Polygenic risk scores for the prediction of common cancers in East Asians: A population-based prospective cohort study

  1. Peh Joo Ho
  2. Iain BeeHuat Tan
  3. Dawn Qingqing Chong
  4. Chiea Chuen Khor
  5. Jian-Min Yuan
  6. Woon-Puay Koh
  7. Rajkumar Dorajoo  Is a corresponding author
  8. Jingmei Li  Is a corresponding author
  1. Genome Institute of Singapore (GIS), Agency for Science, Technology and Research (A*STAR), Singapore
  2. Saw Swee Hock School of Public Health, National University of Singapore and National University Health System, Singapore
  3. Department of Surgery, Yong Loo Lin School of Medicine, National University of Singapore, Singapore
  4. Program in Cancer and Stem Cell Biology, Duke-National University of Singapore Medical School, Singapore
  5. Division of Medical Oncology, National Cancer Centre Singapore, Singapore
  6. Duke-NUS Medical School Singapore, Singapore
  7. UPMC Hillman Cancer Center, United States
  8. Department of Epidemiology, University of Pittsburgh Graduate School of Public Health, United States
  9. Healthy Longevity Translational Research Programme; Yong Loo Lin School of Medicine, National University of Singapore, Singapore
  10. Singapore Institute for Clinical Sciences, Agency for Science Technology and Research (A*STAR), Singapore

Abstract

Background:

To evaluate the utility of polygenic risk scores (PRSs) in identifying high-risk individuals, different publicly available PRSs for breast (n=85), prostate (n=37), colorectal (n=22), and lung cancers (n=11) were examined in a prospective study of 21,694 Chinese adults.

Methods:

We constructed PRS using weights curated in the online PGS Catalog. PRS performance was evaluated by distribution, discrimination, predictive ability, and calibration. Hazard ratios (HR) and corresponding confidence intervals (CI) of the common cancers after 20 years of follow-up were estimated using Cox proportional hazard models for different levels of PRS.

Results:

A total of 495 breast, 308 prostate, 332 female-colorectal, 409 male-colorectal, 181 female-lung, and 381 male-lung incident cancers were identified. The area under receiver operating characteristic curve for the best-performing site-specific PRS were 0.61 (PGS000873, breast), 0.70 (PGS00662, prostate), 0.65 (PGS000055, female-colorectal), 0.60 (PGS000734, male-colorectal), 0.56 (PGS000721, female-lung), and 0.58 (PGS000070, male-lung), respectively. Compared to the middle quintile, individuals in the highest cancer-specific PRS quintile were 64% more likely to develop cancers of the breast, prostate, and colorectal. For lung cancer, the lowest cancer-specific PRS quintile was associated with 28–34% decreased risk compared to the middle quintile. In contrast, the HR observed for quintiles 4 (female-lung: 0.95 [0.61–1.47]; male-lung: 1.14 [0.82–1.57]) and 5 (female-lung: 0.95 [0.61–1.47]) were not significantly different from that for the middle quintile.

Conclusions:

Site-specific PRSs can stratify the risk of developing breast, prostate, and colorectal cancers in this East Asian population. Appropriate correction factors may be required to improve calibration.

Funding:

This work is supported by the National Research Foundation Singapore (NRF-NRFF2017-02), PRECISION Health Research, Singapore (PRECISE) and the Agency for Science, Technology and Research (A*STAR). WP Koh was supported by National Medical Research Council, Singapore (NMRC/CSA/0055/2013). CC Khor was supported by National Research Foundation Singapore (NRF-NRFI2018-01). Rajkumar Dorajoo received a grant from the Agency for Science, Technology and Research Career Development Award (A*STAR CDA - 202D8090), and from Ministry of Health Healthy Longevity Catalyst Award (HLCA20Jan-0022).

The Singapore Chinese Health Study was supported by grants from the National Medical Research Council, Singapore (NMRC/CIRG/1456/2016) and the U.S. National Institutes of Health (NIH) (R01 CA144034 and UM1 CA182876).

Editor's evaluation

Utilizing one of the largest prospective Asian cohorts with long-term follow-up data, the important study reveals the utility of polygenic risk scores in identifying high-risk individuals from specific cancers. The convincing results reached by the high-quality data highlight the translational significance of this study, which may be important for future exploration in the field of cancer epidemiology.

https://doi.org/10.7554/eLife.82608.sa0

eLife digest

Although humans contain the same genes, the sequence within these DNA sites can vary from person to person. These small variations, also known as genetic variants, can increase the risk of developing certain diseases. While each variant will only have a weak effect, if multiple variations are present the odds of developing the disease becomes significantly higher.

To determine which variants are linked to a disease, researchers carry out genome-wide association studies which involve analyzing the genomes of individuals with and without the condition and comparing their genetic codes. This data is then used to calculate how different combinations of variants impact a person’s chance of getting the disease, also known as a polygenic risk score.

Currently, most genome-wide association studies only incorporate genetic data from people with European ancestry. Consequently, polygenic risk scores performed using this information may not accurately predict the risk of developing the disease for individuals with other ethnicities, such as people with Asian ancestry.

Here, Ho et al. evaluated how well previously calculated polygenic risk scores for the four most common cancers (breast, colorectal, prostate and lung) worked on individuals of East Asian descent. The scores were tested on a dataset containing the genetic sequence, medical history, diet and activity levels of over 21,000 people living in Singapore in the 1990s. Ho et al. found that the polygenic risk scores for breast, prostate and colorectal cancer were able to predict disease risk. However, the score for lung cancer did not perform as well. The polygenic risk score for breast cancer was the most accurate, and was able to stratify individuals into distinct risk bands at an earlier age than other scores.

These findings shed light on which existing polygenic risk scores will be effective at assessing cancer risk in individuals with East Asian ancestry. Indeed, Ho et al. have already incorporated the polygenic risk score for breast cancer into a pilot study screening individuals in a comparable population in Singapore. However, the polygenic risk scores tested still performed better on individuals with European ancestry, highlighting the need to address the lack of Asian representation in genome-wide association studies.

Introduction

Polygenic risk scores (PRSs) for a range of health traits and conditions have been developed in recent years. These scores, which are based on summary statistics from genome-wide association studies (GWAS), can be used to stratify people depending on their genetic risk of acquiring various diseases, to improve screening and preventative interventions, as well as patient care (Polygenic Risk Score Task Force of the International Common Disease Alliance, 2021; Lambert et al., 2019). Precision risk assessment may help develop tailored screening strategies targeting individuals at higher risk of disease of interest (Clift et al., 2022).

The contributions of heritable genetic factors are different for different cancers. Twin studies have highlighted statistically significant effects of heritable genetic risk factors for cancers of the prostate, colorectal, and breast (Lichtenstein et al., 2000). The amount of phenotypic variance explained by the common genetic variants found by GWAS is also known to vary (Cano-Gamez and Trynka, 2020), suggesting that PRS derived from GWAS findings may perform to varying degrees for different cancers.

The area under receiver operating characteristic curve (AUC) is an important discrimination index for evaluating the performance of PRS. The greater the AUC, the better the discriminatory ability to separate cases from non-cases. A value of 0.5 suggests that the tool is performing no better than chance, while a value of 1 is obtained when cases and non-cases are perfectly separated. The range of reported AUC associated with published PRS ranged from 0.584 to 0.678 for breast cancer (Mavaddat et al., 2019; Ho et al., 2020; Kachuri et al., 2020; Du et al., 2021; Jia et al., 2020; Lacaze et al., 2021; Zhang et al., 2018), 0.591–0.769 for prostate cancer (Kachuri et al., 2020; Jia et al., 2020; Fritsche et al., 2020), 0.609–0.708 for colorectal cancer (Kachuri et al., 2020; Jia et al., 2020; Gafni et al., 2021; Archambault et al., 2022), and 0.52–0.846 for lung cancer (Kachuri et al., 2020; Jia et al., 2020; Fritsche et al., 2020; Hung et al., 2021). In a study by Jia et al. looking at eight common cancers in the UK Biobank population-based cohort study (n=400,812 participants of European descent), the observed AUC ranged from 0.567 to 0.662 (Jia et al., 2020).

While prediction of individual cancer risks through PRS remains moderate, emerging data supports the use of PRS for population-based cancer risk stratification. In previous work, Ho et al. examined the overlap of women identified to be at high risk of developing breast cancer based on family history for the disease, a non-genetic breast cancer risk prediction model, a breast cancer PRS, and carriership of rare pathogenic variants in established breast cancer predisposition genes (Ho et al., 2022). The overlap of individuals found to be at elevated risk of developing breast cancer based on the genetic and non-genetic models was low. PRS was also found to be able to identify high-risk individuals among young women who were not yet eligible to attend mammography screening. The findings suggest that a genetic tool that is feasible to be deployed for population-based screening may complement current screening programs.

Disparities in the genetic risk of cancer among various ancestry populations are poorly understood. Ideally, selected genetic variants that make up PRS should be relevant to the population being screened. The development of training datasets of PRS are dominated by samples of European ancestry, resulting in ancestry bias and issues with transferability to other populations (Lambert et al., 2019; Fritsche et al., 2021). The mismatch between the ancestries of the GWAS samples and the target populations for PRS application is a limiting factor (Fritsche et al., 2021). In this study, we evaluated the utility of common PRS, curated in the Polygenic Score (PGS) Catalog, in predicting the risk of the commonly diagnosed cancers with high genetic predisposition (breast, prostate, colorectal, and lung) in a prospective cohort comprising 21,694 participants of East Asian descent in Singapore. The reporting framework recommended for the interpretation and evaluation of PRS detailed in Wand et al. is used (Wand et al., 2021).

Materials and methods

Singapore Chinese Health Study

The Singapore Chinese Health Study (SCHS) is a population-based prospective cohort study of ethnic Chinese men and women recruited between April 1993 and December 1998 (Hankin et al., 2001). Participants were 45–74 years of age at recruitment and were restricted to the two major dialect groups of Chinese adults in Singapore, who were the Hokkiens and the Cantonese that had originated from Fujian and Guangdong provinces in Southern China, respectively. All our study participants were residents of government housing flats, which were built to accommodate approximately 86% of the resident population in Singapore during the enrolment period. A total of 63,257 individuals (35,298 women and 27,959 men) provided written informed consent (Hankin et al., 2001). The study was approved by the Institutional Review Boards of the National University of Singapore, University of Pittsburgh, and the Agency for Science, Technology and Research (A*STAR, reference number 2022-042). Written, informed consent was obtained from all study participants.

Baseline

Request a detailed protocol

An in-person baseline interview was performed at recruitment to collect data on diet using a validated 165-item food frequency questionnaire, smoking, alcohol, physical activity, medical history, and menstrual and reproductive history from women.

Selection of common cancers

Request a detailed protocol

In Singapore, between 2015 and 2019, colorectal cancer, the most prevalent cancer in men, accounted for nearly 17% of cancer diagnoses, while breast cancer, the most common cancer in women, accounted for about three out of ten cancer diagnoses (National Registry of Diseases Office, 2021). During this time, cancers of the breast, prostate, colorectal, and lung accounted for approximately half of the total cancer diagnoses. These four most common cancers were selected for inclusion in this study. We further stratified the analysis by sex as differences in colorectal and lung cancer incidence by sex have been reported in Singapore (de Kok et al., 2008).

A unique National Registration Identity Card (NRIC) number for every Singaporean enables the compilation and linkage of data from national register data to the same individual (Emmanuel, 1993). Identification of incident cases of cancer was accomplished by record linkage of all surviving cohort participants with the database of the nationwide Singapore Cancer Registry (Emmanuel, 1993). The Singapore Cancer Register was founded in 1968. Prior to 2009, reporting of neoplasms by all medical practitioners and pathology laboratories to the registry is voluntary (Fung et al., 2016). The registry’s staff compares cancer patient hospital discharges and death certificates to registered cases for verification. Completeness of reporting in the 1970s is 96% and in the 1990s, it was close to 100% (Sim et al., 2006). Cancers that developed among SCHS participants were identified using International Classification of Diseases (ICD) codes ICD-O-3 (breast: C50, prostate: C61, colorectal: C18, C19, and C20, lung: C34).

Follow-up

Request a detailed protocol

Death date was obtained by record linkage with the database Birth and Death Registry of Singapore (Emmanuel, 1993). The data on migration in our cohort was collected during our subsequent follow-up interviews, and we were informed about the migration by the family members of cohort members who had migrated. To date, only 47 (<1%) of the entire cohort participants were known to be lost to follow up due to migration out of Singapore, suggesting that the ascertainment of cancer and death incidences among the cohort participants was virtually complete.

Genotyping and imputation

Request a detailed protocol

Between 1999 and 2004, a total of 28,346 subjects contributed blood samples. A total of 25,273 SCHS participants were genotyped between the years 2017–2018 with the Illumina Infinium Global Screening Array (GSA) v1.0 and v2.0 (Chang et al., 2021).

Details on the sample quality control (QC) processes are previously described (Chang et al., 2021). Briefly, samples with a call rate of 95% or below (n=176) or heterozygosity extremes (>3 standard deviation [SD], n=236) were removed. Identity-by-state measurements were performed by pairwise comparisons of samples to detect related samples (first and second degree). One sample from each identified pair with the lower call rate was eliminated from further analysis (n=2746). To identify any ethnic outliers, principal component analysis was used in conjunction with 1000 Genomes Project reference populations and within the SCHS samples, which resulted in the further removal of 287 samples. Of the 21,828 samples that passed genotyping QC, 134 participants who were diagnosed with cancer before recruitment or had missing cancer outcomes and were excluded from the study, resulting in a final analytical dataset of 21,694 (Supplementary file 1a).

Alleles for all SNPs were coded to the forward strand and mapped to hg19. SNP QC steps included the exclusion of sex-linked and mitochondrial variants, gross Hardy–Weinberg equilibrium (HWE) outliers (p<1 × 10–6), monomorphic SNPs, or those with a minor allele frequency (MAF) < 1.0%, and SNPs with low call rates (<95.0%). We imputed for additional autosomal SNPs using IMPUTE v2 (Marchini et al., 2007) and with a two reference panel imputation approach by including (1) the cosmopolitan 1000 Genomes reference panels (Phase 3, representing 2504 samples) and (2) an Asian panel comprising 4810 Singaporeans (2780 Chinese, 903 Malays, 1127 Indians) (Chang et al., 2021). SNPs with imputation quality score INFO < 0.8, MAF  < 1.0%, or HWE p < 1  ×  10−6, as well as non-biallelic SNPs were excluded.

Polygenic risk scores

Request a detailed protocol

Published PRSs were retrieved from the PGS Catalog, an open database of polygenic scores (retrieved on 26 February 2022) (Supplementary file 1b; Lambert et al., 2021). Of the 2166 PRSs available in the resource, 1706 PRSs comprising less than 100,000 predictors were downloaded. Only PRSs with odds ratios or log odds ratios as weights were included; we excluded PRSs which used odds ratio over expected risk, inverse-variance weighting, and unweighted. A total of 85, 37, 22, and 11 PRSs were available for breast, prostate, colorectal, and lung cancers, respectively. Supplementary file 1c shows the number of individual variants comprising each PRS and proportion of variants missing in the SCHS cohort. Individual PRSs were calculated using the allelic scoring (–score sum) functions with default parameters in PLINK (v1.90b5.2) (Chang et al., 2015). The formula used was,

PRS= β1x1+ β2x2++ βkxk++ β313x313

where xk is the dosage of risk allele (0–2) for SNP k, βk is the corresponding weight.

PRS distribution

Request a detailed protocol

Two-sided, two-sample t-tests with a type I error of 0.05 were used to examine whether there was a difference in the distribution of standardised PRS (subtraction of mean value followed by the division by the SD) between site-specific cancer cases and non-cancer controls.

PRS discrimination

Request a detailed protocol

Discrimination was quantified by the area under the receiver operating characteristic curve (AUC), using logistic regression models, and their corresponding 95% CI. An AUC of 0.9–1.0 is considered excellent, 0.8–0.9 very good, 0.7–0.8 good, 0.6–0.7 sufficient, and 0.5–0.6 insufficient (Šimundić, 2009). The site-specific PRS with the highest AUC (logistic regression models) was selected. To test the sensitivity of the PRS selection, we obtained a time-to-event metric for AUC at 5 years, we used AUC.cd() from the ‘survAUC’ package in R (Kamarudin et al., 2017).

Associations between PRS and risk of developing cancers

Request a detailed protocol

Subjects were classified into PRS percentile groups. Person-years of follow-up were calculated for each subject from the date of enrolment to the date of cancer diagnosis, death, or 31 December 2015 (the date of linkage with the Singapore Cancer Registry), whichever came first. Follow-up time was censored at 20 years after recruitment. The associations between cancer-specific PRS quintiles (where individuals ranked by PRS were categorised into quintiles, using the middle quintile [40–60%] as a reference to reflect the average risk of the population) and the incidence of site-specific cancers were investigated using Cox proportional hazards modelling to estimate hazard ratios (HR) and corresponding 95% confidence intervals (CI), using time since recruitment as the time scale, and adjusted for age at recruitment. Tests for trends were conducted using two-sided Wald tests with a type I error of 0.05. Assumptions for proportional hazards were checked using the cox.zph() function in the ‘survival’ package in R, where a formal score test is done to test if a time-dependent variable is required.

HR and corresponding 95% CI were also estimated for every SD increase in PRS. PRS is known to have ‘portability’ issues related to genetic ancestry and demographics (Martin et al., 2019; Mostafavi et al., 2020). Hence, we adjusted for variables in the models, including age at recruitment, dialect group (Hokkien or Cantonese), highest level of education (no formal education, primary school, or secondary or higher), body mass index (continuous, kg/m2), cigarette smoking (non-smoker, ex-smoker, current smoker), alcohol consumption (never, weekly, daily), moderate physical activity (none, 1–3 hr/week, ≥3 hr/week), vigorous work/strenuous physical activity at least once a week (no or yes), and familial history of cancer (no or yes).

PRS absolute risk association

Request a detailed protocol

The 5-year absolute risks of developing breast, prostate, colorectal, and lung cancers were computed for PRS groups of increasing five percentiles over the follow-up period. Incidence (between 2013 and 2017) and mortality (the year 2016) statistics in Singapore (reported in National Registry of Disease Office, 1968 and Singapore Statistics, 2023, respectively) were used for the absolute risk estimations. We estimated the cancer-specific 5-year absolute risk based on PRS an iterative method detailed by Mavaddat et al., 2015.

PRS calibration

Request a detailed protocol

Calibration was studied by comparing the expected proportion of cases in the 5 years after recruitment to the observed proportion of cases that occurred in that 5 years, within each decile of PRS. Linear regression of the 10 points (pairs of expected and observed proportion) was used to study the overall calibration. A curve close to the diagonal indicates that predicted cancer risks correspond well to observed proportions. A slope above 1 implies that the model underestimates the absolute risk. Conversely, a slope below 1 implies that the model overestimates the absolute risk. In addition, we used the Hosmer-Lemeshow test to check the goodness-of-fit.

Results

Characteristics of the study population

Table 1 shows the characteristics of the 21,694 participants who were cancer-free at recruitment. The median follow-up time for the cohort was 20 years (interquartile range [IQR]: 18–22). As of December 2015, 495 women developed breast cancer, 308 men developed prostate, 774 (332 women and 409 men) colorectal cancer, and 562 (181 women and 381) lung cancer. The median age at recruitment was 54 years (IQR: 49–61). The median age at diagnosis was 65 years (IQR: 59–70) for female breast cancers, 72 years (IQR: 67–77) for prostate cancers, 71 years (IQR: 65–76) for male colorectal cancers, 71 years (IQR: 64–78) for female colorectal cancers, 74 years (IQR: 68–78) for male lung cancers, and 74 years (IQR: 66–79) for female lung cancers.

Table 1
Demographics of our study population by gender and cancer site.

Demographics variables were collected using structured questionnaire at recruitment. Family history for lung cancer was not available. Information on cancer occurrence (number of cancer and age at cancer occurrence) was obtained through linkage with the Singapore Cancer Registry in December 2015. Follow-up time was calculated from age at recruitment. IQR: Interquartile range.

Entire cohortIndividuals who developed cancer
BreastProstateColorectalLung
AllFemaleMaleFemaleMaleFemaleMaleFemaleMale
n21,69412,0849610495308332409181381
Age at recruitment in years, median (IQR)54 (49–61)54 (48–60)55 (49–62)53 (48–59)59 (54–64)58 (52–64)59 (52–65)59 (55–64)60 (55–64)
Number of cancers developed
0 (did not develop cancer)19633 (90)11096 (92)8537 (89)
12013 (9)968 (8)1045 (11)476 (96)293 (95)317 (95)387 (95)175 (97)362 (95)
248 (0)20 (0)28 (0)19 (4)15 (5)15 (5)22 (5)6 (3)19 (5)
Age at diagnosis among individuals who develop cancer(s) (earliest age for those with multiple cancers) in years, median (IQR)70 (64–77)68 (62–76)72 (67–77)65 (59–70)72 (67–77)71 (64–78)71 (65–6)74 (66–79)74 (68–78)
Length of follow-up (longest follow-up for those with multiple cancers) in years, median (IQR)20 (18–22)20 (18–22)19 (17–21)11 (6–16)13 (9–17)13 (8–17)11 (7–16)14 (9–17)14 (10–17)
Dialect group (%)
Hokkien10663 (49)6132 (51)4531 (47)260 (53)153 (50)185 (56)164 (40)95 (52)162 (43)
Cantonese11031 (51)5952 (49)5079 (53)235 (47)155 (50)147 (44)245 (60)86 (48)219 (57)
Highest education (%)
No4629 (21)3878 (32)751 (8)128 (26)20 (6)123 (37)46 (11)85 (47)57 (15)
Primary level9760 (45)5082 (42)4678 (49)206 (42)146 (47)138 (42)232 (57)62 (34)228 (60)
Secondary or above7305 (34)3124 (26)4181 (44)161 (33)142 (46)71 (21)131 (32)34 (19)96 (25)
Body mass index in kg/m2, median (IQR)23 (21–25)23 (21–25)23 (21–25)23 (21–25)23 (21–25)23 (21–24)23 (21–25)23 (20–24)23 (20–24)
Smoking status (%)
Never15553 (72)11235 (93)4318 (45)472 (95)166 (54)296 (89)153 (37)129 (71)63 (17)
Ex-smoker2374 (11)261 (2)2113 (22)8 (2)66 (21)14 (4)108 (26)9 (5)74 (19)
Current smoker3767 (17)588 (5)3179 (33)15 (3)76 (25)22 (7)148 (36)43 (24)244 (64)
Number of cigarettes smoked (%)
Does not smoke15553 (72)11235 (93)4318 (45)472 (95)166 (54)296 (89)153 (37)129 (71)63 (17)
<122408 (11)581 (5)1827 (19)14 (3)54 (18)26 (8)85 (21)36 (20)81 (21)
13–222344 (11)206 (2)2138 (22)6 (1)53 (17)9 (3)108 (26)15 (8)135 (35)
≥231389 (6)62 (1)1327 (14)3 (1)35 (11)1 (0)63 (15)1 (1)102 (27)
Alcohol consumption (%)
Never/ occasionally19079 (88)11506 (95)7573 (79)470 (95)253 (82)315 (95)303 (74)174 (96)296 (78)
Weekly1885 (9)437 (4)1448 (15)20 (4)44 (14)10 (3)66 (16)5 (3)49 (13)
Daily730 (3)141 (1)589 (6)5 (1)11 (4)7 (2)40 (10)2 (1)36 (9)
Moderate physical activity (%)
No16584 (76)9446 (78)7138 (74)380 (77)208 (68)269 (81)295 (72)143 (79)294 (77)
1–3 hr/week3274 (15)1679 (14)1595 (17)69 (14)62 (20)43 (13)68 (17)23 (13)53 (14)
≥ 3 hr/week1836 (8)959 (8)877 (9)46 (9)38 (12)20 (6)46 (11)15 (8)34 (9)
Vigorous physical activity/ strenuous sports at least once a week (%)
No18467 (85)11221 (93)7246 (75)452 (91)239 (78)311 (94)342 (84)175 (97)314 (82)
Yes3227 (15)863 (7)2364 (25)43 (9)69 (22)21 (6)67 (16)6 (3)67 (18)
Family history of any cancer in first-degree relatives (%)
No18193 (84)10141 (84)8052 (84)404 (82)236 (77)281 (85)336 (82)165 (91)333 (87)
Yes3501 (16)1943 (16)1558 (16)91 (18)72 (23)51 (15)73 (18)16 (9)48 (13)

Lack of Asian representation in PRS development

Among PRS for breast (n=85), prostate (n=37), colorectal (n=22), and lung cancers (n=11) examined, the reported source of variant associations or GWAS used to build PRS were from predominantly European ancestry populations (Supplementary file 1c). Only six PRS for breast cancer (PGS000028, PGS000029, PGS000050, PGS000345, PGS0001336, and PGS001804), three for prostate cancer (PGS000878, PGS001291, and PGS001805), three PRS for colorectal cancer (PGS000055, PGS000802, and PGS001802), and one for lung cancer (PGS000070) were based on GWAS that included some non-European participants. For PRS development training, all but two PRSs were based on samples of non-European ancestry (PGS000733 for prostate cancer and PGS000802 for colorectal cancer). No significant association (p>0.05) was found between number of variants included in the various PRSs evaluated for each cancer and discriminatory ability (Supplementary file 1d).

PRS distribution

Figure 1 depicts the (A) distribution, (B) discrimination, (C) absolute risk association, and (D) calibration of the best-performing PRS (based on AUC) (Supplementary file 1d) for the four cancers studied: breast (PGS000873; Brentnall et al., 2020), prostate (PGS000662; Conti et al., 2021), colorectal (female: PGS000055; Schmit et al., 2019; male: PGS000734; Archambault et al., 2020), and lung (female: PGS000721; Jia et al., 2020; male: PGS000070; Dai et al., 2019). All PRSs were normally distributed, with a right shift observed in the distribution curves for cancer cases (Figure 1A). The mean value of each site-specific cancer PRS was significantly higher in cancer patients compared to controls (pt-test <0.00273).

Site-specific polygenic risk scores (PRSs) performance assessment.

(A) Distribution, (B) discrimination, (C) absolute risk association, and (D) calibration for each of the four common cancers studied (from left to right: breast, prostate, lung [female], lung [male], colorectal [female], and colorectal [male]). Two-sided, two-sample t-tests with a type I error of 0.05 were used to examine whether there was a difference in the distribution of standardised PRS (subtraction of mean value followed by the division by the standard deviation) between site-specific cancer cases and non-cancer controls (A). The PRSs showcased are the best-performing scores based on area under the receiver operator characteristic curve (AUC) values in the female and male populations, (i) unadjusted [solid line], and (ii) adjusted for age at recruitment [dashed line] (B). Each colored line in the plots for absolute risk association denotes a five percentile increase in the standardised PRS score in (C). Calibration calculated based on 5-year absolute risk by PRS deciles in (D). A prediction tool is considered more accurate when the AUC is larger. An AUC of 0.9–1.0 is considered excellent, 0.8–0.9 very good, 0.7–0.8 good, 0.6–0.7 sufficient, 0.5–0.6 bad, and less than 0.5 considered not useful (PMID: 27683318).

Figure 1—source data 1

Tables on absolute risk for breast cancer.

https://cdn.elifesciences.org/articles/82608/elife-82608-fig1-data1-v2.zip
Figure 1—source data 2

Tables on absolute risk for colorectal cancer.

https://cdn.elifesciences.org/articles/82608/elife-82608-fig1-data2-v2.zip
Figure 1—source data 3

Tables on absolute risk for lung cancer.

https://cdn.elifesciences.org/articles/82608/elife-82608-fig1-data3-v2.zip
Figure 1—source data 4

Tables on absolute risk for prostate cancer.

https://cdn.elifesciences.org/articles/82608/elife-82608-fig1-data4-v2.zip
Figure 1—source data 5

Tables on polygenic risk scores (PRS) performance assessment.

https://cdn.elifesciences.org/articles/82608/elife-82608-fig1-data5-v2.zip

PRS discriminatory ability

The highest AUC obtained from logistic models was observed for prostate cancer (0.70, 95% CI: [0.66–0.73]), followed by female breast cancer (0.61 [0.58–0.63]), male colorectal cancer (0.60, 95% CI=0.58–0.63), female colorectal cancer (0.56 [0.52–0.60]), male lung cancer (0.58 [0.55–0.61]), and female lung cancer (0.55 [0.50–0.59]) (Figure 1B).

Associations between PRS and the relative hazard of developing cancers

During the follow-up period of 20 years, the risk of acquiring breast, colorectal, or lung cancer increased significantly with higher PRS after adjusting for age at recruitment. Compared to the first PRS quintile, individuals in the highest quintile were more likely to develop the four cancers studied. The highest HR observed was for prostate cancer (8.99 [95%CI: 5.27–15.35]) and lowest for female lung cancer (1.69 [1.03–2.79]), adjusted for age at recruitment (Supplementary file 1e). Significant trends were found for the associations between PRS quintiles and site-specific cancers (p-trend ranges from 1.35×10–25 for prostate cancer to 0.008 for female lung cancer, Supplementary file 1e).

Compared to the middle cancer-specific PRS quintile, individuals in the highest PRS quintile were 64% more likely to develop cancers of the breast, prostate, and colorectal (Table 2). Individuals in the lowest PRS quintile were associated with a 16–72% reduction in risk of developing these cancers. For female lung cancer, the PRS quintiles 1 and 2 were associated with 44% and 45% decreased risk compared to the middle quintile, respectively. However, the HR observed for PRS quintiles 4 (0.89 [0.57–1.39]) and 5 (0.95 [0.61–1.47]) were not significantly different when compared to the middle quintile.

Table 2
Hazard ratios (HR) and corresponding 95% confidence intervals (CI) associated with polygenic risk score quintiles (Q) compared to the population median, using the Cox proportional hazards model and censored at 20 years after recruitment.

Individuals were categorised into cancer-specific quintiles based on their cancer-specific polygenic risk score (PRS). All models were adjusted for age at recruitment.

Cancer site – genderQ1Q2Q3Q4Q5
Breast – female
 Number of cases557386107145
 HR (95% CI)0.61 (0.44–0.86)0.80 (0.59–1.09)1.00 (Referent)1.25 (0.94–1.66)1.64 (1.26–2.14)
Prostate – male
 Number of cases15315559129
 HR (95% CI)0.28 (0.16–0.50)0.57 (0.37–0.88)1.00 (Referent)1.11 (0.77–1.60)2.52 (1.84–3.46)
Colorectal – female
 Number of cases47435366101
 HR (95% CI)0.84 (0.57–1.25)0.80 (0.53–1.20)1.00 (Referent)1.27 (0.88–1.82)1.91 (1.37–2.67)
Colorectal – male
 Number of cases36707187114
 HR (95% CI)0.51 (0.34–0.77)1.00 (0.72–1.39)1.00 (Referent)1.29 (0.94–1.76)1.67 (1.24–2.25)
Lung – female
 Number of cases2526413640
 HR (95% CI)0.56 (0.34–0.92)0.55 (0.34–0.91)1.00 (Referent)0.89 (0.57–1.39)0.95 (0.61–1.47)
Lung – male
 Number of cases51586880103
 HR (95% CI)0.72 (0.50–1.04)0.79 (0.56–1.13)1.00 (Referent)1.14 (0.82–1.57)1.46 (1.07–1.98)

Every SD increase in PRS is associated with 39–108% elevated risks of breast, prostate, and colorectal cancers (p<1.06 × 10–8), adjusted for age, dialect group, education, BMI, smoking status, physical activity, alcohol consumption, and family history (Table 3). The increased risk for female and male lung cancer was lower than the other three cancers (HRfemale: 1.21 [1.04–1.40], p=0.011; HRmale: 1.35 [1.22–1.49], p=1.01 × 10–8). Age at recruitment is significantly associated with elevated risks of developing all cancers, with the exception of female breast cancer (HR: 1.00 [0.99–1.02], p=0.571). Highest education level and BMI were positively correlated with breast cancer risk. Smoking was significantly associated with an ~30% reduction in risk of prostate cancer but increased the risk of lung cancer by approximately two- and fivefold for past and current smokers, compared to non-smokers, respectively. Alcohol consumption increased the risk of both female and male colorectal cancer by approximately 60% but was only significant for male colorectal cancer. Family history of cancer was only significantly associated with an increased risk for prostate cancer (HR: 1.53 [1.16–2.02], p=7.59 × 10–4).

Table 3
Associations between per standard deviation (SD) increase in site-specific polygenic risk scores and cancer occurrence.

Hazard ratios (HR) and corresponding 95% confidence intervals (CI) were estimated using Cox proportional hazard models, adjusted for age at recruitment, dialect group, highest education attained, body mass index, smoking status, alcohol consumption, and physical activity. Follow-up time was censored at 20 years after recruitment. Significant results are shown in bold.

Cancer site
BreastProstateColorectal – femaleColorectal – maleLung – femaleLung – male
HR (95% CI)p-ValueHR (95% CI)p-ValueHR (95% CI)p-ValueHR (95% CI)p-ValueHR (95% CI)p-ValueHR (95% CI)p-Value
Site-specific polygenic risk score, per SD increase1.47 (1.34–1.60)5.80E-172.08 (1.85–2.34)1.56E-331.39 (1.24–1.55)1.06E-081.44 (1.30–1.59)5.41E-121.21 (1.04–1.40)1.10E-021.35 (1.22–1.49)1.01E-08
Age at recruitment, years1.00 (0.99–1.02)5.82E-011.09 (1.07–1.10)6.34E-231.07 (1.05–1.09)7.24E-171.06 (1.05–1.08)9.53E-181.07 (1.05–1.10)1.65E-101.09 (1.07–1.10)1.46E-27
Dialect group (Cantonese vs Hokkien)0.88 (0.73–1.05)1.61E-010.98 (0.78–1.24)8.86E-010.78 (0.62–0.99)3.96E-021.22 (0.99–1.50)6.78E-020.92 (0.67–1.25)5.78E-011.07 (0.87–1.33)5.21E-01
Highest education (primary vs no)1.21 (0.95–1.53)1.20E-011.32 (0.81–2.14)2.65E-011.08 (0.83–1.41)5.60E-010.98 (0.70–1.37)8.91E-010.83 (0.58–1.19)3.11E-010.87 (0.64–1.18)3.67E-01
Highest education
(secondary or above vs no)
1.54 (1.18–2.01)1.57E-031.60 (0.98–2.63)6.17E-021.06 (0.76–1.48)7.46E-010.80 (0.55–1.16)2.33E-011.10 (0.69–1.74)6.87E-010.63 (0.44–0.90)1.16E-02
Body mass index, kg/m21.04 (1.02–1.07)1.28E-031.01 (0.98–1.05)5.15E-010.99 (0.96–1.02)5.33E-011.02 (0.98–1.05)3.19E-010.97 (0.92–1.01)1.58E-010.97 (0.93–1.00)5.60E-02
Smoking status
(ex-smoker vs non-smoker)
0.90 (0.45–1.83)7.81E-010.68 (0.50–0.92)1.32E-021.51 (0.86–2.66)1.55E-011.17 (0.90–1.52)2.36E-012.16 (1.04–4.48)3.86E-021.99 (1.41–2.83)1.09E-04
Smoking status
(current smoker vs non-smoker)
0.83 (0.49–1.39)4.72E-010.70 (0.52–0.93)1.52E-021.10 (0.69–1.75)6.85E-011.22 (0.96–1.56)1.08E-015.78 (3.98–8.38)2.69E-205.15 (3.83–6.91)1.17E-27
Alcohol consumption
(weekly vs never/ occasionally)
1.04 (0.65–1.67)8.74E-010.98 (0.70–1.39)9.29E-010.76 (0.38–1.54)4.46E-011.31 (1.00–1.73)5.39E-020.72 (0.27–1.96)5.23E-010.89 (0.65–1.22)4.81E-01
Alcohol consumption
(daily vs never/ occasionally)
0.71 (0.27–1.91)5.00E-010.74 (0.40–1.36)3.32E-011.55 (0.69–3.49)2.89E-011.64 (1.15–2.34)6.54E-030.66 (0.16–2.68)5.63E-011.21 (0.85–1.73)2.81E-01
Moderate physical activity
(1–3 hr/week vs no)
0.98 (0.75–1.28)8.80E-011.17 (0.87–1.57)2.97E-010.88 (0.63–1.24)4.74E-011.02 (0.78–1.35)8.79E-010.99 (0.62–1.58)9.73E-010.90 (0.67–1.23)5.17E-01
Moderate physical activity
(≥3 hr/week vs no)
1.17 (0.86–1.60)3.20E-010.99 (0.68–1.45)9.78E-010.59 (0.37–0.96)3.33E-021.10 (0.80–1.52)5.45E-010.96 (0.54–1.70)8.78E-010.86 (0.59–1.26)4.36E-01
Vigorous physical activity/ strenuous sports at least once a week (yes vs no)1.24 (0.90–1.70)1.89E-011.05 (0.79–1.41)7.16E-011.09 (0.68–1.74)7.25E-010.75 (0.57–1.00)5.06E-020.58 (0.24–1.42)2.30E-010.95 (0.72–1.26)7.37E-01
Family history (yes vs no)1.14 (0.90–1.45)2.67E-011.53 (1.16–2.02)2.47E-031.08 (0.79–1.48)6.20E-011.24 (0.95–1.62)1.09E-010.67 (0.40–1.13)1.33E-010.97 (0.71–1.33)8.63E-01

All Cox models presented in Tables 2 and 3 did not violate the proportionality assumption for the PRS studied (p-values of cox-zph() for PRS were >0.05).

Association of PRS with absolute risk

In terms of the 5-year absolute risk of developing site-specific cancers, the largest difference between the highest and lowest PRS categories was observed for prostate cancer, followed by breast cancer (Figure 1C). A separation of the absolute risk curves was observed for female breast cancer already at age 30 years. For prostate cancer, the separation of curves was observed only after age 50 years. Slight separation of the curves began after 50 years of age for colorectal and lung cancer.

PRS calibration

In general, predicted risks for the higher PRS categories did not correspond well to the observed proportions for female breast, male prostate, and female lung cancers (Figure 1D); in particular, predicted risks were overestimated for the higher risk categories. Overestimation of risk was observed for all PRS categories for male lung cancer. In contrast, predicted risks were underestimated for both female and male colorectal cancers. Nonetheless, the CI associated with the calibration slopes for all cancers included 1, with the exception of female (0.32 [-0.31 to 0.95]) and male lung cancers (0.26 [-0.25, 0.78]).

Discussion

Precision prevention in oncology is based on the idea that an individual’s risk, which is influenced by genetics, environment, and lifestyle factors, is linked to the amount of benefit achieved through cancer screening (Roberts, 2018). Risk stratification for cancer screening can be used in this framework to identify and recommend screening for persons with a high enough cancer risk that the benefits outweigh the risks. Several PRS prediction models have been established for site-specific cancers, each with its own set of strengths and limitations, and different risk models may produce different results for the same individual.

In an increasingly inclusive world, genetic studies fall short on diversity. According to a 2009 study, an overwhelming 96% of people who took part in genome-wide association studies (GWAS) were of European ancestry (Need and Goldstein, 2009). GWAS results are the backbone on which PRS is developed. A concern raised was that, without representation from a broader spectrum of populations, genomic medicine may be limited to benefitting ‘a privileged few’ (Popejoy and Fullerton, 2016).

Genetic studies in 2016 showed that the proportion of people not of European ancestry included in GWAS has increased to approximately 20% (Popejoy and Fullerton, 2016). Most of this rise can be attributed to more research on Asian ancestry communities in Asia (Popejoy and Fullerton, 2016). With increasing interest worldwide in using a risk-based approach to screening programs over the current age-based paradigm, this progress raises questions on whether selected established PRS shown to perform well in European-based populations has equal utility in Asians. Nonetheless, as our results show, most of the populations from which PRS were developed are still predominantly of European ancestry.

In accordance with published Polygenic Risk Score Reporting Standards, we reported PRS distribution, discrimination, absolute risk association, and calibration for each of the four common cancers studied (Wand et al., 2021). Our results show that cancer cases were associated with higher PRS compared to non-cancer controls. In the age-adjusted models, a constant trend between PRS percentile rank and observed cancer risk in our study population supports the validity of PRS for breast, prostate, and colorectal cancers, but not for lung cancer. The best-performing PRS for female breast cancer was able to stratify women into distinct bands of breast cancer risk at an earlier age, and across all ages, suggesting that it could be a useful prediction tool in risk-based breast cancer screening in combination with other risk factors specific to breast cancer (Ho et al., 2022). This PRS has been incorporated into a pilot risk-based breast cancer screening study in a comparable study population (Liu et al., 2022). The best-performing PRS for prostate and male colorectal cancers in this study appeared to exhibit sufficient discriminatory ability and predictive value, especially for older participants.

PRS may be of limited use in predicting female colorectal and female/male lung cancer. The least predictive value was in lung cancer, which could be related to the higher prevalence of EGFR mutant lung cancer which has an Asian predilection, thus less amenable to PRS developed in Caucasian population (Shigematsu et al., 2005). It is reassuring to see tobacco smoking is a strong risk factor for lung cancer in our dataset. However, smoking appeared to be associated with a protective effect for prostate cancer. While smoking is a well-known risk factor for many cancers (Jacob et al., 2018), in particular lung cancer, observational studies frequently show that smokers are associated with a lower incidence of prostate cancer (Rohrmann et al., 2013; Watters et al., 2009; Adami et al., 1996; Lund Nilsen et al., 2000; Engeland et al., 1996; Islami et al., 2014; Ordóñez-Mena et al., 2016; Giovannucci et al., 2007). However, a Mendelian randomisation study did not support the association (Larsson et al., 2020).

There is room for improvement in the discriminatory ability of PRS (Lewis and Green, 2021). As noted by Lambert et al. in a review, a wider divergence between the average scores of cases and non-cases (quantified by AUC) and associated effect sizes (odds ratio and SD) is expected when PRS explains more of the heredity for each trait (Lambert et al., 2019). Larger GWAS sample sizes of appropriate ancestries and the inclusion of rarer genetic variants, obtained through other methods such as whole-genome sequencing, would likely be required to boost explained heritability (Lambert et al., 2019). In addition, group-wise estimates, which arbitrarily classify the top 10%, 5%, or 1% of samples as the at-risk group, are not optimal for decisions at the individual level (Lewis and Green, 2021). Emerging new methodologies that estimate probability values for hypothetically assigning an individual as at risk or not at risk, thus providing individuals with more clarity, may help to overcome this limitation (Sun et al., 2021). At this point, PRS may not have yet reached the standards as a clinical tool by itself. However, it is still helpful in guiding screening decisions and supplementing established protocols (Polygenic Risk Score Task Force of the International Common Disease Alliance, 2021).

As highlighted by Wei et al., the reliability of score values is necessary for application at the individual level (Wei et al., 2022). Even when a PRS has adequate discrimination, estimated risks can be unreliable (Van Calster et al., 2019). Our results show that cancer risk estimates based on PRS developed using populations of European ancestry are not optimally calibrated for our Asian study population. However, the CI associated with the calibration slopes for all cancers except for female and male lung included the value of 1, suggesting that the overall calibration is not poor. For female and male lung cancers associated with low AUC values, poor calibration is not unexpected. All PRSs except PGS000662 (prostate cancer) passed the formal Hosmer-Lemeshow goodness-of-fit test. Males in the first 5 deciles of PGS000662 did not develop prostate cancer, suggesting that a linear fit may not be appropriate. A hard threshold beginning from the 6th decile may perform better at identifying males at elevated risk of developing prostate cancer.

Poorly calibrated PRS can be misleading and have clinical repercussions (Van Calster et al., 2019; Van Calster and Vickers, 2015). Underestimation of risk may result in a false sense of security. Overestimation of risk may cause unnecessary anxiety, misguided interventions, and overtreatment. In a population-wide screening setting, however, where the return of PRS results can be designed such that only high-risk individuals are highlighted, underestimation of risk may be less of an issue. Arguably, with parallel input from other risk factors and evaluation by healthcare specialists, the overestimation of risk that results in a higher number of at-risk individuals identified may increase the number of cancers potentially detected early. Nonetheless, suitable correction factors will be required to ensure the reliability of PRS prior to clinical implementation.

While the study population used in this analysis comprises less than a thousand cases of the most common cancers examined, the SCHS, established between April 1993 and December 1998, is one of the largest population-based Asian cohorts in the world with high-quality prospective data on exposure and comprehensive capture of morbidity and mortality. All cancer cases are incident cases diagnosed over three decades of follow-up. Blood samples were collected from a subset of SCHS participants who were alive and contactable between 1999 and 2004 (after the recruitment period 1993–1998). While we attempted to adjust associations between PRSs and incident cancers in the study by including multiple related risk factors as covariates in Cox proportional hazards models, we acknowledge the potential of survival biases in the study. This is one of the best resources to evaluate the utility of PRS in a prospective manner. The findings open a window in our current understanding of which PRS is relevant and ready to be deployed in risk-based cancer screening studies. Nonetheless, it should be noted that among the PRS interrogated from the PGS Catalog, the ‘best’ PRS selected in this study may be only superficially superior in terms of AUC (i.e. trailing decimal places) over the ‘next best’ PRS. We further tested the sensitivity of the PRS selection using a time-to-event metric for AUC (at 5 years), and the differences found were non-informative from the logistic regression (Supplementary file 1f). To increase the number of events, we combined the males and females for lung and colorectal cancers. The resulting AUCs (from the logistic regression) were not appreciably different from the sex-specific analysis (Supplementary file 1g).

Ethnic representation in PRS model development, PRS validation, limited discriminative ability in the general population, ill calibration, insufficient healthcare professional and patient education, and healthcare system integration are all hurdles that must be crossed before PRS can be implemented responsibly as a public health instrument (Lewis and Vassos, 2020; Slunecka et al., 2021). While nationwide screening programs have helped to raise cancer awareness, there is still a need to improve the effectiveness and efficiency of cancer screening in Asian countries such as Singapore, given the steadily rising incidence rates. Despite the challenges, a risk-based screening strategy that includes the use of PRS should be actively examined for research and implementation.

Data availability

All polygenic risk scores used in this study are publicly available in the PGS Catalog (https://www.pgscatalog.org; Lambert et al., 2021). The data that support the findings of our study are available from the corresponding authors of the study upon reasonable request (Dr Rajkumar s/o Dorajoo, dorajoor@gis.a-star.edu.sg and Dr Jingmei Li, lijm1@gis.a-star.edu.sg). More information regarding the data access to SCHS can be found at: https://sph.nus.edu.sg/research/cohort-schs/. The data are not publicly available due to Singapore laws. Figure 1—source data 1 contains the numerical data used to generate Figure 1. The code for the study is uploaded as Source code 1.

References

    1. Archambault AN
    2. Su YR
    3. Jeon J
    4. Thomas M
    5. Lin Y
    6. Conti DV
    7. Win AK
    8. Sakoda LC
    9. Lansdorp-Vogelaar I
    10. Peterse EFP
    11. Zauber AG
    12. Duggan D
    13. Holowatyj AN
    14. Huyghe JR
    15. Brenner H
    16. Cotterchio M
    17. Bézieau S
    18. Schmit SL
    19. Edlund CK
    20. Southey MC
    21. MacInnis RJ
    22. Campbell PT
    23. Chang-Claude J
    24. Slattery ML
    25. Chan AT
    26. Joshi AD
    27. Song M
    28. Cao Y
    29. Woods MO
    30. White E
    31. Weinstein SJ
    32. Ulrich CM
    33. Hoffmeister M
    34. Bien SA
    35. Harrison TA
    36. Hampe J
    37. Li CI
    38. Schafmayer C
    39. Offit K
    40. Pharoah PD
    41. Moreno V
    42. Lindblom A
    43. Wolk A
    44. Wu AH
    45. Li L
    46. Gunter MJ
    47. Gsur A
    48. Keku TO
    49. Pearlman R
    50. Bishop DT
    51. Castellví-Bel S
    52. Moreira L
    53. Vodicka P
    54. Kampman E
    55. Giles GG
    56. Albanes D
    57. Baron JA
    58. Berndt SI
    59. Brezina S
    60. Buch S
    61. Buchanan DD
    62. Trichopoulou A
    63. Severi G
    64. Chirlaque MD
    65. Sánchez MJ
    66. Palli D
    67. Kühn T
    68. Murphy N
    69. Cross AJ
    70. Burnett-Hartman AN
    71. Chanock SJ
    72. de la Chapelle A
    73. Easton DF
    74. Elliott F
    75. English DR
    76. Feskens EJM
    77. FitzGerald LM
    78. Goodman PJ
    79. Hopper JL
    80. Hudson TJ
    81. Hunter DJ
    82. Jacobs EJ
    83. Joshu CE
    84. Küry S
    85. Markowitz SD
    86. Milne RL
    87. Platz EA
    88. Rennert G
    89. Rennert HS
    90. Schumacher FR
    91. Sandler RS
    92. Seminara D
    93. Tangen CM
    94. Thibodeau SN
    95. Toland AE
    96. van Duijnhoven FJB
    97. Visvanathan K
    98. Vodickova L
    99. Potter JD
    100. Männistö S
    101. Weigl K
    102. Figueiredo J
    103. Martín V
    104. Larsson SC
    105. Parfrey PS
    106. Huang WY
    107. Lenz HJ
    108. Castelao JE
    109. Gago-Dominguez M
    110. Muñoz-Garzón V
    111. Mancao C
    112. Haiman CA
    113. Wilkens LR
    114. Siegel E
    115. Barry E
    116. Younghusband B
    117. Van Guelpen B
    118. Harlid S
    119. Zeleniuch-Jacquotte A
    120. Liang PS
    121. Du M
    122. Casey G
    123. Lindor NM
    124. Le Marchand L
    125. Gallinger SJ
    126. Jenkins MA
    127. Newcomb PA
    128. Gruber SB
    129. Schoen RE
    130. Hampel H
    131. Corley DA
    132. Hsu L
    133. Peters U
    134. Hayes RB
    (2020) Cumulative burden of colorectal cancer-associated genetic variants is more strongly associated with early-onset vs late-onset cancer
    Gastroenterology 158:1274–1286.
    https://doi.org/10.1053/j.gastro.2019.12.012
    1. Conti DV
    2. Darst BF
    3. Moss LC
    4. Saunders EJ
    5. Sheng X
    6. Chou A
    7. Schumacher FR
    8. Olama AAA
    9. Benlloch S
    10. Dadaev T
    11. Brook MN
    12. Sahimi A
    13. Hoffmann TJ
    14. Takahashi A
    15. Matsuda K
    16. Momozawa Y
    17. Fujita M
    18. Muir K
    19. Lophatananon A
    20. Wan P
    21. Le Marchand L
    22. Wilkens LR
    23. Stevens VL
    24. Gapstur SM
    25. Carter BD
    26. Schleutker J
    27. Tammela TLJ
    28. Sipeky C
    29. Auvinen A
    30. Giles GG
    31. Southey MC
    32. MacInnis RJ
    33. Cybulski C
    34. Wokołorczyk D
    35. Lubiński J
    36. Neal DE
    37. Donovan JL
    38. Hamdy FC
    39. Martin RM
    40. Nordestgaard BG
    41. Nielsen SF
    42. Weischer M
    43. Bojesen SE
    44. Røder MA
    45. Iversen P
    46. Batra J
    47. Chambers S
    48. Moya L
    49. Horvath L
    50. Clements JA
    51. Tilley W
    52. Risbridger GP
    53. Gronberg H
    54. Aly M
    55. Szulkin R
    56. Eklund M
    57. Nordström T
    58. Pashayan N
    59. Dunning AM
    60. Ghoussaini M
    61. Travis RC
    62. Key TJ
    63. Riboli E
    64. Park JY
    65. Sellers TA
    66. Lin H-Y
    67. Albanes D
    68. Weinstein SJ
    69. Mucci LA
    70. Giovannucci E
    71. Lindstrom S
    72. Kraft P
    73. Hunter DJ
    74. Penney KL
    75. Turman C
    76. Tangen CM
    77. Goodman PJ
    78. Thompson IM Jr
    79. Hamilton RJ
    80. Fleshner NE
    81. Finelli A
    82. Parent M-É
    83. Stanford JL
    84. Ostrander EA
    85. Geybels MS
    86. Koutros S
    87. Freeman LEB
    88. Stampfer M
    89. Wolk A
    90. Håkansson N
    91. Andriole GL
    92. Hoover RN
    93. Machiela MJ
    94. Sørensen KD
    95. Borre M
    96. Blot WJ
    97. Zheng W
    98. Yeboah ED
    99. Mensah JE
    100. Lu Y-J
    101. Zhang H-W
    102. Feng N
    103. Mao X
    104. Wu Y
    105. Zhao S-C
    106. Sun Z
    107. Thibodeau SN
    108. McDonnell SK
    109. Schaid DJ
    110. West CML
    111. Burnet N
    112. Barnett G
    113. Maier C
    114. Schnoeller T
    115. Luedeke M
    116. Kibel AS
    117. Drake BF
    118. Cussenot O
    119. Cancel-Tassin G
    120. Menegaux F
    121. Truong T
    122. Koudou YA
    123. John EM
    124. Grindedal EM
    125. Maehle L
    126. Khaw K-T
    127. Ingles SA
    128. Stern MC
    129. Vega A
    130. Gómez-Caamaño A
    131. Fachal L
    132. Rosenstein BS
    133. Kerns SL
    134. Ostrer H
    135. Teixeira MR
    136. Paulo P
    137. Brandão A
    138. Watya S
    139. Lubwama A
    140. Bensen JT
    141. Fontham ETH
    142. Mohler J
    143. Taylor JA
    144. Kogevinas M
    145. Llorca J
    146. Castaño-Vinyals G
    147. Cannon-Albright L
    148. Teerlink CC
    149. Huff CD
    150. Strom SS
    151. Multigner L
    152. Blanchet P
    153. Brureau L
    154. Kaneva R
    155. Slavov C
    156. Mitev V
    157. Leach RJ
    158. Weaver B
    159. Brenner H
    160. Cuk K
    161. Holleczek B
    162. Saum K-U
    163. Klein EA
    164. Hsing AW
    165. Kittles RA
    166. Murphy AB
    167. Logothetis CJ
    168. Kim J
    169. Neuhausen SL
    170. Steele L
    171. Ding YC
    172. Isaacs WB
    173. Nemesure B
    174. Hennis AJM
    175. Carpten J
    176. Pandha H
    177. Michael A
    178. De Ruyck K
    179. De Meerleer G
    180. Ost P
    181. Xu J
    182. Razack A
    183. Lim J
    184. Teo S-H
    185. Newcomb LF
    186. Lin DW
    187. Fowke JH
    188. Neslund-Dudas C
    189. Rybicki BA
    190. Gamulin M
    191. Lessel D
    192. Kulis T
    193. Usmani N
    194. Singhal S
    195. Parliament M
    196. Claessens F
    197. Joniau S
    198. Van den Broeck T
    199. Gago-Dominguez M
    200. Castelao JE
    201. Martinez ME
    202. Larkin S
    203. Townsend PA
    204. Aukim-Hastie C
    205. Bush WS
    206. Aldrich MC
    207. Crawford DC
    208. Srivastava S
    209. Cullen JC
    210. Petrovics G
    211. Casey G
    212. Roobol MJ
    213. Jenster G
    214. van Schaik RHN
    215. Hu JJ
    216. Sanderson M
    217. Varma R
    218. McKean-Cowdin R
    219. Torres M
    220. Mancuso N
    221. Berndt SI
    222. Van Den Eeden SK
    223. Easton DF
    224. Chanock SJ
    225. Cook MB
    226. Wiklund F
    227. Nakagawa H
    228. Witte JS
    229. Eeles RA
    230. Kote-Jarai Z
    231. Haiman CA
    (2021) Trans-Ancestry genome-wide association meta-analysis of prostate cancer identifies new susceptibility loci and informs genetic risk prediction
    Nature Genetics 53:65–75.
    https://doi.org/10.1038/s41588-020-00748-0
    1. Emmanuel S
    (1993)
    Quality assurance in medicine: research and evaluation activities towards quality control in Singapore
    Annals of the Academy of Medicine, Singapore 22:129–133.
    1. Mavaddat N
    2. Pharoah PDP
    3. Michailidou K
    4. Tyrer J
    5. Brook MN
    6. Bolla MK
    7. Wang Q
    8. Dennis J
    9. Dunning AM
    10. Shah M
    11. Luben R
    12. Brown J
    13. Bojesen SE
    14. Nordestgaard BG
    15. Nielsen SF
    16. Flyger H
    17. Czene K
    18. Darabi H
    19. Eriksson M
    20. Peto J
    21. Dos-Santos-Silva I
    22. Dudbridge F
    23. Johnson N
    24. Schmidt MK
    25. Broeks A
    26. Verhoef S
    27. Rutgers EJ
    28. Swerdlow A
    29. Ashworth A
    30. Orr N
    31. Schoemaker MJ
    32. Figueroa J
    33. Chanock SJ
    34. Brinton L
    35. Lissowska J
    36. Couch FJ
    37. Olson JE
    38. Vachon C
    39. Pankratz VS
    40. Lambrechts D
    41. Wildiers H
    42. Van Ongeval C
    43. van Limbergen E
    44. Kristensen V
    45. Grenaker Alnæs G
    46. Nord S
    47. Borresen-Dale A-L
    48. Nevanlinna H
    49. Muranen TA
    50. Aittomäki K
    51. Blomqvist C
    52. Chang-Claude J
    53. Rudolph A
    54. Seibold P
    55. Flesch-Janys D
    56. Fasching PA
    57. Haeberle L
    58. Ekici AB
    59. Beckmann MW
    60. Burwinkel B
    61. Marme F
    62. Schneeweiss A
    63. Sohn C
    64. Trentham-Dietz A
    65. Newcomb P
    66. Titus L
    67. Egan KM
    68. Hunter DJ
    69. Lindstrom S
    70. Tamimi RM
    71. Kraft P
    72. Rahman N
    73. Turnbull C
    74. Renwick A
    75. Seal S
    76. Li J
    77. Liu J
    78. Humphreys K
    79. Benitez J
    80. Pilar Zamora M
    81. Arias Perez JI
    82. Menéndez P
    83. Jakubowska A
    84. Lubinski J
    85. Jaworska-Bieniek K
    86. Durda K
    87. Bogdanova NV
    88. Antonenkova NN
    89. Dörk T
    90. Anton-Culver H
    91. Neuhausen SL
    92. Ziogas A
    93. Bernstein L
    94. Devilee P
    95. Tollenaar RAEM
    96. Seynaeve C
    97. van Asperen CJ
    98. Cox A
    99. Cross SS
    100. Reed MWR
    101. Khusnutdinova E
    102. Bermisheva M
    103. Prokofyeva D
    104. Takhirova Z
    105. Meindl A
    106. Schmutzler RK
    107. Sutter C
    108. Yang R
    109. Schürmann P
    110. Bremer M
    111. Christiansen H
    112. Park-Simon T-W
    113. Hillemanns P
    114. Guénel P
    115. Truong T
    116. Menegaux F
    117. Sanchez M
    118. Radice P
    119. Peterlongo P
    120. Manoukian S
    121. Pensotti V
    122. Hopper JL
    123. Tsimiklis H
    124. Apicella C
    125. Southey MC
    126. Brauch H
    127. Brüning T
    128. Ko Y-D
    129. Sigurdson AJ
    130. Doody MM
    131. Hamann U
    132. Torres D
    133. Ulmer H-U
    134. Försti A
    135. Sawyer EJ
    136. Tomlinson I
    137. Kerin MJ
    138. Miller N
    139. Andrulis IL
    140. Knight JA
    141. Glendon G
    142. Marie Mulligan A
    143. Chenevix-Trench G
    144. Balleine R
    145. Giles GG
    146. Milne RL
    147. McLean C
    148. Lindblom A
    149. Margolin S
    150. Haiman CA
    151. Henderson BE
    152. Schumacher F
    153. Le Marchand L
    154. Eilber U
    155. Wang-Gohrke S
    156. Hooning MJ
    157. Hollestelle A
    158. van den Ouweland AMW
    159. Koppert LB
    160. Carpenter J
    161. Clarke C
    162. Scott R
    163. Mannermaa A
    164. Kataja V
    165. Kosma V-M
    166. Hartikainen JM
    167. Brenner H
    168. Arndt V
    169. Stegmaier C
    170. Karina Dieffenbach A
    171. Winqvist R
    172. Pylkäs K
    173. Jukkola-Vuorinen A
    174. Grip M
    175. Offit K
    176. Vijai J
    177. Robson M
    178. Rau-Murthy R
    179. Dwek M
    180. Swann R
    181. Annie Perkins K
    182. Goldberg MS
    183. Labrèche F
    184. Dumont M
    185. Eccles DM
    186. Tapper WJ
    187. Rafiq S
    188. John EM
    189. Whittemore AS
    190. Slager S
    191. Yannoukakos D
    192. Toland AE
    193. Yao S
    194. Zheng W
    195. Halverson SL
    196. González-Neira A
    197. Pita G
    198. Rosario Alonso M
    199. Álvarez N
    200. Herrero D
    201. Tessier DC
    202. Vincent D
    203. Bacot F
    204. Luccarini C
    205. Baynes C
    206. Ahmed S
    207. Maranian M
    208. Healey CS
    209. Simard J
    210. Hall P
    211. Easton DF
    212. Garcia-Closas M
    (2015) Prediction of breast cancer risk based on profiling with common genetic variants
    Journal of the National Cancer Institute 107:djv036.
    https://doi.org/10.1093/jnci/djv036
    1. Mavaddat N
    2. Michailidou K
    3. Dennis J
    4. Lush M
    5. Fachal L
    6. Lee A
    7. Tyrer JP
    8. Chen T-H
    9. Wang Q
    10. Bolla MK
    11. Yang X
    12. Adank MA
    13. Ahearn T
    14. Aittomäki K
    15. Allen J
    16. Andrulis IL
    17. Anton-Culver H
    18. Antonenkova NN
    19. Arndt V
    20. Aronson KJ
    21. Auer PL
    22. Auvinen P
    23. Barrdahl M
    24. Beane Freeman LE
    25. Beckmann MW
    26. Behrens S
    27. Benitez J
    28. Bermisheva M
    29. Bernstein L
    30. Blomqvist C
    31. Bogdanova NV
    32. Bojesen SE
    33. Bonanni B
    34. Børresen-Dale A-L
    35. Brauch H
    36. Bremer M
    37. Brenner H
    38. Brentnall A
    39. Brock IW
    40. Brooks-Wilson A
    41. Brucker SY
    42. Brüning T
    43. Burwinkel B
    44. Campa D
    45. Carter BD
    46. Castelao JE
    47. Chanock SJ
    48. Chlebowski R
    49. Christiansen H
    50. Clarke CL
    51. Collée JM
    52. Cordina-Duverger E
    53. Cornelissen S
    54. Couch FJ
    55. Cox A
    56. Cross SS
    57. Czene K
    58. Daly MB
    59. Devilee P
    60. Dörk T
    61. Dos-Santos-Silva I
    62. Dumont M
    63. Durcan L
    64. Dwek M
    65. Eccles DM
    66. Ekici AB
    67. Eliassen AH
    68. Ellberg C
    69. Engel C
    70. Eriksson M
    71. Evans DG
    72. Fasching PA
    73. Figueroa J
    74. Fletcher O
    75. Flyger H
    76. Försti A
    77. Fritschi L
    78. Gabrielson M
    79. Gago-Dominguez M
    80. Gapstur SM
    81. García-Sáenz JA
    82. Gaudet MM
    83. Georgoulias V
    84. Giles GG
    85. Gilyazova IR
    86. Glendon G
    87. Goldberg MS
    88. Goldgar DE
    89. González-Neira A
    90. Grenaker Alnæs GI
    91. Grip M
    92. Gronwald J
    93. Grundy A
    94. Guénel P
    95. Haeberle L
    96. Hahnen E
    97. Haiman CA
    98. Håkansson N
    99. Hamann U
    100. Hankinson SE
    101. Harkness EF
    102. Hart SN
    103. He W
    104. Hein A
    105. Heyworth J
    106. Hillemanns P
    107. Hollestelle A
    108. Hooning MJ
    109. Hoover RN
    110. Hopper JL
    111. Howell A
    112. Huang G
    113. Humphreys K
    114. Hunter DJ
    115. Jakimovska M
    116. Jakubowska A
    117. Janni W
    118. John EM
    119. Johnson N
    120. Jones ME
    121. Jukkola-Vuorinen A
    122. Jung A
    123. Kaaks R
    124. Kaczmarek K
    125. Kataja V
    126. Keeman R
    127. Kerin MJ
    128. Khusnutdinova E
    129. Kiiski JI
    130. Knight JA
    131. Ko Y-D
    132. Kosma V-M
    133. Koutros S
    134. Kristensen VN
    135. Krüger U
    136. Kühl T
    137. Lambrechts D
    138. Le Marchand L
    139. Lee E
    140. Lejbkowicz F
    141. Lilyquist J
    142. Lindblom A
    143. Lindström S
    144. Lissowska J
    145. Lo W-Y
    146. Loibl S
    147. Long J
    148. Lubiński J
    149. Lux MP
    150. MacInnis RJ
    151. Maishman T
    152. Makalic E
    153. Maleva Kostovska I
    154. Mannermaa A
    155. Manoukian S
    156. Margolin S
    157. Martens JWM
    158. Martinez ME
    159. Mavroudis D
    160. McLean C
    161. Meindl A
    162. Menon U
    163. Middha P
    164. Miller N
    165. Moreno F
    166. Mulligan AM
    167. Mulot C
    168. Muñoz-Garzon VM
    169. Neuhausen SL
    170. Nevanlinna H
    171. Neven P
    172. Newman WG
    173. Nielsen SF
    174. Nordestgaard BG
    175. Norman A
    176. Offit K
    177. Olson JE
    178. Olsson H
    179. Orr N
    180. Pankratz VS
    181. Park-Simon T-W
    182. Perez JIA
    183. Pérez-Barrios C
    184. Peterlongo P
    185. Peto J
    186. Pinchev M
    187. Plaseska-Karanfilska D
    188. Polley EC
    189. Prentice R
    190. Presneau N
    191. Prokofyeva D
    192. Purrington K
    193. Pylkäs K
    194. Rack B
    195. Radice P
    196. Rau-Murthy R
    197. Rennert G
    198. Rennert HS
    199. Rhenius V
    200. Robson M
    201. Romero A
    202. Ruddy KJ
    203. Ruebner M
    204. Saloustros E
    205. Sandler DP
    206. Sawyer EJ
    207. Schmidt DF
    208. Schmutzler RK
    209. Schneeweiss A
    210. Schoemaker MJ
    211. Schumacher F
    212. Schürmann P
    213. Schwentner L
    214. Scott C
    215. Scott RJ
    216. Seynaeve C
    217. Shah M
    218. Sherman ME
    219. Shrubsole MJ
    220. Shu X-O
    221. Slager S
    222. Smeets A
    223. Sohn C
    224. Soucy P
    225. Southey MC
    226. Spinelli JJ
    227. Stegmaier C
    228. Stone J
    229. Swerdlow AJ
    230. Tamimi RM
    231. Tapper WJ
    232. Taylor JA
    233. Terry MB
    234. Thöne K
    235. Tollenaar RAEM
    236. Tomlinson I
    237. Truong T
    238. Tzardi M
    239. Ulmer H-U
    240. Untch M
    241. Vachon CM
    242. van Veen EM
    243. Vijai J
    244. Weinberg CR
    245. Wendt C
    246. Whittemore AS
    247. Wildiers H
    248. Willett W
    249. Winqvist R
    250. Wolk A
    251. Yang XR
    252. Yannoukakos D
    253. Zhang Y
    254. Zheng W
    255. Ziogas A
    256. ABCTB Investigators
    257. kConFab/AOCS Investigators
    258. NBCS Collaborators
    259. Dunning AM
    260. Thompson DJ
    261. Chenevix-Trench G
    262. Chang-Claude J
    263. Schmidt MK
    264. Hall P
    265. Milne RL
    266. Pharoah PDP
    267. Antoniou AC
    268. Chatterjee N
    269. Kraft P
    270. García-Closas M
    271. Simard J
    272. Easton DF
    (2019) Polygenic risk scores for prediction of breast cancer and breast cancer subtypes
    American Journal of Human Genetics 104:21–34.
    https://doi.org/10.1016/j.ajhg.2018.11.002
    1. Schmit SL
    2. Edlund CK
    3. Schumacher FR
    4. Gong J
    5. Harrison TA
    6. Huyghe JR
    7. Qu C
    8. Melas M
    9. Van Den Berg DJ
    10. Wang H
    11. Tring S
    12. Plummer SJ
    13. Albanes D
    14. Alonso MH
    15. Amos CI
    16. Anton K
    17. Aragaki AK
    18. Arndt V
    19. Barry EL
    20. Berndt SI
    21. Bezieau S
    22. Bien S
    23. Bloomer A
    24. Boehm J
    25. Boutron-Ruault M-C
    26. Brenner H
    27. Brezina S
    28. Buchanan DD
    29. Butterbach K
    30. Caan BJ
    31. Campbell PT
    32. Carlson CS
    33. Castelao JE
    34. Chan AT
    35. Chang-Claude J
    36. Chanock SJ
    37. Cheng I
    38. Cheng Y-W
    39. Chin LS
    40. Church JM
    41. Church T
    42. Coetzee GA
    43. Cotterchio M
    44. Cruz Correa M
    45. Curtis KR
    46. Duggan D
    47. Easton DF
    48. English D
    49. Feskens EJM
    50. Fischer R
    51. FitzGerald LM
    52. Fortini BK
    53. Fritsche LG
    54. Fuchs CS
    55. Gago-Dominguez M
    56. Gala M
    57. Gallinger SJ
    58. Gauderman WJ
    59. Giles GG
    60. Giovannucci EL
    61. Gogarten SM
    62. Gonzalez-Villalpando C
    63. Gonzalez-Villalpando EM
    64. Grady WM
    65. Greenson JK
    66. Gsur A
    67. Gunter M
    68. Haiman CA
    69. Hampe J
    70. Harlid S
    71. Harju JF
    72. Hayes RB
    73. Hofer P
    74. Hoffmeister M
    75. Hopper JL
    76. Huang S-C
    77. Huerta JM
    78. Hudson TJ
    79. Hunter DJ
    80. Idos GE
    81. Iwasaki M
    82. Jackson RD
    83. Jacobs EJ
    84. Jee SH
    85. Jenkins MA
    86. Jia W-H
    87. Jiao S
    88. Joshi AD
    89. Kolonel LN
    90. Kono S
    91. Kooperberg C
    92. Krogh V
    93. Kuehn T
    94. Küry S
    95. LaCroix A
    96. Laurie CA
    97. Lejbkowicz F
    98. Lemire M
    99. Lenz H-J
    100. Levine D
    101. Li CI
    102. Li L
    103. Lieb W
    104. Lin Y
    105. Lindor NM
    106. Liu Y-R
    107. Loupakis F
    108. Lu Y
    109. Luh F
    110. Ma J
    111. Mancao C
    112. Manion FJ
    113. Markowitz SD
    114. Martin V
    115. Matsuda K
    116. Matsuo K
    117. McDonnell KJ
    118. McNeil CE
    119. Milne R
    120. Molina AJ
    121. Mukherjee B
    122. Murphy N
    123. Newcomb PA
    124. Offit K
    125. Omichessan H
    126. Palli D
    127. Cotoré JPP
    128. Pérez-Mayoral J
    129. Pharoah PD
    130. Potter JD
    131. Qu C
    132. Raskin L
    133. Rennert G
    134. Rennert HS
    135. Riggs BM
    136. Schafmayer C
    137. Schoen RE
    138. Sellers TA
    139. Seminara D
    140. Severi G
    141. Shi W
    142. Shibata D
    143. Shu X-O
    144. Siegel EM
    145. Slattery ML
    146. Southey M
    147. Stadler ZK
    148. Stern MC
    149. Stintzing S
    150. Taverna D
    151. Thibodeau SN
    152. Thomas DC
    153. Trichopoulou A
    154. Tsugane S
    155. Ulrich CM
    156. van Duijnhoven FJB
    157. van Guelpan B
    158. Vijai J
    159. Virtamo J
    160. Weinstein SJ
    161. White E
    162. Win AK
    163. Wolk A
    164. Woods M
    165. Wu AH
    166. Wu K
    167. Xiang Y-B
    168. Yen Y
    169. Zanke BW
    170. Zeng Y-X
    171. Zhang B
    172. Zubair N
    173. Kweon S-S
    174. Figueiredo JC
    175. Zheng W
    176. Marchand LL
    177. Lindblom A
    178. Moreno V
    179. Peters U
    180. Casey G
    181. Hsu L
    182. Conti DV
    183. Gruber SB
    (2019) Novel common genetic susceptibility loci for colorectal cancer
    Journal of the National Cancer Institute 111:146–157.
    https://doi.org/10.1093/jnci/djy099
    1. Šimundić AM
    (2009)
    Measures of diagnostic accuracy: basic definitions
    EJIFCC 19:203–211.

Article and author information

Author details

  1. Peh Joo Ho

    1. Genome Institute of Singapore (GIS), Agency for Science, Technology and Research (A*STAR), Singapore, Singapore
    2. Saw Swee Hock School of Public Health, National University of Singapore and National University Health System, Singapore, Singapore
    3. Department of Surgery, Yong Loo Lin School of Medicine, National University of Singapore, Singapore, Singapore
    Contribution
    Conceptualization, Formal analysis, Visualization, Methodology, Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-3017-4023
  2. Iain BeeHuat Tan

    1. Genome Institute of Singapore (GIS), Agency for Science, Technology and Research (A*STAR), Singapore, Singapore
    2. Saw Swee Hock School of Public Health, National University of Singapore and National University Health System, Singapore, Singapore
    3. Program in Cancer and Stem Cell Biology, Duke-National University of Singapore Medical School, Singapore, Singapore
    4. Division of Medical Oncology, National Cancer Centre Singapore, Singapore, Singapore
    Contribution
    Writing – review and editing, Interpretation of data
    Competing interests
    No competing interests declared
  3. Dawn Qingqing Chong

    1. Division of Medical Oncology, National Cancer Centre Singapore, Singapore, Singapore
    2. Duke-NUS Medical School Singapore, Singapore, Singapore
    Contribution
    Writing – review and editing
    Competing interests
    No competing interests declared
  4. Chiea Chuen Khor

    Genome Institute of Singapore (GIS), Agency for Science, Technology and Research (A*STAR), Singapore, Singapore
    Contribution
    Data curation, Funding acquisition, Writing – review and editing
    Competing interests
    No competing interests declared
  5. Jian-Min Yuan

    1. UPMC Hillman Cancer Center, Pittsburgh, United States
    2. Department of Epidemiology, University of Pittsburgh Graduate School of Public Health, Pittsburgh, United States
    Contribution
    Resources, Data curation, Funding acquisition, Writing – review and editing
    Competing interests
    No competing interests declared
  6. Woon-Puay Koh

    1. Healthy Longevity Translational Research Programme; Yong Loo Lin School of Medicine, National University of Singapore, Singapore, Singapore
    2. Singapore Institute for Clinical Sciences, Agency for Science Technology and Research (A*STAR), Singapore, Singapore
    Contribution
    Resources, Data curation, Funding acquisition, Writing – review and editing
    Competing interests
    No competing interests declared
  7. Rajkumar Dorajoo

    Genome Institute of Singapore (GIS), Agency for Science, Technology and Research (A*STAR), Singapore, Singapore
    Contribution
    Conceptualization, Resources, Data curation, Funding acquisition, Methodology, Writing – original draft, Writing – review and editing
    For correspondence
    dorajoor@gis.a-star.edu.sg
    Competing interests
    received a grant from the Agency for Science, Technology and Research Career Development Award (A*STAR CDA - 202D8090), and from Ministry of Health Healthy Longevity Catalyst Award (HLCA20Jan-0022). The author has no other competing interests to declare
  8. Jingmei Li

    1. Genome Institute of Singapore (GIS), Agency for Science, Technology and Research (A*STAR), Singapore, Singapore
    2. Department of Surgery, Yong Loo Lin School of Medicine, National University of Singapore, Singapore, Singapore
    Contribution
    Conceptualization, Resources, Data curation, Supervision, Funding acquisition, Methodology, Writing – original draft, Writing – review and editing
    For correspondence
    lijm1@gis.a-star.edu.sg
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-8587-7511

Funding

National Research Foundation Singapore (PRECISE)

  • Jingmei Li

National Medical Research Council (NMRC/CSA/0055/2013)

  • Woon-Puay Koh

National Research Foundation Singapore (NRF-NRFI2018-01)

  • Chiea Chuen Khor

Ministry of Health -Singapore (HLCA20Jan-0022)

  • Rajkumar Dorajoo

National Institutes of Health (R01 CA144034)

  • Jian-Min Yuan

National Institutes of Health (UM1 CA182876)

  • Jian-Min Yuan

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank the Singapore Cancer Registry for the identification of incident cancer cases among participants of the SCHS and Siew-Hong Low of the National University of Singapore for supervising the fieldwork of the SCHS.

Ethics

Human subjects: The study was approved by the Institutional Review Boards of the National University of Singapore, University of Pittsburgh, and the Agency for Science, Technology and Research (A*STAR, reference number 2022-042). Written, informed consent was obtained from all study participants.

Version history

  1. Received: August 10, 2022
  2. Preprint posted: September 15, 2022 (view preprint)
  3. Accepted: March 24, 2023
  4. Accepted Manuscript published: March 27, 2023 (version 1)
  5. Version of Record published: May 4, 2023 (version 2)

Copyright

© 2023, Ho et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 695
    Page views
  • 113
    Downloads
  • 2
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Peh Joo Ho
  2. Iain BeeHuat Tan
  3. Dawn Qingqing Chong
  4. Chiea Chuen Khor
  5. Jian-Min Yuan
  6. Woon-Puay Koh
  7. Rajkumar Dorajoo
  8. Jingmei Li
(2023)
Polygenic risk scores for the prediction of common cancers in East Asians: A population-based prospective cohort study
eLife 12:e82608.
https://doi.org/10.7554/eLife.82608

Share this article

https://doi.org/10.7554/eLife.82608

Further reading

    1. Ecology
    2. Epidemiology and Global Health
    Aleksandra Kovacevic, David RM Smith ... Lulla Opatowski
    Research Article

    Non-pharmaceutical interventions implemented to block SARS-CoV-2 transmission in early 2020 led to global reductions in the incidence of invasive pneumococcal disease (IPD). By contrast, most European countries reported an increase in antibiotic resistance among invasive Streptococcus pneumoniae isolates from 2019 to 2020, while an increasing number of studies reported stable pneumococcal carriage prevalence over the same period. To disentangle the impacts of the COVID-19 pandemic on pneumococcal epidemiology in the community setting, we propose a mathematical model formalizing simultaneous transmission of SARS-CoV-2 and antibiotic-sensitive and -resistant strains of S. pneumoniae. To test hypotheses underlying these trends five mechanisms were built into the model and examined: (1) a population-wide reduction of antibiotic prescriptions in the community, (2) lockdown effect on pneumococcal transmission, (3) a reduced risk of developing an IPD due to the absence of common respiratory viruses, (4) community azithromycin use in COVID-19 infected individuals, (5) and a longer carriage duration of antibiotic-resistant pneumococcal strains. Among 31 possible pandemic scenarios involving mechanisms individually or in combination, model simulations surprisingly identified only two scenarios that reproduced the reported trends in the general population. They included factors (1), (3), and (4). These scenarios replicated a nearly 50% reduction in annual IPD, and an increase in antibiotic resistance from 20% to 22%, all while maintaining a relatively stable pneumococcal carriage. Exploring further, higher SARS-CoV-2 R0 values and synergistic within-host virus-bacteria interaction mechanisms could have additionally contributed to the observed antibiotic resistance increase. Our work demonstrates the utility of the mathematical modeling approach in unraveling the complex effects of the COVID-19 pandemic responses on AMR dynamics.

    1. Epidemiology and Global Health
    Olivera Djuric, Elisabetta Larosa ... The Reggio Emilia Covid-19 Working Group
    Research Article

    Background:

    The aim of our study was to test the hypothesis that the community contact tracing strategy of testing contacts in households immediately instead of at the end of quarantine had an impact on the transmission of SARS-CoV-2 in schools in Reggio Emilia Province.

    Methods:

    We analysed surveillance data on notification of COVID-19 cases in schools between 1 September 2020 and 4 April 2021. We have applied a mediation analysis that allows for interaction between the intervention (before/after period) and the mediator.

    Results:

    Median tracing delay decreased from 7 to 3.1 days and the percentage of the known infection source increased from 34–54.8% (incident rate ratio-IRR 1.61 1.40–1.86). Implementation of prompt contact tracing was associated with a 10% decrease in the number of secondary cases (excess relative risk –0.1 95% CI –0.35–0.15). Knowing the source of infection of the index case led to a decrease in secondary transmission (IRR 0.75 95% CI 0.63–0.91) while the decrease in tracing delay was associated with decreased risk of secondary cases (1/IRR 0.97 95% CI 0.94–1.01 per one day of delay). The direct effect of the intervention accounted for the 29% decrease in the number of secondary cases (excess relative risk –0.29 95%–0.61 to 0.03).

    Conclusions:

    Prompt contact testing in the community reduces the time of contact tracing and increases the ability to identify the source of infection in school outbreaks. Although there are strong reasons for thinking it is a causal link, observed differences can be also due to differences in the force of infection and to other control measures put in place.

    Funding:

    This project was carried out with the technical and financial support of the Italian Ministry of Health – CCM 2020 and Ricerca Corrente Annual Program 2023.