Research ArticleOriginal Article
Open Access

Effectiveness of colorectal cancer screening integrating non-genetic and genetic risk: a prospective study based on UK Biobank data

Yu Zhang, Chao Sheng, Zhangyan Lyu, Hongji Dai, Fangfang Song, Fengju Song, Yubei Huang and Kexin Chen
Cancer Biology & Medicine August 2024, 21 (8) 712-723; DOI: https://doi.org/10.20892/j.issn.2095-3941.2024.0096

Abstract

Objective: Few studies have evaluated the benefits of colorectal cancer (CRC) screening integrating both non-genetic and genetic risk factors. Here, we aimed to integrate an existing non-genetic risk model (QCancer-10) and a 139-variant polygenic risk score to evaluate the effectiveness of screening on CRC incidence and mortality.

Methods: We applied the integrated model to calculate 10-year CRC risk for 430,908 participants in the UK Biobank, and divided the participants into low-, intermediate-, and high-risk groups. We calculated the screening-associated hazard ratios (HRs) and absolute risk reductions (ARRs) for CRC incidence and mortality according to risk stratification.

Results: During a median follow-up of 11.03 years and 12.60 years, we observed 5,158 CRC cases and 1,487 CRC deaths, respectively. CRC incidence and mortality were significantly lower among screened than non-screened participants in both the intermediate- and high-risk groups [incidence: HR: 0.87, 95% confidence interval (CI): 0.81–0.94; 0.81, 0.73–0.90; mortality: 0.75, 0.64–0.87; 0.70, 0.58–0.85], which composed approximately 60% of the study population. The ARRs (95% CI) were 0.17 (0.11–0.24) and 0.43 (0.24–0.61), respectively, for CRC incidence, and 0.08 (0.05–0.11) and 0.24 (0.15–0.33), respectively, for mortality. Screening did not significantly reduce the relative or absolute risk of CRC incidence and mortality in the low-risk group. Further analysis revealed that screening was most effective for men and individuals with distal CRC among the intermediate to high-risk groups.

Conclusions: After integrating both genetic and non-genetic factors, our findings provided priority evidence of risk-stratified CRC screening and valuable insights for the rational allocation of health resources.

keywords

Introduction

Colorectal cancer (CRC) is the third most frequently diagnosed cancer and the second leading cause of cancer-related death worldwide1. Screening is a critical preventive approach that can decrease both CRC mortality and incidence2. Nonetheless, the prevailing CRC screening strategy, based solely on age, faces several challenges, including suboptimal participation and diagnostic yield; escalating demands on colonoscopy services; and potential risks associated with colonoscopy procedures, such as perforation and bleeding3–5. Consequently, selecting an appropriate strategy is increasingly important to enhance CRC screening participation and effectiveness.

Although age is widely recognized as the primary risk factor for CRC development, multiple non-genetic and genetic risk factors have also been identified and measured6. Several studies have developed risk prediction models to identify individuals at elevated risk of colorectal neoplasms7,8. Risk-stratified screening strategies have been endorsed because of their ability to enhance the benefit-to-harm ratio of screening and to optimize selection of the screening target population3–5.

Among non-genetic risk prediction models, the QCancer-10 model notably has shown superior predictive performance in external validation cohorts9. This model is a weighted, gender-specific, 15-year CRC risk prediction model developed on the basis of data from nearly 5 million participants 25–84 years of age registered at QResearch between 1998 and 201310. Because the risk factors incorporated in QCancer-10 can easily be extracted from electronic health records, this risk-stratified tool has been endorsed to facilitate shared decision-making in CRC screening11. Additionally, multiple genome-wide association studies (GWAS) have identified more than 100 single-nucleotide polymorphisms (SNPs) associated with CRC risk12–14. These SNPs can be used to generate composite measures called polygenic risk scores (PRSs), which can be applied to assess the cumulative effects of genetic variants and predict elevated risk of CRC12.

Although a randomized controlled trial (RCT) evaluating risk-stratified CRC screening remains ongoing, 2 cohort studies have examined its potential long-term effects on CRC incidence and mortality15–17. These studies have suggested that using a single genetic or non-genetic risk stratification can alter the magnitude of CRC incidence and mortality reduction during screening15,16. Previous studies have shown that incorporating PRSs with age and other non-genetic risk factors can improve model discrimination8,18. However, to date, no study has evaluated the decrease in CRC incidence and mortality in screening by integrating both non-genetic and genetic risk. Furthermore, using an appropriate weighted model and cut-off values for long-term risk may offer a more efficient and accurate approach than those used in previous research to identify individuals at higher risk of CRC15,17.

Therefore, our study using data from the UK Biobank was aimed at categorizing individuals into low-, intermediate-, and high-risk groups by using a risk tool integrating an existing non-genetic risk model (QCancer-10) and a 139-variant PRS. We then assessed the relative and absolute risk of CRC incidence and mortality associated with CRC screening in these risk groups.

Materials and methods

Study population

The UK Biobank is a large prospective cohort study designed to investigate genetic and lifestyle risk factors for a variety of chronic diseases. Between March 2006 and July 2010, approximately 502,000 individuals 37 to 73 years of age from England, Wales, and Scotland participated in the baseline assessment19. Participants provided information on their demographic characteristics, screening history, medical conditions, and lifestyle factors through touchscreen questionnaires and face-to-face interviews. Specifically, information on CRC screening was acquired by inquiring whether participants had undergone any CRC screenings, including tests for blood in the stool, sigmoidoscopy, or colonoscopy. Participants who answered “yes” were classified as the screening group, whereas those who answered “no” were classified as the non-screening group. Blood samples from these participants were collected at recruitment and used for genotyping. The collection and use of UK Biobank data were approved by the North West Multicenter Research Ethics Committee (16/NW/0274). All participants in the UK Biobank provided written informed consent to participate in the study.

In this study, the initial cohort comprised 487,259 participants with imputed genetic data. We excluded 28,098 participants with non-white British ancestry, because the PRS was constructed by using GWAS conducted in individuals of European descent. We further excluded 351 participants because of gender mismatch, 7,297 participants with missing data regarding CRC screening history, 2,812 participants with prevalent CRC at recruitment, and 17,793 participants with incomplete data for predictors in the QCancer-10 model. Therefore, a total of 430,908 participants were included in our analysis (Figure 1).

Figure 1
Figure 1

Flowchart of participant selection in the UK Biobank. CRC, colorectal cancer; PRS, polygenic risk score.

QCancer-10 model

The QCancer-10 model can be found at https://www.QCancer.org/15yr/colorectal/. This model incorporates predictors such as age, ethnicity, family history of CRC, alcohol consumption, smoking status, and disease history (diabetes, colorectal polyps, and ulcerative colitis). Additionally, the model includes history of other cancers, including lung cancer, blood cancer, and oral cancer for men, and uterine cancer, ovarian cancer, and cervical cancer for women. Notably, for men, the model also includes the Townsend deprivation index and body mass index. The detailed categorization of each predictor and its corresponding coding in the UK Biobank is presented in Table S1.

Polygenic risk score

Detailed information regarding genotyping and imputation in the UK Biobank has been described in a previous study20. In brief, samples were genotyped by using 2 arrays sharing 95% marker content: the Applied Biosystems UK BiLEVE Axiom Array and the Applied Biosystems UK Biobank Axiom Array. The imputation was conducted with SHAPEIT3 and IMPUTE3, on the basis of the merged UK10K and 1,000 Genomes phase 3 panels. We estimated PRS by using 139 of 140 CRC-associated SNPs recently identified in a GWAS among populations of European descent (rs377429877 was not measured in the UK Biobank and therefore was not included in the analysis)12. The list of SNPs and their corresponding beta values are presented in Table S2. The PRS was calculated by summation of the number of risk alleles weighted by their effect size, using the -score command in PLINK version 1.9021.

Covariates

Education level was categorized into 2 groups (less than college education, or college education or above). Ideal physical activity was defined as ≥ 150 min/week moderate intensity, ≥ 75 min/week vigorous intensity, or a combination thereof; otherwise, physical activity was defined as non-ideal22. High intake of fruits and vegetables was defined as ≥ 3 servings/day; otherwise, intake was defined as low. High intake of processed meats and unprocessed red meats was defined as > 1.5 servings/week and > 1 serving/week, respectively; otherwise intake was defined as low. Regular use of multivitamins and aspirin was categorized into 2 groups (yes or no). Self-rated health status was categorized into 4 levels (poor, fair, good, or excellent). For each categorical variable, missing values were added as an independent category.

Outcomes

The primary outcomes of this study were CRC incidence and mortality, determined on the basis of the first record of hospitalization for CRC, registration in the CRC registry, or CRC death, as classified by International Classification of Diseases (10th revision) codes C18–C20. The CRC registry and death data were sourced from the National Health Service (NHS) Digital for England and Wales, the NHS Central Register, and the National Records of Scotland. Hospital admission data were obtained from the Hospital Episode Statistics and Scottish Morbidity Records. Proximal colon cancers were defined as codes C18.0–C18.5. Distal colon cancers and rectal cancers were defined as codes C18.6–C18.9, C19, and C20. Because follow-up ended on February 29, 2020, for England; February 28, 2018, for Wales; and January 31, 2021, for Scotland, the outcomes were right-censored at the date of CRC diagnosis (for CRC incidence only), death, loss to follow-up, or the end of the follow-up period, whichever occurred first. Survival analysis methods, such as the Kaplan-Meier (K-M) estimator and the Cox regression model, were used to properly handle the censored data.

Statistical analysis

Categorical variables are presented as frequency (n) and percentage (%), and were compared with the chi-square (χ2) test or Fisher’s exact test, as appropriate. The integrated model (QCancer-10 + PRS) was developed from the Cox regression model, which included the QCancer-10 score and PRS as variables. We used Schoenfeld residuals to test the proportional hazard (PH) assumption. The QCancer-10 + PRS model satisfied the PH assumption (Table S3). The predictive performance of the model was measured in terms of discrimination and calibration. Discrimination was assessed with the area under the time-dependent receiver operating characteristic curve (AUC) (perfect discrimination = 1.0, no discrimination = 0.5) and the net reclassification index. Calibration was evaluated with calibration plots. Participants were categorized into low-risk, intermediate-risk, and high-risk groups according to the 10-year CRC risk estimated by the model. The cut-off values were determined in X-tile software (version 3.6.1)23. The cumulative incidence and mortality curves for CRC were illustrated with the K-M method, and compared with the confounder-adjusted log-rank test with inverse probability weighting. Multivariable Cox regression models were used to estimate hazard ratios (HRs) with 95% confidence intervals (CIs) of CRC screening on CRC incidence and mortality. Model 1 adjusted for education level, physical activity, fruit intake, vegetable intake, processed meat intake, unprocessed red meat intake, regular use of multivitamins, regular use of aspirin, and self-rated health status. Model 2 further adjusted for risk predictors included in QCancer-10. To indicate public health priorities, we calculated absolute risk reductions (ARRs) associated with CRC screening by subtracting the cumulative risk in the screening group from that in the non-screening group. We obtained the 95% CIs of the ARRs by applying a bootstrapping procedure with 1,000 repeated samplings. We conducted subgroup analyses according to gender, and explored risk-stratified screening with the incidence and mortality for site-specific CRC. To assess the robustness of our results, we conducted sensitivity analysis selecting participants at the recommended age (≥ 45 years or ≥ 50 years) for population-wide screening to evaluate the effectiveness of risk-stratified CRC screening. We used the Benjamini-Hochberg procedure to control the false discovery rate in multiple comparisons.

All data analyses were performed in R (version 4.0.3, R Foundation for Statistical Computing, Vienna, Austria). A P value < 0.05 was considered statistically significant. This study adhered to the Strengthening the Reporting of Observational Studies in Epidemiology guidelines24.

Results

Of the 430,908 participants, 45.02% were men. The mean age of the participants was 56.72 years. Overall, 136,682 (31.72%) participants had undergone CRC screening before enrollment, whereas 294,226 (68.28%) were classified into the non-screening group. As shown in Table 1, participants in the screening group were more likely to be male, to be older, and to have a family history of CRC (all P < 0.01).

View this table:
Table 1

Characteristics of study participants according to screening status in the UK Biobank

Performance and risk stratification of the QCancer-10 + PRS model

After a median follow-up of 11.03 years and 12.60 years, 5,158 incident CRC cases and 1,487 CRC deaths were identified in the cohort, respectively. The 10-year AUC for the QCancer-10 + PRS model was 0.70, and the model was well calibrated (Figure S1). The performance of the model was also assessed separately for men and women, as shown in Figure S1. Comparison of the QCancer-10 + PRS model with the single QCancer-10 model indicated that approximately 4.3% of CRC cases gained improved risk reclassification (net reclassification index = 0.06, z = 13.02, P < 0.01, Table S4). The cut-off value of the predicted 10-year CRC risk according to the QCancer-10 + PRS model was 0.96 and 2.19 for men, and 0.66 and 1.21 for women. These cut-off points divided the population into low-, intermediate-, and high-risk groups, respectively. On the basis of different cut-off values for men and women, all participants were categorized into low-risk (40%), intermediate-risk (47%), or high-risk (13%) groups. The intermediate- and high-risk groups had significantly higher risk of CRC incidence and mortality than the low-risk group (Figure S2), with HRs of 2.17 (95% CI: 1.97–2.38) and 3.77 (95% CI: 3.35–4.24), respectively, for incidence, and 2.10 (95% CI: 1.75–2.52) and 4.19 (95% CI: 3.35–5.24), respectively, for mortality (Figure S2). Similar trends were observed for men and women (Figure S3). The baseline characteristics of participants between the screening and non-screening groups across risk groups are presented in Table S5.

Association of screening with CRC incidence and mortality across risk groups

Adjusted K-M curve analysis revealed that the screening group had significantly lower cumulative incidence and mortality rates for CRC than the non-screening group (both log-rank P < 0.01, Figure S4). When participants were categorized into the 3 groups, screening was associated with CRC incidence, and the adjusted HRs were 1.04 (95% CI: 0.87–1.24), 0.87 (95% CI: 0.81–0.94), and 0.81 (95% CI: 0.73–0.90) in the low-, intermediate-, and high-risk groups, respectively (P for trend < 0.01). Similar results were found for CRC mortality: the adjusted HRs were 0.87 (95% CI: 0.60–1.26), 0.75 (95% CI: 0.64–0.87), and 0.70 (95% CI: 0.58–0.85) in the low-, intermediate-, and high-risk groups, respectively (P for trend = 0.52, Figure 2 and Table S6). In the intermediate-high risk groups, which included both intermediate-risk and high-risk individuals, participants who underwent screening had a lower risk of CRC incidence (adjusted HR = 0.85, 95% CI: 0.80–0.91) and mortality (adjusted HR = 0.73, 95% CI: 0.65–0.83) than those who were not screened (Table S6). Furthermore, the ARRs associated with screening increased from the low-risk (0.00, 95% CI: −0.04 to 0.04) to the intermediate-risk (0.17, 95% CI: 0.11–0.24), to the high-risk (0.43, 95% CI: 0.24–0.61) groups. A similar trend was observed for CRC mortality: the corresponding ARRs were 0.01 (95% CI: −0.01 to 0.03), 0.08 (95% CI: 0.05–0.11), and 0.24 (95% CI: 0.15–0.33) (Figure 2 and Table S6). Among participants in the intermediate-high risk group, the ARRs for CRC incidence and mortality were 0.23 (95% CI: 0.16–0.29) and 0.11 (95% CI: 0.08–0.14), respectively (Table S6). The combination of PRS and QCancer-10 for risk stratification showed improvements in effectiveness for decreasing CRC incidence and mortality over the QCancer-10 model alone. These improvements were quantified as 3.41% for CRC incidence and 2.67% for CRC mortality in the intermediate-high risk group (Table S7).

Figure 2
Figure 2

Multivariable-adjusted cumulative incidence and mortality for colorectal cancer according to screening status and risk stratification. The non-screening group is shown as a solid line, and the screening group is shown as a dashed line. Cox proportional hazards models were adjusted for education level, physical activity, fruit intake, vegetable intake, processed meat intake, unprocessed red meat intake, regular use of multivitamins, regular use of aspirin, self-rated health status, and risk predictors included in the QCancer-10 model. For CRC incidence, P for trend = 0.005; for CRC mortality, P for trend = 0.522. HR, hazard ratio; CI, confidence interval; ARR, absolute risk reduction.

The association between screening and CRC incidence and mortality did not change substantially in the sensitivity analysis. As shown in Table S8, among participants at the recommended age (≥ 45 years or ≥ 50 years) for population-wide screening, similar decreases in CRC incidence and mortality with screening were observed only in the intermediate-risk and high-risk groups, but not the low-risk group. Among those with intermediate-high risk, in participants ≥ 45 years of age, screening was associated with adjusted HRs of 0.85 (95% CI: 0.80–0.91) and 0.74 (95% CI: 0.65–0.83) for CRC incidence and mortality, respectively. The corresponding ARRs were 0.24 (95% CI: 0.17–0.30) and 0.11 (95% CI: 0.08–0.14), respectively. Similar results were observed in participants ≥ 50 years of age.

Association of screening with CRC incidence and mortality in men and women across risk groups

In both men and women, CRC mortality significantly decreased with screening. For women, however, no significant association was observed between screening and the risk of CRC incidence (log-rank P = 0.26, Figure S4). Because no decrease in CRC mortality was found in the low-risk group, we conducted subgroup analyses by gender only in the intermediate-risk and high-risk groups (Figure 3 and Table S9). For men, screening significantly decreased CRC incidence and mortality in both the intermediate-risk and high-risk groups. In the intermediate-high risk group, screening was associated with decreased CRC incidence and mortality, with adjusted HRs of 0.81 (95% CI: 0.74–0.88) and 0.74 (95% CI: 0.64–0.87), respectively. The corresponding ARRs were 0.43 (95% CI: 0.32–0.55) and 0.12 (95% CI: 0.06–0.17), respectively. In women, screening decreased CRC incidence in only the high-risk group, with an adjusted HR of 0.83 (95% CI: 0.70–0.98), but decreased CRC mortality in both the intermediate-risk and high-risk groups, with adjusted HRs of 0.75 (95% CI: 0.59–0.95) and 0.67 (95% CI: 0.50–0.90), respectively. In the intermediate-high risk group, screening significantly decreased CRC mortality, with an adjusted HR of 0.72 (95% CI: 0.60–0.86), but not CRC incidence (adjusted HR = 0.92, 95% CI: 0.83–1.01). The corresponding ARR for CRC mortality was 0.11 (95% CI: 0.07–0.14).

Figure 3
Figure 3

Associations of screening with colorectal cancer incidence and mortality according to gender and risk stratification. Cox proportional hazards models were adjusted for education level, physical activity, fruit intake, vegetable intake, processed meat intake, unprocessed red meat intake, regular use of multivitamins, regular use of aspirin, self-rated health status, and risk predictors included in the QCancer-10 model. HR, hazard ratio; CI, confidence interval; CRC, colorectal cancer.

Association of screening with site-specific CRC incidence and mortality across risk groups

Assessment by CRC subsite indicated that screening significantly decreased the incidence and mortality for distal CRC (both log-rank P < 0.01), whereas no significant reduction was observed for proximal colon cancer (for incidence: log-rank P = 0.57; for mortality: log-rank P = 0.18, Figure S5). Figure 4 and Table S10 illustrate similar reduction trends in both the intermediate-risk and high-risk groups for distal CRC incidence and mortality. In the intermediate-high risk group, screening showed adjusted HRs of 0.79 (95% CI: 0.72–0.87) and 0.69 (95% CI: 0.60–0.80) for distal CRC incidence and mortality, respectively. The corresponding ARRs were 0.07 (95% CI: 0.01–0.12) and 0.08 (95% CI: 0.05–0.11), respectively. However, no significant decrease in incidence or mortality was observed for proximal colon cancer, analyzed separately or in the group at intermediate-high risk.

Figure 4
Figure 4

Associations of screening with site-specific colorectal cancer incidence and mortality according to risk stratification. Cox proportional hazards models were adjusted for education level, physical activity, fruit intake, vegetable intake, processed meat intake, unprocessed red meat intake, regular use of multivitamins, regular use of aspirin, self-rated health status, and risk predictors included in the QCancer-10 model. HR, hazard ratio; CI, confidence interval; CRC, colorectal cancer.

Discussion

In this large population-based cohort study, screening was associated with a 27% decrease in CRC mortality and a 15% decrease in CRC incidence among individuals at intermediate to high risk, who accounted for approximately 60% of the population. Nevertheless, this effectiveness was not observed in the low-risk group. Use of a risk-stratified tool incorporating an existing non-genetic risk model (QCancer-10) and a 139-variant PRS appeared to be more suitable for prioritizing screening for intermediate-risk and high-risk populations than a non-genetic risk model alone.

Effective CRC screening strategies are crucial for decreasing incidence and mortality rates. A prior cost-effectiveness analysis has indicated that a risk prediction model should achieve an AUC of at least 0.65, and that risk-stratified screening is more cost-effective than population-wide screening25. In this study, we evaluated the QCancer-10 + PRS model, which not only met this AUC criterion but also demonstrated similar performance to that reported in a previous study integrating the QCancer-10 model with a 50-variant PRS18. To further distinguish our approach, we compared it with the Asia-Pacific Colorectal Screening (APCS) score-based risk-stratified screening strategy used in an ongoing RCT in China17. Unlike the APCS strategy, which relies on traditional risk factors, our strategy including genetic risk factors enhanced the effectiveness of screening by approximately 3.41% and 2.67% for CRC incidence and mortality, respectively. We also determined ARRs, an important decision-making indicator for prioritizing health-care services26. In our study, decreases in CRC incidence and mortality were observed only in the intermediate-risk and high-risk groups (60% of the population), but not in the low-risk group. This difference might be attributable to several factors. Patients with a family history of CRC may start screening at relatively earlier ages and prefer colonoscopy to stool testing, thus potentially contributing to the observed decrease in these higher risk groups. Additionally, the roles of both PRS and most predictors in QCancer-10 throughout the entire CRC tumorigenesis pathway indicate a probable convergence of high-risk populations for CRC and its precancerous lesions6,27. Consequently, implementing risk-stratified screening combining genetic and non-genetic risk factors might lead to more precise CRC screening for both primary and secondary prevention.

With scientific advancements, precision medicine has been gradually introduced into cancer prevention and screening, and precision screening of CRC has attracted increasing attention28,29. We found that the benefits of CRC screening persisted in the intermediate-high risk group at the recommended screening ages. As urbanization and industrialization lead to lifestyle changes and affect CRC risk30, conducting long-term follow-up and less frequent screening in low-risk populations might be more prudent than implementing population-wide screening5. Furthermore, previous studies have suggested that the PRS may be a more accurate indicator of CRC risk in younger individuals31. The combination of PRS and the QCancer-10 model may offer a more comprehensive and accurate assessment of CRC risk to address the rising incidence of early-onset CRC diagnosed in people younger than 50 years32, because the QCancer-10 covers a broad age range (25–84 years). Further research remains needed.

Gender-specific screening recommendations are generally not available in CRC screening guidelines worldwide. CRC screening has been shown to be less effective and less cost-effective in women than men33,34. However, a prior decision analysis has found no difference in optimal screening strategies between men and women35. In the current study, the screening benefit persisted for men in the intermediate-high risk group, whereas for women, this benefit was observed only in the high-risk group. This difference might have been due to the varying effects of age on CRC risk between men and women. A previous study comparing age-specific CRC incidence and mortality has found that women reached similar rates to those in men approximately 4–8 years later36. Moreover, the known healthy volunteer bias in the UK Biobank, which is particularly strong in women, might further weaken the performance of the model and the effectiveness of risk stratification in women37. Because men have lower screening participation rates than women, highlighting the specific advantages of screening for different risk groups according to gender might help increase gender-specific screening participation17,38. However, given the higher risk of proximal colon cancer in women than men, which is challenging to detect during screening, and that women have lower levels of fecal hemoglobin than men39, further exploration of gender-specific risk factors or biomarkers and evaluation of the effects of specific screening methods is necessary.

Proximal colon cancer, whose incidence increases with age, has a poorer prognosis than distal tumors32. Previous studies have indicated that non-genetic risk factors and genetic architectures differ for cancers along the colorectum40,41. Thus, precision screening strategies should consider the anatomical subregion of CRC. Our study revealed that risk-stratified screening was more successful in detecting distal CRC than proximal colon cancer. Our results are consistent with most findings from population-wide fecal occult blood tests or sigmoidoscopy screening. This finding is attributable to the colon transition time influencing hemoglobin degradation and the limited length of sigmoidoscopy42,43. Furthermore, observational results regarding the efficacy of colonoscopy screening for proximal colon cancer remain inconsistent, and an RCT remains ongoing42,44–46. To improve screening efficacy for proximal colon cancer, further research is needed, building upon the evaluation of risk-stratified screening for overall CRC.

To our knowledge, this study is one of the few studies providing evidence of the long-term effects of risk-stratified CRC screening by integrating non-genetic and genetic factors. The strengths of this study include its cohort study design with a large sample size and long-term follow-up, as well as the availability of extensive data on demographic and lifestyle factors, which allowed us to carefully consider potential confounding factors. Several limitations warrant attention. First, data on the type, frequency, and date of each participant’s CRC screening were not available. This limitation prevented further investigation of the effectiveness of risk-stratified screening with respect to specific screening methods. However, our results showed significant differences of screening effectiveness between different risk groups. A previous study has suggested that no reason exists to believe that this effect would vary depending on the screening methods16. Second, although the benefits of CRC screening were seen only in intermediate- and high-risk populations, this decrease in CRC incidence and mortality must be weighed against the costs, harmful effects during screening, and potential ethical issues of blood collection for PRS determination. Third, information on environmental factors and CRC screening history relied on self-reported exposure, which inevitably caused recall bias. Fourth, the population-specific cut-off values based on the QCancer-10 + PRS model must be validated in other populations. Finally, the UK Biobank study is a general population cohort with a lower incidence of CRC than observed in populations of individuals at the recommended age for screening. Therefore, risk stratification might be more effective in the actual screening population47.

Conclusions

By integrating an existing non-genetic risk model (QCancer-10) and a 139-variant PRS, we provided preliminary evidence for prioritizing CRC screening across risk groups. Additionally, these findings provided new insights for formulating public health policy and rationally allocating health resources in real-world CRC screening settings.

Supporting Information

Conflict of interest statement

No potential conflicts of interest are disclosed.

Author contributions

Conceived and designed the analysis: Kexin Chen, Yubei Huang, Zhangyan Lyu, Hongji Dai.

Collected the data: Yu Zhang, Chao Sheng.

Contributed data or analysis tools: Fangfang Song, Fengju Song.

Performed the analysis: Yu Zhang, Chao Sheng.

Wrote the paper: Yu Zhang.

Data availability statement

The UK Biobank is an open access resource and bona fide researchers can apply to use the UK Biobank dataset by registering and applying at http://ukbiobank.ac.uk/register-apply/.

Acknowledgements

We thank all UK Biobank participants and staff. This research was performed with the UK Biobank Resource under approval application number 76092.

  • Received March 14, 2024.
  • Accepted May 20, 2024.

References

  1. 1.↵
    2. Sung H,
    3. Ferlay J,
    4. Siegel RL,
    5. Laversanne M,
    6. Soerjomataram I,
    7. Jemal A, et al.
    Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2021; 71: 209–49.
    OpenUrlCrossRefPubMed
  2. 2.↵
    2. Zheng S,
    3. Schrijvers JJA,
    4. Greuter MJW,
    5. Kats-Ugurlu G,
    6. Lu W,
    7. de Bock GH.
    Effectiveness of colorectal cancer (CRC) screening on all-cause and CRC-specific mortality reduction: a systematic review and meta-analysis. Cancers (Basel). 2023; 15: 1948.
    OpenUrl
  3. 3.↵
    2. Hull MA,
    3. Rees CJ,
    4. Sharp L,
    5. Koo S.
    A risk-stratified approach to colorectal cancer prevention and diagnosis. Nat Rev Gastroenterol Hepatol. 2020; 17: 773–80.
    OpenUrlPubMed
  4. 4.
    2. Lansdorp-Vogelaar I,
    3. Meester R,
    4. de Jonge L,
    5. Buron A,
    6. Haug U,
    7. Senore C.
    Risk-stratified strategies in population screening for colorectal cancer. Int J Cancer. 2022; 150: 397–405.
    OpenUrl
  5. 5.↵
    2. Kastrinos F,
    3. Kupfer SS,
    4. Gupta S.
    Colorectal cancer risk assessment and precision approaches to screening: brave new world or worlds apart? Gastroenterology. 2023; 164: 812–27.
    OpenUrlCrossRef
  6. 6.↵
    2. Keum N,
    3. Giovannucci E.
    Global burden of colorectal cancer: emerging trends, risk factors and prevention strategies. Nat Rev Gastroenterol Hepatol. 2019; 16: 713–32.
    OpenUrlCrossRefPubMed
  7. 7.↵
    2. Peng L,
    3. Weigl K,
    4. Boakye D,
    5. Brenner H.
    Risk scores for predicting advanced colorectal neoplasia in the average-risk population: a systematic review and meta-analysis. Am J Gastroenterol. 2018; 113: 1788–800.
    OpenUrlPubMed
  8. 8.↵
    2. Saunders CL,
    3. Kilian B,
    4. Thompson DJ,
    5. McGeoch LJ,
    6. Griffin SJ,
    7. Antoniou AC, et al.
    External validation of risk prediction models incorporating common genetic variants for incident colorectal cancer using UK Biobank. Cancer Prev Res (Phila). 2020; 13: 509–20.
    OpenUrlAbstract/FREE Full Text
  9. 9.↵
    2. Usher-Smith JA,
    3. Harshfield A,
    4. Saunders CL,
    5. Sharp SJ,
    6. Emery J,
    7. Walter FM, et al.
    External validation of risk prediction models for incident colorectal cancer using UK Biobank. Br J Cancer. 2018; 118: 750–9.
    OpenUrlCrossRefPubMed
  10. 10.↵
    2. Hippisley-Cox J,
    3. Coupland C.
    Development and validation of risk prediction algorithms to estimate future risk of common cancers in men and women: prospective cohort study. BMJ Open. 2015; 5: e007825.
  11. 11.↵
    2. Helsingen LM,
    3. Vandvik PO,
    4. Jodal HC,
    5. Agoritsas T,
    6. Lytvyn L,
    7. Anderson JC, et al.
    Colorectal cancer screening with faecal immunochemical testing, sigmoidoscopy or colonoscopy: a clinical practice guideline. Br Med J. 2019; 367: l5515.
  12. 12.↵
    2. Thomas M,
    3. Sakoda LC,
    4. Hoffmeister M,
    5. Rosenthal EA,
    6. Lee JK,
    7. van Duijnhoven FJB, et al.
    Genome-wide modeling of polygenic risk score in colorectal cancer risk. Am J Hum Genet. 2020; 107: 432–44.
    OpenUrlCrossRefPubMed
  13. 13.
    2. Law PJ,
    3. Timofeeva M,
    4. Fernandez-Rozadilla C,
    5. Broderick P,
    6. Studd J,
    7. Fernandez-Tajes J, et al.
    Association analyses identify 31 new risk loci for colorectal cancer susceptibility. Nat Commun. 2019; 10: 2154.
    OpenUrlCrossRefPubMed
  14. 14.↵
    2. Huyghe JR,
    3. Bien SA,
    4. Harrison TA,
    5. Kang HM,
    6. Chen S,
    7. Schmit SL, et al.
    Discovery of common and rare genetic risk variants for colorectal cancer. Nat Genet. 2019; 51: 76–87.
    OpenUrlCrossRefPubMed
  15. 15.↵
    2. Wang K,
    3. Ma W,
    4. Wu K,
    5. Ogino S,
    6. Giovannucci EL,
    7. Chan AT, et al.
    Long-term colorectal cancer incidence and mortality after colonoscopy screening according to individuals’ risk profiles. J Natl Cancer Inst. 2021; 113: 1177–85.
    OpenUrlPubMed
  16. 16.↵
    2. Choi J,
    3. Jia G,
    4. Wen W,
    5. Long J,
    6. Shu XO,
    7. Zheng W.
    Effects of screenings in reducing colorectal cancer incidence and mortality differ by polygenic risk scores. Clin Transl Gastroenterol. 2021; 12: e00344.
  17. 17.↵
    2. Chen H,
    3. Lu M,
    4. Liu C,
    5. Zou S,
    6. Du L,
    7. Liao X, et al.
    Comparative evaluation of participation and diagnostic yield of colonoscopy vs fecal immunochemical test vs risk-adapted screening in colorectal cancer screening: interim analysis of a multicenter randomized controlled trial (TARGET-C). Am J Gastroenterol. 2020; 115: 1264–74.
    OpenUrlPubMed
  18. 18.↵
    2. Briggs SEW,
    3. Law P,
    4. East JE,
    5. Wordsworth S,
    6. Dunlop M,
    7. Houlston R, et al.
    Integrating genome-wide polygenic risk scores and non-genetic risk to predict colorectal cancer diagnosis using UK Biobank data: population based cohort study. Br Med J. 2022; 379: e071707.
  19. 19.↵
    2. Sudlow C,
    3. Gallacher J,
    4. Allen N,
    5. Beral V,
    6. Burton P,
    7. Danesh J, et al.
    UK Biobank: an open access resource for identifying the causes of a wide range of complex diseases of middle and old age. PLoS Med. 2015; 12: e1001779.
  20. 20.↵
    2. Bycroft C,
    3. Freeman C,
    4. Petkova D,
    5. Band G,
    6. Elliott LT,
    7. Sharp K, et al.
    The UK Biobank resource with deep phenotyping and genomic data. Nature. 2018; 562: 203–9.
    OpenUrlCrossRefPubMed
  21. 21.↵
    2. Purcell S,
    3. Neale B,
    4. Todd-Brown K,
    5. Thomas L,
    6. Ferreira MA,
    7. Bender D, et al.
    Plink: a tool set for whole-genome association and population-based linkage analyses. Am J Hum Genet. 2007; 81: 559–75.
    OpenUrlCrossRefPubMed
  22. 22.↵
    2. Lloyd-Jones DM,
    3. Hong Y,
    4. Labarthe D,
    5. Mozaffarian D,
    6. Appel LJ,
    7. Van Horn L, et al.
    Defining and setting national goals for cardiovascular health promotion and disease reduction: the American Heart Association’s strategic Impact Goal through 2020 and beyond. Circulation. 2010; 121: 586–613.
    OpenUrlAbstract/FREE Full Text
  23. 23.↵
    2. Camp RL,
    3. Dolled-Filhart M,
    4. Rimm DL.
    X-tile: a new bio-informatics tool for biomarker assessment and outcome-based cut-point optimization. Clin Cancer Res. 2004; 10: 7252–9.
    OpenUrlAbstract/FREE Full Text
  24. 24.↵
    2. von Elm E,
    3. Altman DG,
    4. Egger M,
    5. Pocock SJ,
    6. Gotzsche PC,
    7. Vandenbroucke JP, et al.
    The strengthening the reporting of observational studies in epidemiology (STROBE) statement: guidelines for reporting observational studies. Lancet. 2007; 370: 1453–7.
    OpenUrlCrossRefPubMed
  25. 25.↵
    2. Naber SK,
    3. Kundu S,
    4. Kuntz KM,
    5. Dotson WD,
    6. Williams MS,
    7. Zauber AG, et al.
    Cost-effectiveness of risk-stratified colorectal cancer screening based on polygenic risk: current status and future potential. JNCI Cancer Spectr. 2019; 4: pkz086.
  26. 26.↵
    2. Bretthauer M.
    Colorectal cancer screening. J Intern Med. 2011; 270: 87–98.
    OpenUrlCrossRefPubMed
  27. 27.↵
    2. Obon-Santacana M,
    3. Diez-Villanueva A,
    4. Alonso MH,
    5. Ibanez-Sanz G,
    6. Guino E,
    7. Lopez A, et al.
    Polygenic risk score across distinct colorectal cancer screening outcomes: from premalignant polyps to colorectal cancer. BMC Med. 2021; 19: 261.
    OpenUrl
  28. 28.↵
    2. Robert DL,
    3. Thisun J,
    4. Sabrina K,
    5. Fan Z,
    6. Susan JC,
    7. Alex B, et al.
    Pathogenesis and precision medicine for predicting response in inflammatory bowel disease: advances and future directions. eGastroenterology. 2024; 2: e100006.
  29. 29.↵
    2. Loomans-Kropp HA,
    3. Umar A.
    Cancer prevention and screening: the next step in the era of precision medicine. NPJ Precis Oncol. 2019; 3: 3.
    OpenUrl
  30. 30.↵
    2. Chen X,
    3. Ding J,
    4. Li H,
    5. Carr PR,
    6. Hoffmeister M,
    7. Brenner H.
    The power of a healthy lifestyle for cancer prevention: the example of colorectal cancer. Cancer Biol Med. 2022; 19: 1586–97.
    OpenUrlAbstract/FREE Full Text
  31. 31.↵
    2. Archambault AN,
    3. Su YR,
    4. Jeon J,
    5. Thomas M,
    6. Lin Y,
    7. Conti DV, et al.
    Cumulative burden of colorectal cancer-associated genetic variants is more strongly associated with early-onset vs late-onset cancer. Gastroenterology. 2020; 158: 1274–86.e12 .
    OpenUrlCrossRefPubMed
  32. 32.↵
    2. Siegel RL,
    3. Wagle NS,
    4. Cercek A,
    5. Smith RA,
    6. Jemal A.
    Colorectal cancer statistics, 2023. CA Cancer J Clin. 2023; 73: 233–54.
    OpenUrlCrossRefPubMed
  33. 33.↵
    2. Holme Ø,
    3. Loberg M,
    4. Kalager M,
    5. Bretthauer M,
    6. Hernan MA,
    7. Aas E, et al.
    Long-term effectiveness of sigmoidoscopy screening on colorectal cancer incidence and mortality in women and men: a randomized trial. Ann Intern Med. 2018; 168: 775–82.
    OpenUrlCrossRefPubMed
  34. 34.↵
    2. Theuer CP,
    3. Taylor TH,
    4. Brewster WR,
    5. Anton-Culver H.
    Gender and race/ethnicity affect the cost-effectiveness of colorectal cancer screening. J Natl Med Assoc. 2006; 98: 51–7.
    OpenUrlPubMed
  35. 35.↵
    2. Meester RGS,
    3. Peterse EFP,
    4. Knudsen AB,
    5. de Weerdt AC,
    6. Chen JC,
    7. Lietz AP, et al.
    Optimizing colorectal cancer screening by race and sex: microsimulation analysis II to inform the American Cancer Society colorectal cancer screening guideline. Cancer. 2018; 124: 2974–85.
    OpenUrl
  36. 36.↵
    2. Brenner H,
    3. Hoffmeister M,
    4. Arndt V,
    5. Haug U.
    Gender differences in colorectal cancer: implications for age at initiation of screening. Br J Cancer. 2007; 96: 828–31.
    OpenUrlCrossRefPubMed
  37. 37.↵
    2. Fry A,
    3. Littlejohns TJ,
    4. Sudlow C,
    5. Doherty N,
    6. Adamska L,
    7. Sprosen T, et al.
    Comparison of sociodemographic and health-related characteristics of UK Biobank participants with those of the general population. Am J Epidemiol. 2017; 186: 1026–34.
    OpenUrlCrossRefPubMed
  38. 38.↵
    2. Botteri E,
    3. Hoff G,
    4. Randel KR,
    5. Holme O,
    6. de Lange T,
    7. Bernklev T, et al.
    Characteristics of nonparticipants in a randomised colorectal cancer screening trial comparing sigmoidoscopy and faecal immunochemical testing. Int J Cancer. 2022; 151: 361–71.
    OpenUrl
  39. 39.↵
    2. White A,
    3. Ironmonger L,
    4. Steele RJC,
    5. Ormiston-Smith N,
    6. Crawford C,
    7. Seims A.
    A review of sex-related differences in colorectal cancer incidence, screening uptake, routes to diagnosis, cancer stage and survival in the UK. BMC Cancer. 2018; 18: 906.
    OpenUrlCrossRefPubMed
  40. 40.↵
    2. Huyghe JR,
    3. Harrison TA,
    4. Bien SA,
    5. Hampel H,
    6. Figueiredo JC,
    7. Schmit SL, et al.
    Genetic architectures of proximal and distal colorectal cancer are partly distinct. Gut. 2021; 70: 1325–34.
    OpenUrlAbstract/FREE Full Text
  41. 41.↵
    2. Wang L,
    3. Lo CH,
    4. He X,
    5. Hang D,
    6. Wang M,
    7. Wu K, et al.
    Risk factor profiles differ for cancers of different regions of the colorectum. Gastroenterology. 2020; 159: 241–56.e13 .
    OpenUrlCrossRefPubMed
  42. 42.↵
    2. Brenner H,
    3. Stock C,
    4. Hoffmeister M.
    Effect of screening sigmoidoscopy and screening colonoscopy on colorectal cancer incidence and mortality: systematic review and meta-analysis of randomised controlled trials and observational studies. Br Med J. 2014; 348: g2467.
  43. 43.↵
    2. Lu M,
    3. Luo X,
    4. Li N,
    5. Chen H,
    6. Dai M.
    Diagnostic accuracy of fecal occult blood tests for detecting proximal versus distal colorectal neoplasia: a systematic review and meta-analysis. Clin Epidemiol. 2019; 11: 943–54.
    OpenUrlCrossRefPubMed
  44. 44.↵
    2. Steffen A,
    3. Weber MF,
    4. Roder DM,
    5. Banks E.
    Colorectal cancer screening and subsequent incidence of colorectal cancer: results from the 45 and Up Study. Med J Aust. 2014; 201: 523–7.
    OpenUrl
  45. 45.
    2. Morois S,
    3. Cottet V,
    4. Racine A,
    5. Clavel-Chapelon F,
    6. Carbonnel F,
    7. Bastide N, et al.
    Colonoscopy reduced distal colorectal cancer risk and excess cancer risk associated with family history. Cancer Causes Control. 2014; 25: 1329–36.
    OpenUrlCrossRefPubMed
  46. 46.↵
    2. Bretthauer M,
    3. Loberg M,
    4. Wieszczy P,
    5. Kalager M,
    6. Emilsson L,
    7. Garborg K, et al.
    Effect of colonoscopy screening on risks of colorectal cancer and related death. N Engl J Med. 2022; 387: 1547–56.
    OpenUrlCrossRefPubMed
  47. 47.↵
    2. Usher-Smith JA,
    3. Sharp SJ,
    4. Griffin SJ.
    The spectrum effect in tests for risk prediction, screening, and diagnosis. Br Med J. 2016; 353: i3139.
Back to top

In this issue

Print
Download PDF
Email Article
Citation Tools

Jump to section

Keywords