Yarnall, Crouch & Lewis (Division of Genetics and Molecular Medicine, King’s College London, United Kingdom.). Cancer Epidemiology 2013 Jan 29
Background: Epidemiological studies have identified potentially modifiable risks for colorectal cancer, including alcohol intake, diet and a sedentary lifestyle. Modelling these environmental factors alongside genetic risk is critical in obtaining accurate estimates of disease risk and improving our understanding of behavioural modifications. Methods: 14 independent single nucleotide polymorphisms identified though GWAS studies and reported on by the international consortium COGENT were used to model genetic disease risk at a population level. Six well validated environmental risks were selected for modelling together with the genetic risk factors (alcohol intake; smoking; exercise levels; BMI; fibre intake and consumption of red and processed meat). Through a simulation study using risk modelling software, we assessed the potential impact of behavioural modifications on disease risk. Results: Modelling the genetic data alone leads to 24% of the population being classified as reduced risk; 60% average risk; 10% elevated risk and 6% high risk for colorectal cancer. Adding alcohol consumption to the model reduced the elevated and high risk categories to 9% and 5% respectively. The simulation study suggests that a substantial proportion of individuals could reduce their disease risk profile by altering their behaviour, including reclassification of over 62% of heavy drinkers. Conclusion: Modelling lifestyle factors alongside genetic risk can provide useful strategies to select individuals for screening for colorectal cancer risk. Impact: Quantifying the impact of moderating behaviour, particularly related to alcohol intake and obesity levels, is beneficial for informing health campaigns and tailoring prevention strategies.
Over the last 30 years the lifetime risk of colorectal cancer (CRC) for men has almost doubled, from 3.5% to 6.9% in the UK in 2008. For women the increase is more than a quarter, rising from 3.9% to 5.4%. Since both genetic and environmental factors contribute to the susceptibility to colorectal cancer, this trend may be due to a change in the dietary and lifestyle factors of the general population leading to higher levels of obesity and more sedentary pastimes.
The major risk factor for colorectal cancer is age and over 85% of colorectal cancer occurs in people over the age of 60. Other risk factors include the presence of polyps and people having an Ashkenazi Jewish genetic heritage. The use of non-steroidal anti-inflammatory drugs (NSAIDs), hormone replacement therapy and aspirin use have also been associated with disease risk. However, it is estimated that between 52 and 57% of colorectal cancers are associated with lifestyle and environmental factors. Many risk factors for colorectal cancer may be modified by intervention, ranging from known risks, such as increased risk from a sedentary lifestyle and dietary changes. The evidence for dietary factors indicates possible increased risk from diets low in fibre, garlic, calcium, fruit, vegetables and fish and high in red and processed meat. In addition to alcohol, BMI, smoking and exercise, we chose to model the most consistent and well validated dietary findings, which suggest that low levels of fibre and high levels of red and processed meat are both significant risk factors.
The international consortium COGENT (COlorectal cancer GENeTics) have identified many of the known genetic variants that predispose to CRC with the 14 single nucleotide polymorphisms (SNPs) found to be convincingly associated with CRC risk from GWA studies summarized in Houlston et al.’s recent update. Of these 14 SNPs, the mean odds ratio per allele is 1.14, with the highest odds ratio reported for SNP rs16892766 near the EIF3H gene (OR 1.28).
The identification of SNPs that contribute to susceptibility for CRC has raised the prospect of genetic screening. Companies such as DeCODEme and 24andme include panels of SNPs for CRC in their genetic testing panels, yet research suggests that the genetic risk prediction alone is of questionable utility. In this research study, we combined the known genetic risk with data on the environmental risks for CRC, enabling more complete risk prediction. We applied a statistical risk model and to determine the impact of modelling environmental factors alongside the 14 genetic susceptibility loci identified by the COGENT consortium.
Early screening for colorectal cancer can be extremely helpful in identifying individuals with polyps and nonpolypoid lesions and preventing the development of cancer. Regular faecal occult blood tests (FOBT) in the over 50 s for example have been found to reduce the number of deaths due to CRC by 15–33%. In the UK, screening is offered to all men and women aged between 60 and 69 at a cost of £77.3 million and this will be extended to 74 year olds. However, it has been suggested that if individuals are provided with a personalized disease risk assessment from their combined genetic and environmental profile, they are likely to be more motivated to alter their lifestyle as a preventative measure, which would increase the effectiveness of health campaigns. In this study we develop predictions of CRC risk in different sub populations and assess the impact of modifying lifestyle factors on risk levels. By providing predictions of disease risk both before and after a lifestyle change for a given genetic profile, the study illustrates the potential benefits for both selection of candidates for screening programmes and the tailored promotion of healthier lifestyle choices, in high risk groups.
There are several modifiable risk factors for colorectal cancer and building predictive models encompassing both genetic and environmental factors enables us to move in the direction of a complete assessment of disease risk. This paper describes a predictive model which takes account of the known genetic contribution as well as the modifiable risks. There is considerable evidence to suggest that detecting polyps in the early stages can reduce mortality rates for colorectal cancer and whilst the interactions between the genetic and environmental elements are undeniably complex, separating out the inherited risk from the lifestyle factors using this model helps to illustrate the potential gains from modifying lifestyle behaviour and could usefully inform healthy lifestyle campaigns.
Our findings indicate that that cessation of alcohol consumption and reducing obesity levels lead to the most significant changes to the proportion of the population reducing their disease risk category. Whilst this could have been predicted to some extent by the higher odds ratios for these factors, it is the combination of relative risk, together with the prevalence of the factor within the population that determines the overall impact. In addition, being able to create personalized risk predictions in this way, has the potential to motivate greater behavioural change, showing for example, that it is possible to significantly reduce disease risk by moving from a high risk category to an average risk category though increasing fibre levels; cessation of alcohol consumption or weight management, given a particular genetic profile. Further research is required to increase understanding of how individuals respond to risk assessment based on genetic information. This may increases their motivation since the results are personal, or decrease their motivation because they consider that their genetic risk cannot be modified.
Our focus has been on risk categorization, and not on the absolute level of risk estimated from the combination of genetic and environmental risk factors, which is modest for most categories. There are two advantages to this strategy. Firstly it moves away from the strategy used, for example, by direct-to-consumer genetic testing companies such as 23andme and deCODEme (who provide a single figure of risk with no confidence intervals) towards the strategy deployed in genetic counselling of using a qualitative risk level, which can be more easily interpreted for the purpose of risk prediction. Secondly, it puts a stronger statistical framework on the risk model: an assignment to elevated risk implies that the risk is statistically distinct from the risk of the average, baseline, individual, given the uncertainty of the parameters used in the model.
There are several limitations of the model. Firstly, the model is built from estimates in the literature extracted from different studies. This enables researchers to select the best study to capture information on each risk factor, but assumes that information is directly comparable between studies. This limits the precision with which risk estimates can be calculated. A further limitation is that the model assumes all risk factors entered are independent. For known gene and environment interactions, this can be overcome by either modelling the interaction explicitly as an environmental risk factor, or by omitting known genetic loci to prevent over-representation of a risk factor (such as SNPs on the FTO gene which are associated with BMI). Within the genetic component, linkage disequilibrium between SNPs can be tested to confirm no correlation at a population level; few interactions of risk between genetic loci have been identified, so the assumption of independence should not be a major problem. For the environmental component, assumptions of independence are more difficult to assess. Lack of independence may lead to inaccuracies in the population frequencies estimated, but the contribution of environmental factors to the model is based on relative risks that are estimated in the presence of relevant covariates, so levels of risk should not be inflated. Increasing our understanding of the association between lifestyle factors, as well as between genes and the environment, will be important in obtaining more accurate assessments of risk. In addition, the accuracy could be further improved by more specific modelling of the population being targeted. Applying data with relative risks by sex, by population group, or for individuals with a first degree relative with CRC for example, would provide more accurate estimations of disease risk specific to those populations.
Colorectal cancer screening programmes are widespread, but are age-targeted and look for signs of cancer in early development. In contrast, the methods described here can be used to target lifestyle factors, and are relevant for younger age-groups. The approach could encourage behavioural changes and help to reduce CRC rates. Although the model indicates that certain individuals can reduce their CRC risk by changing their behaviour, the time taken for changes in environmental risk factors to have an effect on risk is unknown, and will differ by factor. Additional research is needed to further elucidate the genetic and environmental contributions to disease risk and to measure the longer term impact of behavioural change on disease outcomes.
- Association of the insulin-like growth factor 1 (IGF1) microsatellite with predisposition to colorectal cancer (familyhistorybowelcancer.wordpress.com)