Participants and procedure
We recruited breast cancer survivors from the Yonsei Severance Hospital (Seoul, Korea) between July and October 2018. Among breast cancer survivors who were on outpatient visits to the department of the breast cancer center, women who were approved by the doctor as suitable subjects to participate in the study were recruited. Participants were eligible if they were < 75 years of age, had histologically confirmed stage <Ⅳ breast cancer, completed treatments (i.e., surgery, radiotherapy, chemotherapy) within three months of study enrolment, and understood and provided written informed consent in Korean. Each participant completed a face-to-face survey with a trained interviewer. The study protocol was approved by the Institutional Review Board of Severance Hospital [No. 4-2018-0379]. All participants provided written informed consent.
Sociodemographic and clinical variables
Demographic and clinical variables were collected using a self-reported questionnaire and reviewing electrical medical records. Demographic information included age, education, income, and marital status. Clinical data included body mass index (BMI), breast cancer stage, treatment(s) received, surgery method, and surgery site.
Theory of planned behavior variables
We used TPB questionnaires developed by Courneya et al. [22, 23] and translated the questions resulting in a Korean version of the TPB questionnaire (K-TPB). The original questionnaire was translated into Korean by two independent bilingual experts who have doctoral degrees in public health and exercise physiology. After the translation, a panel (N = 3) consisting of nursing, public health, and exercise science professors reviewed each question and derived the first version of K-TPB. The K-TPB underwent backward translation into English by two independent scholars. Panels of experts compared the original TPB and the back-translated K-TPB. The panel discussion adjusted semantic nuance words or sentences to reflect cultural differences (K-TPB second version). We then conducted a pilot test to examine the validity of the final version with 200 BCS. Exploratory factor analysis indicated each subscale showed adequate internal consistency (Cronbach’s alpha > 0.6) (affective attitude = 0.88, instrument attitude = 0.79, subjective norms = 0.87, PBC = 0.65, intention = 0.75; Kaiser Meyer Olkin = 0.81; p < 0.001).
Attitude reflects the individual’s overall evaluations of performing the behavior. Six items assessed attitude, and participants responded on a seven-point Likert scale with the descriptors of extremely (1 and 7), quite (2 and 6), slightly (3 and 5), and neutral (4). Attitude was measured in two components including affective (i.e., enjoyable/unenjoyable, pleasurable/painful, and fun/boring) and instrumental (i.e., useful/useless, beneficial/harmful, and important/unimportant) attitudes. The statement preceded each item, ‘I think that for me to participate in regular physical activity (or exercise) over the next month would be…’. Internal consistency for affective and instrumental attitude was 0.87 and 0.76, respectively (see Supplementary Table).
Subjective norm reflects the perceptions of social pressure to perform the behavior. Subjective norm includes the more traditionally measured injunctive component (e.g., individual believes important others want them to perform the behavior) and a descriptive component (e.g., whether important others perform the behavior themselves). Four items assessed subjective norm, and participants responded on a seven-point Likert scale with the descriptors of extremely (1 and 7), quite (2 and 6), slightly (3 and 5), and neutral (4). The three injunctive norm items were preceded by ‘I think that if I engaged in regular physical activity (or exercise) over the next month, most people who are important to me would… (a) approve (1 = extremely disapproving to 7 = extremely approving), (b) support (1 = extremely unsupportive to 7 = extremely supportive), and (c) encourage (1 = extremely discouraging to 7 = extremely encouraging)’. The statement preceded the descriptive norm item, ‘I think that over the next month, most people who are important to me will regularly participate in physical activity (or exercise)’ (1 = strongly disagree to 7 = strongly disagree). In this study, we used the injunctive norm items and removed the descriptive norm item due to issues with convergent validity in the confirmatory factor analysis. As measured by Cronbach’s α, internal consistency for the injunctive norm items was 0.88 (see Supplementary Table).
PBC reflects the degree of personal control the individual has over performing the behavior. Three items assessed PBC. The statement preceded all items, ‘If you were really motivated…’, followed by (a) How much control would you have over doing regular physical activity (or exercise) over the next month? (1 = very little control to 7 = complete control), (b) Is it entirely up to you to participate in regular physical activity (or exercise) for the next month?’ (1 = no, not at all to 7 = entirely depends on me), and (c) How confident are you that you could do regular physical activity (or exercise) over the next month?’ (1 = no, not at all confident to 7 = complete confidence). The internal consistency for PBC was 0.62 (see Supplementary Table).
Intention
Intention is the most proximal determinant of behavior. Two items and one planning item measured intentions to engage in physical activity. The two items were (a) Do you intend to do regular PA (or exercise) over the next month? (1 = no, not really to 7 = strongly intend) and b) How motivated are you doing regular physical activity over the next month? (1 = not at all motivated to 7 = extremely motivated). The internal consistency of this scale was 0.82. The planning question was ‘Do you have plans for when, where, and the type of physical activity (or exercise) you will do in the next month?’ (1: no specific plans to 7: very detailed plans). The internal consistency for planning was 0.80 (see Supplementary Table).
Physical activity behavior was measured using the 16-item Korean version of the Global Physical Activity Questionnaire (GPAQ) developed by the World Health Organization [24]. The GPAQ asks participants about physical activities (including exercise) performed for at least 10 minutes during a typical week. Participants report on time spent being physically active at the (a) workplace, (b) during transportation, (c) during leisure, and sedentary behavior. Physical activity is expressed in metabolic equivalent values (METs) in accordance with the GPAQ analysis guideline (e.g., vigorous activity = 8METs, moderate activity = 4METs, transportation activity = 4METs). Among the physical activity parameters, leisure time moderate to vigorous physical activity (MVPA) was used for Examining for TPB model. The Korean version of the GPAQ has previously demonstrated evidence of reliability and validity [25].
Statistical analysis
Participants’ demographic, medical/clinical, physical activity, and TPB variables were summarized using descriptive statistics. Pearson’s correlations examined associations between TPB constructs and physical activity behavior constructs. Medium to large relationships were defined as r = 0.3 to 0.5 [26]. Statistical tests were two-sided (α = 0.05). We used SPSS 22.0 (IBM Corporation, Armonk, NY) for all statistical analyses. We also used PASW Statistics 19 (PASW Inc., Chicago, IL) and AMOS 19.0 (Small Waters Corp., Chicago, IL). The maximum likelihood estimation of SEM was conducted to allow for both an assessment of overall model fit and statistical significance tests for the size of each theoretical relation in the TPB model. Model fit was examined using a number of indices such as chi-square index, goodness-of-fit index (GFI), adjusted goodness-of-fit (AGFI), root mean square of approximation (RMSEA), and comparative fit index (CFI).