Colorectal cancer risk meets precision medicine

Science Spotlight

Colorectal cancer risk meets precision medicine

From the Peters and Hsu Groups, Public Health Sciences Division

April 16, 2018

Colorectal cancer (CRC) incidence and mortality have been declining over the past several decades. These encouraging trends are believed to be largely due to improved treatments, changes in risk factor patterns, and increased screening, with the latter estimated to account for over 50% of the reduction in CRC-related mortality.

Although numerous CRC risk factors are known, the recommended age at which to begin screening does not take into consideration any factors other than family history of CRC. Current guidelines suggest that individuals at an average risk for CRC, which includes those without family history, should start screening at age 50. While the recent CRC mortality statistics highlight the importance and beneficial impact of the enhanced screening efforts, there is still considerable room for improvement, as less than 60% of individuals between ages 50 to 75 underwent any colorectal test in 2013 in the U.S. It is possible that more individuals would undergo screening if they were aware of their specific long-term risk for development of CRC.

A paper recently published in the journal Gastroenterology by researchers in the Public Health Sciences Division from the groups of Drs. Ulrike Peters and Li Hsu, Fred Hutch members and co-senior authors, along with first author Dr. Jihyoun Jeon, a former Fred Hutch scientist now in the Department of Epidemiology at the University of Michigan, conducted modeling analyses to predict individual ten-year risk for CRC. The authors included a comprehensive panel of environmental, lifestyle, and genetic risk factors in their models that allowed for more precise risk estimates than have previously been reported. Based on individual risk levels, recommended ages to start CRC screening were then calculated. These models are in line with the goals of precision medicine and may prove to be first steps toward the development of CRC screening recommendations at the level of the individual, as Dr. Hsu noted, “There is currently no risk-based screening recommendation.”

The authors utilized data collected from approximately 20,000 participants (about half were CRC cases and half were controls) of European descent from two large consortia, the Genetics and Epidemiology of Colorectal Cancer Consortium (GECCO) and the Colorectal Transdisciplinary (CORECT) study. Extensive lifestyle and genetic information had been collected from the participants which allowed the authors to conduct complex risk modeling analyses. “Thanks to these international collaborations, we have the largest collection of CRC case-controls in the world, for which we not only conducted genome-wide scans for millions of genetic variants but we also harmonized the environmental and lifestyle data across participating studies. Given the large sample size, we are able to identify many genetic risk factors that each carry a small risk,” said Dr. Peters.

For an average 50-year-old who has not previously received an endoscopy, the authors determined that the ten-year risk for development of CRC is approximately 1%. Based on that risk threshold, the authors could provide optimized ages to start screening based on individual ten-year CRC risk computed by the models incorporating both environmental/lifestyle and genetic risk factors together, as different combinations of the risk factors yield different levels of long-term CRC risk. Dr. Jeon explained, “Someone who experiences a lot of risk factors increasing CRC risk reaches this risk threshold much earlier so is recommended to have the first screening earlier than age 50, but someone who has few risk factors reaches this risk threshold later and can postpone it to later ages.”

Graphical representation of recommended age to start colorectal cancer screening.

Recommended age to start colorectal cancer (CRC) screening for individuals (a) with or (b) without a family history of CRC based on risk scores (vertical dashed lines) of both environmental/lifestyle and genetic risk factors. The horizontal dashed lines indicate current recommended ages to start CRC screening.

Image provided by Dr. Jihyoun Jeon

Dr. Hsu described their strategy of incorporating the various risk factors (environmental/lifestyle and genetic) into the analyses, “We used our data to estimate the effect size of each risk factor, and we grouped them into E-score and G-score; within the E-score we have the contribution of environmental and lifestyle factors collectively.” Nineteen factors, such as height, body mass index, education, alcohol consumption, smoking status, regular aspirin use, physical activity, and dietary habits, among others, were used to calculate the E-score. The scores also accounted for the fact that different factors do not equally influence the risk for CRC. That is, some factors may have a large impact on risk while other factors may exert a relatively small effect. As explained by Dr. Hsu, “For each E-score variable, we had categories, and given the effect sizes, we added the risks together and came up with the overall environmental risk for the individual.”

The G-score was calculated based on 63 genetic variants, called single nucleotide polymorphisms (SNPs), that are associated with CRC risk. For each genetic risk variant, an individual can have zero, one, or two copies, and the authors added up the number of copies for each individual. As with the E-scores, the overall G-scores accounted for the CRC risk effect size of each genetic variant, wherein SNPs with greater effect on CRC risk were given more weight.

As can be expected, the prediction models provided the best estimate of ten-year CRC risk when the E-score, G-score, and family history were all considered together, as compared to when any of the components were considered alone. The authors found a distribution of CRC risk among the participants. For individuals with an overall low-risk profile, the projected ten-year risk for developing CRC was lower than that of the general population, while individuals with a high-risk profile had a greater ten-year risk, these would benefit from an earlier CRC screening.

The authors then estimated the recommended age at which individuals of different risk levels should start CRC screening (see Figure). The calculated recommended age is based on when an individual would reach a ten-year risk for CRC of around 1%, the risk threshold of the average 50-year-old who had not had a prior endoscopy. As an example, a woman without family history of CRC and with a low overall ten-year risk at the 1st percentile, would be recommended to start screening at age 71. However, a woman in the 99th percentile of risk reaches the risk threshold much earlier and is recommended to start screening at age 46. Currently, individuals without a family history are not recommended to undergo CRC screening until age 50; however, CRC diagnoses in individuals under the age of 50 have increased in recent decades. These results clearly demonstrate the cumulative impact of genetic, environmental, and lifestyle risk factors on the risk for CRC. Furthermore, the recommended age to start CRC screening is greatly influenced by the overall risk profile and spans a wide age range dependent on individual risk level.

Several follow-up projects to extend and confirm the new risk models are planned or already underway. Dr. Hsu indicated that the current models will continue to be a work in progress and will be refined when new data become available, “we’re going to continue to include genetic risk factors as we discover more,” she said. The authors have also started a collaboration with researchers in California where they will validate the models in a large genome-wide association study that has over 100,000 participants. Another goal is to develop models that will perform well in predicting CRC risk for minority populations. Dr. Peters described the current limitations in developing the models, “We need to do more fine-mapping to apply models to minority groups, as we currently do not have enough data from these populations.”