iCARE Breast Cancer Risk Model Development and Validation NO - - PowerPoint PPT Presentation

▶
icare breast cancer risk model development and validation
SMART_READER_LITE
LIVE PREVIEW

iCARE Breast Cancer Risk Model Development and Validation NO - - PowerPoint PPT Presentation

Aug 19, 2022 27 likes •147 views

6/9/2017 iCARE Breast Cancer Risk Model Development and Validation NO DISCLOSURES Montserrat Garcia-Closas, M.D. Dr.P.H. Senior Investigator and Deputy Director Division of Cancer Epidemiology and Genetics 2 1. iCARE risk modelling

slide-1
SLIDE 1

6/9/2017 1

iCARE Breast Cancer Risk Model Development and Validation

Montserrat Garcia-Closas, M.D. Dr.P.H. Senior Investigator and Deputy Director Division of Cancer Epidemiology and Genetics

2

NO DISCLOSURES

3

1. iCARE risk modelling approach 2. PRS development in BCAC 3. Building integrated risk models 4. Validation in prospective cohorts 5. Population risk stratification

4

1. iCARE risk modelling approach 2. PRS development in BCAC 3. Building integrated risk models 4. Validation in prospective cohorts 5. Population risk stratification

slide-2
SLIDE 2

6/9/2017 2

5

Breast cancer risk models

  • Many existing models for different uses and target populations:
  • 10 hormonal/environmental models
  • 12 hereditary models
  • 1 hormonal/environmental and hereditary model
  • Many have not been externally validated in prospective cohorts
  • Many don’t have online tools for risk calculation

Cintolo-Gonzalez et al. Breast Cancer Res Treat 2017

6

For a comprehensive review of risk models see Cintolo-Gonzalez et al. Breast Cancer Res Treat 2017

Model Prediction Density? BBD path? Target population Externally validated in prospective cohorts? Online tool? BCRAT (Gail) Invasive No Yes General Yes Yes BCRAT(Chen) Invasive Yes Yes General No No Rosner–Colditz Invasive No No General Limited No Tworoger Invasive No No General No No BCSC (Barlow) Inv+ DCIS Yes No Screening No No BCSC-BBD (Tice) Inv+ DCIS Yes Yes Screening No Yes BBD-BC Inv+ DCIS No Yes BBD biopsy No No Bodian Inv+ DCIS No Yes LCIS Limited Yes IBIS (Tyrer-Cuzick) Inv+ DCIS No Yes General & high genetic risk Yes Yes

Breast cancer models with hormonal/environmental factors

7

Why a new breast cancer risk model ?

  • Flexible modelling approach:
  • Comprehensive and easier to update
  • Incorporation of polygenetic risk scores
  • Updated rates and risk factor distributions in the target population
  • Accommodate missing data in risk factors
  • Predict subtype-specific risk

8

Development of a synthetic risk model

Models for Absolute Risk Models for Absolute Risk Risk factor RR Risk factor RR Population distribution

  • f risk

factors Population distribution

  • f risk

factors Population Disease Incidence Population Disease Incidence Rate of competing mortality Rate of competing mortality

Chatterjee et al. http://dceg.cancer.gov/tools/analysis/icare

Individualized Coherent Absolute Risk Estimator (iCARE)

Used for model calibration and imputation of missing data

slide-3
SLIDE 3

6/9/2017 3

Breast CAncer STratification

Risk factors Tumor subtypes Prognosis

Questionnaires Breast density Genetics (PRS) Molecular subtypes SD vs Interval Early vs late onset Aggressive Recurrence Mortality

Understanding the determinants of risk and prognosis of molecular subtypes

Risk model development Validation Online tools / implementation

Translating knowledge into risk stratification for precision prevention

iCARE BOACIDEAPlus Prospective cohort studies General population

10

1. iCARE risk modelling approach 2. PRS development in BCAC 3. Building integrated risk models 4. Validation in prospective cohorts 5. Population risk stratification 6. Future work

per-allele log (OR) for risk allele at locus j from logistic regression adjusted for study and 7 PCs number of risk alleles at j locus (0, 1 or 2) xj

Analyses included women of European origin in iCOGS: 33,673 cases (21,365 ER+ and 5,738 ER-) 33,381 controls

βj

Polygenic risk score (PRS) based on 77 SNPs

Mavaddat E et al, JNCI 2015 Mavaddat E et al, JNCI 2015

Distribution of risk alleles for 77 SNPs

(N=33,381) (N=33,673)

slide-4
SLIDE 4

6/9/2017 4

Mavaddat E et al, JNCI 2015

Polygenic risk stratification by family history of breast cancer

Quintiles of 77-PRS Quintiles of 77-PRS

10% 16% 5% 16% 25% 9%

Life-time breast cancer risk

Women without a family history (90% of the population) Women with a family history (10% of the population)

Mavaddat E et al, JNCI 2015

Risk stratification for ER-positive than ER-negative disease

Quintiles of 77-PRS

Life-time breast cancer risk

Risk for ER-positive disease Risk for ER-negative disease

16% 7.5% 4% 1%

Quintiles of 77-PRS Quintiles of 77-PRS Mavaddat E et al, In preparation

Improved PRS based on iCOGS + OncoArray analyses

P-value cut off for SNP selection Number

  • f SNPs

OR per SD (95% CI) AUC 77 SNP PRS (P< 5x10-8 in 2015) 77 1.49 (1.43-1.57) 0.61 P< 10-5 268* 1.65 (1.58-1.73) 0.64 *Selected based on data from 94,094 cases and 75,017 controls (European) from 69 studies in BCAC, divided into training (90%) and test (10%) sets.

16

1. iCARE risk modelling approach 2. PRS development in BCAC 3. Building integrated risk models 4. Validation in prospective cohorts 5. Population risk stratification

slide-5
SLIDE 5

6/9/2017 5

17

iCARE breast cancer risk model for the UK

Models for Absolute Risk Models for Absolute Risk

RR

Reproductive, BMI, OC, HRT, alcohol, BBD, family history; PRS

RR

Reproductive, BMI, OC, HRT, alcohol, BBD, family history; PRS Distribution of risk factors from Health Survey for England, Fertility Tables,

  • thers

Distribution of risk factors from Health Survey for England, Fertility Tables,

  • thers

Breast cancer incidence from Office of National Statistics, UK Breast cancer incidence from Office of National Statistics, UK Rate of competing mortality from Office of National Statistics, UK Rate of competing mortality from Office of National Statistics, UK

Literature review BPC3 (8 cohorts) BCAC (PRS)

18

Log linear and non-parametric risk scores from a model including 77- SNP PRS and hormonal/environmental risk factors (BCAC)

  • 3
  • 2
  • 1

1 2 3

  • 3
  • 2
  • 1

1 2 3 Log linear risk score Fitted log O R 100 200 300 400 500 F requency Non-parametric smoothing Linear logistic

Multiplicative effects of PRS and risk factors

Rudolph et al. In submission

Goodness of fit tests P=0.252 global test P= 0.179 tail-based test

19

Alcohol and breast cancer risk by PRS percentiles

Rudolph et al. In submission

Goodness of fit tests P=0.013 global test P= 0.18 tail-based test OR [95%CI] 77-SNP PRS Percentiles Genetic risk

20

Study Name Cases Controls Bavarian Breast Cancer Case-control study (BBCC) 512 367 Mayo Mammography Health Study (MMHS) 456 1166 Nurses Health Study (NHS) 850 849 European Prospective Investigation into Cancer (EPIC) 86 968 Mayo Clinic Breast Cancer Study (MCBCS) 677 864 Melbourne Case-control study (MCCS) 68 28 Multi-ethnic Cohort (MEC) 110 101 Singapore and Sweden Breast Cancer Study (SASBAC) 869 783 TOTAL 3,628 5,126

Vachon et al In preparation

Studies to evaluate PRS and density joint effects

slide-6
SLIDE 6

6/9/2017 6

21

Study Name Percent density OPERA* (95%CI) 77-SNP PRS OR per SD (95%CI) BBCC 1.18 (1.0, 1.40) 1.30 (1.13, 1.49) MMHS 1.67 (1.45, 1.92) 1.59 (1.42, 1.78) NHS 1.45 (1.32, 1.59) 1.58 (1.43, 1.75) EPIC 1.32 (1.04, 1.67) 1.41 (1.12, 1.77) MCBCS 1.89 (1.65, 2.17) 1.46 (1.32, 1.62) MCCS 1.62 (1.03, 2.56) 1.21 (0.74, 1.96) MEC 1.53 (1.21, 1.92) 1.57 (1.17, 2.11) SASBAC 1.29 (1.17, 1.42) 1.58 (1.43, 1.75) Overall 1.45 (1.38, 1.52) 1.51 (1.44, 1.59)

* Age and BMI adjusted

Vachon et al. In preparation

Association of PD and 77-SNP PRS with breast cancer risk

Vachon et al. In preparation PD OPERA (95%CI)

Percent density and breast cancer risk by PRS percentiles

23

Goodness of fit tests P=0.164 global test P= 0.372 tail-based test

Vachon et al. In preparation

Log linear and non-parametric risk scores from a model including 77- SNP PRS and mammographic density

Multiplicative effects of PRS and mammographic breast density

24

1. iCARE risk modelling approach 2. PRS development in BCAC 3. Building integrated risk models 4. Validation in prospective cohorts 5. Population risk stratification

slide-7
SLIDE 7

6/9/2017 7

Study N Country Year at DNA Collection Age Range at DNA Collection NHS I 32,826 US 1989 43-68 MCCS 41,514 Australia 1990-1994 27-76 EPIC 19,613 Europe 1991-1999 20-85 PLCO 48,059 US 1993- 50-75 NHS II 29,240 US 1996 32-49 CPS II 41,014 US 1998-2002 47-74 MMHS 19,865 US 2003-2006 40-80 SISTERS 50,884 US 2003-2009 35-74 UKBGS 95,059 UK 2003-2011 18-74 UK Biobank 262,127 UK 2006-2010 40-70 PROCAS 58,000 UK 2009 46–73 KARMA 70,877 Sweden 2011-2013 40-74

12 prospective cohorts participating in risk model validation Qx-based risk factors in iCARE model compared to BCRAT and IBIS

Risk factors iCARE BCRAT IBIS Age at menarche ✔ ✔ ✔ Parity ✔ ✔ ✔ Age at first birth ✔ ✔ Oral contraceptive use ✔ Menopause (status)* ✔ ✔ ✔ Age at menopause ✔ ✔ Menopausal replacement therapy use ✔ ✔ Body mass index (BMI) ✔ ✔ Alcohol use ✔ Benign breast disease** ✔ ✔ ✔ Type of benign breast disease Benign breast disease - hyperplasia ✔ ✔ Benign breast disease - atypical hyperplasia ✔ ✔ ✔ Lobular Carcinoma in situ (LCIS) ✔ ✔ Family history of breast and/or ovarian cancer First-degree relatives with breast cancer ✔ ✔ ✔ Second-degree relatives with breast cancer ✔ Third-degree relatives with breast cancer ✔ Age of onset of breast cancer in a relative ✔ Bilateral breast cancer in a relative ✔ Ovarian cancer in a relative ✔ Male breast cancer ✔ *iCARE and BCRAT use age 50 as a surrogate for menopausal status **Assessed as yes/no for iCARE and number of biopsies for BCRAT

Relative risk calibration

P=0.67 P=0.04 P=0.16

5-year risk calibration

P=0.05 P=0.07 P=2x10-7

iCARE BCRAT IBIS

Qx-based risk model calibration in the UKBGS

36,921 women ≥ 50 years old : 601 cases in 5-years

AUC: 0.60 (95% CI 0.58, 0.62) AUC: 0.52 (95% CI 0.50, 0.54) AUC: 0.60 95% CI (0.58, 0.62)

iCARE risk model validation in multiple cohorts (ages ≥50)

BGS (UK) CPS II (US) KARMA (Sweden) MMCS (Australia) MMHS (US) PLCO (US) Sisters (US) UK Biobank (UK)

Relative risk calibration

slide-8
SLIDE 8

6/9/2017 8

iCARE risk model validation in multiple cohorts (ages ≥50)

BGS (UK) CPS II (US) KARMA (Sweden) MMCS (Australia) MMHS (US) PLCO (US) Sisters (US) UK Biobank (UK)

Absolute risk calibration

30

G-Meta-analysis to combine RR estimates from multiple risk models

Runlong Tang, Prosejit Kundu and Nilanjan Chatterjee

Log RR from BCDDP and BPC3 breast cancer risk models

31

Runlong Tang, Prosejit Kundu and Nilanjan Chatterjee

Log RR from BCDDP and BPC3 breast cancer risk models

G-Meta-analysis to combine RR estimates from multiple risk models

32

Runlong Tang, Prosejit Kundu and Nilanjan Chatterjee

G-Meta-analysis to combine RR estimates from multiple risk models

Log RR from BCDDP and BPC3 breast cancer risk models

slide-9
SLIDE 9

6/9/2017 9

33

Runlong Tang, Prosejit Kundu and Nilanjan Chatterjee

Meta-analysis of BCDDP and BPC3 models

Log RR from BCDDP and BPC3 breast cancer risk models

34

1. iCARE risk modelling approach 2. PRS development in BCAC 3. Building integrated risk models 4. Validation in prospective cohorts 5. Population risk stratification

35

Improvement of risk stratification with SNPs

5-year risk of breast cancer for 50 years old women, UK European ancestry

0.00 0.25 0.50 0.75 1.00 1.25 1 2 3 4 5 6 7 8 9 10 Five−Year Absolute Risk of Breast Cancer (%) Density

Environmental factors, AUC=0.61

36

Improvement of risk stratification with SNPs

0.00 0.25 0.50 0.75 1.00 1.25 1 2 3 4 5 6 7 8 9 10 Five−Year Absolute Risk of Breast Cancer (%) Density

Environmental factors, AUC=0.61 PRS, AUC=0.64

5-year risk of breast cancer for 50 years old women, UK European ancestry

slide-10
SLIDE 10

6/9/2017 10

37

Improvement of risk stratification with SNPs

0.00 0.25 0.50 0.75 1.00 1.25 1 2 3 4 5 6 7 8 9 10 Five−Year Absolute Risk of Breast Cancer (%) Density

Environmental factors, AUC=0.61 PRS, AUC=0.64 All factors, AUC=0.67

5-year risk of breast cancer for 50 years old women, UK European ancestry

38

Performance of integrated risk model

Moderate risk (5-Year risk > 3%) High risk (5-Year Risk > 4%) Models AUC % pop % cases % pop % cases Qx-only 0.61 1.6% 3.9% 0.2% 0.6% PRS-only 0.64 5.6% 13.9% 1.5% 4.9% All factors 0.67 10.1% 25.9% 4.2% 13.5%

Percentage of 50 year old women at different levels of risk and corresponding percentage of cases expected at those levels

39

Improvement of risk stratification with SNPs and density

Distribution of five-year risk of breast cancer, UK Caucasians

0.00 0.25 0.50 0.75 1.00 1.25 1 2 3 4 5 6 7 8 9 10 Five−Year Absolute Risk of Breast Cancer (%) Density

Environmental factors, AUC=0.61

40

Distribution of five-year risk of breast cancer, UK Caucasians

0.00 0.25 0.50 0.75 1.00 1.25 1 2 3 4 5 6 7 8 9 10 Five−Year Absolute Risk of Breast Cancer (%) Density

Environmental factors, AUC=0.61 Breast density, AUC=0.61

Improvement of risk stratification with SNPs and density

slide-11
SLIDE 11

6/9/2017 11

41

Distribution of five-year risk of breast cancer, UK Caucasians

0.00 0.25 0.50 0.75 1.00 1.25 1 2 3 4 5 6 7 8 9 10 Five−Year Absolute Risk of Breast Cancer (%) Density

Environmental factors, AUC=0.61 Breast density, AUC=0.61 PRS, AUC=0.64

Improvement of risk stratification with SNPs and density

42

Distribution of five-year risk of breast cancer, UK Caucasians

0.00 0.25 0.50 0.75 1.00 1.25 1 2 3 4 5 6 7 8 9 10 Five−Year Absolute Risk of Breast Cancer (%) Density

Environmental factors, AUC=0.61 Breast density, AUC=0.61 PRS, AUC=0.64 All factors, AUC=0.70

Improvement of risk stratification with SNPs and density

43

Performance of integrated risk model

Moderate risk (5-year risk >3%) High risk (5-year risk >4%) Models AUC % pop % cases % pop % cases Qx-only 0.61 1.6% 3.9% 0.2% 0.6% Density 0.61 2.0% 4.8% 0.3% 0.8% PRS-only 0.64 5.6% 13.9% 1.5% 4.9% All factors 0.70 13.9% 36.6% 7.0% 23.3%

Percentage of 50 year old women at different levels of risk and corresponding percentage of cases expected at those levels

w/o density 0.67 10.1% 25.9% 4.2% 13.5%

44

Summary

  • SNPs alone provide more risk stratification than “environmental” risk

factors or breast density alone.

  • Substantial improvements in breast cancer risk models can be attained

by integrating multiple risk factors.

  • Calibration in populations with large sample sizes is critical
  • Further model building and validation work is needed.
slide-12
SLIDE 12

6/9/2017 12

45

Acknowledgments

NCI/DCEG Amber Wilcox Tom Ahearn Johns Hopkins University Nilanjan Chatterjee Parichoy Choudhury Haoyu Zhang Runlong Tang Prosejit Kundu ICR, London Mark Brook Nick Orr Tony Swerdlow BCAC Doug Easton Nasim Mavaddat Kyriaki Michailidou Per Hall and Mikael Eriksson Celine Vachon Mia Gaudet Dale Sander, C Weinberg, J Taylor Antonis Antoniou Paul Pharoah Marjanka Schmidt Jenny Chang-Claude and Anja Rudolph Roger Milne Jaques Simard … and many more

Breast Cancer Association Consortium