Development of a composite score to assess colorectal cancer risk based on both genetic and environmental factors

Introduction

Colorectal cancer (CRC) is a common cause of cancer-related deaths worldwide (1,2). In the United Kingdom (UK), approximately 42,900 new cases are reported annually (3-5). Early detection of CRC through screening is crucial to its prevention and treatment (6). Although CRC screening is typically recommended to begin at age 50 years (7,8), there has been an increase in the incidence rate of CRC among adults under 50 years over the last few decades (9-12).

Studies have identified several risk factors that increase the likelihood of developing CRC, including obesity (13), diabetes (14), genetic predisposition (15), a family history of CRC (16), and unhealthy lifestyle factors such as smoking, alcohol intake, and red and processed meat take (17,18). Most existing studies use a combination score of lifestyle and environmental risk factors (19-21). However, incorporating genetic factors into this combination can provide a more comprehensive assessment of CRC risk by considering both inherited genetic traits and environmental influences. Previous studies have typically focused on genetic and environmental risk factors separately, indicating a need for integrated research (13-16). Determining the CRC risk by using a combination score of genetic and environmental risk factors may improve risk stratification compared to traditional age-based screening (22).

Several studies have developed risk scores for CRC using various methods. For example, a risk score for CRC was developed using simple factors, such as age, alcohol consumption, waist circumference, occupational sitting time, and diabetes (23). The total score for each participant was calculated by summing the adjusted risk factors, with higher coefficients assigned more score points. Other studies have generated weighted risk scores using lifestyle and environmental risk factors (19-21), which were calculated by summing all risk factors weighted by their log-odds ratio estimates.

Cho et al. developed a lifestyle risk score using five lifestyle factors, with each factor assigned a risk score of 0 for healthy and 1 for unhealthy (24). The total risk score was calculated by summing the risk score of all factors. Conversely, Choi et al. included eight factors in the calculation of a healthy lifestyle score (HLS) (25). Each factor was given a score of 0 or 1, with 1 representing the healthy behavior category. The HLS was divided into unhealthy (score 0–1), intermediate (score 2–3), and healthy (score ≥4). Similarly, Wu et al. created HLS using six lifestyle factors, with participants accumulating 1 point for each of the six healthy lifestyle patterns (26). The final HLS ranged from 0 to 6 and was further categorized into unfavorable (score 0–2 points), intermediate (score 3–4 points), and favorable (score 5–6 points).

In this study, we used data from the UK Biobank to estimate the CRC risk based on a novel composite risk score (C-score) that includes both environmental and genetic factors. We developed a framework to combine the polygenic risk score (PRS) with other identified risk factors. Similar to previous studies, the C-score was calculated by summing all identified risk factors. To shed light on the growing incidence rate among adults younger than 50 years (9-12), we used logistic regression to assess CRC risk for both the older (ages 50 years old and above) and the younger (<50 years old) groups separately. We present this article in accordance with the STROBE reporting checklist (available at https://ace.amegroups.com/article/view/10.21037/ace-24-19/rc).

Methods

Study participants

We used data from the UK Biobank, including adults aged 40–70 years between 2006 and 2010. Our study included 10,966 participants (2,387 cases and 8,579 controls) in the older group and 981 participants (198 cases and 783 controls) in the younger group. CRC cases were identified using ICD-10 codes of C18.0–C18.9, C19, C20, and C26.0 (25,27). Participants with any other types of cancer were excluded. Controls were selected by matching with cases within a 5-year age difference and living in the same residential area, and were not diagnosed with any type of cancer. For quality control, we only included White British individuals with completed imputed genotype information and a genetic relationship greater than the second degree in the analyses (28). Ethics approval was not required for this analysis, as the UK Biobank data is accessible to all researchers. The data had been fully de-identified before we accessed it. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

PRS calculation

We utilized a PRS using 140 single nucleotide polymorphisms (SNPs) that have been identified as common risk variants in European individuals. The list of the SNPs and their corresponding effect sizes on the risk of CRC is available from a previous genome-wide association study (29). To calculate the PRS for each participant, we used the scoring function in the PLINK 2.0 software (30). Initially, we extracted all risk SNPs from the imputed genotyping data for each CRC case and control in the UK Biobank. Subsequently, we computed the PRS by summing the risk alleles of the respective variants (imputed dosages for imputed SNPs; 0, 1 or 2 copies of the risk alleles for genotyped SNPs). We adopted the approach from Jia et al., classifying individuals with PRS scores in the top 5% as the high-risk group (coded as 1), while the remaining individuals were categorized as the low-risk group (coded as 0) (31).

C-score calculation

We generated a series of C-scores using the UK Biobank database by summing the values of PRS with all other CRC associated risk factors that identified in our previous study (27), including body mass index (BMI), sex, PRS, family history of CRC, alcohol intake frequency, and employment status (Table 1). To calculate the C-scores, the BMI value was normalized to a range between 0 and 1 to ensure a consistent scale with other factors.

For the younger group, we considered significant risk factors identified based on the analysis results of our previous study (27), including PRS and family history of CRC. In the older group, we incorporated various significant risk factors, including PRS, sex, family history of CRC, employment status, BMI, and alcohol intake frequency, identified through both univariate and multivariate regression analyses. To examine how the results differ when only considering participants with PRS scores in the top 5% and the middle range of 41–60%, we conducted a comparative analysis using this subset of older participants from the original case-control dataset. In this selected subset of the older group, we included significant risk factors of PRS, sex, family history of CRC, and BMI (27).

Statistical analysis

We used logistic regression to explore the association between the C-score and CRC for both the older and the younger groups separately. Initially, we evaluated models with PRS alone to determine the contribution of PRS to CRC risk stratification. We then constructed models with C-scores, which included PRS and all other significant risk factors. To assess the relative contributions of genetic and environmental factors to the prediction of the risk of CRC development, we conducted a dominance analysis using the {domir} package in R. An environmental score was calculated for each group by summing up all identified risk factors except PRS, representing the environmental aspect in the dominance analysis.

Results

In the older group, participants were aged 50 to 70 years, with BMI values ranging from 15.27 to 54.52 kg/m², and 5,435 (49.6%) were male. In the younger group, participants were aged 40 to 49 years, with BMI values between 15.84 and 53.57 kg/m², and 536 (54.6%) were female. The older group had a slightly higher percentage of participants with a family history of CRC (15.3% vs. 12.8%) compared to the younger group. In contrast, the younger group had a slightly higher percentage of participants with a high PRS (6.0% vs. 5.0%) compared to the older group (Table 1). Table 2 presents the results of the analysis for the older group. Model 2 calculated the C-score, considering four identified risk factors from multivariate analysis: PRS, sex, family history of CRC, and employment status. The C-score value of two [odds ratio (OR) =2.84; 95% confidence interval (CI): 2.46–3.27; P<0.001] indicates that participants with any two risk factors had a slightly higher risk of developing CRC compared to the PRS alone (OR =2.69; 95% CI: 2.26–3.21; P<0.001) in model 1. A C-score value of three indicates a 3.70 times greater risk of developing CRC (OR =3.70; 95% CI: 2.53–5.40; P<0.001) than the reference group (C-score equals zero). This risk is also higher than that associated with the PRS alone in the model. However, a C-score value of four showed a 5.16-fold elevated risk of developing CRC (OR =5.16; 95% CI: 0.32–82.58; P=0.25), but it is not statistically significant due to the low numbers of participants in this group.

Models 3 and 4 calculated the C-score using six identified risk factors derived from both univariate and multivariate analyses: PRS, sex, family history of CRC, employment status, BMI, and alcohol intake frequency. A higher C-score was associated with greater risk of developing CRC (OR =1.43; 95% CI: 1.35–1.50; P<0.001). Model 4 revealed that participants in the fourth quartile of the C-score had a higher risk of developing CRC (OR =3.95; 95% CI: 2.11–7.40; P<0.001) compared to the PRS alone in model 1. Despite the C-score considering the integrated influence from both environmental and genetic factors, the relative contribution of each remains unclear in the C-score. Dominance analysis indicated that environmental factors contribute 53.2%, while the genetic factor contributes 46.8% to the predictive power of CRC risk.

In the analysis of the selected subset of the older group (Table 3), comprising participants with the top 5% and middle 41–60% PRS, we found that a C-score value of two (OR =3.42; 95% CI: 2.63–4.45; P<0.001) or three (OR =5.43; 95% CI: 3.27–9.01; P<0.001) had a higher OR than PRS (OR =2.82; 95% CI: 2.31–3.45; P<0.001) alone in model 1. This suggests a greater risk of developing CRC when participants have two or three CRC-associated risk factors. The third (OR =3.41; 95% CI: 2.62–4.43; P<0.001) and fourth (OR =5.43; 95% CI: 3.27–9.01; P<0.001) quartiles of C-score exhibited higher OR than PRS alone in the model. These findings align with the earlier analysis involving all older participants, indicating that a higher C-score is associated with an increased CRC risk. Furthermore, having two or more risk factors can contribute to a higher risk of developing CRC compared to PRS alone.

Table 4 presents the results of the analysis for the younger group. A C-score value of 1, indicating participants with a risk factor of either a high PRS or a family history of CRC, had a 2.46 times greater risk of CRC development (OR =2.46; 95% CI: 1.69–3.58; P<0.001). However, due to a small number of participants with two risk factors simultaneously, the result showed an insignificant and lower OR for a C-score value of 2 compared to the C-score value of 1. The PRS had a 71.0% relative contribution, while the family history of CRC had a 29.0% relative contribution in terms of prediction power. This may help explain why the C-score value of one had slightly lower OR than PRS alone in model 1 (2.46 vs. 3.18).

Discussion

Our study demonstrated that a higher C-score is associated with an increased risk of CRC. Older participants with two or more risk factors had a higher risk of developing CRC compared to those with a high PRS alone (Tables 2,3). Younger participants with a risk factor of either a high PRS or a family history of CRC had a 2.46 times greater risk of developing CRC (Table 4). These results highlight the interactive impact of genetic and environmental factors on CRC risk, and the effectiveness of using a C-score that incorporates both aspects in assessing CRC risk.

Previous studies treated genetic and non-genetic factors separately when assessing CRC risk (21,31,32). For example, Hsu et al. developed a model to determine CRC risk using 27 CRC susceptibility loci (32), while Jia et al. constructed a PRS based on genome-wide associated studies for eight common cancers, including CRC (31). Both studies have indicated the potential utility of genetic risk scores (G-score) in the development of personalized CRC screening and risk prediction strategies. They have also consistently found that a higher PRS value corresponds to an increased CRC risk. These findings align with the results of our study well when we considered the PRS alone in the analysis, demonstrating a clear association between PRS and CRC risk.

Wang et al. calculated an environmental risk score (E-score) based on 17 environmental factors to estimate the absolute risk of CRC over 10 and 30 years (21). Their findings showed that a higher E-score was associated with an increased risk of CRC. In contrast to this study, our approach differs in that we integrated both genetic and environmental risk factors in the current analysis.

In other previous studies, researchers explored the relationship between CRC risk and HLS containing different potential risk factors (25,26,33). Choi et al. (25) and Xin et al. (33) constructed an HLS using 8 lifestyle factors, and Wu et al. (26) used 16 lifestyle factors to construct an HLS. The studies categorized the HLS into unhealthy, intermediate, and healthy/favorable categories and found that a favorable lifestyle and a low PRS was associated with a lower risk of CRC. However, instead of creating a C-score by combining HLS and PRS, previous studies aligned HLS categories with different PRS groups to explore their association with CRC risk.

Other studies suggested that combining genetic and E-score can more accurately provide risk stratification for CRC (19,20,34). Jeon et al. modeled the 10-year absolute risk of CRC based on family history, E-score with 19 lifestyle and environmental factors, and G-score with 63 CRC-associated SNPs (20). The accuracy of determining CRC risk for models with E-score and G-score was higher than that of the model with family history only. The model that included both E-score, G-score, and family history had slightly different accuracy for estimating CRC risk between men and women.

Archambault et al. developed an E-score based on 16 lifestyle and environmental factors, and a PRS using 141 variants (19). The results indicated that increasing values of E-score and PRS were associated with an increasing risk of CRC for individuals younger than 50 years old. This finding aligns with the results of the analysis conducted on young adults in this current study, which demonstrated a significant association between PRS and C-score values and an increased risk of CRC.

Ren et al. categorized the E-score and PRS into low, intermediate, and high groups and investigated how these groups interact to affect CRC risk, taking into account the screening status of individuals (34). The study found that a higher E-score was associated with an increased risk of CRC incidence and mortality, regardless of genetic risk. However, PRS and E-score were separate from each other in these studies. In contrast, our study combined PRS and environmental risk factors into a C-score in the analysis to examine CRC risk.

This study stands out from previous research by integrating genetic and environmental factors into a comprehensive risk score for assessing the risk of developing CRC (19,20,34). Additionally, our study examines risk stratification for both the older and younger groups. However, a limitation of our study is the small sample size of participants in the younger group, which prevented us from conducting comparative analyses using participants with the top 5% and middle 41–60% PRS scores. To further explore CRC risk factors for young adults, future studies should collect larger samples involving participants younger than 50 years. Another limitation is that the results are limited to White British individuals only since the PRS was calculated based on the identified SNPs from European ancestry. Future research is needed to determine if the results are consistent across other population groups. In addition, we realize that additional research is needed to identify environmental or behavioral risk factors associated with the risk of developing CRC and determine how these risk factors can be incorporated into the calculation of the C-scores. Moreover, future research can refine the C-score and explore its potential applications.

Conclusions

In this study, we developed a risk stratification system using the C-score to assess the risk of CRC for both older and younger groups. Our findings indicate that higher C-score values were associated with greater risk of developing CRC compared to the model utilizing PRS alone in the older group. This highlights the importance of including environmental factors in CRC risk stratification. The findings support a proposal to use a C-score that takes into account both genetic risk as well as other risk factors when assessing the risk of developing CRC.

Acknowledgments

This article was based on one chapter of Mei Yang’s dissertation completed at Texas State University. Our gratitude goes to the participants and staff of the UK Biobank for their dedication and valuable contributions to this research.

Footnote

Reporting Checklist: The authors have completed the STROBE reporting checklist. Available at https://ace.amegroups.com/article/view/10.21037/ace-24-19/rc

Peer Review File: Available at https://ace.amegroups.com/article/view/10.21037/ace-24-19/prf

Funding: This work was supported by a grant from the Allen Discovery Center program, a Paul G. Allen Frontiers Group advised program of the Paul G. Allen Family Foundation, and a Good Systems for Ethical AI grant from the University of Texas at Austin (to V.M.N.). Compute resources were supported by a Director’s Discretionary Award from the Texas Advanced Computing Cluster. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. There was no additional external funding received for this study.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://ace.amegroups.com/article/view/10.21037/ace-24-19/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. Ethics approval was not required for this analysis, as the UK Biobank data is accessible to all researchers. The data had been fully de-identified before we accessed it. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Gunter MJ, Alhomoud S, Arnold M, et al. Meeting report from the joint IARC-NCI international cancer seminar series: a focus on colorectal cancer. Ann Oncol 2019;30:510-9. [Crossref] [PubMed]
2. Gutlic I, Schyman T, Lydrup ML, et al. Increasing colorectal cancer incidence in individuals aged < 50 years-a population-based study. Int J Colorectal Dis 2019;34:1221-6. [Crossref] [PubMed]
3. Cancer Research UK. What is bowel cancer? [accessed April 26, 2022]. Available online: https://www.cancerresearchuk.org/about-cancer/bowel-cancer/about-bowel-cancer
4. Cowling TE, Cromwell DA, Bellot A, et al. Logistic regression and machine learning predicted patient mortality from large sets of diagnosis codes comparably. J Clin Epidemiol 2021;133:43-52. [Crossref] [PubMed]
5. Young C, Wood HM, Seshadri RA, et al. The colorectal cancer-associated faecal microbiome of developing countries resembles that of developed countries. Genome Med 2021;13:27. [Crossref] [PubMed]
6. US Preventive Services Task Force. Screening for Colorectal Cancer: US Preventive Services Task Force Recommendation Statement. JAMA 2021;325:1965-77. [Crossref] [PubMed]
7. Provenzale D, Ness RM, Llor X, et al. NCCN Guidelines Insights: Colorectal Cancer Screening, Version 2.2020. J Natl Compr Canc Netw 2020;18:1312-20. [Crossref] [PubMed]
8. Qaseem A, Crandall CJ, Mustafa RA, et al. Screening for Colorectal Cancer in Asymptomatic Average-Risk Adults: A Guidance Statement From the American College of Physicians. Ann Intern Med 2019;171:643-54. [Crossref] [PubMed]
9. Brenner DR, Ruan Y, Shaw E, et al. Increasing colorectal cancer incidence trends among younger adults in Canada. Prev Med 2017;105:345-9. [Crossref] [PubMed]
10. Feletto E, Yu XQ, Lew JB, et al. Trends in Colon and Rectal Cancer Incidence in Australia from 1982 to 2014: Analysis of Data on Over 375,000 Cases. Cancer Epidemiol Biomarkers Prev 2019;28:83-90. [Crossref] [PubMed]
11. Siegel RL, Jemal A, Ward EM. Increase in incidence of colorectal cancer among young men and women in the United States. Cancer Epidemiol Biomarkers Prev 2009;18:1695-8. [Crossref] [PubMed]
12. Siegel RL, Torre LA, Soerjomataram I, et al. Global patterns and trends in colorectal cancer incidence in young adults. Gut 2019;68:2179-85. [Crossref] [PubMed]
13. Renehan AG, Tyson M, Egger M, et al. Body-mass index and incidence of cancer: a systematic review and meta-analysis of prospective observational studies. Lancet 2008;371:569-78. [Crossref] [PubMed]
14. Larsson SC, Orsini N, Wolk A. Diabetes mellitus and risk of colorectal cancer: a meta-analysis. J Natl Cancer Inst 2005;97:1679-87. [Crossref] [PubMed]
15. Huyghe JR, Bien SA, Harrison TA, et al. Discovery of common and rare genetic risk variants for colorectal cancer. Nat Genet 2019;51:76-87. [Crossref] [PubMed]
16. Chen X, Guo F, Hoffmeister M, et al. Non-steroidal anti-inflammatory drugs, polygenic risk score and colorectal cancer risk. Aliment Pharmacol Ther 2021;54:167-75. [Crossref] [PubMed]
17. Rawla P, Sunkara T, Barsouk A. Epidemiology of colorectal cancer: incidence, mortality, survival, and risk factors. Prz Gastroenterol 2019;14:89-103. [Crossref] [PubMed]
18. Thélin C, Sikka S. Epidemiology of colorectal cancer—incidence, lifetime risk factors statistics and temporal trends. In: Ettarh R, editor. Screening for colorectal cancer with colonoscopy. InTech; 2015:61-77.
19. Archambault AN, Jeon J, Lin Y, et al. Risk Stratification for Early-Onset Colorectal Cancer Using a Combination of Genetic and Environmental Risk Scores: An International Multi-Center Study. J Natl Cancer Inst 2022;114:528-39. [Crossref] [PubMed]
20. Jeon J, Du M, Schoen RE, et al. Determining Risk of Colorectal Cancer and Starting Age of Screening Based on Lifestyle, Environmental, and Genetic Factors. Gastroenterology 2018;154:2152-2164.e19. [Crossref] [PubMed]
21. Wang X, O'Connell K, Jeon J, et al. Combined effect of modifiable and non-modifiable risk factors for colorectal cancer risk in a pooled analysis of 11 population-based studies. BMJ Open Gastroenterol 2019;6:e000339. [Crossref] [PubMed]
22. Hull MA, Rees CJ, Sharp L, et al. A risk-stratified approach to colorectal cancer prevention and diagnosis. Nat Rev Gastroenterol Hepatol 2020;17:773-80. [Crossref] [PubMed]
23. Guo L, Chen H, Wang G, et al. Development of a risk score for colorectal cancer in Chinese males: A prospective cohort study. Cancer Med 2020;9:816-23. [Crossref] [PubMed]
24. Cho YA, Lee J, Oh JH, et al. Genetic Risk Score, Combined Lifestyle Factors and Risk of Colorectal Cancer. Cancer Res Treat 2019;51:1033-40. [Crossref] [PubMed]
25. Choi J, Jia G, Wen W, et al. Healthy lifestyles, genetic modifiers, and colorectal cancer risk: a prospective cohort study in the UK Biobank. Am J Clin Nutr 2021;113:810-20. [Crossref] [PubMed]
26. Wu E, Ni JT, Chen X, et al. Genetic risk, incident colorectal cancer, and the benefits of adhering to a healthy lifestyle: A prospective study using data from UK Biobank and FinnGen. Front Oncol 2022;12:894086. [Crossref] [PubMed]
27. Yang M, Narasimhan VM, Zhan FB. High polygenic risk score is a risk factor associated with colorectal cancer based on data from the UK Biobank. PLoS One 2023;18:e0295155. [Crossref] [PubMed]
28. Bycroft C, Freeman C, Petkova D, et al. The UK Biobank resource with deep phenotyping and genomic data. Nature 2018;562:203-9. [Crossref] [PubMed]
29. Thomas M, Sakoda LC, Hoffmeister M, et al. Genome-wide Modeling of Polygenic Risk Score in Colorectal Cancer Risk. Am J Hum Genet 2020;107:432-44. [Crossref] [PubMed]
30. Choi SW, Mak TS, O'Reilly PF. Tutorial: a guide to performing polygenic risk score analyses. Nat Protoc 2020;15:2759-72. [Crossref] [PubMed]
31. Jia G, Lu Y, Wen W, et al. Evaluating the Utility of Polygenic Risk Scores in Identifying High-Risk Individuals for Eight Common Cancers. JNCI Cancer Spectr 2020;4:pkaa021. [Crossref] [PubMed]
32. Hsu L, Jeon J, Brenner H, et al. A model to determine colorectal cancer risk using common genetic susceptibility loci. Gastroenterology 2015;148:1330-9.e14. [Crossref] [PubMed]
33. Xin J, Du M, Gu D, et al. Risk assessment for colorectal cancer via polygenic risk score and lifestyle exposure: a large-scale association study of East Asian and European populations. Genome Med 2023;15:4. [Crossref] [PubMed]
34. Ren J, Zhang P, Li Z, et al. Association of screening status, polygenic risk score and environmental risk factors with colorectal cancer incidence and mortality risks. Int J Cancer 2023;152:1778-88. [Crossref] [PubMed]