Noninvasive multicancer liquid biopsy tests are rapidly emerging for early detection, but implementation will require risk stratification to enhance risk-benefit balance. Using data from the UK Biobank study, we trained and validated a model that allows estimation of absolute risk of developing at least one of the eight common cancers among women. We found that pan-cancer risk scores (PCRSs), defined by the combination of body mass index, smoking, family history of cancer, and cancer-specific polygenic risk scores, were strongly associated with the risk of developing at least one cancer (hazard ratio = 1.40 per sd unit, 95% CI: 1.35 - 1.46, 5-year AUC = 0.61). Projections using recently reported diagnostic accuracy for a multicancer screening test show that positive predictive values of it is expected to vary widely across strata defined by age and PCRS due to differences in underlying absolute risks. Our study suggests that risk-stratification, based on age, emerging polygenic risk scores, and other important cancer risk factors, have significant potential to increase risk-benefit balance of early detection through multicancer liquid biopsy tests.
Brief Communication
Promise of Risk Stratification for Multicancer Screening with Liquid Biopsy Tests
https://doi.org/10.21203/rs.3.rs-400074/v1
This work is licensed under a CC BY 4.0 License
Version 1
posted
You are reading this latest preprint version
Developing effective screening tools for early cancer detection has long been a pressing interest for the oncology community due to the poor prognosis and survival associated with advancing cancer stage1. Identifying individuals at the subclinical or asymptomatic stage provides the opportunity for early intervention that has shown to improve survival2. In a general population with relatively low prevalence of cancer, a clinically useful screening test should be broadly accessible as well as specific and sensitive. A screening test should be specific to minimize overdiagnosis-related psychological and financial burden and risks associated with unnecessary follow-up treatments and sensitive to prevent missed or interval cases. For these reasons, to date, only a handful of age-based single-cancer screening modalities such as colonoscopy for colorectal cancer3, mammogram for breast cancer4, low-dose computerized tomography (LDCT) scan for lung cancer5, and pap test for cervical cancer6 are supported by United States Preventative Taskforce. The current single-cancer screening tools, albeit effective, are not without limitations. Issues including lack of adherence to screening recommendations7–9, low positive predictive value (PPV) or high false-positives10, and missed or interval cancer cases11 are of concern. Additionally, for many of the cancers with poor prognosis or high late-stage diagnosis rate, there are no presently accepted screening tools used for cancer detection in asymptomatic individuals.
In this context, blood-based multicancer liquid biopsy tests using analytes such as cell-free DNA (cfDNA) for population-based early cancer screening is gaining traction12–21. Liquid biopsy tests are considered minimally invasive and easily accessible, with perceivably better patient compliance12,13. Several recent studies have started to explore the feasibility of such approaches for early multicancer detection in a limited clinical setting14–21. Of note is a recent 12-month prospective interventional cohort study involving 10,006 females aged 65-75 years old without a personal history of cancer20. DETECT-A (Detecting cancers Earlier Through Elective Mutation-based blood Collection and Testing), their multicancer liquid biopsy test, had an overall sensitivity of 27.1% at 98.9% specificity with a corresponding PPV of 19.4% and negative predictive value (NPV) of 99.3%20. Although the future of the multicancer liquid biopsy tests is promising, such approaches are unlikely to be implemented for the general population with fairly low risk of cancer due to costs and limited performance of current methods. To improve the risk-benefit balance, one can employ a risk-modeling approach and target individuals for screening who exceed certain risk thresholds22. In the multicancer scenario, however, population risk stratification becomes more challenging as one needs to consider risk factors across a number of common cancers.
In this article, we use data from the prospective UK Biobank (UKBB) cohort study and the US population cancer incidence rate to develop a multicancer model that allows estimation of future risk for developing at least one of the eight most common cancers in females. First, from the from the 2013-2017 United States Cancer Statistics (USCS) database, we identified the top ten incident cancer types for females23. We retained eight (breast, colorectum, endometrium, kidney, lung, melanoma, non-Hodgkin's lymphoma, ovary) cancer types with sufficiently large and publicly available genome-wide association studies (GWAS) for developing polygenic risk scores (PRSs). Our analysis involved 160,586 UKBB unrelated female participants of White British ancestry aged 40-71 with 6,886 incident cancer cases over the course of follow-up. Data were split into 2/3 training and 1/3 test set. We used the training sample to fit a Cox proportional hazard model with the outcome as the first incidence of any cancer.
The model specified a baseline hazard as a function of age and assumed multiplicative effects of the risk factors24. The set of risk factors include two major lifestyle related exposures, namely smoking (status and pack years of smoking) and body mass index (BMI), that are known to influence risk across multiple cancers; family history of the most common cancers (breast, colorectum, lung, and prostate) in first-degree relatives and the PRSs for each of the individual cancers. We further calibrated the estimate of the age-specific baseline hazard function so that overall incidence rates match those reported for the US population in the NCI-SEER cancer registry25. We then computed pan-cancer risk scores (PCRSs) as the weighted sum of the predictors included in the model. Performance of the PCRSs were evaluated using standardized hazard ratio for instantaneous risk and area under the curve (AUC) using up to 5-years follow-up data.
We assumed that the probability of an individual carrying an asymptomatic but screen detectable cancer is proportional to the risk of incident cancer over a small time interval. We observe that use of a window of 11-month for calculation of underlying absolute risk led to a projected PPV for the DETECT-A test in the age group 65-75 to match with what has been reported for this group from a recent prospective study20. Thus, we calculate 11-month absolute risk of incident cancer across PCRS percentiles and age groups and use them to approximate the underlying probabilities of screen detectable asymptomatic cancers. Then using Bayes theorem, and the reported sensitivity (27.1% overall and 23.5% for breast, colorectal, and lung cancer) and specificity (98.9%) of the DETECT-A test, we compute the expected PPVs and NPVs across the PCRS percentiles and different ages. We also projected PPVs and NPVs for the test for three individual cancer types (postmenopausal breast, colorectum, and lung) based on corresponding cancer-specific risk models.
PCRSs were strongly associated with risk of developing at least one cancer during the follow-up of the UKBB study in females (HR: 1.40 per 1 SD, 95% CI: 1.35 - 1.46) (Supplementary Table S4). AUC associated with the PCRS at 5-years follow-up was 0.61 (Supplementary Table S4). Fig. 1 illustrates the estimated 11-month risk of developing at least one of the eight cancers for females by age and PCRS strata. Overall, we observe a strong degree of stratification by the combined effects of age and PCRS. The range of risk between a 50-year-old person who is at the lowest centile of PCRS and an 80-year-old who is at the highest centile of PCRS was 0.14%-3.25% for females. The level of stratification of absolute risk due to PCRS was stronger with increasing age and vice versa due to multiplicative nature of their joint effects26. The projected PPVs for the liquid biopsy test also varied substantially according to the level of underlying risk of the population strata. For example, 75-year-old females in the 90th-95th PCRS percentile (AR: 1.91%) will have a 2.7-fold risk compared to the same-aged female in the 5th-10th PCRS percentile (AR: 0.72%) (Fig. 1a). This corresponds to a PPV value of 32.4% and 15.2%, respectively, translating to a 17.2% (95% CI: 16.7%-17.5%) PPV difference (Fig. 1b). NPV across all risk percentiles was reasonably high (min NPV: 97.6%) across all strata (Fig. 1c).
If a fixed threshold for PPV is used to recommend multicancer screening, then the eligibility will strongly vary by both age and PCRS. For example, at a threshold for 25% PPV, females could be eligible for the test as early as age 58 if they were at the highest centile of PCRS. On the other hand, women who were at or below the 40th and 20th percentiles of respective PCRS distributions, will not achieve the required PPV even at the oldest age groups. If a higher PPV threshold is used, such as 40%, we observe that only a small number of risk groups will be eligible for screening. No women who are younger than 71 year or have PCRS value lower than 99th percentile of PCRS will meet the 40% PPV threshold.
To further elucidate how multicancer liquid biopsy tests could be useful for persons with a high risk of a specific type of cancer, we built cancer-specific prediction models for breast, colorectal, and lung cancers including additional cancer specific risk factors (see Supplementary Figure S3-S5). As expected, individual cancer models have higher discriminatory ability compared to the multicancer model (Supplementary Table S4). However, as the absolute risk of any individual cancer is lower compare to that for any cancer, the level of the PPV values for individual cancers are correspondingly lower (Fig. 2). Nevertheless, we do observe strong stratification of absolute risk and PPV by age and underlying risk score for each individual cancer. The range of PPV between the lowest and highest risk groups were 0.9%-27.6%, 0.3-11.9% and 0.2%-52.3% for cancers of breast, colon, and lung in females. Very strong stratification for PPV is seen for lung cancer due to the underlying stratification of lung cancer by smoking.
In summary, we conducted a first of a kind study to highlight the potential for a risk-based approach in early multicancer screening. Specifically, we demonstrated that for females, the combination of age and PCRS can provide substantial stratification of absolute risk of developing at least one of several common cancers within short time intervals. As a result, one would expect the PPV for multicancer screening tests, such as that of DETECT-A, to vary substantially across different risk categories. In particular, we observe that while age is an important determinant of PPV due to higher incidence of cancer cases with increasing age, the other risk factors are also expected to play an important synergistic role. For example, we estimated that the average PPV for all cancers for a test like DETECT-A to be about 13.2% and 19.7% for a 60- and 70-year-old female, respectively. The range of PPV between the 1st-99th percentile of PCRS for these two age groups were 5.7%-28.2% and 10.7%-44.1%, indicating that the PPV yield of the test can be lower (or higher) for the older (or younger) age group for a substantial fraction of the population due to risk associated with PRSs, BMI, smoking, and family history of cancer. As the risk-benefit balance for population screening depends heavily on underlying PPV (and NPV)27,28, our analysis indicates that multicancer risk models are likely to play an important role in the implementation of population multicancer liquid biopsy tests.
Our analysis has several limitations. We assumed that the reported sensitivity and specificity of a test like DETECT-A would be applicable across all age, sex and risk groups. Given the increase in observed sequence alterations in cfDNA resulting from clonal hematopoiesis in older individuals29–32, improvements in cfDNA analyses will be needed to overcome these challenges, potentially through use of mutation- agnostic methods15,16,32. Additional empirical data are needed to explore potential heterogeneity of the diagnostic accuracy of liquid biopsy tests by age and other risk factors. Second, in our multicancer risk prediction models, we included a limited set of risk-factors, namely two major life-style related factors that influence the risk of multiple cancers, family history of the most common cancers, and PRSs for each individual cancer types. We also assumed a proportional hazard model for all cancer risk, assuming multiplicative effects of age and all the other risk factors. Additional efforts are needed to build and validate more refined multicancer risk models in prospective cohort studies by including additional risk factors and interaction effects. Further, models33,34 that incorporate extensive family history information and carrier status for rare high-penetrant mutations would be important to consider for individuals in high-risk families that show strong clustering of related cancers. Third, our model building effort was restricted to women of White British ancestry in the UK Biobank study, due to not only limited sample size of other groups in the UK Biobank itself, but also lack of large-scale GWAS for other groups required to build PRS that are as predictive as those for European origin populations. Large studies of diverse populations are urgently needed, both to study the accuracy of multicancer liquid biopsy tests and to build and validate robust multicancer risk prediction models across different racial and ethnic groups. Finally, we have focused on exploring the potential utility of a multicancer risk model to determine initial screening eligibility for liquid biopsy test. A secondary application of the individual cancer specific risk models could be to improve identification of target sites to select follow-up procedures among individuals who test positive in the initial multicancer screening. Additional studies are merited to explore these opportunities.
In conclusion, our study shows the promise for the potential of risk-based approach to increase the risk-benefit balance of multicancer screening using liquid biopsy tests. In the future, well powered empirical studies are needed in diverse populations to prospectively evaluate the PPV and NPV of these tests, overall and stratified by risk groups.
Literature Search and List of Cancer Types
Based on the report from the 2013-2017 United States Cancer Statistics (USCS) database, we identified the top 10 incident cancer types for females, after excluding non-melanoma skin cancer23. First, we surveyed the NHGRI-EBI Catalog of Published Genome-Wide Association Studies (GWAS Catalog)35 and the Polygenic Risk Score (PGS) Catalog36 to select the largest European ancestry-based GWAS as of May 2020 for each cancer type. We additionally browsed PubMed37 for large cancer specific GWASs that were not included in the GWAS Catalog or PGS Catalog. For breast and colorectal cancer, we searched for prior European sample based large-scale polygenic risk score (PRS) studies as of July 2020 and selected for studies reporting the best performing PRS (Supplementary Table S2). We did not consider pleiotropic GWAS. We filtered to cancer types with at least ten independent genome-wide significant SNPs after LD clumping at a genome-wide significant (GWS) p-value, 5E-8, threshold. Ultimately, 8 cancer types (breast, colorectum, endometrium, kidney, lung, melanoma, Non-Hodgkin’s lymphoma, ovary) were included in our analysis. For the full list of source literature and GWAS summary statistics included in our analysis, see Supplementary Table S2.
UK Biobank Study Population
UK Biobank (UKBB) is a prospective epidemiological cohort study with over 500,000 participants38–40. Individuals aged 40-69 at baseline were recruited across the United Kingdom (UK) from 2006-201038–40. A wide range of genotypic and phenotypic information including personal medical and family history and lifestyle data were collected at enrollment38–40. UKBB data is regularly updated by completion of follow-up questionnaires, linkage to national cancer and mortality registries and hospital inpatient electronic medical records systems38–40. With linkage to the national cancer registry data, cancer diagnosis date, type (coded based on International Classification of Disease 9 (ICD-9) or International Classification of Disease 10 (ICD-10)) were available for participants diagnosed with cancer38–40. For our analysis, we used ICD-9 and ICD-10 codes for cancer classification (see Supplementary Table S3). The UK Biobank Ethics Advisroy Committee (EAC) provides oversight to the UK Biobank Study.
We then filtered to unrelated UKBB participants of White British ancestry with available imputed genotype data. We excluded individuals who were lost to follow-up, with genetic sex and self-reported sex mismatch, or those with any cancer diagnosis prior to baseline assessment (prevalent cancers). These quality control procedures resulted in a study population involving 160,586 females.
SNP Selection Criteria
After determining the source literatures (Supplementary Table S2) for each cancer type, we reviewed the manuscript and any relevant additional resources. We extracted all autosomal SNPs from each cancer GWAS along with their summary statistics such as RSIDs, observed effect size estimates (OR or beta), effective (or risk) allele, risk allele frequency (RAF), and p-value. We excluded variants with minor allele frequency (MAF) < 0.01 and ambiguous SNPs (A/T or G/C allele) with MAF > 0.40. We filtered to variants with MAF difference of less than 0.10 relative to the UK Biobank data. We removed variants with allele mismatches that could not be resolved by strand or dosage flips and/or SNPs with complete information mismatch, based on RSID, chromosome number, and position, to the European 1000 Genome reference panel41 or the UK Biobank data. We filtered to variants with information score 0.90 based on the UK Biobank imputed genotype data. Finally, using Plink42 we performed LD clumping at a p-value threshold of 5E-8, r2 of 0.1, and 1000 kb window with the European 1000 Genome reference panel41 as the reference panel to remove SNPs in linkage disequilibrium within each cancer type.
Polygenic Risk Score
Then, PRS for UK Biobank participants were computed using PRSice243.
Formula used for PRS calculation in PRSice2:
Statistical Analysis
We developed a pan-cancer risk prediction model to estimate the risk of developing at least one cancer over the course of follow-up. The multicancer model included 8 cancer types (breast, colorectum, endometrium, kidney, lung, melanoma, Non-Hodgkin’s lymphoma, and ovary) for females. Data was split into ⅔ training set and ⅓ of test set—independent validation dataset used for model performance evaluation and subsequent analysis.
Cox proportional hazard regression (Cox) model44 was fitted to the training set with the outcome as incidence to any first cancer included in the analysis. The models specified a baseline hazard as a function of age and assumed multiplicative effects of the risk factors44:
Polygenic risk scores for each cancer, family history of cancer (breast, colorectum, lung, and prostate) in any first-degree relatives (nonadopted), body mass index, and pack years of smoking were included as predictors in the model (see Supplementary Figure S2). We also adjusted for the first ten principal components. Also, as UKBB is a left-truncated and right censored cohort, we used age as timescale for the Cox model—that is, participants enter the model at recruitment age and exit at cancer incidence age, censoring age, or death age–whichever occurs first. We used the censoring date for the cancer registry data provided by UKBB45.
In addition, we developed cancer-specific risk prediction models for three cancer types: breast (postmenopausal), colorectum, and lung cancer. In the individual models, we included risk factors with convincing evidence for individual cancer risk based on literature review, drawing heavily from National Cancer Institute’s Cancer Prevention Overview46–48. List of risk factors included in each cancer model is shown in Supplementary Figure S3-S5. As complete data is required for all the predictors included in the analysis, missing risk factor values were imputed using non-parametric random forest based multiple imputation49.
We computed pan-cancer risk scores (PCRS) or cancer-specific risk scores for all UKBB participants as the weighted sum of the predictors with weights for each predictor as the estimated log hazard ratio (HRs) from the fitted Cox model. Then, in the test set, we assessed the discriminatory accuracy of the pan-cancer risk score (PCRS) or the cancer-specific risk score (for individual cancer models) using Harrel’s concordance index (C-statistic) and area under the curve (AUC) at 5 years follow-up.
Absolute Risk Estimation, Positive Predictive Value, and Negative Predictive Value Calculations
Absolute Risk Estimation using iCARE (Individualized Coherent Absolute Risk Estimation)50:
We used iCARE to build our absolute risk model. Detailed methodology for absolute risk model building is described in Choudhury et al 202050. Briefly, risk estimates for each individual in the test set was obtained by feeding age-specific cancer incidence rate by 1-year strata, log HR parameter from the Cox model, and the reference dataset into the model. We used 2016 cancer incidence rates in white individuals of the SEER*Stat database25. Cancer incidence rates for a given age and sex was determined by the following year’s cancer incidence rates. For instance, cancer incidence rates for females aged 50-51 in our study will correspond to SEER*Stat’s cancer incidence rates for females aged 51-52. This is to account for the fact that DETECT-A test was performed at study enrollment and the individuals were followed up over the course of 12 months. DETECT-A was used to detect cancers early, prior to conventional diagnosis. Reference dataset was obtained by simulating 10,000 samples representative of the underlying UKBB population using the normal distribution with PCRS or cancer-specific risk score mean and standard deviation.
Time window Selection:
DETECT-A study reported an empirical PPV value of 19.4% (95% CI: 13.1%-27.1%)20. We wanted to select a time window for absolute risk estimation so that the PPV for females aged 65-75 is equal to the point estimate of 19.4% reported in the DETECT-A study20. We varied the time window my 1 month around 1 year and calculated the weighted average PPV for females aged 65-75 based on the UKBB PCRS distribution and age distribution as reported by the US Census Bureau51. We found a time window of 11 months provides the best match for the overall PPV for the 65-75 group to the value of 19.4%. Thus, subsequently we calculated PPV and NPV for different age and PCRS risk groups based on underlying 11-month absolute risk.
Given the absolute risk estimate, x , the positive predictive value and negative predictive value of the multicancer liquid biopsy test can be calculated using the formula below:
The absolute risk estimate can be written as a function of age and the risk factors. We assumed that the sensitivity and specificity of the multicancer liquid biopsy test does not depend on the underlying risk factors and we used value of these as reported from the DETECT-A study. In the multicancer scenario, the reported sensitivity and specificity is 27.1% and 98.9%, respectively (Supplementary Table S6)20. For breast, colorectal, and lung cancer, the reported sensitivity was 23.5% and specificity was 98.9% (Supplementary Table S6)20.
ACKNOWLEDGMENTS
This research has been conducted using the UK Biobank Resource under Application Number 6 17712. This work was supported in part by was supported R01 grant from the National Human Genome Research Institute [1 R01 HG010480-01] and the National Cancer Institute [1 1U01CA249866-01], the Dr. Miriam and Sheldon G. Adelson Medical Research Foundation, the Gray Foundation, and the Commonwealth Foundation. R.B.S. is a founder and consultant of Delfi Diagnostics. We thank Dr. Bert Vogelstein (Ludwig Center at Johns Hopkins University School of Medicine) for comments on the manuscript.
AUTHOR CONTRIBUTIONS
S.K and N.C conceived the study design and drafted the manuscripts. S.K did all the data analyses under supervision of N.C. All other co-authors reviewed and approved the final version of the paper.
COMPETING INTERESTS STATEMENT
K.V currently receives research funding from Cepheid and collaborates with Optra Health. V.E.V. is a founder of Delfi Diagnostics and Personal Genome Diagnostics, serves on the Board of Directors and as a consultant for both organizations, and owns Delfi Diagnostics and Personal Genome Diagnostics stock, which are subject to certain restrictions under university policy. Additionally, Johns Hopkins University owns equity in Delfi Diagnostics and Personal Genome Diagnostics. V.E.V. is an advisor to Bristol-Myers Squibb and Takeda Pharmaceuticals. Within the last five years, V.E.V. has been an advisor to Genentech, Merck, Daiichi Sankyo, Janssen Diagnostics, and Ignyta. These arrangements have been reviewed and approved by the Johns Hopkins University in accordance with its conflict-of-interest policies.
DATA AVAILABILITY
Researcher can apply for data from the UK Biobank data through the UK Biobank Access Management System (AMS) https://www.ukbiobank.ac.uk/enable-your-research/register.
CODE AVAILABILITY
Code used in the study is available at https://github.com/eswk-im/PCRSanalysis-.git.
- Siegel, R. L., Miller, K. D. & Jemal, A. Cancer statistics, 2017. CA. Cancer J. Clin. 67, 7–30 (2017).
- Hiom, S. C. Diagnosing cancer earlier: reviewing the evidence for improving cancer survival. Br. J. Cancer 112 Suppl 1, S1-5 (2015).
- US Preventive Services Task Force et al. Screening for Colorectal Cancer: US Preventive Services Task Force Recommendation Statement. JAMA 315, 2564 (2016).
- Nelson, H. D. et al. Screening for Breast Cancer: A Systematic Review to Update the 2009 U.S. Preventive Services Task Force Recommendation. (Agency for Healthcare Research and Quality (US), 2016).
- Humphrey, L. et al. Screening for Lung Cancer: Systematic Review to Update the U.S. Preventive Services Task Force Recommendation. (Agency for Healthcare Research and Quality (US), 2013).
- Melnikow, J. et al. Screening for Cervical Cancer With High-Risk Human Papillomavirus Testing: Updated Evidence Report and Systematic Review for the US Preventive Services Task Force. JAMA 320, 687–705 (2018).
- Brown, M. L. et al. Challenges in meeting Healthy People 2020 objectives for cancer-related preventive services, National Health Interview Survey, 2008 and 2010. Prev. Chronic. Dis. 11, E29 (2014).
- Lopez-Olivo, M. A. et al. Patient Adherence to Screening for Lung Cancer in the US: A Systematic Review and Meta-analysis. JAMA Netw. Open 3, e2025102 (2020).
- Vijan, S. Adherence to Colorectal Cancer Screening: A Randomized Clinical Trial of Competing Strategies. Arch. Intern. Med. 172, 575 (2012).
- Nelson, H. D. et al. Harms of Breast Cancer Screening: Systematic Review to Update the 2009 U.S. Preventive Services Task Force Recommendation. Ann. Intern. Med. 164, 256–267 (2016).
- Irvin, V. L. et al. Comparison of Mortality Among Participants of Women’s Health Initiative Trials With Screening-Detected Breast Cancers vs Interval Breast Cancers. JAMA Netw. Open 3, e207227 (2020).
- Mattox, A. K. et al. Applications of liquid biopsies for cancer. Sci. Transl. Med. 11, (2019).
- Cescon, D. W., Bratman, S. V., Chan, S. M. & Siu, L. L. Circulating tumor DNA and liquid biopsy in oncology. Nat. Cancer 1, 276–290 (2020).
- Cohen, J. D. et al. Detection and localization of surgically resectable cancers with a multi-analyte blood test. Science 359, 926–930 (2018).
- Cristiano, S. et al. Genome-wide cell-free DNA fragmentation in patients with cancer. Nature 570, 385–389 (2019).
- Mouliere, F. et al. Enhanced detection of circulating tumor DNA by fragment size analysis. Sci. Transl. Med. 10, (2018).
- Liu, M. C. et al. Sensitive and specific multi-cancer detection and localization using methylation signatures in cell-free DNA. Ann. Oncol. 31, 745–759 (2020).
- Jiang, T., Ren, S. & Zhou, C. Multi-cancer blood testing combined with PET-CT: road for hope to screen for cancer and guide intervention. Signal Transduct. Target. Ther. 5, 95 (2020).
- Chen, X. et al. Non-invasive early detection of cancer four years before conventional diagnosis using a blood test. Nat. Commun. 11, 3475 (2020).
- Lennon, A. M. et al. Feasibility of blood testing combined with PET-CT to screen for cancer and guide intervention. Science 369, eabb9601 (2020).
- Phallen, J. et al. Direct detection of early-stage cancers using circulating tumor DNA. Sci. Transl. Med. 9, (2017).
- Usher-Smith, J., Emery, J., Hamilton, W., Griffin, S. J. & Walter, F. M. Risk prediction tools for cancer in primary care. Br. J. Cancer 113, 1645–1650 (2015).
- U.S. Cancer Statistics Working Group. U.S. Cancer Statistics Data Visualizations Tool, based on 2019 submission data (1999-2017): U.S. Department of Health and Human Services, Centers for Disease Control and Prevention and National Cancer Institute; www.cdc.gov/cancer/dataviz, released in June 2020.
- Cox, D. R. Regression Models and Life-Tables. J. R. Stat. Soc. Ser. B Methodol. 34, 187–202 (1972).
- Surveillance, Epidemiology, and End Results (SEER) Program (www.seer.cancer.gov) SEER*Stat Database: Incidence - SEER 21 Regs Limited-Field Research Data + Hurricane Katrina Impacted Louisiana Cases, Nov 2018 Sub (2000-2016) <Katrina/Rita Population Adjustment> - Linked To County Attributes - Total U.S., 1969-2017 Counties, National Cancer Institute, DCCPS, Surveillance Research Program, released April 2019, based on the November 2018 submission.
- Maas, P. et al. Breast Cancer Risk From Modifiable and Nonmodifiable Risk Factors Among White Women in the United States. JAMA Oncol. 2, 1295 (2016).
- Pashayan, N., Morris, S., Gilbert, F. J. & Pharoah, P. D. P. Cost-effectiveness and Benefit-to-Harm Ratio of Risk-Stratified Screening for Breast Cancer: A Life-Table Model. JAMA Oncol. 4, 1504 (2018).
- Vilaprinyo, E. et al. Cost-Effectiveness and Harm-Benefit Analyses of Risk-Based Screening Strategies for Breast Cancer. PLoS ONE 9, e86858 (2014).
- Xie, M. et al. Age-related mutations associated with clonal hematopoietic expansion and malignancies. Nat. Med. 20, 1472–1478 (2014).
- McKerrell, T. et al. Leukemia-associated somatic mutations drive distinct patterns of age-related clonal hemopoiesis. Cell Rep. 10, 1239–1245 (2015).
- Genovese, G. et al. Clonal hematopoiesis and blood-cancer risk inferred from blood DNA sequence. N. Engl. J. Med. 371, 2477–2487 (2014).
- Leal, A. et al. White blood cell and cell-free DNA analyses for detection of residual disease in gastric cancer. Nat. Commun. 11, 525 (2020).
- Chen, S. et al. Prediction of germline mutations and cancer risk in the Lynch syndrome. JAMA 296, 1479–1487 (2006).
- Lee, A. et al. BOADICEA: a comprehensive breast cancer risk prediction model incorporating genetic and nongenetic risk factors. Genet. Med. Off. J. Am. Coll. Med. Genet. 21, 1708–1718 (2019).
- Buniello, A. et al. The NHGRI-EBI GWAS Catalog of published genome-wide association studies, targeted arrays and summary statistics 2019. Nucleic Acids Res. 47, D1005–D1012 (2019).
- Lambert, S. A. et al. The Polygenic Score Catalog: an open database for reproducibility and systematic evaluation. http://medrxiv.org/lookup/doi/10.1101/2020.05.20.20108217 (2020) doi:10.1101/2020.05.20.20108217.
- PubMed. PubMed https://pubmed.ncbi.nlm.nih.gov/.
- Ollier, W., Sprosen, T. & Peakman, T. UK Biobank: from concept to reality. Pharmacogenomics 6, 639–646 (2005).
- Bycroft, C. et al. The UK Biobank resource with deep phenotyping and genomic data. Nature 562, 203–209 (2018).
- Sudlow, C. 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. 12, e1001779 (2015).
- 1000 Genomes Project Consortium et al. A global reference for human genetic variation. Nature 526, 68–74 (2015).
- Purcell, S. et al. PLINK: a tool set for whole-genome association and population-based linkage analyses. Am. J. Hum. Genet. 81, 559–575 (2007).
- Choi, S. W. & O’Reilly, P. F. PRSice-2: Polygenic Risk Score software for biobank-scale data. GigaScience 8, (2019).
- Cox, D. R. Regression Models and Life-Tables. J. R. Stat. Soc. Ser. B Methodol. 34, 187–202 (1972).
- UKB : External Info : Data_providers_and_dates. https://biobank.ndph.ox.ac.uk/ukb/exinfo.cgi?src=Data_providers_and_dates.
- PDQ Screening and Prevention Editorial Board. Breast Cancer Prevention (PDQ®): Health Professional Version. in PDQ Cancer Information Summaries (National Cancer Institute (US), 2002).
- PDQ Screening and Prevention Editorial Board. Colorectal Cancer Prevention (PDQ®): Health Professional Version. in PDQ Cancer Information Summaries (National Cancer Institute (US), 2002).
- PDQ Screening and Prevention Editorial Board. Lung Cancer Prevention (PDQ®): Health Professional Version. in PDQ Cancer Information Summaries (National Cancer Institute (US), 2002).
- Stekhoven, D. J. & Buhlmann, P. MissForest--non-parametric missing value imputation for mixed-type data. Bioinformatics 28, 112–118 (2012).
- Pal Choudhury, P. et al. iCARE: An R package to build, validate and apply absolute risk models. PLOS ONE 15, e0228198 (2020).
- Bureau, U. C. Age and Sex Composition in the United States: 2019. The United States Census Bureau https://www.census.gov/data/tables/2019/demo/age-and-sex/2019-age-sex-composition.html.
Yes there is potential Competing Interest.
K.V currently receives research funding from Cepheid and collaborates with Optra Health. V.E.V. is a founder of Delfi Diagnostics and Personal Genome Diagnostics, serves on the Board of Directors and as a consultant for both organizations, and owns Delfi Diagnostics and Personal Genome Diagnostics stock, which are subject to certain restrictions under university policy. Additionally, Johns Hopkins University owns equity in Delfi Diagnostics and Personal Genome Diagnostics. V.E.V. is an advisor to Bristol-Myers Squibb and Takeda Pharmaceuticals. Within the last five years, V.E.V. has been an advisor to Genentech, Merck, Daiichi Sankyo, Janssen Diagnostics, and Ignyta. These arrangements have been reviewed and approved by the Johns Hopkins University in accordance with its conflict-of-interest policies.
- SupplementaryInfoBC.xlsx
Supplementary Tables
- SupplementaryInfoBC.docx
Supplementary Figures and Figure/Table Legends
- flatChatterjeers.pdf
Reporting Summary
- flatChatterjeeepc.pdf
Editorial Policy Checklist