Do high risk patients alter their lifestyle to reduce risk of colorectal cancer?

Patients who underwent colonoscopy surveillance between 1996 and 2012 at Dunedin Hospital, New Zealand for a family history of CRC were eligible for inclusion. Participants were excluded if they were aged over 75 years at recruitment (the age at which surveillance colonoscopy is discontinued, in keeping with national guidelines), had a personal history of CRC, symptoms of CRC or inflammatory bowel disease. Participants were randomly selected from a clinical endoscopy database of 1,086 patients.

Potential participants were provided with an information sheet and visual aid by mail, then telephoned to obtain verbal consent. The survey was administered using a labeled visual scale to assist in risk estimation. The telephone interview was standardized across ten interviewers. The methods are reported previously elsewhere [24]. Briefly, demographic and clinical details were collected during the interview, and additional clinical details were obtained from hospital records.

Demographic information is shown in Table 1. Participants reported family history of cancer for first- and second-degree relatives, including type of cancer and age at diagnosis, confirmed by review of hospital records. The number and year of surveillance colonoscopies were recorded from hospital records including colonoscopy outcomes. NZDep2006 is a measure of socioeconomic deprivation derived from census indices, based on the street address of each participant [25].

Each participant was asked for their perception to what degree CRC is genetic, how knowledgeable they were about CRC lifestyle risk factors, and to what degree they had altered the following lifestyle risk factors (reducing weight, red/processed meat, alcohol and smoking; increasing dietary fiber, fruit/veges, and physical activity) because of concerns about CRC. Participants were prompted with regard to each lifestyle factor (i.e. “how much have you reduced tobacco/cigarette consumption because of thoughts about CRC”). We used a four point, ordinal scale (Not at all, A little, Somewhat, Very much so). The questions regarding behavioural change enquired about subjective, self-perceived change since the initiation of surveillance. There was no formal validation of patient responses to measured behaviours. For patients who did not smoke/consume alcohol, their change in behaviour was labelled as “Not applicable” We asked participants to estimate their personal lifetime risk of CRC without colonoscopy surveillance on a 0 – 100% scale. The Generalized Anxiety Disorder 7 scale (GAD-7) was administered as per authors recommendations [26].

The Lower South Regional Ethics Committee approved the study.

Risk of CRC was estimated by New Zealand Guidelines Group (NZGG) recommendations [27]. These guidelines classify risk by the presence of family history and features of genetic CRC syndromes. Briefly, NZGG categorizes patients into four risk categories on the basis of the number of first degree relatives with histories of CRC and their age at first diagnosis [27]. NZGG category was determined by a team of three investigators (GPT, AC, JR) and from clinical documentation, and agreement between the two was good (Kappa 0.76).

Statistical analysis was performed using Stata 10.0 (StataCorp) and StatView version 5.01 (SAS Institute). Results are shown as mean (95% confidence interval) or number (percentage) unless otherwise stated. Continuous variables were compared between categories using ANOVA, and discrete variables were compared using the Chi-square test. To assess the relationship between anxiety, worry due to family history, and perceived knowledge and behavioural change, ordinal logistic regression including adjustments for age, sex and socioeconomic status was performed. A P value of < 0.05 was considered statistically significant. This was an exploratory study, and there was no previous data to base a sample size calculation on. Pearsons correlation was used to measure the association between different the number of healthy lifestyles adopted, and anxiety, worry and knowledgeability.

The primary outcome was the proportion of participants reporting changes in each healthy behaviour. Secondary outcomes included whether risk perception, generalized anxiety, worry due to family history and risk due to family history were associated with changes in a composite of all recorded healthy behaviours. To adjust these analyses for confounding factors, ordinal logistic regression was used. Odds ratios compared either a one-point increase, or SD increase of exposure variable with the number of healthy lifestyle behaviours changed “Somewhat” or “Very much so” compared to those changed “Not at all” or “A little”.

Full details of the study population including demographics are reported elsewhere [24]. Of 250 people initially sampled, 34 were ineligible, 47 could not be contacted and 21 declined consent. The exclusions were as follows: previous CRC n = 11, inflammatory bowel disease n = 4, not in surveillance program = 1, duplicate n = 1, deceased n = 4, aged over 75 years n = 13. The remaining 148 (69%) participated. Those excluded were younger than respondents (55.3 years old [95% CI 53.0 to 57.5] vs. 57.9 [95% CI 56.5 to 59.3]; p < 0.05), but similar by sex (42.4% vs. 37.8% male; P = 0.63). The clinical and demographic details including NZGG-based risk groups of the included population are given in Table 1. The majority were middle aged (range 35 to 74 years), female, married or in a de-facto relationship, had undergoing a single colonoscopy and were at least three times the population risk of developing colorectal cancer in their lifetimes.

The majority of participants (77.7%) believed that CRC risk was genetic and relatively uninfluenced by lifestyle behaviours (Table 2). Two-thirds of patients worried “Sometimes”, “Often” or “Always” about their risk based on their family history. The majority felt they were “Somewhat” or “Very much so” aware of lifestyle risk factors for CRC.

Most (79.7%, n = 118) reported altering at least one lifestyle variable to reduce their CRC risk (Figure 1). For most, this meant dietary change. Thirty participants (20.2%) reported no changes. The median number of healthy behaviours adopted was two. A small number of participants (4.1%, n = 6) reported adopting all healthy behaviours at least somewhat. Relatively few reduced their weight, alcohol intake or tobacco use due to concern over CRC risk.

The reporting of changes in dietary changes were correlated with each other (reducing red meat/increasing fiber, fruit and vegetables (r = 0.36 to 0.5) Reporting reducing weight was correlated with decreasing red meat (r = 0.55) and alcohol (r = 0.2), increasing fiber (r = 0.38), physical activity (r = 0.40) and fruit/vegetables (r = 0.48).

Increased physical activity was correlated with reducing alcohol (r = 0.19) and weight (r = 0.40), increasing fruit and vegetables (r = 0.55) and physical activity (r = 0.55). Decreased alcohol intake was correlated with increasing fruit and vegetable intake (r = 0.25) and decreased weight (r = 0.20).

No other variables were correlated with reported changes in tobacco consumption.

Global GAD-7 score and sub-scale items are shown in Table 2. The average anxiety levels in our study population correlate with ‘Mild to Moderate’ anxiety [26]. The influences of perceived knowledge, risk perception and anxiety on reported behaviour change after adjustment for age, sex and socioeconomic status are shown in Table 3.

Perceived knowledge of CRC risk factors and specific worry about CRC due to family history were positively associated with reported healthy behaviour change. Higher generalized anxiety symptoms were associated with a lower number of healthy behavioural changes.

The odds of adopting at least one behaviour “Somewhat or Very much so” were 4.71 for those who worried about CRC due to family history “Often or Always”, and 9.27 for those who perceived they were “Very much so” knowledgeable about CRC risk factors. The odds of adopting each behaviour “Somewhat or Very much so” were 0.66 for each SD increase in GAD-7 score.

Worry due to family history was modestly correlated with agreeing that CRC was due to genetics (r = 0.33) and perceived knowledgeability about CRC risk (r = 0.21). Thoughts about the role of genetics and perceived knowledgeability not correlated.

There was no association with either NZGG-based risk, or the concordance of perceived risk and NZGG-based risk, and behaviour change. There was no association of reported adoption of healthy behaviours with CRC risk perception or whether participants believed genetics to be the primary cause of CRC.

Both GAD-7 and worry due to family history remained significantly associated with reported behavioural change when both variables were included in the model, indicating they are independent predictors of reported behavioural change.

The association between either GAD-7 or worry due to family history and healthy behaviours was not diminished when perceived knowledge of risk factors was included in a multivariate model, suggesting the mechanism by which anxiety relates to healthy behaviours is not mediated through increased knowledge of risk factors.

We prompted participants by enquiring whether they had adopted specific healthy behaviours, which may have led to over-reporting of healthy choices. Furthermore, a number of our questions were about subjective perception. While this provides insight into patients’ thoughts, it also decreases the external validity of the findings.

This was a cross sectional study. Future studies should ideally assess the components of risk perception, anxiety and behavioural change longitudinally before, during and after participation in a surveillance program. Our study focused on patients undergoing surveillance, future studies could compare adoption of health behaviours with low risk groups.

Our study population was predominantly NZ European, potentially limiting generalization to people of other cultures.

There is no model for assessing the risk of CRC that includes measures of lifestyle risk factors that has been validated in our population or is in regular use. Calculated risk due to family history as assessed by NZGG category may not completely reflect the true risk for each individual although is the tool used in clinical practice.

Participant recall has been shown to underestimate the prevalence of CRC in the family history [34]. We validated the family history from the clinical records of each participant, but cannot be certain that we have not under-estimated their true CRC risk attributable to family history.

We had a relatively good response rate, although those excluded were slightly but significantly younger than those included, raising the possibility of selection bias although it would appear unlikely to have influenced our conclusions.

Clinicians should recognize that behavioural risk factor awareness and reported attempts to change are widespread in this high-risk population and support may be indicated. Further research is required to establish risk markers for patients who are less likely to improve lifestyle factors.