
Cancer type: Black, multiple cancer types; Red, breast cancer; Blue, prostate cancer; Purple, head and neck cancer; Orange, lymphoma; Green, colorectal cancer.
Abbreviations: VO 2 peak, peak rate of oxygen consumption during incremental exercise; BMI, body mass index. summarized in Table 3. Adherence rates ranged from 61.9% to 91.0%.
The R 2 , defined as the percentage of variance explained by the model, was reported in 9 studies [46][47][48][49][50][51]55,[57][58][59]  Awareness of the importance of physical exercise not only in cancer treatment but also in other chronic diseases, such as chronic obstructive pulmonary disease and diabetes, has increased over the past years. 72,73 Consequently, the number of RCTs investigating exercise interventions during and after cancer treatment has increased.
However, data on predictors of adherence to the exercise intervention are often not described in these RCTs, particularly in those performed in a home-based setting. 74 One of the strengths of this systematic review is that all studies were of "high" methodological quality. This is in contrast to methodological quality assessment of a previous systematic review, assessing determinants of exercise adherence and maintenance. 75 This difference may be due to the use of a methodological quality assessment tool that was adapted from existing quality criteria lists compiled by Kampshoff et al, 75 whereas we pursued the PRISMA guidelines for reporting systematic reviews and used the PEDro scale, which is especially designed for assessment of clinical trials. 40,43 The internal validity of our review is partially warranted by limiting the inclusion to randomized studies. 76 A systematic review by Husebø et al demonstrated that several psychological factors predicted exercise intervention adherence. However, socio-demographic, medical or physical, and physiological factors were not investigated. 69 In contrast, our review indicated that psychological factors only partially predicted exercise intervention adherence and suggest a more important role for socio-demographic, medical or physical, and physiological factors, such as fewer exercise limitations due to cancer treatment, pretreatment fatigue or high VO 2 peak levels.

| Study limitations
A limitation of our review was the relatively low number of RCTs included despite the extensive literature search. Few RCTs that investigated predictors of exercise intervention adherence during and after cancer treatment and met our inclusion criteria were identified. Grey literature was not considered in the literature search. The possibility that an RCT fulfilling our inclusion and exclusion criteria was conducted but not reported in the scientific literature was estimated to be very small. We were unable to perform a quantitative analysis or a best-evidence synthesis, due to the heterogeneity of the data.

| Clinical implications and conclusions
In summary, recommendations for future trials include the use of equivalent measuring instruments in future RCTs to facilitate a more homogeneous analysis across studies. We recommend future RCTs to report predictors of exercise intervention adherence and to use objective measurement instruments such as attendance records and validated wearable activity trackers (eg, accelerometers). This facilitates the comparison of studies investigating predictors of exercise intervention adherence during and after multimodality cancer treatment. 74 Hence, the power of generated data in the field of exercise oncology will increase. Furthermore, we recommend the analysis and reporting of potential preexistent factors that may impede adherence to and participation in an exercise intervention in clinical practice. In this manner, patients less likely to adhere can be offered a personalized exercise intervention and extra guidance, by means of, eg, prolonged coaching to facilitate exercise adherence. 77,78 These approaches might result in optimizing participation in exercise interventions and retaining the less motivated, less fit patients who will potentially benefit most. 25 Since it is increasingly recognized that exercise interventions should be included in the treatment of cancer patients, predictors of exercise intervention adherence should be taken into account when composing these interventions.