Data
This cross-sectional study used data collected as part of the baseline assessment for the Advancing Survivorship after Cancer Outcomes Trial (ASCOT) [36]. ASCOT is a randomised controlled trial of a health behaviour intervention for people LWBC.
Participants
Participants were recruited from ten NHS trusts across London and Essex. These hospital sites were asked to send out a ‘Health and Lifestyle after Cancer’ survey to all patients diagnosed with breast, prostate and colorectal cancer between 2012 and 2015. However, hospitals did not always correctly identify diagnosis dates, so ethical approval was obtained to include individuals diagnosed outside of these dates. Patients completed the questionnaire on paper or online and then returned it to the research team. Of 13,546 surveys sent, 5,835 were returned (response rate = 42.8%). At the end of the questionnaire, patients had the option to leave their contact details to learn more about a trial of a lifestyle intervention Individuals who expressed interest were assessed for eligibility. A total of 3,354 individuals indicated interest (57.5% of surveys returned), of which 1,348 were eligible to participate (40.1% of those who expressed interest). Individuals were eligible to participate if they were aged 18 years and over, were diagnosed with non-metastatic breast, prostate or colorectal cancer (Stage I-III – primarily assessed by patient report during screening), and were not currently receiving active anti-cancer treatment (with the exception of oral anti-cancer treatments taken at home). Eligible participants were then provided with the full trial information and asked to provide informed consent to participate. National cancer registry data were collected for the majority of participants and when this was received it was discovered that 14 participants had Stage IV cancer at diagnosis and 28 had Stage 0 cancer, but these participants were still included in analyses. Ethical approval for the ASCOT was obtained through the National Research Ethics Service Committee South Central—Oxford B (reference number 14/SC/1369), and all participants provided informed consent.
Measures
Physical activity
Physical activity was assessed using an Omron pedometer (Omron, Kyoto, Japan) with the count reader covered [37]. Omron pedometers have established validity and reliability at different walking and running speeds [37, 38]. The method for how the pedometer data were processed is described in detail elsewhere [39]. Participants were asked to wear the pedometer all day for six days, on their waist or in their pocket, except when showering, bathing, swimming, doing water sports, or doing contact sports. Participants were also asked to complete a log-book indicating the dates they wore the pedometer, the time they put the pedometer on and took it off each day, and any physical activity they performed when they took the pedometer off. The pedometer data was cleaned using the log-books so that physical activity reported to have been performed when the pedometers were off was included. If data were not available for two days or more, then pedometer data were discarded to ensure there were enough days to provide a meaningful estimate of average daily steps [40].
Data from pedometers were uploaded using the Omron software Bi-link gateway (Omron). This provided the number of average daily steps, and the number of average weekly steps classified as aerobic (steps walked at a pace of 60 steps/min or higher for bouts of 10 min or more) [41]. For average daily steps, a cut-off of 10,000 was used to denote meeting physical activity guidelines [42]. For weekly aerobic steps, a cut-off of 15,000 was used to indicate meeting physical activity guidelines. This cut-off was chosen based on the assumption that when participants walk at an aerobic pace, they on average walk at a pace of 100 steps per minute [43].
Diet
Diet was assessed using 24 hour dietary recalls. The process for collecting and processing the recalls is described in detail elsewhere [44]. In brief, participants used Myfood24® online software to search a database for food and drink items they have consumed the previous day, select the most appropriate option, and determine portion size by selecting one of a range of pictures or by inputting data from household measures or weights. Participants were asked to complete recalls on one weekday and one weekend day.
Participants were sent letters with dates they were due to complete their weekday and weekend day recalls. On the day of their first scheduled recall, participants were sent emails with instructions on how to self-complete their recall and a link to Myfood24®. Participants who did not use email were contacted by telephone by a researcher who collected dietary information and inputted this into Myfood24® on their behalf. These individuals were sent a booklet before the call containing food portion images taken (with permission) from A Photographic Atlas of Food Portion Sizes to help with portion size estimation [45]. If participants had any questions or queries when completing their recalls, or if researchers noticed any unusual data entries, participants were contacted to resolve issues.
When the recalls were complete, data from Myfood24® were exported as an Excel file, and cleaned by experienced researchers who were Registered Dietitians or individuals with a post-graduate qualification in nutrition. Any unusually small or large data entries were inspected and only changed if two researchers agreed this was an error. After cleaning the dietary data, weighted average daily intake was calculated, with the weekday recall given a weighting of five and the weekend recall a weighting of two. Percentage energy from fat was calculated as 9kcal/g, and percentage energy from sugar was calculated as 3.75kcal/g.
To assess adherence to the five WCRF recommendations for diet, the following cut-offs were used to denote adherence: (1) fruit and vegetables – at least five portions (400g) per day [20], (2) fibre – at least 30g per day [20], (3) red meat – less than 500g per week [20], (4) processed meat – 0g per day [20], and (5) high calorie food – total calories from fat ≤33% of total energy intake [46] and free sugar percentage of daily calories ≤5% of total energy intake [47].
Alcohol
Alcohol consumption was assessed using two questions, adapted from the AUDIT Alcohol Consumption Questions [48]. The first item was “How often do you have a drink containing alcohol?” with response options “never”/”monthly or less”/”2-4 times per month”/”2-3 times per week”/”4-5 times per week”/”every day”. The second item was “how many units of alcohol do you drink on a typical day when you are drinking?” with response options “never”/”1-2”/”3-4”/”5-6”/”7-9”/”10+”. These two responses were converted to numerical scores and multiplied to estimate the total number of units consumed on average per week. The total score ranged from 0 to 70 units per week. Given that national UK guidelines for alcohol consumption recommend not drinking more than 14 units of alcohol per week [49], this was used as the cut-off to denote meeting versus not meeting recommendations.
Smoking
Smoking status was assessed using a single item from the Health Survey for England which indicated whether participants were a current smoker or non-smoker [50]. Smokers were classified as not meeting national guidelines for smoking whereas non-smokers were classified as meeting guidelines.
CHBRI index
The composite health behaviour risk index (CHBRI) was calculated based on nine health behaviours recommended by the WCRF for people LWBC (average daily steps, weekly aerobic steps, fruit and vegetables, fibre, red meat, processed meat, high calorie food, alcohol and tobacco). Table 1 shows the cut-offs used to determine whether participants were/were not meeting the guidelines. Participants were given a score of 1 if they were meeting guidelines, and a score of 0 if they were not. To calculate the CHBRI, these scores for each of the nine behaviours were summed. The CHBRI ranged from 0 (not meeting any recommendations) to 9 (meeting all recommendations).
Table 1. Health behaviour cut-off points determining whether ASCOT participants are meeting WCRF recommendations.
Behaviour
|
Meeting (score = 1)
|
Not meeting (score = 0)
|
Daily physical activity
|
≥10,000 average steps/day
|
<10,000 average steps/day
|
Fitness
|
≥6,000 weekly aerobic steps
|
<6,000 weekly aerobic steps
|
Fruit and vegetables
|
≥400g/day (one portion = 80g)
|
<400g/day
|
Fibre
|
≥30g per day
|
<30g/day
|
Red meat
|
<500g/week
|
≥500 g/week
|
Processed meat
|
0g/day
|
>0g/day
|
High calorie food
|
Fat: ≤33% of total energy intake
Sugar: free sugar percentage of daily calories ≤5% of energy intake
|
Fat: >33% of total energy intake
Sugar: free sugar percentage of daily calories >5% of energy intake
|
Alcohol
|
≤14 units/week
|
>14 units/week
|
Tobacco
|
Non-smoker
|
Smoker
|
Psychological distress
Psychological distress was assessed using the Anxiety/Depression dimension of the five-level EuroQol-5D questionnaire (EQ-5D-5L) [51]. The EQ-5D-5L has been validated for use in people LWBC [52]. The Anxiety/Depression dimension consists of one item asking participants to report if they were “not”/”slightly”/”moderately”/”severely”/”extremely” anxious or depressed on that day and is scored from 1 (no problems) to 5 (severe problems). In this study, due to issues with skewness, anxiety/depression scores were dichotomised into no problems (score = 1) versus any problems (score = 2-5). This method of dichotomising EQ-5D-5L index scores has been used previously in large samples of people LWBC [53, 54].
Confounders
Participants reported their age in years, sex (male/female), ethnicity (dichotomised into white/non-white due to small numbers in some ethnic groups) and marital status (dichotomised into married/not married) and highest level of education (none/GCSE or vocational/A level/degree or above). Participants were also asked to report if they had any of the following comorbidities: osteoporosis, diabetes, asthma, stroke, Parkinson’s disease, Alzheimer’s disease or dementia, lung disease, arthritis, angina, heart attack, heart murmur, irregular head rhythm, any other heart problem or hypertension. The total number of comorbidities participants reported was summed. Height and weight were self-reported, and body mass index was calculated using the formula weight(kg)/(height(m))2.
Cancer type, stage at diagnosis, and date of diagnosis were all self-reported and if consent was given, these data were also provided by the National Cancer Registration and Analysis Service (NCRAS). NCRAS data were used if available, but if not available, then self-report data were used. For some people, NCRAS data suggested that they had been diagnosed with another cancer since their breast/prostate/colorectal cancer diagnosis. Hence, in this study, cancer type was categorised into most recent diagnosis of breast, prostate, colorectal, or breast/prostate/colorectal plus one other. The number of days between this most recent cancer diagnosis and baseline assessments was calculated. Participants also self-reported on the treatment received for their most recent cancer, which was categorised into surgery only, surgery plus any other treatment, other treatments, and no treatment/active surveillance.
Analysis
Missing data
Multiple imputation (MI) by chained equations was used to impute missing data on predictors, outcomes and covariates given recommendations to impute all three [55]. Twenty imputed datasets were generated and pooled using Rubin’s rules [56].
Descriptive statistics
Descriptive statistics for the observed and imputed datasets were calculated. Means and standard deviations (SDs) were calculated for continuous variables, and frequencies and percentages were computed for categorical variables.
Main analyses
Multiple linear regression was conducted to assess the association between anxiety/depression and the CHBRI index. Logistic regression was conducted to assess associations between anxiety/depression and meeting WCRF recommendations for each health behaviour. All assumptions were tested for and met. Analyses were adjusted for all covariates. The results for binary outcomes (meeting WCRF recommendations/not) are reported as adjusted odds ratios (ORs) and 95% confidence intervals (CIs). The results for continuous outcomes (CHBRI) are reported as beta (B) coefficients and 95% confidence intervals. Two models were run for each analysis. Model 1 included age and sex, and Model 2 included age, sex, ethnicity, marital status, highest level of education, total number of comorbidities, cancer type, cancer stage, treatment, and time between cancer diagnosis and baseline assessments. Stata version 18.0 was used for all analyses.
Sensitivity analyses
Two sensitivity analyses were conducted. First, the analyses were repeated on a sample of participants with no missing data on the exposure, outcome and covariates. Second, the analyses were repeated with body mass index added to Model 2, as it was uncertain if body mass index was on the causal pathway (e.g., depression -> weight gain -> lower fitness behaviours) or acted as a confounder.