ABSTRACT
Purpose
There is great promise in breast cancer risk stratification to target screening and prevention. It is unclear whether adding gene panels to other risk tools improves breast cancer risk stratification and adds discriminatory benefit on a population basis.
Methods
In total, 10,025 of 57,902 women aged 46 to 73 years in the Predicting Risk of Cancer at Screening study provided DNA samples. A case–control study was used to evaluate breast cancer risk assessment using polygenic risk scores (PRSs), cancer gene panel (n = 33), mammographic density (density residual [DR]), and risk factors collected using a self-completed 2-page questionnaire (Tyrer-Cuzick [TC] model version 8). In total, 525 cases and 1410 controls underwent gene panel testing and PRS calculation (18, 143, and/or 313 single-nucleotide polymorphisms [SNPs]).
Results
Actionable pathogenic variants (PGVs) in BRCA1/2 were found in 1.7% of cases and 0.55% of controls, and overall PGVs were found in 6.1% of cases and 1.3% of controls. A combined assessment of TC8-DR-SNP313 and gene panel provided the best risk stratification with 26.1% of controls and 9.7% of cases identified at <1.4% 10-year risk and 9.01% of controls and 23.3% of cases at ≥8% 10-year risk. Because actionable PGVs were uncommon, discrimination was identical with/without gene panel (with/without: area under the curve = 0.67, 95% CI = 0.64-0.70). Only 7 of 17 PGVs in cases resulted in actionable risk category change. Extended case (n = 644)–control (n = 1779) series with TC8-DR-SNP143 identified 18.9% of controls and only 6.4% of stage 2+ cases at <1.4% 10-year risk and 20.7% of controls and 47.9% of stage 2+ cases at ≥5% 10-year risk.
Conclusion
Further studies and economic analysis will determine whether adding panels to PRS is a cost-effective strategy for risk stratification.
Keywords
Introduction
Breast cancer is the most commonly diagnosed female cancer globally. In familial breast cancer, around half of the cases are explained by a known genetic component,
1
, 2
, 3
predominantly by pathogenic variants (PGVs) in BRCA1 or BRCA2 and single-nucleotide polymorphisms (SNPs). SNPs account for a greater proportion of the familial risk component than all PGVs in high- or moderate-risk breast cancer genes.1
, 2
, 3
SNPs also explain a large proportion of inherited genetic risk in women developing sporadic breast cancer and having no familial history of the disease. Therefore, at population level, SNPs are likely to be more informative for risk stratification than screening for moderate- and high-risk gene variants in a gene panel.4
,5
Genotyping for the number of risk-associated breast cancer alleles at each site (ie, 0, 1, 2) gives an individual odds ratio (OR) for each risk SNP, and multiplying the ORs gives a polygenic risk score (PRS).6
Currently, several breast cancer risk prediction models incorporate classical risk factors, including acquired age, family history of breast (and ovarian) cancer, age of menses, first full-term pregnancy (and parity) and menopause, body mass index, type and number of breast biopsies, and use of hormone replacement therapy (dose, type, duration, and time since last use).
7
, 8
, 9
In addition, high mammographic density has been established as an important breast cancer risk factor, and incorporation of mammographic density greatly improves the accuracy of risk prediction models.10
,11
Recent studies have considered the value of including an SNP-based PRS into risk prediction algorithms showing promising results.12
, 13
, 14
, - Vachon C.M.
- Pankratz V.S.
- Scott C.G.
- et al.
The contributions of breast density and common genetic variation to breast cancer risk.
J Natl Cancer Inst. 2015; 107dju397https://doi.org/10.1093/jnci/dju397
15
We collected data for classical breast cancer risk factors and mammographic density of 57,902 women attending screening, who were aged 46 to 73 years at entry to the Predicting Risk of Cancer at Screening (PROCAS) study.
16
More than 10,000 women also provided saliva DNA samples for initial validation of 18-breast cancer susceptibility SNP PRS (SNP18).17
Recently, we showed that by combining mammographic density and SNP18 with the Tyrer-Cuzick (TC) risk prediction model version 7, women aged 46 to 73 years could be accurately divided into four 10-year risk groups (<2%, low; 2%-3.49%, average; 3.5%-4.99%, above average; and ≥5%, moderate/high).16
, 17
, 18
This was further improved using 143-SNP PRS.- McIntosh A.
- Evans D.G.
- et al.
Clinical Guidelines and Evidence Review for the Classification and Care of Women At Risk of Familial Breast Cancer. London: National Collaborating Centre for Primary Care/University of Sheffield. NICE Guideline CG014; 2004 (updated 2006 CG41, 2013/2017 CG184).
https://www.nice.org.uk/guidance/cg164
Date accessed: June 17, 2018
12
However, additional improvements in risk stratification are required to refine groups more precisely and to discriminate the large number at average risk so that appropriate personalized risk reduction/early detection strategies can be put in place. It is possible that a number of heterozygotes for PGVs in actionable breast cancer genes could be erroneously placed in a low-risk group (with potential reduction in screening) in the absence of carrying out a gene panel alongside an SNP PRS analysis. Although undertaking such testing on a population basis just for breast cancer would be expensive, it is certainly possible in the foreseeable future that in high income countries a germline genome could be used to risk stratify across a number of health conditions, including a range of cancers and potentially also cardiovascular disease and diabetes. Until now the addition of a breast cancer panel of moderate and high-risk genes has not been fully evaluated alongside other risk factors. In this article, we report on the evaluation of a combined risk derived from standard risk factors, mammographic density, PRS of up to 313 SNPs, and a gene panel in the PROCAS study and assess the potential added value of the gene panel.Materials and Methods
A total of 57,902 women aged 46 to 73 years from the Greater Manchester area were recruited to the PROCAS study, between October 2009 and June 2015, at the time of their mammography screening within the National Health Service Breast Screening Programme. Standard breast cancer risk factors were collected using a self-completed 2-page questionnaire. Saliva samples were collected from 10,025 women after their prevalent study mammogram on drop-in days at sites across Greater Manchester. In addition, saliva was collected from women with breast cancer (invasive or ductal carcinoma in situ [DCIS]) diagnosed before or after recruitment to the study. We call women with breast cancer diagnosed at or after entry to the cohort prospective cases in the analysis.
Saliva samples were collected to extract DNA for SNP genotyping and further genetic research. DNA samples were stored at –20 °C. The 18 SNPs from the study by Turnbull et al
19
were genotyped as previously described,12
,17
using a custom designed Sequenom MassARRAY iPLEX assay or TaqMan SNP Genotyping Assay. SNP143 was derived using a custom Illumina genotyping platform, which was specifically designed for the Collaborative Oncological Gene-Environment Study consortium (OncoArray: https://epi.grants.cancer.gov/gameon/) (Supplemental Table 1). Samples were assayed in 2 batches and included 31 internal controls. Genotyping and quality control were performed as previously described.12
SNP313 was also derived from the OncoArray but required additional imputations of SNPs as described previously.14
Per-allele OR were derived from published OR and allele frequency as described previously by normalizing around a relative risk of 1.0.- Vachon C.M.
- Pankratz V.S.
- Scott C.G.
- et al.
The contributions of breast density and common genetic variation to breast cancer risk.
J Natl Cancer Inst. 2015; 107dju397https://doi.org/10.1093/jnci/dju397
12
,17
In brief, PRS for SNP18 and SNP143 were calculated by multiplying each allele’s OR for each SNP (when a single SNP failed, a score of 1.0 was given). SNP313 was derived from the Breast Cancer Association Consortium data set.14
Each PRS was used in further statistical analyses. All women who self-reported that they were not of White European origin, including Ashkenazi Jewish, were excluded from the assessment of SNP PRS because we have previously identified (data not presented) that each PRS is not well-calibrated for these ethnicities in our cohort.- Vachon C.M.
- Pankratz V.S.
- Scott C.G.
- et al.
The contributions of breast density and common genetic variation to breast cancer risk.
J Natl Cancer Inst. 2015; 107dju397https://doi.org/10.1093/jnci/dju397
A custom panel of genes was screened in the Breast Cancer Risk after Diagnostic Gene Sequencing (BRIDGES) study.
20
The BRIDGES study performed sequencing of 33 genes (Supplemental Table 2).20
Although BRIDGES only directly classified truncating variants as pathogenic, we identified and included all missense variants in the main positive breast and breast/ovarian cancer-associated study genes (ATM, BARD1, BRCA1, BRCA2, CDH1, CHEK2, NF1, PALB2, PTEN, RAD50, RAD51C, RAD51D, and TP53) and assessed pathogenicity using American College of Medical Genetics and Genomics/Association for Molecular Pathology guidelines.21
The following ORs were used when a PGV was identified: OR = 6 for BRCA1, BRCA2, and TP53; OR = 5 for PTEN, PALB2, STK11, and CDH1; OR = 2 for NF1, ATM, CHEK2, RAD51C, RAD51D, and BARD1, reflecting the lower end of the 95% CI from BRIDGES.20
A reduced risk of 0.965 was applied to those testing negative for all PGVs. The reduced risk for nonheterozygotes was calculated to provide an overall OR of 1.0 once the ORs from PGVs were included in controls. All available cases as of September 2016 (n = 340) and 4 to 5 unaffected controls matched for age (±12 months), date (within 1 month), and entry mammogram type (analog/digital) underwent gene panel testing (Table 1).Table 1Risk factors in cases and controls who underwent or did not undergo panel testing
| Attribute | Breast Cancers With Panel | Breast Cancers Without Panel | P Value | Controls With Panel | Controls Without Panel | P Value |
|---|---|---|---|---|---|---|
| Number | 340 | 281 | 1410 | 7026 | ||
| Median (IQR) age at entry, y | 60.0 (54.0-64.8) | 59.8 (53.0-65.0) | .823 | 60.0 (53.8-65.0) | 59.0 (53.0-65.0) | .542 |
| Median (IQR) age at cancer, y | 62.0 (54.6-66.8) | 63.7 (58.8-69.5) | <.001 | – | – | |
| FDR with breast cancer, n (%) | 58 (17.1) | 52 (18.5) | .716 | 194 (13.8) | 937 (13.3) | .702 |
| NICE criteria, n (%) | 26 (7.6) | 19 (6.8) | .789 | 69 (4.9) | 344 (4.9) | .949 |
| Personal history of ovarian cancer | 0 | 0 | 2 (0.14%) | 10 (0.14%) | 1.0 | |
| FDR ovarian cancer | 6 (1.8%) | 9 (3.2%) | .29 | 40 (2.8%) | 150 (2.1%) | .115 |
| Median (IQR) BMI kg/m2 | 26.2 (23.6-30.2) | 27.2 (24.1-30.5) | .122 | 25.9 (23.2-29.9) | 26.1 (23.4-29.8) | .246 |
| Missing BMI, n (%) | 25 (7.4) | 23 (8.2) | 58 (4.1) | 382 (5.4) | ||
| Median (IQR) DR | 1.08 (0.84-1.34) | 1.11 (0.89-1.38) | .306 | 0.94 (0.75-1.17) | 0.95 (0.75-1.19) | .488 |
| Median (IQR) TC8 | 3.18 (2.59-4.44) | 3.34 (2.49-4.67) | .592 | 3.06 (2.46-4.03) | 3.06 (2.48-4.02) | .901 |
NICE criteria are based on NICE algorithm. BMI, body mass index; DR, density residual; FDR, first-degree relative; IQR, interquartile range; NICE, National Institute for Health and Care Excellence; TC, Tyrer-Cuzick.
The probability of BRCA1/2 PGV was determined using the Manchester scoring system for each affected (cases only) individual.
22
In brief, each cancer in the family that included a pathology adjustment for the proband was scored, with Manchester scoring system score of 15 to 19 equating to the National Institute for Health and Care Excellence (NICE) 10% threshold for testing in the United Kingdom.18
- McIntosh A.
- Evans D.G.
- et al.
Clinical Guidelines and Evidence Review for the Classification and Care of Women At Risk of Familial Breast Cancer. London: National Collaborating Centre for Primary Care/University of Sheffield. NICE Guideline CG014; 2004 (updated 2006 CG41, 2013/2017 CG184).
https://www.nice.org.uk/guidance/cg164
Date accessed: June 17, 2018
Mammographic density was estimated by 2 readers using a visual analog scale (VAS), as previously described.
10
,11
Density was adjusted for body mass index and age and reported as density residual (DR), essentially an adjusted predictive OR for density in combination with other factors. A total of 11 women with bilateral breast cancer on prevalent study screen or breast implants had no assessable VAS score and were excluded.Ten-year risk based on questionnaire risk factors was estimated using the TC model. This included age, age at first full-term pregnancy, weight (in kilograms), height (in meters), number, and age of breast cancer diagnoses in affected first- and second-degree relatives, and previous breast biopsy. Clinical end points examined were obtained from histopathology reports, which included invasive breast cancer, DCIS, and invasive tumor grade and stage.
Statistical methods
Primary analysis considered all participants with a gene panel. Secondary analysis used a larger number set with SNP143 PRS. SNP313 was the primary PRS of interest because this has an abundance of data external to this study to support it. We considered SNP143 and SNP18 as well because we have previously reported their performance in related analysis. For SNP143 and SNP18, predefined per-allele risks were from an independent meta-analysis.
12
Projected 10-year risk was evaluated assuming independence between the factors, as justified by earlier analysis in a wider cohort.
12
The main focus was on high-risk (>8% 10-year risk) and moderate-risk (5%-7.9%) groups, which are clinically relevant in the United Kingdom, and the lowest-risk group (<1.4%). Area under the curve (AUC) statistics measured discrimination. Differences between 2 categorical variables were tested using χ2 test and 2-sample Wilcoxon tests for continuous data.Results
Of 57,902 women comprising the full PROCAS study, 10,025 with saliva DNA samples represent a subset. Women who provided DNA samples were slightly older (mean age 60.0 years vs 56.9 years; P < .001) and had a greater family history (first-degree relative n = 194 [13.8%] vs n = 6016 [11.6%] P < .0001) and overall TC8 risk than those who did not (mean 10-year risk of 3.06% vs 2.80%; P < .0001) (Supplemental Table 3).
DNA cohort
Of the 10,025 women with saliva DNA samples (aged 46-73 years at entry), 553 had been diagnosed with breast cancer before entry to PROCAS study (Supplemental Figure 1). Of the remaining 9472 women unaffected at study entry, 270 diagnosed at the time of their mammography on entry to PROCAS were prospective cases, with a further 401 being diagnosed after their mammography with a median follow up of 9.86 years. Although 415 women without breast cancer were excluded from the full cohort on the basis of ethnicity, only 27 women with breast cancer and 173 women without breast cancer within the SNP-OncoArray genotyping subcohort were excluded because they self-reported as not White European, and the additional 215 excluded women only had SNP18 performed. SNP143 for these women showed a mean and median PRS well above 1 for controls with unrealistically high 10-year combined risks (Supplemental Table 4). As part of the Breast Cancer Association Consortium OncoArray study, 340 White European prospective cancer cases and 1410 controls were selected in 2016 to undergo SNP OncoArray
14
(SNP313) and underwent a sequencing of 33 genes as part of the BRIDGES study- Vachon C.M.
- Pankratz V.S.
- Scott C.G.
- et al.
The contributions of breast density and common genetic variation to breast cancer risk.
J Natl Cancer Inst. 2015; 107dju397https://doi.org/10.1093/jnci/dju397
20
(Tables 2 and 3). An additional 304 White Europeans with prospective breast cancer (all breast cancers, most diagnosed after 2016) and 379 without breast cancer were analyzed using OncoArray for SNP143. The main combined analysis was based on women without cancer at entry, which consisted of 340 cases who developed breast cancer and 1410 controls.Table 2Actionable moderate and high-risk genes in PROCAS breast cancer cases by MSS
| MSS | Number | BRCA1/BRCA2 | % | PALB2 | % | CHEK2/ATM | % | PTEN/TP53 NF1 | % |
|---|---|---|---|---|---|---|---|---|---|
| MSS >15 | 29 | 3 | 10.3 | 2 | 6.9 | 0 | 0.0 | 0 | 0.0 |
| MSS 10-14 | 50 | 1 | 2.0 | 0 | 0.0 | 1 | 2.0 | 0 | 0.0 |
| MSS <10 | 446 | 5 | 1.1 | 2 | 0.45 | 15 | 23.4 | 3 | 0.7 |
| Total | 525 | 9 | 1.7 | 4 | 0.8 | 16 | 23.0 | 3 | 0.6 |
All PGVs were truncating except PTEN c.367C>T (p.His123Tyr) 3 times on ClinVar as likely pathogenic/pathogenic and TP53 c.455C>T (p.Pro152Leu) multiple times on ClinVar as pathogenic and 2 in ATM, c.8122G>A; p.(Asp2708Asn) and c.9022C>T; p.(Arg3008Cys) both also pathogenic on ClinVar.
MSS, Manchester Scoring System; PGV, pathogenic variant; PROCAS, Predicting Risk of Cancer at Screening.
Table 3Increased identification of low risk (<1.4% 10-year risk), moderate/high risk (≥5% 10-year risk), and high risk (≥8% 10-year risk) by addition of DR, SNP PRS (18,143,313), and a gene panel to TC8 in 340 breast cancer cases and 1410 controls
| 10-Year Risk Level | TC8 | TC8-DR | TC8-DR-SNP18 | TC8-DR-SNP143 | TC8-DR-SNP313 | TC8-DR-SNP143 + Gene Panel | TC8-DR-SNP313+ Gene Panel |
|---|---|---|---|---|---|---|---|
| <1.4%Controls | 7 0.50% (0.02-1.02) | 66 4.68% (3.64-5.91) | 158 11.21% (9.61-12.97) | 273 19.36% (17.33-21.52) | 351 24.89% (22.45-27.02) | 288 20.43% (18.35-22.63) | 370 26.24% (23.82-28.46) |
| <1.4% Cases | 1 0.29% (0.01-1.63) | 10 2.94% (1.41-5.34) | 18 5.29% (3.17-8.24) | 28 8.24% (5.54-11.68) | 31 9.12% (6.28-12.69) | 28 8.24% (5.54-11.68) | 33 9.71% (6.78-13.36) |
| Cases-to-controls ratio | 0.59 (0.07-4.80) | 0.63 (0.33-1.21) | 0.47 (0.29-0.76) | 0.43 (0.29-0.62) | 0.37 (0.26-0.52) | 0.40 (0.28-0.58) | 0.37 (0.26-0.52) |
| ≥5%Controls | 200 14.18% (12.40-16.12) | 261 18.51% (16.52-20.64) | 278 19.72% (17.67-21.89) | 281 19.93% (17.87-22.11) | 305 21.63% (19.58-23.95) | 274 19.43% (17.40-21.60) | 286 20.28% (18.14-22.41) |
| ≥5% Cases | 64 18.82% (14.81-23.39) | 95 27.94% (23.24-33.04) | 113 33.24% (28.25-38.52) | 148 43.53% (38.19-49.98) | 144 42.35% (37.04-47.80) | 149 43.82% (38.48-49.28) | 147 43.24% (37.90-48.69) |
| Cases-to-controls ratio | 1.33 (1.03-1.71) | 1.51 (1.23-1.85) | 1.69 (1.40 -2.03) | 2.18 (1.86-2.56) | 1.96 (1.67-2.30) | 2.26 (1.92-2.65) | 2.13 (1.82-2.50) |
| OR ≥5% vs <1.4% | 2.24 (0.27-18.55) | 2.40 (1.18-4.86) | 3.57 (2.09-6.09) | 5.14 (3.32-7.95) | 5.35 (3.52-8.11) | 5.59 (3.62-8.65) | 5.76 (3.83-8.67) |
| Overall AUC (95% CI) | 0.536 (0.502-0.571) | 0.599 (0.565-0.632) | 0.634 (0.602-0.667) | 0.677 (0.646-0.709) | 0.665 (0.633-0.697) | 0.684 (0.652-0.715) | 0.672 (0.641-0.704) |
| ≥8%Controls | 37 2.62% (1.85-3.50) | 80 5.67% (4.52-7.01) | 102 7.23% (5.94-8.71) | 116 8.23% (6.85-9.79) | 126 8.94% (7.50-10.55) | 116 8.23% (6.85-9.79) | 127 9.01% (7.56-10.62) |
| ≥8% Cases | 16 4.71% (2.71-7.53) | 29 8.53% (5.79-12.02) | 35 10.29% (7.28-14.03) | 57 16.76% (12.95-21.17) | 75 22.06% (17.76-26.85) | 63 18.53% (14.54-23.07) | 80 23.53% (19.12-28.41) |
| Cases:controls ratio | 1.79 (1.01-3.19) | 1.50 (1.00-2.26) | 1.42 (0.99-2.05) | 2.04 (1.52-2.73) | 2.47 (1.90-3.20) | 2.25 (1.70-2.99) | 2.61 (2.03-3.37) |
| OR ≥8% vs <1.4% | 3.03 (0.34-26.66) | 2.39 (1.09-5.27) | 3.01 (1.62-5.60) | 4.79 (2.90-7.91) | 6.74 (4.23-10.73) | 5.59 (3.41-9.16) | 7.06 (4.49-11.11) |
AUC, area under the receiver operating characteristic curve; DR, density residual; OR, odds ratio; PRS, polygenic risk score; SNP, single-nucleotide polymorphism; TC, Tyrer-Cuzick.
Panel testing
Among the 525 women with breast cancer undergoing panel testing (Modertae And high risk Gene Panel [MAGP]) (range = 28.9-78.6 years; median = 59.0 years; interquartile range [IQR] = 52.8-65.0) (Table 2, Supplemental Tables 5 and 6), 9 (1.7%) BRCA1/2 PGVs were identified (1 in BRCA1 and 8 in BRCA2) with only BRCA2 (n = 5) in the prospective cohort comprising patients who had cancers and were aged 46 to 79 years. In addition, there were 20 (3.8%) PGVs in PALB2 (n = 4), ATM (n = 6), and CHEK2 (n = 10) and 1 each in PTEN, NF1, and TP53. Of 9 BRCA1/2 PGVs, only 3 (33%) were found using NICE criteria including the recent extended testing of all triple negative breast cancers to age 60 years. An additional 2 PALB2 PGVs were identified using NICE criteria, but not the 16 ATM/CHEK2 PGVs or the 3 PGVs were identified in syndromic genes. Neither of the syndromic cases were known to the Manchester Centre for Genomic Medicine including the highly ascertained NF1 genetic register nor were they known to have clinical features supportive of these diagnoses.
23
None of the individuals with a BRCA PGV were known to the Manchester Centre for Genomic Medicine. In 1410 women without breast cancer at mean age 69.2 years (median = 70.0 years; IQR = 63.3-74.4), there were 2 BRCA1, 6 BRCA2, 3 PALB2, 4 ATM, and 3 CHEK2 PGVs (total = 18/1410 [1.3%], BRCA = 8/1410 [0.57%]). There were no PGVs identified in RAD51C, RAD51D, STK11, or CDH1 in cases or controls, but 1 control had a BARD1 PGV. In total, 4 women with BRCA2 PGVs had first-degree relatives and 1 had a second-degree relative with breast cancer. Both women with BRCA1 and BRCA2 had no breast/ovarian cancer family history and none had a family history consistent with being at high-risk in accordance with NICE criteria. None of the controls or their relatives would have qualified for testing under the National Health Service in the United Kingdom. Although 71 of 340 (20.9%) prospective cases were carcinoma in situ, 5 of 71 (7% [3/40] high-grade carcinoma in situ) had a PGV (1 in BRCA2, 1 in PALB2, 2 in CHEK2, and 1 in ATM) slightly higher than the rate in invasive cases, ie, 14 of 269 (5.2%).PRS
Combined 10-year risks are shown in Table 3 for cases and controls with TC8: TC8 with DR (TC8-DR) and the addition of SNP18, SNP143, and SNP313 and finally the gene panel (MAGP) (this excludes 185 prevalent cases). As can be seen, the proportion of controls (population) at ≥5% 10-year risk rose from around 1 in 7 (14%) with TC8 alone to around 20% when DR was added. Little further increase was seen when SNPs were added. However, the proportion of cases identified at ≥5% 10-year risk rose above 40% when SNP143 and SNP313 PRS were added from only 19% and 28% with TC8 and TC8-DR, respectively. Importantly, there was an incremental rise in the ratio of cases to controls, indicating that high SNP scores were particularly useful in identifying high-risk individuals. In addition, the addition of more SNPs was particularly valuable in identifying those at potentially actionable low risk (<1.4%). Very few were identified using TC8 alone (0.5%), which rose to 4.7% when DR was added but increased substantially to 26.1% in the full assessment with the TC8-DR-SNP313 panel. Most of this difference could be attributed to the PRS with little additional proportional benefit added by the MAGP. Importantly, despite more controls being identified, there was a lower ratio of cancers per control. The best overall ratio between the <1.4% group and those at ≥5% and ≥8% 10-year risk was for the full assessment group with an OR of 5.76 and 7.06, respectively.
A small number of women appear to benefit from molecular diagnosis via the gene panel (MAGP), although the only significant difference between the <5% and ≥5% categories was related to SNP313 controls reducing from 0.216 to 0.201 (difference = 0.014; 95% CI = 0.005-0.023; P = .002) (Supplemental Table 7). Of 340 breast cancer cases, 17 (5%) diagnosed after entry had an actionable PGV identified, and most of these were already in an actionable high-/moderate-risk group (Table 4). One CHEK2 PGV heterozygote moved out of the low-risk group, and 6 other PGV heterozygotes moved up into the actionable high-/moderate-risk group, with 10 in the TC8-DR-SNP143 group and 9 in the TC8DRSNP313 group already being at that level. Of the controls, a lower proportion were already in the actionable moderate-/high-risk group (7/19 [37%]). However, 1 BRCA1 heterozygote with no family history would have had screening reduced by application of TC8-DR-SNP313 and TC8-DR-SNP143 as well as a CHEK2 and ATM PGV heterozygote with no family history. A single PALB2 PGV heterozygote was also in the low-risk category with TC8-DR-SNP313. The lower combined 10-year risk was particularly evident on TC8-DR-SNP313 for moderate-risk genes in controls (ATM/CHEK2) with a mean of OR = 4.93 (median = 3.97; IQR = 1.58-7.94) compared with cases (mean = 6.50; median = 6.11; IQR = 4.00-8.91). Overall, for breast cancer identification, the best performing combination was TC8-DR-SNP143 + MAGP with an AUC of 0.680 (95% CI = 0.652-0.715). However, SNP313 appeared better at identifying women in the tails of the distribution with a case-to-control ratio of 7.07 when assessing the <1.4% vs ≥8% groups. All additions of PRS, including SNP18, significantly improved the AUC compared with TC8-DR, although additions of MAGP panels did not significantly improve the AUC compared with PRS because power was limited owing to MAGP being a relatively rare factor. We also assessed the movement of women across risk groups to join an actionable group (Supplemental Table 8). When comparing TC8-DR with the TC8-DR-SNP313 group in breast cancer cases, 74 women moved up into the actionable moderate/high-risk group with 22 moving down (discordant-pairs, P < .0001). In the high-risk (≥8% 10-year risk) group, 57 women moved up, whereas only 5 moved down (P < .0001). On the contrary, 29 women moved down into the low-risk group, whereas 6 of 10 identified by TC8 as low risk moved upward (P < .0001).
Table 4Movement of moderate- and high-risk gene groups by addition of PGVs to and from actionable risk categories by addition of a gene panel to TC8DRSNP143 and TC8DRSNP313
| Changes in Risk Groups | Breast Cancers (n = 340) | Controls (n = 1410) | ||||||
|---|---|---|---|---|---|---|---|---|
| TC8-DR-SNP143-genes | TC8-DR-SNP313 | TC8-DR-SNP143+ | TC8-DR-SNP313+ | |||||
| <1.4% to average | 1 | CHEK2 | 1 | CHEK2 | 2 | 1 CHEK2 1 ATM | 2 | 1 CHEK2 1 ATM |
| <1.4% to moderate | 0 | 0 | 1 | 1 BRCA1 | 2 | 1 BRCA1 1 PALB2 | ||
| Below average 1.4%-1.99% to high | 0 | 0 | 3 | 1 BRCA2 2 PALB2 | 0 | |||
| Average to moderate | 1 | NF1 | 1 | NF1 | 0 | 0 | ||
| Average to high | 4 | 3 BRCA2 1 TP53 | 3 | 2 BRCA2 1 PALB2 | 2 | 1 BRCA1 1 PALB2 | 3 | 1 BRCA2 2 PALB2 |
| Above average to moderate | 1 | ATM | 0 | 1 | ATM | 1 | ATM | |
| Above average to high | 1 | CHEK2 | 3 | 1 BRCA2 1 TP53 1 CHEK2 | 2 | 1 BRCA2 1 ATM | 3 | 2 BRCA2 1 ATM |
| Moderate or high already | 10 | 2 BRCA2 3 ATM 4 CHEK2 1 PALB2 | 9 | 2 BRCA2 3 ATM 4 CHEK2 | 7 | 4 BRCA2 1 CHEK2 2 ATM | 7 | 1 BRCA1 3 BRCA2 1 CHEK2 2 ATM |
DR, density residual; PGV, pathogenic variant; SNP, single-nucleotide polymorphism; TC, Tyrer-Cuzick.
a One ATM did not change to an actionable category from average risk.+1 CHEK2 did not change to an actionable category from average risk.
Extended analysis
We then analyzed an extended set of the full prospective breast cancer cases (n = 644) and additional controls (n = 1779) occurring in White Europeans using SNP143. Results were similar to the smaller case–control study. In particular, for around 19% of the controls at low risk (<1.4% 10-year risk), there were only 7.4% of cancer cases and only 6.4% of cancers at stage 2 or greater (stage 2+). In contrast, for only 20.8% of the controls, TC8-DR-SNP143 (≥5% 10-year risk) identified 47.9% of stage 2+ cancers (Table 5, Supplemental Figures 2-4).
Table 5Extended case–control study of 644 cases and 1797 controls using TC8-DR-SNP143
| TC8-DR-SNP143 10 Year Risk | Total | BC % BC (Risk Group) (Percentage of All BC) | No BC (Percentage of Controls) | Stage 2+ (Percentage of All) | DCIS (Percentage of BC in Risk Group) | Grade 1 Invasive (Percentage of BC in Risk Group) |
|---|---|---|---|---|---|---|
| <1% | 168 | 15 8.92 (2.33) | 153 (8.60) | 2 (1.2) | 3 (20.0) | 4 (26.7) |
| 1%-2% | 585 | 98 16.75 (15.2) | 487 (27.37) | 21 (12.2) | 19 (19.4) | 16 (16.) |
| 2%-3.5% | 676 | 149 22.04 (23.14) | 527 (29.62) | 36 (20.93) | 23 (15.4) | 30 (19.0) |
| 3.5%-5% | 350 | 108 30.86 (16.77) | 242 (13.60) | 32 (18.6) | 22 (20.4) | 17 (15.45) |
| 5%-7.9% | 358 | 148 41.34 (22.98) | 210 (11.90) | 46 (26.7) | 24 (16.2) | 21 (13.5) |
| >8% | 286 | 126 44.06 (19.57) | 160 (8.99) | 38 (21.84) | 22 (17.46) | 19 (14.4) |
| Total | 2423 | 644 | 1779 | 169 | 113 | 107 (16.0) |
| <1.4 | 385 | 48 12.24 (7.45) | 337 (18.94) | 9 (5.5) | 10 (20.83) | 10 (9.4) |
Stage 2+: 22 of 113 (19.1%) of <2% cancers versus 84 of 274 (30.3%) of ≥5%, P = .011.
DCIS: 22 of 113 (19.5%) of <2% cancers versus 46 of 274 (16.8%) of ≥5%, P = .45.
Grade 1: 20 of 113 (17.5%) of <2% cancers versus 40 of 274 (14.6%) of ≥5%, P = .45.
Grade 1 or DCIS: 42 of 115 (37.2%) of <2% cancers versus 86 of 274 (31.4%) of ≥5%, P = .19.
Grade 1 or DCIS: 20 of 48 (41.7%) of <1.4% cancers versus 86 of 274 (31.4%) of ≥5%, P = .13.
BC, Breast cancer; DCIS, ductal carcinoma in situ; DR, density residual; SNP, single-nucleotide polymorphism; TC, Tyrer-Cuzick.
High-stage cancers were more frequent as a proportion of all cancers in the ≥5% (84/274 [30.3%]) group than in the <2% group (22/113 [19.1%], difference = 11.4%; 95% CI = 1.15-19.9; P ≤ .011). In contrast, there was weak evidence that DCIS and grade 1 cancers were more frequent as a proportion of all cancers in the lower risk group (42/115 [37.2%]) than in the higher risk group (86/274 [31.4%]), although the difference (5.8%; 95% CI = 0.34-1.82; P = .174) was not significant.
Discussion
This study has shown the potential impact of using all available risk stratification tools currently used in risk algorithms to stratify breast cancer risk in a case–control study from a large screening cohort. To our knowledge, all the available tools have not been used together within a published report of a prospective cohort. In particular, although panel testing of high-risk genes associated with breast cancer does make a difference for a small group of women and does add discrimination, it appeared to have a much less substantial effect than using a PRS on the proportion of women in the population in actionable groups and in those diagnosed with cancer. Of the prospective breast cancer cases, 5% had an actionable PGV, but most of these were in the moderate-risk CHEK2 and ATM genes, which confer only a 2- to 3-fold relative risk.
20
,24
Only 1.7% had a BRCA1 or BRCA2 PGV in the overall 525 cases, which included breast cancer cases diagnosed from age 29 years and had an age structure fairly close to population testing age. The lower OR of approximately 3.0 between cases and controls than that observed in other reports potentially reflects the higher proportions of cases at older ages than in many previous studies. In contrast, the proportion of cancers in the ≥5% 10-year risk category rose from 29% using TC8 with density to 42.6% when SNP143 was added. SNP143 added substantially to SNP18, but further gains were only seen when using SNP313 for ≥8%. Nonetheless, the highest ratio of cases to controls in the ≥8% vs <1.4% group was for the TC8-DR-SNP313 + MAGP at 7.06 (95% CI = 4.49-11.11), whereas ratios for TC8 and TC8-DR/TC8-SNP18 were only around 3.0 (Table 3). Clearly, additional SNPs add more value than a gene panel (MAGP) in improving discrimination at the top and bottom end of the risk spectrum.Interestingly, a number of large studies in unselected Western populations have shown that only 1.8% to 2.6% of breast cancers have a BRCA1/2 PGV.
20
,24
,25
If resources were limited for risk stratification, then PRS should be prioritised as it clearly added more in a population setting than a gene panel (MAGP), and SNP143 was as informative as SNP313 overall, although less informative at identifying the important tails of the distribution. Nonetheless, with the cost of genetic testing dropping constantly, the addition of a panel combined with PRS genotyping may still be optimal. Certainly, very few of even the BRCA1/2 PGV heterozygotes would have been identified in routine practice with only 3 of 9 cancers and none of the 8 controls meeting testing criteria even in the extended family.18
A number of publications are now advocating population testing of women for breast cancer genes.- McIntosh A.
- Evans D.G.
- et al.
Clinical Guidelines and Evidence Review for the Classification and Care of Women At Risk of Familial Breast Cancer. London: National Collaborating Centre for Primary Care/University of Sheffield. NICE Guideline CG014; 2004 (updated 2006 CG41, 2013/2017 CG184).
https://www.nice.org.uk/guidance/cg164
Date accessed: June 17, 2018
26
However, if this were to be implemented, it would logically need to occur before most population screening programs commence, ie, at around age 50 years. This is because much of the risk in BRCA1/2 heterozygotes is before age 50 years and the risk of ovarian cancer for BRCA1 heterozygotes also predates 50 years. Even if early testing of breast cancer genes is implemented in the population at or before age 30 years, there would still be a hiatus for those of screening age, and testing could be added in both populations alongside PRS.In this study, we found a slightly higher rate of PGVs in prospective DCIS cases (5/71 [7%]) than in those with invasive breast cancer (14/269 [5.2%]), but with wide uncertainty. There have been very few studies assessing the detection rate using gene panels (MAGP) in isolated DCIS. One study of 665 DCIS <50 years found a PGV in a 5 gene panel of BRCA1/2, PALB2, CHEK2, and TP53 in 7.7%.
27
This rate could have been higher with a larger gene panel and most notably omitted ATM. As such, DCIS should essentially be treated as invasive breast cancer in determining eligibility for panel testing because this will affect treatment to manage contralateral and ovarian cancer risks. These risks can now be evaluated in the CanRisk model.9
There is some concern in Europe with regard to risk stratification in breast screening programs regarding the need to pay for additional screening and therefore potentially reducing screening in a low-risk group. We selected a <1.4% 10-year risk because it is below the risk for average women aged 40 years,
12
who are not offered screening for 10-year risk in most countries. Implementing a reduction in screening at this level of risk will require persuading policy makers, those in charge of screening programs, and women who would no longer be offered mammography as a result that this is safe. Low-risk does not mean no risk, rather a reliably low-risk for a woman presenting with a potentially lethal breast cancer at an advanced stage.28
We have previously observed in this cohort that those in the low-risk group using TC8-DR-SNP18 were less likely to have stage 2 cancers, and overall cancer occurrences were more likely to have favorable pathology, including a high rate of slower growing grade 1 cancers.29
The analysis in this article, in a larger assessment, also found that SNP143 identified a low-risk group with a low proportion of stage 2 cancers.We have now expanded on the utility of a model including a continuous measure of density (DR-VAS) that we have previously shown is superior to 3 automatic measures (Volpara, Densitas, and Quantra) as well as the visual thresholding method Cumulus.
30
With a prospective follow up of median 9.9 years and now 60% of cancers occurring postprevalence mammogram, this adds to our initial assessments that largely predicted risks in prevalent PROCAS cases.10
,11
Importantly, the use of a continuous scale has advantages over the Breast Imaging Reporting and Data System classification system, which creates artificially large changes in risks between categories, reducing successful incorporation in prediction models.31
We have shown that artificial intelligence methods can learn density assessment from reader VAS and that they are as predictive of breast cancer risk.32
, - Ionescu G.V.
- Fergie M.
- Berks M.
- et al.
Prediction of reader estimates of mammographic density using convolutional neural networks.
J Med Imaging (Bellingham). 2019; 6031405https://doi.org/10.1117/1.JMI.6.3.031405
32
Other deep learning methods for assessing mammographic density also show promise for risk prediction.- Ionescu G.V.
- Fergie M.
- Berks M.
- et al.
Prediction of reader estimates of mammographic density using convolutional neural networks.
J Med Imaging (Bellingham). 2019; 6031405https://doi.org/10.1117/1.JMI.6.3.031405
33
,- Haji Maghsoudi O.
- Gastounioti A.
- Scott C.
- et al.
Deep-LIBRA: an artificial-intelligence method for robust quantification of breast density with independent validation in breast cancer risk assessment.
Med Image Anal. 2021; 73102138https://doi.org/10.1016/j.media.2021.102138
34
There are some limitations to this study. We did not carry out assessments for large rearrangements that can account for up to 20% of BRCA1 PGVs in our population,35
but for the other main breast cancer genes, large rearrangements are generally below 5%. Not all cancers had a gene panel assayed, and there were some differences between those that were tested with the MAGP panel and those not tested. The TC8 model was not very discriminatory between breast cancer cases and controls among those who had panel testing, and this was partly because the controls who provided DNA samples had higher TC8 scores than those who did not (P < .0001), whereas this was less evident for cases (P = .024). We also excluded those who were not White European in origin, including Ashkenazi Jewish. We have previously shown that PRS in these populations need to be recalibrated, although there still appears to be some discrimination between cases and controls.36
In conclusion, this study has shown good risk stratification from a combined TC8-DR-SNP143/313 model with or without an added gene panel. Whereas PRS adds usefully to the risk stratification compared with a model based on density and TC8 alone, gene panels appear to add less, and most of the PRS information was in SNP143. However, for the individual women identified with PGVs in moderate- or high-risk genes, the implications for management may be substantial. Further studies are required to determine whether current gene panels, which can only identify PGVs in around 1.7% of women of screening age without cancer, are cost-effective for use in screening programs.
Data Availability
All raw data are available for review on request.
Conflict of Interest
The authors declare no conflicts of interest.
Acknowledgments
The authors would like to thank the women who agreed to take part in the Predicting Risk of Cancer at Screening study and other members of the Predicting Risk of Cancer at Screening group, including the study radiologists, advanced radiographic practitioners, and study staff, for recruitment and data collection.
This work was supported by the NIHR under its Programme Grants for Applied Research Programme (reference number RP-PG-0707-10031: “Improvement in risk prediction, early detection and prevention of breast cancer”) and the Prevent Breast Cancer (references GA10-033 and GA13-006). D.G.R.E, E.F.H, S.M.A., S.J.H., A.J.M., W.G.N., H.B., and A.H. are supported by the NIHR Manchester Biomedical Research Centre (IS-BRC-1215-20007). The sequencing and analysis for this project was funded by the European Union’s Horizon 2020 Research and Innovation Programme (Breast Cancer Risk after Diagnostic Gene Sequencing: grant number 634935) and the Wellcome Trust (grant no: v203477/Z/16/Z). Breast Cancer Association Consortium co-ordination was additionally funded by the European Union’s Horizon 2020 Research and Innovation Programme (Breast Cancer Risk after Diagnostic Gene Sequencing: grant number 634935, BCAST: grant number 633784) and by Cancer Research UK (C1287/A16563). The views expressed in this article are those of the authors and not necessarily those of Prevent Breast Cancer, the National Health Service, NIHR, or the Department of Health.
Author Information
Conceptualization: D.G.R.E., A.H.; Data Curation: E.F.H., A.R.B., S.S.; Formal Analysis: E.F.H., A.R.B., D.G.R.E.; Writing-original draft: D.G.R.E.; Writing-review and editing: D.G.R.E., E.M.v.V., E.F.H., A.R.B., S.M.A., H.B., E.R.W., S.S., J.S., S.J.H., A.J.M., W.G.N., J.C., A.H.
Ethics Declaration
All ethical standards were followed. The study was approved by the North Manchester Research Ethics Committee (ref. 09/H1008/81). Informed consent was obtained from all individual participants included in the study.
Supplementary Material
- Supplemental Figure and Table
References
- Critical research gaps and translational priorities for the successful prevention and treatment of breast cancer.Breast Cancer Res. 2013; 15: R92https://doi.org/10.1186/bcr3493
- Genome-wide association analysis of more than 120,000 individuals identifies 15 new susceptibility loci for breast cancer.Nat Genet. 2015; 47: 373-380https://doi.org/10.1038/ng.3242
- Association analysis identifies 65 new breast cancer risk loci.Nature. 2017; 551: 92-94https://doi.org/10.1038/nature24284
- Multigene panel testing detects equal rates of pathogenic BRCA1/2 mutations and has a higher diagnostic yield compared to limited BRCA1/2 analysis alone in patients at risk for hereditary breast cancer.Ann Surg Oncol. 2015; 22: 3282-3288https://doi.org/10.1245/s10434-015-4754-2
- Panel testing for familial breast cancer: calibrating the tension between research and clinical care.J Clin Oncol. 2016; 34: 1455-1459https://doi.org/10.1200/JCO.2015.63.7454
- The impact of a panel of 18 SNPs on breast cancer risk in women attending a UK familial screening clinic: a case-control study.J Med Genet. 2017; 54: 111-113https://doi.org/10.1136/jmedgenet-2016-104125
- A breast cancer prediction model incorporating familial and personal risk factors. Stat Med. 2004;23(7):1111-1130. Published correction appears in.Stat Med. 2005; 24: 156https://doi.org/10.1002/sim.1668
- Projecting individualized probabilities of developing breast cancer for white females who are being examined annually.J Natl Cancer Inst. 1989; 81: 1879-1886https://doi.org/10.1093/jnci/81.24.1879
- CanRisk tool-a web interface for the prediction of breast and ovarian cancer risk and the likelihood of carrying genetic pathogenic variants.Cancer Epidemiol Biomarkers Prev. 2021; 30: 469-473https://doi.org/10.1158/1055-9965.EPI-20-1319
- Mammographic breast density refines Tyrer-Cuzick estimates of breast cancer risk in high-risk women: findings from the placebo arm of the International Breast Cancer Intervention Study I.Breast Cancer Res. 2014; 16: 451https://doi.org/10.1186/s13058-014-0451-5
- Mammographic density adds accuracy to both the Tyrer-Cuzick and Gail breast cancer risk models in a prospective UK screening cohort.Breast Cancer Res. 2015; 17: 147https://doi.org/10.1186/s13058-015-0653-5
- A case-control evaluation of 143 single nucleotide polymorphisms for breast cancer risk stratification with classical factors and mammographic density.Int J Cancer. 2020; 146: 2122-2129https://doi.org/10.1002/ijc.32541
- Breast cancer risk prediction using clinical models and 77 independent risk-associated SNPs for women aged under 50 years: Australian Breast Cancer Family Registry.Cancer Epidemiol Biomarkers Prev. 2016; 25: 359-365https://doi.org/10.1158/1055-9965.EPI-15-0838
- The contributions of breast density and common genetic variation to breast cancer risk.J Natl Cancer Inst. 2015; 107dju397https://doi.org/10.1093/jnci/dju397
- Polygenic risk scores for prediction of breast cancer and breast cancer subtypes.Am J Hum Genet. 2019; 104: 21-34https://doi.org/10.1016/j.ajhg.2018.11.002
- Improvement in Risk Prediction, Early Detection and Prevention of Breast Cancer in the NHS Breast Screening Programme and Family History Clinics: A Dual Cohort Study.NIHR Journals Library, 2016
- Use of single-nucleotide polymorphisms and mammographic density plus classic risk factors for breast cancer risk prediction.JAMA Oncol. 2018; 4: 476-482https://doi.org/10.1001/jamaoncol.2017.4881
- Clinical Guidelines and Evidence Review for the Classification and Care of Women At Risk of Familial Breast Cancer. London: National Collaborating Centre for Primary Care/University of Sheffield. NICE Guideline CG014; 2004 (updated 2006 CG41, 2013/2017 CG184).(Accessed June 17, 2018.)https://www.nice.org.uk/guidance/cg164Date accessed: June 17, 2018
- Genome-wide association study identifies five new breast cancer susceptibility loci.Nat Genet. 2010; 42: 504-507https://doi.org/10.1038/ng.586
- Breast cancer risk genes - association analysis in more than 113,000 women.N Engl J Med. 2021; 384: 428-439https://doi.org/10.1056/NEJMoa1913948
- Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology.Genet Med. 2015; 17: 405-424https://doi.org/10.1038/gim.2015.30
- Pathology update to the Manchester Scoring System based on testing in over 4000 families.J Med Genet. 2017; 54: 674-681https://doi.org/10.1136/jmedgenet-2017-104584
- Breast cancer in neurofibromatosis 1: survival and risk of contralateral breast cancer in a five country cohort study. Genet Med. 2020;22(2):398-406. Published correction appears in.Genet Med. 2020; 22: 242https://doi.org/10.1038/s41436-019-0651-6
- A population-based study of genes previously implicated in breast cancer.N Engl J Med. 2021; 384: 440-451https://doi.org/10.1056/NEJMoa2005936
- Prevalence of BRCA1 and BRCA2 pathogenic variants in a large, unselected breast cancer cohort.Int J Cancer. 2019; 144: 1195-1204https://doi.org/10.1002/ijc.31841
- Economic evaluation of population-based BRCA1/BRCA2 mutation testing across multiple countries and health systems.Cancers (Basel). 2020; 12: 1929https://doi.org/10.3390/cancers12071929
- Frequency of pathogenic germline variants in BRCA1, BRCA2, PALB2, CHEK2 and TP53 in ductal carcinoma in situ diagnosed in women under the age of 50 years.Breast Cancer Res. 2019; 21: 58https://doi.org/10.1186/s13058-019-1143-y
- Risk stratified breast cancer screening: UK healthcare policy decision-making stakeholders’ views on a low-risk breast screening pathway.BMC Cancer. 2020; 20: 680https://doi.org/10.1186/s12885-020-07158-9
- Breast cancer pathology and stage are better predicted by risk stratification models that include mammographic density and common genetic variants.Breast Cancer Res Treat. 2019; 176: 141-148https://doi.org/10.1007/s10549-019-05210-2
- A comparison of five methods of measuring mammographic density: a case-control study.Breast Cancer Res. 2018; 20: 10https://doi.org/10.1186/s13058-018-0932-z
- Baseline mammographic breast density and the risk of invasive breast cancer in postmenopausal women participating in the NSABP study of tamoxifen and raloxifene (STAR).Cancer Prev Res (Phila). 2012; 5: 1321-1329https://doi.org/10.1158/1940-6207.CAPR-12-0273
- Prediction of reader estimates of mammographic density using convolutional neural networks.J Med Imaging (Bellingham). 2019; 6031405https://doi.org/10.1117/1.JMI.6.3.031405
- Deep-LIBRA: an artificial-intelligence method for robust quantification of breast density with independent validation in breast cancer risk assessment.Med Image Anal. 2021; 73102138https://doi.org/10.1016/j.media.2021.102138
- Unsupervised deep learning applied to breast density segmentation and mammographic risk scoring.IEEE Trans Med Imaging. 2016; 35: 1322-1331https://doi.org/10.1109/TMI.2016.2532122
- The contribution of whole gene deletions and large rearrangements to the mutation spectrum in inherited tumor predisposing syndromes.Hum Mutat. 2016; 37: 250-256https://doi.org/10.1002/humu.22938
- The importance of ethnicity: are breast cancer polygenic risk scores ready for women who are not of White European origin?.Int J Cancer. 2022; 150: 73-79https://doi.org/10.1002/ijc.33782
Article info
Publication history
Published online: April 15, 2022
Accepted:
March 15,
2022
Received in revised form:
March 14,
2022
Received:
December 24,
2021
Identification
Copyright
© 2022 The Authors. Published by Elsevier Inc. on behalf of American College of Medical Genetics and Genomics.
User license
Creative Commons Attribution (CC BY 4.0) | How you can reuse 
Elsevier's open access license policy
Creative Commons Attribution (CC BY 4.0)
Permitted
- Read, print & download
- Redistribute or republish the final article
- Text & data mine
- Translate the article
- Reuse portions or extracts from the article in other works
- Sell or re-use for commercial purposes
Elsevier's open access license policy
