Study design
This study is a secondary cross-sectional analysis of data from the ROGY care trial [20].The ROGY care trial is a pragmatic, cluster randomized controlled trial that aimed to assess the effect of a Survivorship Care Plan (SCP) on information provision and post-treatment care for cancer patients [20]. Data from both trial arms were analyzed together, as it is assumed that the SCP did not affect lifestyle because it did provide minimal information on lifestyle. The questionnaire administered at 18 months post-diagnosis was used for the purpose of this study. The study was approved by the Medical Research Ethics Committee [20].
Study population and data collection
Eligibility criteria to participate in the ROGY Care trial were (1) newly diagnosed patients with either primary endometrial or ovarian cancer between April 2011 and October 2012, (2) curative treatment, (3) aged ≥ 18 years, and (4) being able to complete a Dutch questionnaire. In total, 544 participants (N=296 endometrial, N=248 ovarian) were eligible and 73% responded (N=221 endometrial, N=174 ovarian) to the baseline questionnaire (Online Resource 1). A total of 230 participants (N=137 endometrial, N=93 ovarian) completed the questionnaire at 18 months (response of 42% from eligible participants for the study) [20].
Measures
Cancer-related psychosocial factors
The validated Impact of Cancer scale version 2 (IOCv2) was used to assess cancer-related psychosocial factors [21]. For each of the subscales: health awareness, meaning of cancer, appearance concerns, body change concerns, life interferences and worry (including items scored on a five-point Likert scale) a sum score was calculated. The item ‘having had cancer has made me take better care of myself’ from the subscale health awareness was excluded from the analysis as Cronbach’s alpha improved when excluding this item (from 0.79 up to 0.84). Furthermore, it was strongly related to our dependent variables (change in lifestyle). Internal consistency (Cronbach’s alpha) for each of the subscales was good (ranging from/ range 0.83-0.93).
Changes in dietary behavior and physical activity
By use of self-developed questionnaires from the PROFILES registry studies, participants reported if they made changes in diet since they were diagnosed with cancer (yes/no) until 18 months post-diagnosis [22]. If so, they reported their most important changes (more/ less) in diet (fat, meat, fish, vegetables, fruit, milk and dairy products, dietary fibers/whole grain products, sugar, alcohol, water, salt and ‘others’) since diagnosis. Changes were classified in healthy (more consumption of fruit, vegetables, fibers/wholegrain and less consumption of fat, meat, sugar, alcohol and salt) or unhealthy (vice versa) according to the World Cancer Research Fund (WCRF) recommendations [23]. If respondents reported more healthy changes than unhealthy changes, they were classified into the eating healthier (N=44) group and if they reported more unhealthy changes, as many unhealthy as healthy changes or no change, they were classified into the no change/eating unhealthier (N= 172) group. As only few participants made unhealthy choices (N=9) these individuals were grouped with the no change group.
Furthermore, participants reported if they made changes in their physical activity since they completed their cancer treatment (yes/no) compared to pre-diagnosis. If so, they reported if they increased or decreased a certain type of physical activity (walking, cycling, gardening, housekeeping, sports and others). Subsequently, respondents were classified into the group: no change in physical activity (N=133), more physical activity (N=43) or less physical activity (N=36), according to the amount of reported decreases or increases in physical activity (despite differences in intensity of activities).
Socio-demographic and clinical factors
Factors obtained from the Netherlands Cancer Registry were age and socio-economic status (based on postal code of the residence area of the patient, categorized into low/ intermediate vs. high) and clinical factors (cancer type: ovarian/endometrial; tumor stage: I vs. II, III or IV; type of treatment: surgery, radiotherapy and chemotherapy; amount of comorbidities: 0,1, >1) [24]. Self-reported questionnaires included marital status (married/living together vs. divorced, widowed, never married/never lived together), educational level (low, intermediate, high) and physical limitations in daily activities (scored as 100 minus score of the EORTC QLQ-C30 physical functioning scale) [25] .Body Mass Index (BMI; weight in kg/length in m²) at 18 months post-diagnosis was based on self-report. Reasons for change in physical activity included complaints related to cancer or its treatment (not further specified), support of physical recovery, prevention of cancer recurrence and other reasons. Furthermore, participants were asked whether they received advice (yes/no) regarding diet (and/or dietary supplements) and/or physical activity, from diagnosis up to 18 months thereafter (provider or moment of received advice was not further specified). Gastrointestinal (GI) symptoms were obtained from a questionnaire developed by a dietitian from our research team and included unintentional weight loss or gain, poor appetite, cramps, frequent small amounts of stool, constipation, diarrhea, change in taste and/or smell, aversion, acid reflux and others. Unintentional weight loss and weight gain were excluded and a sum score (range 0-11) was based on the number of symptoms (suffered in the past or still suffering).
Statistics
Descriptive data was reported by means and standard deviations (SDs) or medians and interquartile ranges (IQRs) (continuous variables), frequencies and percentages (dichotomous and categorical variables) and compared by use of t-tests, ANOVA with post-hoc procedures, Mann-Whitney U and Kruskal-Wallis tests (continuous variables) and Chi-square/Fisher’s Exact tests (categorical variables).
Statistical analyses were performed by SAS for Windows (version 9.4). Associations between psychosocial factors as independent variables (i.e. sum scores from the scales health awareness, meaning of cancer, appearance concerns, body change concerns, life interferences, worry (range 1-5)), and dietary changes (no change/eating unhealthier (reference)/ eating healthier) or physical activity changes (no change(reference)/less/more) as dependent variables, were analyzed with logistic regression models. Before analyzing the data, assumptions were checked and not violated. P-values <0.05 were considered as statistically significant.
Potential confounders were age, socio-economic status, partner status, cancer type, tumor stage,treatment, number of comorbidities, BMI, gastrointestinal symptoms and physical limitations since these have been associated with psychosocial well-being as well as health behavior [9,19,21,26,27]. Variables gastrointestinal symptoms (sum score, 0-11) and physical limitations were considered as confounders prior to analysis. Other variables were considered to be a confounder when it caused a relevant change (a minimum of 10 percent) in the regression coefficient in the univariate model [28]. By hierarchical regression, the relevant confounders (determined per subscale) were added to the univariate model in the first step. Subsequently, the variable gastrointestinal symptoms was added to all models and the variable physical limitations was added to the physical activity change models.