Combinations of modifiable lifestyle behaviours in relation to colorectal cancer risk in Alberta's Tomorrow Project

Colorectal cancer (CRC) is a significant health concern in Canada, with a substantial portion of cases attributable to modifiable lifestyle factors. These factors include obesity, physical inactivity, high red and processed meat consumption, low fruit and vegetable intake, alcohol use, and smoking. Individuals often exhibit combinations of these behaviours, making targeted prevention challenging. Because resources are limited, interventions focusing on subgroups with shared behaviour patterns could be more cost-effective and impactful. This study leverages Alberta's Tomorrow Project (ATP) cohort data to identify these common behaviour patterns and assess their association with CRC risk using latent class analysis (LCA). LCA is a statistical technique that groups individuals based on their responses across multiple variables, revealing underlying behavioral profiles. While previous LCA studies have examined lifestyle factors in specific populations, this study focuses on CRC-specific risk factors in a Canadian general population and links these patterns to actual disease occurrence. The study aims to identify distinct behavioral groups and their association with CRC risk, exploring the influence of sex and family history of CRC on both behavior and subsequent risk.

Existing research consistently links specific modifiable lifestyle behaviors to increased colorectal cancer risk. Studies have shown associations between obesity, inadequate physical activity, high consumption of red and processed meat, low consumption of fruits and vegetables, alcohol consumption, and tobacco smoking with increased CRC incidence. However, the interplay between these behaviors is complex. For instance, smokers are more likely to be regular alcohol consumers, while physically active individuals may also consume more alcohol. These complex interactions complicate the development of clear and effective prevention messages that address single risk factors. This research gap highlights the need for studies that analyze combinations of risk behaviours and their association with CRC risk, informing more effective prevention strategies.

This study utilized data from Alberta's Tomorrow Project (ATP), a prospective cohort study of 31,121 adults (aged 35-69) with no prior cancer history residing in Alberta, Canada. After excluding participants with missing data (78 participants, 0.3%), the study sample included 26,460 participants with complete information on relevant CRC risk factors and demographic variables. The median follow-up period was 13.23 years, during which 267 CRC cases were identified (119 males, 148 females). Six modifiable lifestyle risk factors for CRC were included: BMI (categorized as normal, overweight, obese), physical activity (low, moderate, high), red meat consumption (low, moderate, high), processed meat consumption (low, moderate, high), fruit and vegetable consumption (low, moderate, high), and smoking status (never, former, current). Exploratory latent class analysis (LCA) was performed using PROC LCA in SAS to identify distinct behavioral patterns. The optimal number of classes was selected using AIC, BIC, and sample size-adjusted BIC. Subsequently, Cox proportional hazards models were used to estimate the hazard ratio (HR) of CRC for each identified latent class, compared to a low-risk reference class. Analyses were adjusted for age, sex, ethnicity, household income, education, and family history of CRC. Sensitivity analyses were conducted to assess the robustness of findings by excluding participants with less than one or two years of follow-up and using more precise measures for smoking and obesity. Sex-stratified LCAs were also conducted to investigate sex-specific patterns.

The LCA identified seven distinct behavioral clusters in the overall sample. Compared to the lowest risk class (Class 2), four classes (Classes 1, 4, 5, and 7) exhibited significantly elevated CRC risk (HRs ranging from 2.34 to 2.87). Class 7 (current smokers, high-risk drinkers, inactive) showed the highest risk (HR 2.87, 95% CI 1.43-5.77), followed by Class 1 (high meat consumers, low fruit and vegetable intake; HR 2.48, 95% CI 1.27-4.83), Class 5 (obese, abstainers from alcohol; HR 2.46, 95% CI 1.28-4.70), and Class 4 (healthy diet, but inactive and overweight or obese; HR 2.34, 95% CI 1.23-4.45). Sex-stratified analyses revealed distinct patterns. In men, the highest risk groups showed HRs of 3.89 (95% CI, 1.42-10.67) and 3.15 (95% CI, 1.22-8.12), while the highest-risk groups in women had HRs of 2.19 (95% CI, 1.20-3.98) and 1.99 (95% CI, 1.14-3.47). The low-risk class in women (24.8% of women) exhibited lower meat consumption, high physical activity, and normal BMI compared to that for men (11% of men) that tended to consist of never smokers with a low family history of CRC. Sensitivity analyses yielded consistent results for overall and female classes but indicated slightly stronger effects for men when restricting analyses to those followed for longer durations. When more precise measures of smoking and obesity were used, the classes and associated risks remained largely similar to the primary analysis, except for the emergence of a new high-risk class in women (HR 3.05, 95% CI 1.64-5.68) characterized by heavy smoking, red meat consumption, low physical activity, obese BMI, and high waist circumference.

This study highlights the importance of considering combinations of lifestyle risk factors when assessing CRC risk. The findings demonstrate that multiple unhealthy behaviours, even without encompassing all identified risk factors, increase CRC risk significantly. While obesity alone is a strong risk factor, the combination of multiple unhealthy behaviors, without compensating healthy behaviors, confers the greatest risk. Sex-specific analyses reveal distinct patterns, with men exhibiting higher overall risks and different dominant behavior combinations compared to women. The identification of specific risk behaviour clusters offers opportunities for developing targeted interventions. For instance, interventions focusing on simultaneously modifying smoking, alcohol consumption, and physical activity in a high-risk group could potentially be highly effective. The identification of specific high-risk groups, based on combinations of modifiable lifestyle factors, offers a more targeted approach to cancer prevention strategies compared to addressing single risk factors in isolation. These strategies could prove more efficient in resource allocation and potentially yield a greater population-level impact.

This study provides valuable insights into the interplay of modifiable lifestyle factors and CRC risk by identifying distinct behavioral clusters and quantifying their association with disease. The findings emphasize the importance of developing and implementing multifactorial interventions targeting specific combinations of risk behaviours. Future research should focus on replicating these findings in larger, diverse populations, exploring the effectiveness of targeted interventions, and investigating the underlying mechanisms driving the observed behavioral clusters and their association with disease risk.

Several limitations should be considered. Self-reported data on lifestyle behaviors may be subject to recall bias and social desirability bias. Using baseline data only might not capture lifetime exposures, potentially impacting risk estimates. The chosen cut-offs for categorizing risk behaviours could influence the clustering, and the sample size, while large for an observational study, may still limit the precision of risk estimates and generalizability beyond the Alberta population. The study sample is potentially not representative of the general Alberta population in terms of socio-economic status, BMI, smoking habits, and alcohol consumption.