Age is the most important risk factor for bowel cancer but many genetic and lifestyle risk factors have been identified. There is interest in whether information about these factors could be used to improve screening programmes for bowel cancer.
In research funded by Bowel Cancer UK and led by Dr Juliet Usher-Smith, scientists at the Universities of Cambridge and Sheffield have been assessing the potential value of identifying people at higher and lower risk of bowel cancer – a process called risk stratification. The research assesses whether targeting screening programmes at those at higher risk might improve their effectiveness without increasing costs.
Why are these questions important?
Bowel cancer is the fourth most common cancer in the UK. One person is diagnosed on average every 15 minutes. It is also the second most common cause of cancer death. Many countries, including the UK, have screening programmes to try to pick up bowel cancers early. These screening programmes work, but at the moment most people are initially invited based only on their age, with any further tests based on their last screening test. There is a limit to the number of people who can be invited for screening in the NHS, mainly because of capacity to follow-up people who have a positive test.
Risk stratification holds promise because it might be a way to increase the detection of cancer at an early stage, largely within existing resources.
Much is already known about the risk factors for bowel cancer. Various ‘risk models’ have been developed to identify Individuals at higher risk of having or developing bowel cancer. The models use different combinations of information about age, body weight, diet, family history and genetic profile. More is being learned all the time, in particular about the genetic risk factors, and it is becoming easier to identify people at higher and lower risk. Consequently, there is growing interest in the possibility of inviting people to screening at different ages and with different intervals between screening tests, based on levels of risk.
Selected risk models could be used to group the general population into risk categories for bowel cancer
In our 2015  and 2018  analyses, we showed that selected risk models had the potential to group the general population into categories for risk of bowel cancer. These risk models could, in theory, be used to target screening and preventive strategies at those most likely to benefit, while leaving those at low risk unexposed to the direct and indirect harms of screening programmes.
Opportunities for accurate risk stratification are improving as we learn more about the genetic variants associated with risk of developing bowel cancer
In 2019, we found that 29 different models incorporating common genetic variants to estimate future incidence of bowel cancer in general populations had been published . Many of these models appeared to be better at identifying individuals at higher and lower risk of bowel cancer than using age alone, family history alone, or models incorporating only lifestyle information. Few had been tested in UK populations.
Genetic risk models can distinguish moderately well between those who do and do not develop bowel cancer
We then assessed the performance of 23 models that used different combinations of genetic, family and lifestyle information . We tested them to see how well they identify those at higher risk of developing bowel cancer amongst 500,000 people within the UK Biobank cohort.
- Genetic risk models can identify those who are more likely to develop bowel cancer moderately well.
- The best performing model (Huyghe et al) identified 2% to 5% more bowel cancers for the same number of people screened, compared to using only age.
- The risk models including only genetic information performed equally well in men and women. The risk models which used lifestyle information as well as genetic information performed better in men than women.
Advances in genetic research and technology mean that it may soon be possible to provide a relatively cheap and quick assessment of an individual’s genetic risk of bowel cancer.
Starting screening for bowel cancer at a younger age for men than women could improve the efficiency of screening programmes and reduce existing inequalities between the sexes in bowel cancer outcomes without substantial additional resource use and costs
We assessed the cost-effectiveness, clinical outcomes and impact on costs and use of resources of using information other than age to determine who should be screened and when. We used a simulation of the English bowel cancer screening programme called MiMiC-Bowel. MiMiC-Bowel simulates the life course of UK NHS patients with individual characteristics that determine their cancer risk and response to screening.
In our first analysis we added sex to age . Men have a higher risk of bowel cancer than women and currently suffer a greater burden of illness and death from bowel cancer. We therefore assessed the impact of inviting men to screening earlier than women.
- Screening men from age 56 but keeping the screening start age at 60 for women may be more cost-effective than screening everyone from age 58
- Greater health benefits are also produced, with an additional reduction of 25 bowel cancers, 28 late stage bowel cancers and 19 bowel cancer mortalities per 100 000 people in the population, compared to screening everyone from 58.
- As expected, all the additional health benefits are gained by men.
- If resources become available to enable all individuals to be screened at a younger age, the improvements in efficiency of sex stratification diminish, becoming marginal if everyone is screened from age 50.
- Earlier screening for men could improve the efficiency of screening programmes and reduce inequalities between the sexes in bowel cancer outcomes without substantial additional resource use or costs.
Using risk scores based on information about lifestyle and genetics to decide at what age individuals should be invited to screening for bowel cancer could improve both the overall effectiveness and cost effectiveness of the screening programme
Most recently, we explored whether risk stratification using risk models including lifestyle and genetic risk factors could improve the clinical outcomes and cost-effectiveness of bowel cancer screening, without using significant additional screening resources . Using MiMiC-Bowel again, we tested the risk models that best predicted bowel cancer in our study in the UK Biobank cohort and compared scenarios that used similar health resources to reflect a resource-constrained system such as the NHS.
- Using an accurate risk model (area under the receiver operating characteristic curve of 0.72) to determine the starting age for screening with two-yearly stool test (faecal immunochemical test, FIT), prevents 218 more bowel cancers and 156 more bowel cancer deaths per 100,000 people compared to a policy of sending the same number of invitations to everyone from age 60.
- It is highly likely (96% probability) that risk-stratification in this comparison is more cost-effective than inviting all individuals at age 60.
- We estimated that the maximum that could be spent on risk stratification and it still be cost-effective is £114 per person.
- The benefits of risk stratification are reduced if the risk model is less accurate, the starting age for screening is lower and the threshold for further testing after the FIT test is higher.. The benefits of risk stratification with currently available models are therefore smaller and will decrease as the starting age of screening is lowered across England.
We concluded that using risk models to decide when to invite people for screening could improve the clinical outcomes and cost-effectiveness of bowel cancer screening programmes without using significant additional screening resources. The magnitude of those benefits will depend on the accuracy of the risk models.
Together this work has shown that several models exist that are better than age alone at identifying those at risk of bowel cancer.
Using risk to determine age at first invitation to screening could represent a resource-neutral means of improving the efficiency of screening and preventing more cancers and deaths from bowel cancer at a lower cost.
Before risk stratification can be implemented within the screening programme, however, there are some important additional questions. These include:
- Can risk scores with high enough accuracy to see sufficient benefit from risk stratification be developed for the English population?
- How well do risk models perform in subgroups of the population, including different ethnic groups?
- How might risk assessment be incorporated into screening programmes and how much would that cost?
- Is risk stratification acceptable to the public?
- Which factors are acceptable to the public to use to estimate risk in this context?
- What effects might introducing risk stratification have on uptake of screening? Are those effects outweighed by any potential benefits?
The Cambridge and Sheffield research team is addressing some of these questions in ongoing and future work.
Express: Exploring the social and ethical implications of risk stratified screening for society as a whole
A multidisciplinary project with two linked studies using a community jury and a discrete choice experiment to explore views and choices on potential stratified screening programmes. Funded by Cancer Research UK’s Early Detection Programme. Read more
STRAT-BCS – Developing a strategy for implementation of risk stratification into bowel cancer screening programmes
This project will explore social and ethical considerations, the views of individuals and the potential impact of a risk stratified approach, and will establish the state-of-the-science and future research and policy requirements for incorporating risk stratification into CRC screening programmes. Funded by National Institute for Health Research Advanced Fellowship Read more
 Risk Prediction Models for Colorectal Cancer: A Systematic Review. Juliet A Usher-Smith, Fiona M Walter, Jon D Emery, Aung K Win, Simon J Griffin. Jan 2016. Cancer Prev Research.
 External validation of risk prediction models for incident colorectal cancer using UK Biobank. J A Usher-Smith, A Harshfield, C L Saunders, S J Sharp, J Emery, F M Walter, K Muir, S J Griffin. Jan 2018. British Journal of Cancer.
 Risk Prediction Models for Colorectal Cancer Incorporating Common Genetic Variants: A Systematic Review. Luke McGeoch, Catherine L. Saunders, Simon J. Griffin, Jon D. Emery, Fiona M. Walter, Deborah J. Thompson, Antonis C. Antoniou, and Juliet A. Usher-Smith. July 2019. Cancer Epidemiology, Biomarkers and Prevention.
 External Validation of Risk Prediction Models Incorporating Common Genetic Variants for Incident Colorectal Cancer Using UK Biobank. Catherine L. Saunders, Britt Kilian, Deborah J. Thompson, Luke J. McGeoch, Simon J. Griffin, Antonis C. Antoniou, Jon D. Emery, Fiona M. Walter, Joe Dennis, Xin Yang, and Juliet A. Usher-Smith. Feb 18 2020 Cancer Prev Research.
 Should colorectal cancer screening start at different ages for men and women? Cost-effectiveness analysis for a resource-constrained service. Chloe Thomas, Olena Mandrik, Sophie Whyte, Catherine L. Saunders, Simon J. Griffin, Juliet A. Usher-Smith. Feb 02 2021. Cancer Reports.
 The costs and benefits of risk-stratification for colorectal cancer screening based on phenotypic and genetic risk: a health economic analysis. Chloe Thomas, Olena Mandrik, Catherine L. Saunders, Deborah Thompson, Sophie Whyte, Simon Griffin, Juliet A. Usher-Smith. May 26 2021. Cancer Prev Research.
This briefing was written by Lucy Lloyd, Dr Juliet Usher-Smith, Dr Chloe Thomas and Professor Simon Griffin. Published May 2021
Contact: Lucy Lloyd, Communications Manager, Primary Care Unit, University of Cambridge