Breast and ovarian cancer risk assessment using multigene panel tests Prof Antonis Antoniou Department of Public Health and Primary Care University of Cambridge, U.K.
Cancer risk prediction in the era of NGS Next Generation Sequencing technologies The specific content of the panel may be a problem Multi-gene panels Sequencing of multiple genes in a single assay to identify disease-associated variants The use of a panel in itself is not a problem
Gene-panel content: >100 genes Breast or ovarian cancer indication: >21 genes CHEK2 ATM PALB2 BRIP1 RAD51C RAD51D BARD1 NBN/NBS1 RINT1 RECQL FANCC RAD50 SLX4 NF-1 MRE11A And so on
What are test characteristics that need to be evaluated before clinical implementation? Analytical validity Clinical validity Clinical utility Centers for Disease Control and Prevention: ACCE
Analytical validity How accurately and reliably the test measures the genotype of interest Not a major issue
Clinical validity Are the genetic variants the test is indented to identify associated with risk for breast or ovarian cancer? Are these breast or ovarian cancer risks well quantified?
Clinical utility Whether or not using the gene panel test improves patient outcome
Easton et al, NEJM (2105) A genomic test should not be offered until its clinical validity has been established
Clinical validity in context of risk prediction Is there an association between gene variant(s) and risk of breast or ovarian cancer? Statistical strength of association: P<0.0001 What is the magnitude of the risk? What is the precision of the risk estimate? What methods have been used to estimate risks?
Genetic architecture of breast cancer risk Easton et al, NEJM 2015
Genetic Variant Chacterisation Study Designs (1) Case-Control Studies Variant frequency in cases VS frequency in controls Very large sample sizes required for rare genetic variants Possible in the context of large international consortia Breast Cancer Association Consortium Ovarian Cancer Association Consortium e.g. CHEK2: 1100delC, ATM
Genetic Variant Chacterisation Study Designs (1) Case-Control Studies Valid inference on association and risk depends on: Unselected cases vs Population matched controls Note of Caution: Studies based on data from commercial clinical testing laboratories Large sample sizes BUT case selection criteria unclear, lack of population matched controls
Retrospective family-based studies Mutation screening in cancer patients (with or without selection criteria) dx 41 Screened for mutations
Retrospective family-based studies Mutation screening in cancer patients (with or without selection criteria) Phenotypes of family members and /or mutation testing in relatives dx 50 dx 41 35 Screened for mutations
Retrospective family-based studies Mutation screening in cancer patients (with or without selection criteria) Phenotypes of family members and /or mutation testing in relatives dx 50 Modified segregation analysis PALB2, ATM dx 41 35 Screened for mutations Ascertainment issues, self-reported family histories
Prospective cohort studies Cohort of unaffected mutation carriers followed over time Estimates free of ascertainment/reporting biases Large numbers of unaffected mutation carriers required Long periods of observation Largest study until recently: BRCA1, BRCA2 <64breast, 31 ovarian cancers Mavaddat JNCI 2013; Evans JMG 2104; Senst Clin Gen 2013
BRCA1, BRCA2 : Prospective cancer risks ~10,000 women Breast Cancer 72% (65-79%) 69% (61-77%) Ovarian Cancer 44% (36-53%) 17% (11-25%) JAMA 2017
Breast cancer risks by family history Analyses stratified by birth cohort and study group BRCA1 BRCA2 Family History Category (1 st or 2 nd degree relatives) HR 95%CI HR 95%CI No Breast cancer 1.0 1 Breast cancer 1.51 1.10-2.11 1.53 0.86-2.7 2 Breast cancers 2.00 1.40-2.82 1.91 1.08-3.4 K. Kuchenbaecker, D. Barnes (JAMA 2017) P-trend=0.0001 P-trend=0.02
Prospective studies BRCA2: Genotype Phenotype correlations Breast cancer risk (%) 100 90 80 70 60 50 40 30 20 10 0 BRCA2: Cumulative breast cancer risk by mutation location (wide) 5 to c.2830 c.2831 to c.6401 3 to c.6402 44% 18% 68% (60-75%) 50% (40-63%) OCCR 20 30 40 50 60 70 80 Age (years) mber at risk 1 to c.6401 12 51 Daniel Barnes- JAMA 2017 117 97 51 23 11 HR = 1.9 P-diff <0.001
PALB2 Interest Group ( http://www.palb2.org ) 154 families with PALB2 mutations 17 participating centres
PALB2 Interest Group ( http://www.palb2.org ) 569 families with PALB2 mutations 41 participating centres Family-based study design P = 10-38 P = 5 x 10-3 Xin Yang, Goksa Leslie, Alicja Doroszuk, Marc Tischkowitz
PALB2 Interest Group ( http://www.palb2.org ) Other cancer risks Preliminary results Cancer PALB2 RR (95% CI) P-value Pancreas 2.83 (1.49-5.39) 0.0015 Male breast 5.17 (1.09-24.47) 0.038 Prostate 0.49 (0.25-0.96) 0.037 Lung 0.79 (0.41-1.52) 0.49 Female colon 1.62 (0.74-3.56) 0.23 Male colon 1.36 (0.65-2.85) 0.41 Xin Yang, Goksa Leslie, Alicja Doroszuk
Schmidt et al, JCO 2016 CHEK2 1100delC BCAC: Case Control study design 44,777 cases / 42,997 controls Odds Ratio: 2.26 (95%CI: 1.90 2.69) P-value association = 2 x 10-20 Stronger association with ER-positive breast cancer Associations similar for all truncating variants in CHEK2 (Decker et al JMG 2017)
Schmidt et al, JCO 2016 CHEK2 1100delC BCAC: 44,777 cases / 42,997 controls
ATM truncating mutations and breast cancer A-T Families:breast cancer incidence in relatives of A-T patients Study Country #families RR (95% CI) No Cases Swift et al. 1991 USA 161 5.1 (1.5-16.9) 23 Easton 1994 review 3.9 (2.1-7.2) 58 Thompson 2005 UK 139 2.3 (1.2-4.3) 23 Olson et al 2005 Nordic 66 2.9 (1.9-4.4) 34 Kin-Cohort: ATM mutation screening in families with 3 cases of breast cancer Study Country Cases/controls RR (95% CI) Renwick et al 2006 UK 12/2 2.4 (1.5-3.8) Case-control: 13,087 cases Vs 5,488 controls Study Country Cases/controls RR (95% CI) Decker 2017 UK 85/11 3.3 (1.8-6.5)
ATM genetic variants and breast cancer risk Truncating / splice-junction mutations Yes, RR: 2.9 (2.3-3.6) Rare missense mutations SNPs Common No Most rare missense mutations Little if any increase in risk e.g. Tavtigian 2009 AJHG; Baynes 2007 BCR; Fletcher 2010 CEBP; Goldgar 2011 BCR; Chenevix-Trench 2002 JNCI; Thompson 2005 HM Bernstein et al, Hum Mutat 2006 Rare, evolutionarily unlikely missense mutations in particular regions of ATM Yes, higher risks than truncating mutations 7271T>G (Val2424Gly) Particularly high risks
icogs analysis of ATM 7271T>G The icogs array of ~200,000 SNPs included ATM 7271 T>G Breast cancer: ~43,000 cases and ~42,000 controls (European) Ovarian cancer: ~14,500 cases and ~23,500 controls Southey et al, JMG (2016)
icogs analysis of ATM 7271T>G Restricting the analysis to the 8 studies in which at least one G allele was seen: 11 carriers / 18921 cases 1 carrier / 18428 controls OR = 11.6 (1.5-89.9) P = 0.0012 Risk on a par with those seen in BRCA1 or BRCA2 mutation carriers, with implications for clinical management. No evidence of association with ovarian cancer risk Southey et al, JMG (2016)
Established breast cancer susceptibility genes Gene Estimated Relative Risk Risk by age 80 % BRCA1 16.6 (15-19) 72 BRCA2 12.9 (11-15) 69 TP53 105 (62-165) PTEN Not reliable CDH1 6.6 (2.2-20) 53 STK11 Not reliable NF1 2.6 (2.1-3.2) 26 PALB2 5.3 (3.0-9.4) 45 ATM 2.9 (2.3-3.6) 27 CHEK2 3.0 (2.6-3.5) 29 NBN 2.7 (1.9-3.7) 23 Rare cancer syndromes Updated from Easton et al NEJM (2015)
Established breast cancer susceptibility genes High Risk High Risk High Risk Gene Estimated Relative Risk Risk by age 80 % BRCA1 16.6 (15-19) 72 BRCA2 12.9 (11-15) 69 TP53 105 (62-165) PTEN Not reliable CDH1 6.6 (2.2-20) 53 STK11 Not reliable NF1 2.6 (2.1-3.2) 26 PALB2 5.3 (3.0-9.4) 45 ATM 2.9 (2.3-3.6) 27 CHEK2 3.0 (2.6-3.5) 29 NBN 2.7 (1.9-3.7) 23 Rare cancer syndromes Updated from Easton et al NEJM (2015)
Established breast cancer susceptibility genes High Risk High Risk High Risk Mod-High Risk Mod-High Risk Gene Estimated Relative Risk Risk by age 80 % BRCA1 16.6 (15-19) 72 BRCA2 12.9 (11-15) 69 TP53 105 (62-165) PTEN Not reliable CDH1 6.6 (2.2-20) 53 STK11 Not reliable NF1 2.6 (2.1-3.2) 26 PALB2 5.3 (3.0-9.4) 45 ATM 2.9 (2.3-3.6) 27 CHEK2 3.0 (2.6-3.5) 29 NBN 2.7 (1.9-3.7) 23 Rare cancer syndromes Updated from Easton et al NEJM (2015)
Established breast cancer susceptibility genes High Risk High Risk High Risk Mod-High Risk Moderate Risk Mod-High Risk Moderate Risk Moderate Risk Gene Estimated Relative Risk Risk by age 80 % BRCA1 16.6 (15-19) 72 BRCA2 12.9 (11-15) 69 TP53 105 (62-165) PTEN Not reliable CDH1 6.6 (2.2-20) 53 STK11 Not reliable NF1 2.6 (2.1-3.2) 26 PALB2 5.3 (3.0-9.4) 45 ATM 2.9 (2.3-3.6) 27 CHEK2 3.0 (2.6-3.5) 29 NBN 2.7 (1.9-3.7) 23 Rare cancer syndromes Updated from Easton et al NEJM (2015)
Established breast cancer susceptibility genes High Risk High Risk High Risk Mod-High Risk Imprecise Imprecise Moderate Risk Mod-High Risk Moderate Risk Moderate Risk Imprecise Gene Estimated Relative Risk Risk by age 80 % BRCA1 16.6 (15-19) 72 BRCA2 12.9 (11-15) 69 TP53 105 (62-165) PTEN Not reliable CDH1 6.6 (2.2-20) 53 STK11 Not reliable NF1 2.6 (2.1-3.2) 26 PALB2 5.3 (3.0-9.4) 45 ATM 2.9 (2.3-3.6) 27 CHEK2 3.0 (2.6-3.5) 29 NBN 2.7 (1.9-3.7) 23 Rare cancer syndromes Updated from Easton et al NEJM (2015)
Breast cancer: other susceptibility genes? BARD1, BRIP1, RAD51C, RAD51D, FANCM? Limited evidence of association with overall breast cancer risk Suggestive associations with breast cancer subtypes : ERnegative, Triple negative Clinical validity for breast cancer, currently questionable Larger studies, by disease subtypes required: Large scale target sequencing in BCAC ~70,000 cases / ~70,000 controls https://bridges-research.eu
Established ovarian cancer susceptibility genes Gene Estimated Relative Risk Risk by age 80% BRCA1 49 (40-61) 44 BRCA2 13.7 (9.1-20.7) 17 BRIP1 3.4 (2.1-5.5) 5.8 RAD51C 5.2-6.3 ~10 RAD51D 6.3 12 ~10-12 MLH1 MSH2 MSH6 Other likely: Not reliable Not reliable Not reliable endometrioid ovarian cancer FANCM 2.5 ~3.8 PALB2 3.0 (1.4-6.7) ~5 P<0.0001 P~0.001
Consensus for genes to be included on cancer panels to be offered by all NHS Genetics Services UK Cancer Genetics Group Taylor et al, J Med Genetics (In press) Breast Cancer Ovarian Cancer BRCA1 BRCA1 BRCA2 BRCA2 PALB2 BRIP1 ATM* RAD51C CHEK2** RAD51D PTEN MLH1 STK11 MSH2 TP53 MSH6 * Truncating & c.7271t>g; ** truncating variants only
Consortium of Investigators of Modifiers of BRCA1/2 Identifying and characterising genetic modifiers Initiated: 2005 >70 groups from Europe, North America, Australia, Asia, Africa and South America >55,000 BRCA1, BRCA2 mutation carriers Milne et al Nat Genet (2017); Phelan et al Nat Genet (2017)
BRCA2 mutation carriers: Breast cancer risk by PRS 268 SNPs Kuchenbaecker et al, JNCI (2017); D. Barnes (In prep) BRCA2-Average Average Breast Cancer Risk 0.0 0.2 0.4 0.6 0.8 Average 86% 5%, 95% 10%, 90% 68% 46% 33% 20% 48% 20 30 40 50 60 70 80 Age
BRIP1 & other risk factors Average ovarian cancer risk Top 20% at highest risk, combined risk factor distribution Ramus et al, JNCI 2015
BOADICEA http://ccge.medschl.cam.ac.uk/boadicea/ Lee et al, BJC, 2014; GiM 2016 BRCA1, BRCA2,PALB2, CHEK2, ATM Polygenic (unobserved genetic effects) Family history breast, ovarian prostate, pancreatic cancer Tumour characteristics - ER/PR/HER2/Cytokeratin markers Population, ethnicity, year of birth
BOADICEA: incorporating the effects of common alleles Predicted risk Predicted by 268 risk SNP by PRS 311 (~21% SNP PRS of (σ polygenic 2 =0.25) variance) BRCA1 BRCA1 and and BRCA2 Negative Female BC 50 BC 50 Breast Cancer Risk 20 0.5 0.4 0.3 Average 5 % PRS 95 % PRS 0.2 0.1 0 20 40 60 80 age Lee et al (In prep) 0.5 0.4 0.3 0.2 0.1 0 20 0.5 0.4 0.3 0.2 0.1 20 40 60 80 age 0 20 BC 41 20 40 60 80 age
Parity Risk factors included Age at first live birth Age at menarche Age at menopause BMI Height Alcohol Use Oral contraceptive use Menopause hormone therapy use Mammographic Density From Garcia-Closas J Natl Cancer Inst. 2014 Lancet Oncol. 2012 Nov;13(11):1141-51
Model validation 10 year risk predictions FHrisk families (10,000 women, 419 incident breast cancers) Calibration Discrimination AUC=0.776 (0.752-0.80) H-L GOF, p=0.41 Family history + risk factors + BRCA1/2 testing
Ovarian cancer risk model Rare Variants Ovarian cancer SNPs PRS Lifestyle, hormonal and other risk factors: Use of oral contraceptives Parity Tubal ligation Endometriosis HRT use BMI A. Lee & X. Yang (in prep)
Web-tool prototype development Simon Hartley, Tim Carver, Chantal Babb de Villiers, Alex Cunningham, Andrew Lee Work stream leads: Fiona Walter, Marc Tischkowitz
Carver et al, Bioinformatics 2017
Acknowledgments University of Cambridge Andrew Lee, Alex Cunningham, Tim Carver, Simon Hartley, Daniel Barnes, Xin Yang, Lesley McGuffog, Goska Leslie, Karoline Kuchenbaecker, Nasim Mavaddat, Joe Dennis, Chantal Babb de Villiers, Tommy Nyberg, Deborah Thompson, Paul Pharoah, Fiona Walter, Marc Tischkowitz, Douglas Easton CanRisk co- Investigators: Gareth Evans BCAC investigators Montse Garcia-Closas CIMBA investigators Georgia Chenevix-Trench IBCCS Investigators PROF-Sc Investigators PALB2 Interest Group BRIDGES programe PERSPECTIVE programme B-CAST programme
BRCA1, BRCA2, PALB2, ATM and CHEK2 average breast cancer risks in BOADICEA Lee et al, GiM 2016
Prospective studies: IBCCS, BCFR, kconfab BRCA1: Genotype Phenotype correlations Breast cancer risk (%) 100 90 80 70 60 50 40 30 20 10 0 BRCA1: Cumulative breast cancer risk by mutation location 5 to c.2281 c.2282 to c.4071 3 to c.4072 70% (64-76%) 55% (46-67%) 20 30 40 50 60 70 80 Age (years) mber at risk 82 Daniel to c.4071 Barnes 17 JAMA 2017 101 116 76 45 9 2 HR = 1.5 P-diff = 0.007
Schmidt et al, JCO 2016 CHEK2 1100delC BCAC: 44,777 cases / 42,997 controls
Breakthrough Generations Study 5 year predictions (Family history and risk factors) Women <50 : N=40,786 Cases=323 P=0.29 C= 0.706 (0.68-0.731) Women >50 : P=0.15 N=38,830 Cases=647 C= 0.60 (0.574-0.618) M. Garcia-Closas A. Wilcox M. Brook A. Lee
Risk modifying loci patterns of association Most SNPs associated with risk in the unselected population also modify breast and ovarian cancer risks for carriers BRCA1 BC modifiers: Primarily associated with ER-negative breast cancer in population Partitioned heritability analysis: Correlation (ER-, BRCA1)=0.72 BRCA2 BC modifiers: Associated with overall breast cancer in the population Couch et al, Plos Genet (2013), Gaudet et al Plos Genet (2013), Kuchenbaecker et al Nat Genet (2015), Couch et al Nat Commun (2016), Bojesen et al, Nat Genet (2013), Milne et al, Nat Genet (2017), Phelan et al, Nat Genet (2017)
77- SNP Polygenic Risk Score PRS = β 1 X 1 + β 2 X 2.+ β n X n Log(Odds Ratio) estimate Number of risk alleles at each SNP Breast Cancer Association Consortium (BCAC) 33, 673 breast cancer cases and 33,381 control women PRS normally distributed in both cases and controls Mavaddat et al JNCI 2015 Mean PRS (cases)=0.69 Mean PRS (controls)=0.49
PRS and lifetime breast cancer risk 29% 3.5% Mavaddat et al JNCI 2015
PRS and breast cancer risk associations in CHEK2*1100delC carriers BCAC Muranen et al, Genet Med (2016)
Integrating all modifiers of risk Risk stratification in BRCA2 mutation carriers Predicted 5-year BC risks by percentile of risk distribution Milne and Antoniou, Endocr Relat Cancer. 2016 5-year breast cancer risk 0.08 0.07 0.06 0.05 0.04 0.03 0.02 0.01 0 0.017 Average 5% 10% 15% 20% 25% 30% 35% 20 28 30 40 43 50 55 60 Age Average risk Bottom 5%
Integrating all modifiers of risk Risk stratification in BRCA2 mutation carriers Predicted 5-year BC risks by percentile of risk distribution Milne and Antoniou, Endocr Relat Cancer. 2016 5-year breast cancer risk 0.08 0.07 0.06 0.05 0.04 0.03 0.02 0.01 0 0.017 Average 5% 10% 15% 20% 25% 30% 35% 20 28 30 40 43 50 55 60 Age Chemoprevention