- Email: [email protected]
Implications of polygenic risk for personalised colorectal cancer screening
Annals of Oncology Advance Access published November 16, 2015 1
Implications of polygenic risk for personalised colorectal cancer screening
M. Frampton1, P. Law1, K. Litchfield1, E.J. Morris2, D. Kerr3, C. Turnbull1,4, I.P. Tomlinson5, R.S. Houlston1
1
Division of Genetics and Epidemiology, The Institute of Cancer Research, London, UK
2
Section of Epidemiology and Biostatistics, Leeds Institute of Cancer and Pathology, University of
3
Oxford Cancer Centre, Department of Oncology, University of Oxford, Churchill Hospital, Old
Road, Headington, Oxford, UK
4
William Harvey Research Institute, Queen Mary University London, Charterhouse Square, London,
UK
5
Molecular and Population Genetics Laboratory, Wellcome Trust Centre for Human Genetics,
University of Oxford, Oxford, UK
Correspondence to: Dr Richard S Houlston; Division of Genetics and Epidemiology, The Institute of Cancer Research, London, UK, Tel: ++44 (0) 208 722 4175; E-mail: [email protected]
© The Author 2015. Published by Oxford University Press on behalf of the European Society for Medical Oncology. All rights reserved. For permissions, please email: [email protected].
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
Leeds, Leeds, UK
2
ABSTRACT
Background: We modelled the utility of applying a personalised screening approach for colorectal cancer (CRC) as compared to standard age-based screening. In this personalised screening approach, eligibility is determined by absolute risk which is calculated from age and polygenic risk score (PRS), where the PRS is relative risk attributable to common genetic variation. By contrast,
Design: We calculated absolute risks of CRC from UK population age structure, incidence and mortality rate data, and a PRS distribution which we derived for the 37 known CRC susceptibility variants. We compared the number of CRC cases potentially detectable by personalised and agebased screening. Using Genome-Wide Complex Trait Analysis to calculate the heritability attributable to common variation, we repeated the analysis assuming all common CRC risk variants were known. Results: Based on the known CRC variants, individuals with a PRS in the top 1% have a 2.9-fold increased CRC risk over the population median. Compared with age-based screening (aged 60: 10year absolute risk 1.96% in men, 1.19% in women, as per the UK NHS National Bowel Screening Programme), personalised screening of individuals aged 55-69 at the same risk would lead to 16% fewer men and 17% fewer women being eligible for screening with 10% and 8% respectively fewer screen-detected cases. If all susceptibility variants were known, individuals with a PRS in the top 1% would have an estimated 7.7-fold increased risk. Personalised screening would then result in 26% fewer men and women being eligible for screening with 7% and 5% fewer screen-detected cases.
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
eligibility in age-based screening is determined only by age.
3
Conclusion: Personalised screening using PRS has the potential to optimise population screening for CRC and to define those likely to maximally benefit from chemoprevention. There are however significant technical and operational details to be addressed before any such programme is introduced.
Keywords: Colorectal cancer, polygenic risk, personalised screening
Personalised screening using PRS has the potential to optimise population screening for colorectal cancer and to define those likely to maximally benefit from chemoprevention. There are however significant technical and operational details to be addressed before any such programme is introduced."
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
Key Message: "Polygenic risk score (PRS) is relative risk attributable to common genetic variation
4
INTRODUCTION
In the Western countries, colorectal cancer (CRC) affects over half a million individuals each year [1]. Despite advances in the management of CRC over the last 25 years, the five-year survival remains only around 55%[2].
adenomatous dysplasia. This, together with the strong correlation between stage and survival, [2] provides the rationale for CRC screening programmes.
Colonoscopy is a highly effective screening tool for CRC but is expensive and can be associated with significant morbidity. Hence population screening within the UK National Health Service (NHS) Bowel Cancer Screening Programme (NHS BCSP)[3] and other similar schemes [4] is primarily based on Faecal Occult Blood Testing (FOBT), which has low sensitivity. While the benefit of population screening in reducing CRC mortality is greater than any ensuing harm from overtreatment, stratifying individuals according to their prior CRC risk offers the possibility of targeting screening at those most likely to benefit, including high-risk individuals currently ineligible for screening.
In addition to diet and other lifestyle factors, inherited susceptibility contributes significantly to CRC risk; specifically individuals with CRC in a first-degree relative have a two-fold increased risk [5]. High-risk mutations in APC and the mismatch repair genes account for <5% of all CRC and most of the heritable risk appears to be polygenic [5]. This model of CRC has been supported by
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
The majority of CRC develop from normal epithelium through sequentially worsening degrees of
5
genome-wide association studies (GWAS) [6] (Supplementary Table S1). As demonstrated for breast and prostate cancer, the combined effect of multiple risk SNPs has the potential to achieve a degree of risk discrimination that is useful for population-based prevention and screening programmes [7-9].
Here we have assessed the efficacy of a personalised screening strategy for CRC based on age and polygenic risk score (PRS), i.e. relative risk (RR) attributable to common genetic variation. The
means additional risk variants should be identifiable by new GWAS. This will inevitably increase the precision of PRS based personalised screening. Hence in addition to evaluating the utility of PRS based on the 37 SNPs documented to influence CRC risk in Europeans, we estimated the heritability attributable to all common variants, to examine the predictive value of all possible risk variants.
MATERIALS AND METHODS
To examine the utility of personalised screening for CRC we adopted the methodology of Pashayan et al in their analysis of breast and prostate cancer [7]. We compared the size of the population eligible for screening and the number of CRC cases potentially detectable under age-based screening where eligibility is determined solely by age, and personalised screening where it is determined by absolute risk which is calculated from age and PRS. Additionally we constructed a model of all common risk variants for CRC and evaluated the impact of different age as entry criteria for screening.
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
over-representation of association signals in existing GWAS after accounting for known risk SNPs
6
Calculation of absolute colorectal cancer risk We obtained incident CRC registrations, deaths from CRC and all causes, and mid-year population estimates in 1-year age bands for England (2001-2012) from the Office for National Statistics. Averaging incidence and mortality rates for this period, we used DevCan 6.4.1 software[10], to derive the age-conditional absolute risk of CRC. We then estimated the age-conditional absolute CRC risk for individuals from PRS and CRC incidence.
Details of the 37 risk loci for CRC in Europeans are summarised in Supplementary Table S1. We estimated the variance of the distribution of PRS in the population using the published allele frequencies and per-allele RR under a log-additive model of interaction between risk alleles (Supplementary Table S1). The distribution of risk on a RR scale in the population is log-normal with mean μ, and variance σ2; the distribution of PRS among cases being displaced to the right by σ2. We set the population μ to be −σ2/2, so that the overall mean PRS is 1.0 [7]. We assessed the discriminatory capability of all risk variants through receiver operator characteristic (ROC) curves.
Heritability of all common risk variants for colorectal cancer We calculated the heritable risk of CRC attributable to all common variants by applying GenomeWide Complex Trait Analysis (GCTA)[11]. To avoid inflation of heritability from inclusion of selected familial or early-onset CRC we analysed data from the VQ58 GWAS which has not previously been used as a discovery dataset [12]. Briefly, VQ58 comprised 1,800 CRC cases from the UK-based VICTOR and QUASAR2 adjuvant chemotherapy clinical trials. VICTOR and QUASAR2 (VQ) cases were genotyped using the Illumina HumanHap300 and HumanHap370 arrays. The 2,690 (58) controls, genotyped on the Illumina Human-1.2M-Duo Custom_v1 array, were from the UK population-based 1958 Birth Cohort. As previously advocated when calculating the heritability
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
Calculation of polygenic risk score
7
of common diseases such as cancer we used the lifetime risk of CRC (0.046) to transform data to the liability scale [13].
Modelling population screening The proportion of individuals with a given PRS in the population and in those affected with CRC was calculated from the mean and variance of the log-normal RR distribution. Using these data we estimated the proportion of the population at a higher than given absolute risk threshold and the
Screening for CRC is currently offered to individuals in England between the ages 60 to 69 years through the NHS BCSP. Individuals aged below 60 are ineligible but those aged 70 and over can request screening. The population median 10-year absolute risk of CRC for 60 year old males and females in England is 1.96% and 1.19% respectively. We used these risk values to compare the current age-based screening programme with a personalised approach in individuals aged 55-69 years.
RESULTS
Polygenic risk score The 37 currently identified susceptibility variants for CRC confer a polygenic variance of 0.21 and account for around 14.5% of the familial risk of CRC (Supplementary Table S1). Figure 1 shows the PRS based on the 37 known risk variants for CRC. Individuals within the top 10% of genetic risk have a 1.8-fold increased risk of CRC and those within the top 1% (i.e. 26 – 53 risk alleles) have a 2.9-fold increased risk of CRC as compared to the population median. The area under the corresponding ROC curve for all of the 37 SNPs is 0.63 (Supplementary Figure S1).
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
proportion of CRC diagnosed in this high-risk group.
8
Performance of personalised screening incorporating PRS In England during the period 2001 to 2012 inclusive, on average 7,375,504 of the population was aged 55-69 years and within this age group 5,876 men and 3,743 women were diagnosed with CRC. Figure 2 shows the age-conditional absolute risks of CRC over 10 years for men and women. Table 1 partitions a simulated population of 100,000 individuals aged 55-69 according to eligibility for age-based and personalised screening. Within this cohort 61% of men (61,202/100,000) and
(76/99) of CRC cases being diagnosed in this subset of the population (Table 1). Under the personalised screening model based on the known risk variants, 45% of men (45,344/100,000) and 45% of women (45,467/100,000) would be eligible for screening with 69% of male cases (113/163) and 69% of female cases (68/99) being identified. In absolute terms this would translate to, 16% fewer men and 17% fewer women being eligible for screening at the cost of detecting 10% and 8% fewer cases respectively.
Impact of identifying all common risk variants on personalised screening Using GCTA, we estimated the heritability of CRC for common variation to be 19% (± 8%), which translates to common variation accounting for 55% of the familial risk of CRC. On the basis of this heritability estimate, Figure 1 shows the PRS model if all possible susceptibility variants for CRC were known. Individuals within the top 10% of genetic risk have a 3.1-fold increased risk of CRC and those within the top 1% have a 7.7-fold increased risk of CRC as compared to the population median. The increased individual risk discrimination is reflected in the ROC with the area under the curve being 0.73 (Supplementary Figure S1). Table 1 shows the performance of personalised screening incorporating PRS where it is assumed all susceptibility variants for CRC are known. Under this model 35% of men and 36% of women would be eligible for screening with 72% of
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
62% of women (62,091/100,000) are eligible for age-based screening with 79% (128/163) and 77%
9
cases being identified. Therefore, 26% fewer men and 26% fewer women would be eligible for screening at the cost of detecting 7% and 5% fewer cases.
Changing entry age and its impact on the performance of personalised screening The Scottish bowel cancer screening program is open to individuals aged 50-74 and there is a move within England to extend the upper age of screening to age 75. In England from 2001-2012, the average population aged 50-74 was 12,684,623, with 9,442 men and 6,216 women diagnosed
then compared the effect of the higher NHS entry age (Scenario 1) and offering screening to higher-risk 50-59 year-olds in addition to those aged 60-69 (Scenario 2). Here a higher risk 50-59 year old is one whose 10-year absolute risk of developing CRC is equal to or greater than the average for an individual aged 60 (1.96% in men and 1.19% in women). Under Scenario 1 the effect of raising the screening entry age from 50 to 60 years results in only 50% of the male population being eligible for screening but 79% of the cases remain potentially detectable (Supplementary Table S2). Under Scenario 2, based on the known risk SNPs, 56% of the population would be eligible and 85% of cases potentially detectable; if all risk SNPs were known, then the respective metrics would be 58% and 89%.
DISCUSSION
Our analysis demonstrates that personalised screening programmes for CRC, in which eligibility is based on genetic risk profile in addition to age, have the potential to greatly reduce the number of individuals screened whilst still detecting nearly as many cases. Moreover, the additional detected cases will include early detection in younger genetically-susceptible individuals, and this will
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
with CRC annually. We simulated populations of 100,000 50-74 year-old males and females and
10
produce a substantial gain in life years. This is not only due to these individuals’ younger age, but also because it appears genetic susceptibility can be associated with tumour aggressiveness.
Currently the NHS BCSP is based on the use of FOBT and participation using this screening tool is only around 60% [14]. There is however a move to replace FOBT with the Faecal Immunochemical Test (FIT) which is acknowledged to be a better initial test and is generally considered to be more acceptable [15]. Implementation of FIT-based screening in conjunction with PRS-based risk
influence detection rates.
Although family history information is used extensively to estimate the risk of CRC in family cancer clinics there is considerable potential for recall bias and inaccuracy in the reporting of CRC [16, 17]. Indeed studies have shown that CRC in relatives is often substantially under-reported [16]. Profiling for CRC susceptibility by genetic testing directly addresses this shortcoming.
It is possible that those identified to be at higher CRC risk by virtue of their genetic profile might be more likely to participate in screening, thereby further enhancing the performance of the programme. There is already well-established screening activity managed by specialist genetic and oncology centres for those individuals aged below 60 years with a family history of CRC [18-20]. However, creating such genetic screening programs does not necessarily mean that they will be utilised by health care providers and their clients. Indeed it is acknowledged that the efficacy of genetic risk profiling has been shown to be limited by the decisions and circumstances of those offered germline testing [21]. Hence empiric data is required to determine the performance of any personalised screening applied to the population.
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
profiling would then translate to both better uptake and better detection which will both
11
We acknowledge that there are limitations to our analysis. Specifically, our modelling has been based on estimating the proportion of the population eligible for screening and the proportion of CRC cases screen-detectable in this subgroup according to their genetic risk. We have assumed that all CRC cases are potentially screen-detectable, and so have not considered screening sensitivity.
The sensitivity of any bowel screening programme will increase with a reduction in the inter-
the duration of a screen-detectable phase impacts on programme performance. This is of special relevance to CRC as a major component of genetic susceptibility is mediated through propensity to develop pre-malignant polyps and an increased rate of transitioning of polyps to overt cancer. Hence it could be anticipated that genetic profiling will improve sensitivity a priori, since polypectomy will have a preventative effect in the longer term. Counter to this is that screening programs such as NHS-BCSP based on FOBT or flexible-sigmoidoscopy tend to miss more rightsided tumours, which have a greater tendency to be genetic, than those within the left side of the bowel [22]. Additionally, assuming the caveat of equivalent or improved screening programme sensitivity, detailed economic evaluation is required to determine if the costs associated with implementing personalised screening would be offset by savings from minimising follow up for individuals with false positive screens. Such economic evaluation will likely require empiric data from pilot studies to be truly meaningful.
Here we have based our analysis on National Cancer Registration Data from England to estimate the numbers of the population aged 55-69 years that are eligible for screening and the number of CRC cases potentially detectable in the eligible population stipulating different risk thresholds. The
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
screening interval and in the duration of the pre-clinical screen-detectable phase. Furthermore,
12
optimum risk threshold for the population of England will be different to other populations, such as those in Asia which have much lower incidences of CRC.
While personalised screening is an attractive concept there are a number of significant issues which would have to be addressed before it could be implemented. First and foremost, there will be the contentious issue of moving from a screening program of universal inclusion (based on age), to one in which screening is only offered to some individuals. Indeed it is notable that the
NHS BCSP or Scottish screening programmes. As with any common cancer, tumours will develop in significant numbers in the swathe of the population deemed to be at ‘low risk’ and for whom screening was deemed not to be required. Indeed the performance of personalised screening for CRC appears on the basis of our analysis to be less profound than that offered by adopting such a programme for prostate cancer [7]. Secondly, genetic risk profiling will inevitably add complexity to any screening programme. Finally, as with all forms of genetic testing there are ethical and legal challenges that would have to be addressed prior to any such programme being implemented [23].
In summary, personalised screening making use of a PRS has the potential to optimise the efficiency of screening programmes for CRC. Clearly as demonstrated the identification of additional risk SNPs and incorporation of Mendelian susceptibility alleles will afford better discrimination. In addition to PRS, there is clearly scope for individualised screening based on incorporating family history and non-genetic risk factors such as lifestyle risk factors as well as biomarker data. Real world ‘road testing’ is however needed to establish true efficiency and to address the technical and operational issues. Eligibility for screening using age as the entry criteria is generally accepted by health care professionals and the public alike, hence it remains to be
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
different profile of CRC risk in men and women has not been taken into account in delivery of the
13
established whether eligibility based on age and other factors would be acceptable. Since personalised screening is likely to afford the benefit of detecting CRC in younger subjects at high risk, its use to extend screening to such individuals would be more logistically attractive and acceptable to the public. Furthermore, the observed level of risk discrimination from PRS is likely to be highly informative in formulating and delivering chemoprevention strategies for CRC. Currently such initiatives have been confined to Mendelian predisposition to CRC with recent reports showing the value of aspirin in carriers of MMR gene mutations [24]. However, given the
identify a relatively high number of young at risk individuals there is a rationale for investing in PRS defined chemoprevention trials.
DISCLOSURE The authors have declared no conflicts of interest.
FUNDING Work was funded by Cancer Research UK, the European Union (258236, FP7 collaborative project SYSCOL). KL is in receipt of an ICR studentship.
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
level of risk discrimination afforded by even the currently known CRC SNPs and the opportunity to
14
REFERENCES
1.
(2013) CRU.) Cancer Statistics—Bowel Cancer. Available: http://www.cancerresearchuk.org/cancer-
info/cancerstats/types/bowel/. : http://www.cancerresearchuk.org/cancer-info/cancerstats/types/bowel/
2.
Network NCI. Colorectal cancer survival by stage. . Avilable:
http://wwwncinorguk/publications/data_briefings/colorectal_cancer_survival_by_stage. 2014. 3.
Programme NBCS. http://www.cancerscreening.nhs.uk/bowel/index.html.
4.
Programme BSSBS. http://www.bowelscreening.scot.nhs.uk/.
5.
Aaltonen L, Johns L, Jarvinen H et al. Explaining the familial colorectal cancer risk associated with
mismatch repair (MMR)-deficient and MMR-stable tumors. Clin Cancer Res 2007; 13: 356-361. 6.
Fletcher O, Houlston RS. Architecture of inherited susceptibility to common cancer. Nat Rev Cancer
2010; 10: 353-361. 7.
Pashayan N, Duffy SW, Chowdhury S et al. Polygenic susceptibility to prostate and breast cancer:
implications for personalised screening. Br J Cancer 2011; 104: 1656-1663. 8.
Garcia-Closas M, Gunsoy NB, Chatterjee N. Combined associations of genetic and environmental
risk factors: implications for prevention of breast cancer. J Natl Cancer Inst 2014; 106. 9.
Mavaddat N, Pharoah PD, Michailidou K et al. Prediction of breast cancer risk based on profiling
with common genetic variants. J Natl Cancer Inst 2015; 107. 10.
Fay MP, Pfeiffer R, Cronin KA et al. Age-conditional probabilities of developing cancer. Stat Med
2003; 22: 1837-1848. 11.
Yang J, Lee SH, Goddard ME, Visscher PM. GCTA: a tool for genome-wide complex trait analysis. Am
J Hum Genet 2011; 88: 76-82.
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
2013.
15 12.
Dunlop MG, Dobbins SE, Farrington SM et al. Common variation near CDKN1A, POLD3 and
SHROOM2 influences colorectal cancer risk. Nat Genet 2012; 44: 770-776. 13.
Lee SH, Harold D, Nyholt DR et al. Estimation and partitioning of polygenic variation captured by
common SNPs for Alzheimer's disease, multiple sclerosis and endometriosis. Hum Mol Genet 2013; 22: 832841. 14.
Moss SM, Campbell C, Melia J et al. Performance measures in three rounds of the English bowel
cancer screening pilot. Gut 2012; 61: 101-107. 15.
Young GP, Symonds EL, Allison JE et al. Advances in Fecal Occult Blood Tests: the FIT revolution. Dig
16.
Mitchell RJ, Brewster D, Campbell H et al. Accuracy of reporting of family history of colorectal
cancer. Gut 2004; 53: 291-295. 17.
Mai PL, Garceau AO, Graubard BI et al. Confirmation of family cancer history reported in a
population-based survey. J Natl Cancer Inst 2011; 103: 788-797. 18.
Houlston RS, Murday V, Harocopos C et al. Screening and genetic counselling for relatives of
patients with colorectal cancer in a family cancer clinic. BMJ 1990; 301: 366-368. 19.
Carney PA, O'Malley JP, Gough A et al. Association between documented family history of cancer
and screening for breast and colorectal cancer. Prev Med 2013; 57: 679-684. 20.
Cairns SR, Scholefield JH, Steele RJ et al. Guidelines for colorectal cancer screening and surveillance
in moderate and high risk groups (update from 2002). Gut 2010; 59: 666-689. 21.
Ward RL, Hicks S, Hawkins NJ. Population-based molecular screening for Lynch syndrome:
implications for personalized medicine. J Clin Oncol 2013; 31: 2554-2562. 22.
Johns LE, Houlston RS. A systematic review and meta-analysis of familial colorectal cancer risk. Am J
Gastroenterol 2001; 96: 2992-3003. 23.
Hall AE, Chowdhury S, Hallowell N et al. Implementing risk-stratified screening for common
cancers: a review of potential ethical, legal and social issues. J Public Health (Oxf) 2014; 36: 285-291.
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
Dis Sci 2015; 60: 609-622.
16 24.
Burn J, Gerdes AM, Macrae F et al. Long-term effect of aspirin on cancer risk in carriers of
hereditary colorectal cancer: an analysis from the CAPP2 randomised controlled trial. Lancet 2011; 378: 2081-2087.
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
17
TABLE AND FIGURE LEGENDS
Table 1: Screening eligibility results for a simulated population of 100,000 men and 100,000 women aged 55-69. Individuals are classified as eligible for screening under three models: (a) age-
on known risk SNPs, where any individual aged 55-69 is eligible if their 10-year absolute risk of CRC (as calculated from age and PRS) is greater than or equal to the average for an individual aged 60 (men = 1.96%, women = 1.19%); (c) as for (b) but using PRS based on all risk SNPs.
Figure 1: Population distribution of PRS ordered by RR (compared to population median risk). (a) based on the known 37 risk SNPs; (b) based on all common variants. Vertical red lines (left to right) correspond to 1%, 10%, 50%, 90%, 99% centile respectively.
Figure 2: Ten-year absolute risk of being diagnosed with CRC in men and women in England (2001-2012).
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
based only, where an individual is eligible if ≥60 years; (b) personalised screening using PRS based
18
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
19
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
Table 1: Screening eligibility results for a simulated population of 100,000 men and 100,000 women aged 55-69. Individuals are classified as eligible for screening under three models: (a) age-based only, where an individual is eligible if ≥60 years; (b) personalised screening using PRS based on known risk SNPs, where any individual aged 55-69 is eligible if their 10-year absolute risk of CRC (as calculated from age and PRS) is greater than or equal to the average for an individual aged 60 (men = 1.96%, women = 1.19%); (c) as for (b) but using PRS based on all risk SNPs.
(a) Age-based screening < 60 years
≥ 60 years
Total
30,030 8,768 38,798
24,626 36,576 61,202
54,656 45,344 100,000
21 14 35
29 99 128
50 113 163
28,899 9,010 37,909
25,634 36,457 62,091
54,533 45,467 100,000
14 9 23
18 58 76
32 68 99
29,657 9,141 38,798
34,883 26,318 61,202
64,541 35,459 100,000
15 20 35
30 98 128
45 118 163
28,717 9,192 37,909
35,751 26,340 62,091
64,468 35,532 100,000
10 13 23
18 58 76
28 71 99
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
(b) Personalised screening using known risk SNPs Male population < 1.96% ≥ 1.96% Total Cases < 1.96% ≥ 1.96% Total Female population < 1.19% ≥ 1.19% Total Cases < 1.19% ≥ 1.19% Total (c) Personalised screening using all risk SNPs Male population < 1.96% ≥ 1.96% Total Cases < 1.96% ≥ 1.96% Total Female population < 1.19% ≥ 1.19% Total Cases < 1.19% ≥ 1.19% Total
Implications of polygenic risk for personalised colorectal cancer screening
M. Frampton1, P. Law1, K. Litchfield1, E.J. Morris2, D. Kerr3, C. Turnbull1,4, I.P. Tomlinson5, R.S. Houlston1
1
Division of Genetics and Epidemiology, The Institute of Cancer Research, London, UK
2
Section of Epidemiology and Biostatistics, Leeds Institute of Cancer and Pathology, University of
3
Oxford Cancer Centre, Department of Oncology, University of Oxford, Churchill Hospital, Old
Road, Headington, Oxford, UK
4
William Harvey Research Institute, Queen Mary University London, Charterhouse Square, London,
UK
5
Molecular and Population Genetics Laboratory, Wellcome Trust Centre for Human Genetics,
University of Oxford, Oxford, UK
Correspondence to: Dr Richard S Houlston; Division of Genetics and Epidemiology, The Institute of Cancer Research, London, UK, Tel: ++44 (0) 208 722 4175; E-mail: [email protected]
© The Author 2015. Published by Oxford University Press on behalf of the European Society for Medical Oncology. All rights reserved. For permissions, please email: [email protected].
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
Leeds, Leeds, UK
2
ABSTRACT
Background: We modelled the utility of applying a personalised screening approach for colorectal cancer (CRC) as compared to standard age-based screening. In this personalised screening approach, eligibility is determined by absolute risk which is calculated from age and polygenic risk score (PRS), where the PRS is relative risk attributable to common genetic variation. By contrast,
Design: We calculated absolute risks of CRC from UK population age structure, incidence and mortality rate data, and a PRS distribution which we derived for the 37 known CRC susceptibility variants. We compared the number of CRC cases potentially detectable by personalised and agebased screening. Using Genome-Wide Complex Trait Analysis to calculate the heritability attributable to common variation, we repeated the analysis assuming all common CRC risk variants were known. Results: Based on the known CRC variants, individuals with a PRS in the top 1% have a 2.9-fold increased CRC risk over the population median. Compared with age-based screening (aged 60: 10year absolute risk 1.96% in men, 1.19% in women, as per the UK NHS National Bowel Screening Programme), personalised screening of individuals aged 55-69 at the same risk would lead to 16% fewer men and 17% fewer women being eligible for screening with 10% and 8% respectively fewer screen-detected cases. If all susceptibility variants were known, individuals with a PRS in the top 1% would have an estimated 7.7-fold increased risk. Personalised screening would then result in 26% fewer men and women being eligible for screening with 7% and 5% fewer screen-detected cases.
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
eligibility in age-based screening is determined only by age.
3
Conclusion: Personalised screening using PRS has the potential to optimise population screening for CRC and to define those likely to maximally benefit from chemoprevention. There are however significant technical and operational details to be addressed before any such programme is introduced.
Keywords: Colorectal cancer, polygenic risk, personalised screening
Personalised screening using PRS has the potential to optimise population screening for colorectal cancer and to define those likely to maximally benefit from chemoprevention. There are however significant technical and operational details to be addressed before any such programme is introduced."
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
Key Message: "Polygenic risk score (PRS) is relative risk attributable to common genetic variation
4
INTRODUCTION
In the Western countries, colorectal cancer (CRC) affects over half a million individuals each year [1]. Despite advances in the management of CRC over the last 25 years, the five-year survival remains only around 55%[2].
adenomatous dysplasia. This, together with the strong correlation between stage and survival, [2] provides the rationale for CRC screening programmes.
Colonoscopy is a highly effective screening tool for CRC but is expensive and can be associated with significant morbidity. Hence population screening within the UK National Health Service (NHS) Bowel Cancer Screening Programme (NHS BCSP)[3] and other similar schemes [4] is primarily based on Faecal Occult Blood Testing (FOBT), which has low sensitivity. While the benefit of population screening in reducing CRC mortality is greater than any ensuing harm from overtreatment, stratifying individuals according to their prior CRC risk offers the possibility of targeting screening at those most likely to benefit, including high-risk individuals currently ineligible for screening.
In addition to diet and other lifestyle factors, inherited susceptibility contributes significantly to CRC risk; specifically individuals with CRC in a first-degree relative have a two-fold increased risk [5]. High-risk mutations in APC and the mismatch repair genes account for <5% of all CRC and most of the heritable risk appears to be polygenic [5]. This model of CRC has been supported by
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
The majority of CRC develop from normal epithelium through sequentially worsening degrees of
5
genome-wide association studies (GWAS) [6] (Supplementary Table S1). As demonstrated for breast and prostate cancer, the combined effect of multiple risk SNPs has the potential to achieve a degree of risk discrimination that is useful for population-based prevention and screening programmes [7-9].
Here we have assessed the efficacy of a personalised screening strategy for CRC based on age and polygenic risk score (PRS), i.e. relative risk (RR) attributable to common genetic variation. The
means additional risk variants should be identifiable by new GWAS. This will inevitably increase the precision of PRS based personalised screening. Hence in addition to evaluating the utility of PRS based on the 37 SNPs documented to influence CRC risk in Europeans, we estimated the heritability attributable to all common variants, to examine the predictive value of all possible risk variants.
MATERIALS AND METHODS
To examine the utility of personalised screening for CRC we adopted the methodology of Pashayan et al in their analysis of breast and prostate cancer [7]. We compared the size of the population eligible for screening and the number of CRC cases potentially detectable under age-based screening where eligibility is determined solely by age, and personalised screening where it is determined by absolute risk which is calculated from age and PRS. Additionally we constructed a model of all common risk variants for CRC and evaluated the impact of different age as entry criteria for screening.
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
over-representation of association signals in existing GWAS after accounting for known risk SNPs
6
Calculation of absolute colorectal cancer risk We obtained incident CRC registrations, deaths from CRC and all causes, and mid-year population estimates in 1-year age bands for England (2001-2012) from the Office for National Statistics. Averaging incidence and mortality rates for this period, we used DevCan 6.4.1 software[10], to derive the age-conditional absolute risk of CRC. We then estimated the age-conditional absolute CRC risk for individuals from PRS and CRC incidence.
Details of the 37 risk loci for CRC in Europeans are summarised in Supplementary Table S1. We estimated the variance of the distribution of PRS in the population using the published allele frequencies and per-allele RR under a log-additive model of interaction between risk alleles (Supplementary Table S1). The distribution of risk on a RR scale in the population is log-normal with mean μ, and variance σ2; the distribution of PRS among cases being displaced to the right by σ2. We set the population μ to be −σ2/2, so that the overall mean PRS is 1.0 [7]. We assessed the discriminatory capability of all risk variants through receiver operator characteristic (ROC) curves.
Heritability of all common risk variants for colorectal cancer We calculated the heritable risk of CRC attributable to all common variants by applying GenomeWide Complex Trait Analysis (GCTA)[11]. To avoid inflation of heritability from inclusion of selected familial or early-onset CRC we analysed data from the VQ58 GWAS which has not previously been used as a discovery dataset [12]. Briefly, VQ58 comprised 1,800 CRC cases from the UK-based VICTOR and QUASAR2 adjuvant chemotherapy clinical trials. VICTOR and QUASAR2 (VQ) cases were genotyped using the Illumina HumanHap300 and HumanHap370 arrays. The 2,690 (58) controls, genotyped on the Illumina Human-1.2M-Duo Custom_v1 array, were from the UK population-based 1958 Birth Cohort. As previously advocated when calculating the heritability
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
Calculation of polygenic risk score
7
of common diseases such as cancer we used the lifetime risk of CRC (0.046) to transform data to the liability scale [13].
Modelling population screening The proportion of individuals with a given PRS in the population and in those affected with CRC was calculated from the mean and variance of the log-normal RR distribution. Using these data we estimated the proportion of the population at a higher than given absolute risk threshold and the
Screening for CRC is currently offered to individuals in England between the ages 60 to 69 years through the NHS BCSP. Individuals aged below 60 are ineligible but those aged 70 and over can request screening. The population median 10-year absolute risk of CRC for 60 year old males and females in England is 1.96% and 1.19% respectively. We used these risk values to compare the current age-based screening programme with a personalised approach in individuals aged 55-69 years.
RESULTS
Polygenic risk score The 37 currently identified susceptibility variants for CRC confer a polygenic variance of 0.21 and account for around 14.5% of the familial risk of CRC (Supplementary Table S1). Figure 1 shows the PRS based on the 37 known risk variants for CRC. Individuals within the top 10% of genetic risk have a 1.8-fold increased risk of CRC and those within the top 1% (i.e. 26 – 53 risk alleles) have a 2.9-fold increased risk of CRC as compared to the population median. The area under the corresponding ROC curve for all of the 37 SNPs is 0.63 (Supplementary Figure S1).
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
proportion of CRC diagnosed in this high-risk group.
8
Performance of personalised screening incorporating PRS In England during the period 2001 to 2012 inclusive, on average 7,375,504 of the population was aged 55-69 years and within this age group 5,876 men and 3,743 women were diagnosed with CRC. Figure 2 shows the age-conditional absolute risks of CRC over 10 years for men and women. Table 1 partitions a simulated population of 100,000 individuals aged 55-69 according to eligibility for age-based and personalised screening. Within this cohort 61% of men (61,202/100,000) and
(76/99) of CRC cases being diagnosed in this subset of the population (Table 1). Under the personalised screening model based on the known risk variants, 45% of men (45,344/100,000) and 45% of women (45,467/100,000) would be eligible for screening with 69% of male cases (113/163) and 69% of female cases (68/99) being identified. In absolute terms this would translate to, 16% fewer men and 17% fewer women being eligible for screening at the cost of detecting 10% and 8% fewer cases respectively.
Impact of identifying all common risk variants on personalised screening Using GCTA, we estimated the heritability of CRC for common variation to be 19% (± 8%), which translates to common variation accounting for 55% of the familial risk of CRC. On the basis of this heritability estimate, Figure 1 shows the PRS model if all possible susceptibility variants for CRC were known. Individuals within the top 10% of genetic risk have a 3.1-fold increased risk of CRC and those within the top 1% have a 7.7-fold increased risk of CRC as compared to the population median. The increased individual risk discrimination is reflected in the ROC with the area under the curve being 0.73 (Supplementary Figure S1). Table 1 shows the performance of personalised screening incorporating PRS where it is assumed all susceptibility variants for CRC are known. Under this model 35% of men and 36% of women would be eligible for screening with 72% of
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
62% of women (62,091/100,000) are eligible for age-based screening with 79% (128/163) and 77%
9
cases being identified. Therefore, 26% fewer men and 26% fewer women would be eligible for screening at the cost of detecting 7% and 5% fewer cases.
Changing entry age and its impact on the performance of personalised screening The Scottish bowel cancer screening program is open to individuals aged 50-74 and there is a move within England to extend the upper age of screening to age 75. In England from 2001-2012, the average population aged 50-74 was 12,684,623, with 9,442 men and 6,216 women diagnosed
then compared the effect of the higher NHS entry age (Scenario 1) and offering screening to higher-risk 50-59 year-olds in addition to those aged 60-69 (Scenario 2). Here a higher risk 50-59 year old is one whose 10-year absolute risk of developing CRC is equal to or greater than the average for an individual aged 60 (1.96% in men and 1.19% in women). Under Scenario 1 the effect of raising the screening entry age from 50 to 60 years results in only 50% of the male population being eligible for screening but 79% of the cases remain potentially detectable (Supplementary Table S2). Under Scenario 2, based on the known risk SNPs, 56% of the population would be eligible and 85% of cases potentially detectable; if all risk SNPs were known, then the respective metrics would be 58% and 89%.
DISCUSSION
Our analysis demonstrates that personalised screening programmes for CRC, in which eligibility is based on genetic risk profile in addition to age, have the potential to greatly reduce the number of individuals screened whilst still detecting nearly as many cases. Moreover, the additional detected cases will include early detection in younger genetically-susceptible individuals, and this will
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
with CRC annually. We simulated populations of 100,000 50-74 year-old males and females and
10
produce a substantial gain in life years. This is not only due to these individuals’ younger age, but also because it appears genetic susceptibility can be associated with tumour aggressiveness.
Currently the NHS BCSP is based on the use of FOBT and participation using this screening tool is only around 60% [14]. There is however a move to replace FOBT with the Faecal Immunochemical Test (FIT) which is acknowledged to be a better initial test and is generally considered to be more acceptable [15]. Implementation of FIT-based screening in conjunction with PRS-based risk
influence detection rates.
Although family history information is used extensively to estimate the risk of CRC in family cancer clinics there is considerable potential for recall bias and inaccuracy in the reporting of CRC [16, 17]. Indeed studies have shown that CRC in relatives is often substantially under-reported [16]. Profiling for CRC susceptibility by genetic testing directly addresses this shortcoming.
It is possible that those identified to be at higher CRC risk by virtue of their genetic profile might be more likely to participate in screening, thereby further enhancing the performance of the programme. There is already well-established screening activity managed by specialist genetic and oncology centres for those individuals aged below 60 years with a family history of CRC [18-20]. However, creating such genetic screening programs does not necessarily mean that they will be utilised by health care providers and their clients. Indeed it is acknowledged that the efficacy of genetic risk profiling has been shown to be limited by the decisions and circumstances of those offered germline testing [21]. Hence empiric data is required to determine the performance of any personalised screening applied to the population.
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
profiling would then translate to both better uptake and better detection which will both
11
We acknowledge that there are limitations to our analysis. Specifically, our modelling has been based on estimating the proportion of the population eligible for screening and the proportion of CRC cases screen-detectable in this subgroup according to their genetic risk. We have assumed that all CRC cases are potentially screen-detectable, and so have not considered screening sensitivity.
The sensitivity of any bowel screening programme will increase with a reduction in the inter-
the duration of a screen-detectable phase impacts on programme performance. This is of special relevance to CRC as a major component of genetic susceptibility is mediated through propensity to develop pre-malignant polyps and an increased rate of transitioning of polyps to overt cancer. Hence it could be anticipated that genetic profiling will improve sensitivity a priori, since polypectomy will have a preventative effect in the longer term. Counter to this is that screening programs such as NHS-BCSP based on FOBT or flexible-sigmoidoscopy tend to miss more rightsided tumours, which have a greater tendency to be genetic, than those within the left side of the bowel [22]. Additionally, assuming the caveat of equivalent or improved screening programme sensitivity, detailed economic evaluation is required to determine if the costs associated with implementing personalised screening would be offset by savings from minimising follow up for individuals with false positive screens. Such economic evaluation will likely require empiric data from pilot studies to be truly meaningful.
Here we have based our analysis on National Cancer Registration Data from England to estimate the numbers of the population aged 55-69 years that are eligible for screening and the number of CRC cases potentially detectable in the eligible population stipulating different risk thresholds. The
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
screening interval and in the duration of the pre-clinical screen-detectable phase. Furthermore,
12
optimum risk threshold for the population of England will be different to other populations, such as those in Asia which have much lower incidences of CRC.
While personalised screening is an attractive concept there are a number of significant issues which would have to be addressed before it could be implemented. First and foremost, there will be the contentious issue of moving from a screening program of universal inclusion (based on age), to one in which screening is only offered to some individuals. Indeed it is notable that the
NHS BCSP or Scottish screening programmes. As with any common cancer, tumours will develop in significant numbers in the swathe of the population deemed to be at ‘low risk’ and for whom screening was deemed not to be required. Indeed the performance of personalised screening for CRC appears on the basis of our analysis to be less profound than that offered by adopting such a programme for prostate cancer [7]. Secondly, genetic risk profiling will inevitably add complexity to any screening programme. Finally, as with all forms of genetic testing there are ethical and legal challenges that would have to be addressed prior to any such programme being implemented [23].
In summary, personalised screening making use of a PRS has the potential to optimise the efficiency of screening programmes for CRC. Clearly as demonstrated the identification of additional risk SNPs and incorporation of Mendelian susceptibility alleles will afford better discrimination. In addition to PRS, there is clearly scope for individualised screening based on incorporating family history and non-genetic risk factors such as lifestyle risk factors as well as biomarker data. Real world ‘road testing’ is however needed to establish true efficiency and to address the technical and operational issues. Eligibility for screening using age as the entry criteria is generally accepted by health care professionals and the public alike, hence it remains to be
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
different profile of CRC risk in men and women has not been taken into account in delivery of the
13
established whether eligibility based on age and other factors would be acceptable. Since personalised screening is likely to afford the benefit of detecting CRC in younger subjects at high risk, its use to extend screening to such individuals would be more logistically attractive and acceptable to the public. Furthermore, the observed level of risk discrimination from PRS is likely to be highly informative in formulating and delivering chemoprevention strategies for CRC. Currently such initiatives have been confined to Mendelian predisposition to CRC with recent reports showing the value of aspirin in carriers of MMR gene mutations [24]. However, given the
identify a relatively high number of young at risk individuals there is a rationale for investing in PRS defined chemoprevention trials.
DISCLOSURE The authors have declared no conflicts of interest.
FUNDING Work was funded by Cancer Research UK, the European Union (258236, FP7 collaborative project SYSCOL). KL is in receipt of an ICR studentship.
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
level of risk discrimination afforded by even the currently known CRC SNPs and the opportunity to
14
REFERENCES
1.
(2013) CRU.) Cancer Statistics—Bowel Cancer. Available: http://www.cancerresearchuk.org/cancer-
info/cancerstats/types/bowel/. : http://www.cancerresearchuk.org/cancer-info/cancerstats/types/bowel/
2.
Network NCI. Colorectal cancer survival by stage. . Avilable:
http://wwwncinorguk/publications/data_briefings/colorectal_cancer_survival_by_stage. 2014. 3.
Programme NBCS. http://www.cancerscreening.nhs.uk/bowel/index.html.
4.
Programme BSSBS. http://www.bowelscreening.scot.nhs.uk/.
5.
Aaltonen L, Johns L, Jarvinen H et al. Explaining the familial colorectal cancer risk associated with
mismatch repair (MMR)-deficient and MMR-stable tumors. Clin Cancer Res 2007; 13: 356-361. 6.
Fletcher O, Houlston RS. Architecture of inherited susceptibility to common cancer. Nat Rev Cancer
2010; 10: 353-361. 7.
Pashayan N, Duffy SW, Chowdhury S et al. Polygenic susceptibility to prostate and breast cancer:
implications for personalised screening. Br J Cancer 2011; 104: 1656-1663. 8.
Garcia-Closas M, Gunsoy NB, Chatterjee N. Combined associations of genetic and environmental
risk factors: implications for prevention of breast cancer. J Natl Cancer Inst 2014; 106. 9.
Mavaddat N, Pharoah PD, Michailidou K et al. Prediction of breast cancer risk based on profiling
with common genetic variants. J Natl Cancer Inst 2015; 107. 10.
Fay MP, Pfeiffer R, Cronin KA et al. Age-conditional probabilities of developing cancer. Stat Med
2003; 22: 1837-1848. 11.
Yang J, Lee SH, Goddard ME, Visscher PM. GCTA: a tool for genome-wide complex trait analysis. Am
J Hum Genet 2011; 88: 76-82.
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
2013.
15 12.
Dunlop MG, Dobbins SE, Farrington SM et al. Common variation near CDKN1A, POLD3 and
SHROOM2 influences colorectal cancer risk. Nat Genet 2012; 44: 770-776. 13.
Lee SH, Harold D, Nyholt DR et al. Estimation and partitioning of polygenic variation captured by
common SNPs for Alzheimer's disease, multiple sclerosis and endometriosis. Hum Mol Genet 2013; 22: 832841. 14.
Moss SM, Campbell C, Melia J et al. Performance measures in three rounds of the English bowel
cancer screening pilot. Gut 2012; 61: 101-107. 15.
Young GP, Symonds EL, Allison JE et al. Advances in Fecal Occult Blood Tests: the FIT revolution. Dig
16.
Mitchell RJ, Brewster D, Campbell H et al. Accuracy of reporting of family history of colorectal
cancer. Gut 2004; 53: 291-295. 17.
Mai PL, Garceau AO, Graubard BI et al. Confirmation of family cancer history reported in a
population-based survey. J Natl Cancer Inst 2011; 103: 788-797. 18.
Houlston RS, Murday V, Harocopos C et al. Screening and genetic counselling for relatives of
patients with colorectal cancer in a family cancer clinic. BMJ 1990; 301: 366-368. 19.
Carney PA, O'Malley JP, Gough A et al. Association between documented family history of cancer
and screening for breast and colorectal cancer. Prev Med 2013; 57: 679-684. 20.
Cairns SR, Scholefield JH, Steele RJ et al. Guidelines for colorectal cancer screening and surveillance
in moderate and high risk groups (update from 2002). Gut 2010; 59: 666-689. 21.
Ward RL, Hicks S, Hawkins NJ. Population-based molecular screening for Lynch syndrome:
implications for personalized medicine. J Clin Oncol 2013; 31: 2554-2562. 22.
Johns LE, Houlston RS. A systematic review and meta-analysis of familial colorectal cancer risk. Am J
Gastroenterol 2001; 96: 2992-3003. 23.
Hall AE, Chowdhury S, Hallowell N et al. Implementing risk-stratified screening for common
cancers: a review of potential ethical, legal and social issues. J Public Health (Oxf) 2014; 36: 285-291.
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
Dis Sci 2015; 60: 609-622.
16 24.
Burn J, Gerdes AM, Macrae F et al. Long-term effect of aspirin on cancer risk in carriers of
hereditary colorectal cancer: an analysis from the CAPP2 randomised controlled trial. Lancet 2011; 378: 2081-2087.
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
17
TABLE AND FIGURE LEGENDS
Table 1: Screening eligibility results for a simulated population of 100,000 men and 100,000 women aged 55-69. Individuals are classified as eligible for screening under three models: (a) age-
on known risk SNPs, where any individual aged 55-69 is eligible if their 10-year absolute risk of CRC (as calculated from age and PRS) is greater than or equal to the average for an individual aged 60 (men = 1.96%, women = 1.19%); (c) as for (b) but using PRS based on all risk SNPs.
Figure 1: Population distribution of PRS ordered by RR (compared to population median risk). (a) based on the known 37 risk SNPs; (b) based on all common variants. Vertical red lines (left to right) correspond to 1%, 10%, 50%, 90%, 99% centile respectively.
Figure 2: Ten-year absolute risk of being diagnosed with CRC in men and women in England (2001-2012).
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
based only, where an individual is eligible if ≥60 years; (b) personalised screening using PRS based
18
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
19
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
Table 1: Screening eligibility results for a simulated population of 100,000 men and 100,000 women aged 55-69. Individuals are classified as eligible for screening under three models: (a) age-based only, where an individual is eligible if ≥60 years; (b) personalised screening using PRS based on known risk SNPs, where any individual aged 55-69 is eligible if their 10-year absolute risk of CRC (as calculated from age and PRS) is greater than or equal to the average for an individual aged 60 (men = 1.96%, women = 1.19%); (c) as for (b) but using PRS based on all risk SNPs.
(a) Age-based screening < 60 years
≥ 60 years
Total
30,030 8,768 38,798
24,626 36,576 61,202
54,656 45,344 100,000
21 14 35
29 99 128
50 113 163
28,899 9,010 37,909
25,634 36,457 62,091
54,533 45,467 100,000
14 9 23
18 58 76
32 68 99
29,657 9,141 38,798
34,883 26,318 61,202
64,541 35,459 100,000
15 20 35
30 98 128
45 118 163
28,717 9,192 37,909
35,751 26,340 62,091
64,468 35,532 100,000
10 13 23
18 58 76
28 71 99
Downloaded from http://annonc.oxfordjournals.org/ at Monash University on November 27, 2015
(b) Personalised screening using known risk SNPs Male population < 1.96% ≥ 1.96% Total Cases < 1.96% ≥ 1.96% Total Female population < 1.19% ≥ 1.19% Total Cases < 1.19% ≥ 1.19% Total (c) Personalised screening using all risk SNPs Male population < 1.96% ≥ 1.96% Total Cases < 1.96% ≥ 1.96% Total Female population < 1.19% ≥ 1.19% Total Cases < 1.19% ≥ 1.19% Total