September 29, 2020
Dr. John S. Jiao, MD
Despite progress in reducing colorectal cancer (CRC) incidence and mortality in recent decades in the US, CRC remains the third leading cause of cancer death. CRC is oneof the most preventable and treatable cancers if detected early.In 2013, screening forCRC was recommended for adults between age 50 and 75, in 2013 However, only 58% were compliant with the program.
Currently, screening guidelines are based only on age and family history despite the fact that over 80% of CRC cases have no family history. By evaluating the influence of multiple lifestyles, environmental, and genetic risk factors, risk prediction models can be used to define low- and high-risk populations.This is the core of precision medicine and is a vital advancement since genetic information will increasingly become a routine part of the medical record.
Improved risk stratification may also increase screening adherence and uptake, particularly for individuals at higher risk. They may be more likely to follow recommendations for prevention when aware of their heightened risk. Furthermore, the stratification can optimize the appropriate use of invasive technology.
The researchers developed risk prediction models for CRC based on 19 lifestyle and environmental factors and 63 common genetic variants known to be associated with CRC risk using data from 14 population-based studies. They expanded the risk prediction analysis to define the optimal starting age for screening, demonstrating the potential utility of using a model to tailor screening recommendations according to one’s personal risk profile.
The research is designed to develop models to determine risk of CRC, based on lifestyle and environmental factors and genetic variants, and to identify an optimal age to begin screening.
It is hoped that the work will provide guidelines for initiating CRC screening which are based on family history but do not consider lifestyle, environmental, or genetic risk factors.
In this study, data from two large consortia (9,748 CRC cases and 10,590 controls): The Genetics and Epidemiology of Colorectal Cancer Consortium (GECCO) and the Colorectal Transdisciplinary study (CORECT) were randomly split into two equal halves, with one half for building risk prediction models and the other for evaluating the models. The data consists of 6 case-control studies and 8 cohort-based nested case-control studies.
Lifestyle and environmental risk factor information including demographics, behavioral factors, anthropometric traits, diet, pharmacological factors, and medical history were collected by in-person interviews and/or structured questionnaires. All factors were collected at the study reference time, which was defined as study entry or blood collection for cohort studies and one to two years before sample ascertainment for case-control studies to ensure exposures assessed before cancer diagnoses.
To model the harmonized lifestyle and environmental risk factors, a score(E-score) was calculated based on these factors: height in centimeters, body mass index (kg/m2), education (less than high school graduate, high school graduate or completed General Equivalency Diploma, some college or technical school, college graduate or more), history of type 2 diabetes mellitus (no/yes), smoking status (ever/never), alcohol consumption (< 1g/day, 1-28 g/day, >28 g/day; one standard drink is 14g), regular aspirin use (no/yes), regular NSAIDs use (no/yes), regular use of post-menopausal hormones (no/yes, women only), sex- and study-specific quartiles of smoking pack-years, and dietary factors (intake of fiber, calcium, folate, processed meat, red meat, fruit, vegetable), total-energy, and physical activity (no/yes).
Family history was coded as a yes/no variable for presence or absence of a first-degree relative with CRC, and endoscopy history was coded as yes, no, or missing, depending on whether a participant had sigmoidoscopy or colonoscopy screening before the study reference time, or such information was missing.
For Genetic risk scoring, a total of 63 single-nucleotide polymorphisms (SNPs) at 49 known CRC loci have been identified by Genome-Wide Association Study (GWAS). Similar to E-score, a weighted genetic risk score (G-score) was created. This accounted for the strength of CRC-association with each SNP.
The weights were estimated regression coefficients obtained from a multivariable logistic regression that included all 63 SNPs adjusting for age, sex, genotype platform, and up to 6 principal components (PCs) that were previously determined by the genome-wide association studies for each genotype platform 30 to account for population substructure. A G-score for each individual was constructed by taking the weighted sum of risk alleles over all 63 SNPs, recoded the G-score as percentile based on cut points in controls, and modeled as an ordinal variable.
The 10-year absolute risk of developing CRC and corresponding 95% CIs for a given risk profile was established.
All E-score, G-score, and family history were significantly associated with CRC risk after adjusting for study, age, endoscopy history, genotype platform, and PCs. The E-score was associated with about 1.36-fold increase risk (Men: Odds Ratio (OR)=1.36 per quartile, 95% Confidence Interval (CI), 1.29 to 1.44, Women: OR=1.35 per quartile, 95%CI, 1.28 to 1.42).
Interestingly, the G-score increases the CRC risk with a similar magnitude as the E-score (Men: OR= 1.34 per quartile, 95% CI, 1.27 to 1.42, Women: OR=1.30 per quartile, 95% CI, 1.23 to 1.36). A positive family history increased the CRC risk by about 1.5-fold (Men: OR=1.67, 95% CI, 1.38 to 2.03, Women: OR=1.46, 95% CI, 1.24 to 1.72). All three factors had slightly stronger associations for men compared with women.
By utilizing the risk prediction model (including family history, E-score, and G-score), the researchersestimatedthe recommended age for initiating screening for individuals without having a prior endoscopy according to their risk profiles.
For those with high-risk of CRC as determined by a positive family history and 90th percentile of the combined risk score of E-score and G-score, the recommended age tostart screening is 40 for men and 46 for women, respectively.
On the other hand, for those with a positive family history but in the 10th percentile of the combined risk score, the recommended age to begin in men is 51 and in women is 59 (i.e., 11 years later for men and 13 years later for women).
Most people with positive family history do not reach the risk threshold until after the age of 40, when screening is currently recommended to begin in those with a positive family history. Under the model, about 62% of women and 15% of men with a positive family history do not reach the risk threshold until the age of 50 years.
The recommended starting ages for screening in people who do not have any first-degree relatives with CRC show consistent patterns but they are shifted upwards due to the overall lower risk in those with no family history.
Based on the combined E-score and G-score, for people with no family history and in the 90th percentile of the score, the starting age is 44 for men and 50 for women, respectively. But for those with no family history but in the 10th percentile of the score, the starting age in men is 56 and in women 64 (i.e., 12 years and 14 years later for men and women, respectively).
If the recommended ages for the first screening in those with no family history at the extremes (i.e., 1st percentile versus 99th percentile of the risk score) were compared, the difference in first screening age is 20 years or more. Furthermore, there is a fraction of the population that reaches the risk threshold for starting screening well before age 50. For example, about 15% of men with no family history would reach the risk threshold before age 45.
Both lifestyle and environmental factors and common GWAS variants are independent risk predictors for CRC and improved the discriminatory accuracy significantly compared with models that used only family history information.
The models yield a wide range of clinically actionable variation in risk stratification as demonstrated by the recommendation on when to start screening. These models may be useful to prioritize those at high risk for targeted prevention or intervention and to reduce emphasis on those at low risk of developing CRC, thereby optimizing utilization of screening in clinical practice with individually tailored prevention strategies.
Various aspectsshould be considered when utilizing the models for real clinical use such as difficulty of collecting data for some variables like total caloric intake. The developed model incorporating the most established or likely associated environmental/lifestyle risk factors and the known common genetic variants discovered to datecould serve as the reference for any subsequent parsimonious model.
Models that incorporate both environmental/lifestyle risk factors and common genetic variants may serve as the first step toward developing individually tailored CRC prevention strategies.
Jihyoun J. et al.Determining Risk of Colorectal Cancer and Starting Age of Screening Based on Lifestyle, Environmental, and Genetic Factors. Gastroenterology. 2018