Study design and population
We used data from the COLON study, an ongoing prospective multicenter cohort study among CRC patients [17]. From 2010 onwards, newly diagnosed patients with colon or rectal cancer were recruited in 11 hospitals in the Netherlands. Hospital staff invited eligible patients during a routine clinical visit before scheduled surgery. Patients were not eligible when they had a history of CRC, a previous (partial) bowel resection, known hereditary CRC, inflammatory bowel disease, dementia or another mental condition limiting their ability to fill out surveys, or were non-Dutch speaking. Data were collected at baseline (shortly after diagnosis, before treatment started) and at 6 months and 2 years after diagnosis. All study participants provided written informed consent and the study was approved by the local review board.
This study was performed using data of all participants diagnosed with stages I–III CRC between 2010 and 2015 (n = 1241). Participants were excluded when information on lifestyle was available for < 2 time points (n = 169). Thus, data of 1072 participants remained for analyses. Patients with stage IV disease were excluded a priori, because survival for these patients is generally poor and changes in diet and lifestyle may reflect poor health.
Data collection
Habitual dietary intake was assessed with a 204-item semi-quantitative food frequency questionnaire (FFQ) at baseline and 6 months and 2 years after CRC diagnosis. The FFQ was developed by the Division of Human Nutrition and Health, Wageningen University & Research, the Netherlands. The reference period for the FFQ was the month before diagnosis at baseline and the previous month during follow-up. To assess amounts of food intake, we combined frequencies of intake with standard portion sizes and household measures [18]. The FFQ was previously validated [19] and slightly adapted to be able to distinguish meat intake with respect to red, processed, and white meat. Self-reported dietary intake data from the FFQ were converted into fiber and alcohol intake based on the 2011 Dutch food composition table [20]. Items of interest included fruits, vegetables, dietary fiber, ultra-processed foods, red and processed meat, sugary drinks, and alcohol.
In addition to the FFQ, participants filled out other lifestyle questionnaires. These questionnaires included questions on weight, waist circumference, physical activity, and smoking status. Patients reported weight at diagnosis and at 6 months and 2 years after diagnosis, while height was only reported at diagnosis. BMI was computed in kg/m2. Waist circumference (midway between the lowest rib and the iliac crest) was measured with a tape sent to participants. Moderate-to-vigorous physical activity was self-reported by the validated SQUASH questionnaire [21,22,23]. Moderate-to-vigorous physical activity included all activities (walking, cycling, gardening, odd jobs, sports, household activities, and work) with a metabolic equivalent value ≥ 3 [24]. To ensure quality of the data, we checked each questionnaire after completion and contacted participants by telephone for clarification if needed.
Information was obtained on demographics, side-effects of treatment, and clinical factors. Demographic information, including level of education and living situation, was self-reported at diagnosis. Furthermore, participants reported if they changed their diet before diagnosis due to bowel complaints and if they experienced side-effects of treatment at 6 months and 2 years after diagnosis. Clinical factors were retrieved from the Dutch ColoRectal Audit [25] and included disease stage, tumor site, receipt of neo-adjuvant treatment, stoma placement after surgery, receipt of adjuvant chemotherapy, and presence of comorbidities. Recurrence data (loco-regional or distant recurrence) were retrieved from the medical records by the Netherlands Cancer Registry.
WCRF/AICR lifestyle score
We quantified the degree of concordance between participants’ lifestyles and the 2018 WCRF/AICR recommendations for cancer prevention using the standard WCRF/AICR score developed by Shams-White et al. [26] as a measure of overall lifestyle. The score included 7 recommendations (Table 1), as the recommendation on breastfeeding was not applicable to our study population. The recommendations about dietary supplement use and cancer survivors were not included, since they were not operationalized in the standard WCRF/AICR score [26]. We assigned, for each component, 1 point when the recommendation was met (full concordance), 0.5 points when it was partially met (moderate concordance), and 0 points otherwise (low concordance). Quantitative criteria were used as cut-off points, except for the recommendation on ultra-processed foods where cut-offs were based on tertiles calculated as a percentage of total energy intake from ultra-processed foods. Two recommendations (healthy weight and diet rich in wholegrains, vegetables, fruit, and beans) included sub-recommendations. For these recommendations, the recommendation score was the sum of each sub-recommendation score (meaning that plausible scores were 0, 0.25, 0.5, 0.75, and 1). The overall score ranged from 0 to 7, with higher scores indicating greater concordance with the 2018 WCRF/AICR recommendations.
Table 1 Description of the standardized WCRF/AICR score based on the 2018 WCRF/AICR recommendations for cancer prevention Statistical analyses
To describe the study population, we used descriptive analyses of demographic, clinical, and lifestyle characteristics of the participants. Furthermore, we calculated concordance with the 7 WCRF/AICR recommendations at diagnosis and 6 months and 2 years after diagnosis.
To describe changes over time in health behaviors in the first 2 years after CRC diagnosis, we used linear mixed models. Linear mixed models take into account both the individual trajectories of change and population averages by using all available measurements and including participants with incomplete data [27]. Each health behavior was modelled separately by using the 3 repeated measurements of that dependent variable. Time was scaled in years (continuous) and calculated as date of survey completion minus the date of study enrolment (i.e., shortly after diagnosis). All models included a random intercept, while a random slope was only included when this resulted in a better fit of the model (i.e., for BMI and ultra-processed foods). Inclusion of a random slope in the model means that the change over time can vary between participants. Changes were considered to be in concordance with lifestyle recommendations when the changes were as follows: an increase in physical activity, dietary fiber, fruit and vegetable intake or a decrease in BMI, waist circumference, red and processed meat, ultra-processed foods, sugary drinks, or alcohol intake.
To assess if multiple changes in different health behaviors led to a change in overall lifestyle, we modelled the 3 repeated measures of the WCRF/AICR lifestyle score as a dependent variable in a linear mixed model with random slope (in the same way as described above). To assess if changes in overall lifestyle varied between subgroups, we included a grouping factor and its interaction term with time in the mixed models. As grouping factors, baseline demographic determinants (sex, age, education, and living situation), clinical characteristics (stage, tumor site, stoma, neo-adjuvant treatment, adjuvant chemotherapy, and comorbidities) and self-reported side-effects of treatment were included, each in a separate model.
To further assess the interrelatedness between changes in multiple health behaviors, we assessed change in concordance to the 7 components of the WCRF/AICR lifestyle score. We assessed the proportion of participants who did change concordance to ≥ 1 component(s), who only improved or only deteriorated concordance to ≥ 1 component(s), and who both improved and decreased concordance to components of the lifestyle score. Furthermore, we assessed Pearson correlations between changes in health behaviors.
By using two separate sensitivity analyses, we evaluated the robustness of our reported changes in lifestyle. The potential influence of recurrent CRC or pre-diagnosis illness on changes in lifestyle was determined by excluding participants diagnosed with a recurrence within 2 years of follow-up (n = 98) and by excluding those who reported pre-diagnosis changes in diet due to bowel complaints (n = 129), respectively. All statistical analyses were conducted using SAS 9.4 software (SAS Institute, Cary NC). A p value < 0.05 was considered statistically significant.