This document discusses using genomic and lifestyle data for cardiovascular disease risk stratification. It summarizes that genome-wide studies have provided unprecedented information about genetic backgrounds of complex diseases. Large-scale prospective data combining genomic screening, lifestyle factors, and long-term health outcomes could be used to develop personalized risk algorithms and apps to communicate risk. Pilot studies in Finland have shown genetic risk scores can reclassify risk and identify additional individuals who could benefit from prevention.

1. Risk stratification using genomes and lifestyle
Samuli Ripatti, Prof, PhD
Public Health, Faculty of Medicine, University of Helsinki
Institute for Molecular Medicine Finland (FIMM)
Wellcome Trust Sanger Institute, UK

2. Causes of death in Finland
Suomen virallinen tilasto: Kuolemansyyt 2010
Cardiovascular
diseases
Tumors
Dementia,
Alzheimer’s disease

3. www.fimm.fi 3

4. www.fimm.fi
CHD prevention based on traditional risk
factors
2008 guidelines: statin
treatment for those >20% risk
Problem: 83% of cases have
risk < 20%
74% of those with risk >20%
do not get the disease (in 10
years)
4
Risk distribution
Frequency
0.0 0.2 0.4 0.6 0.8 1.0
01000200030004000

5. www.fimm.fi 5
Risk distribution
Frequency
0.0 0.2 0.4 0.6 0.8 1.0
01000200030004000

6. www.fimm.fi
40% of cases have risk
< 7.5%
83% of those with risk
>7.5% do not get the
disease (in 10 years)
6
Risk distribution
Frequency
0.0 0.2 0.4 0.6 0.8 1.0
01000200030004000

7. www.fimm.fi

8. Genome-wide studies have provided
unprecedented information about genetic
background of complex diseases and traits
ModifiedfromNHGRI/EBIGWASDiagramBrowser

9. www.fimm.fi

10. www.fimm.fi
How to use the genome in prevention?
1. Database: Large-scale prospective data with genomic screen
and follow-up recordings of health
2. Algorithms: to estimate the personalized risks
3. Apps: to communicate the risk to individuals
› Possibility to lower the risks through intervention
1

11. Aarno Palotie
Sequencing Initiative Suomi

12. Type 2 Diabetes
> 10 000 individuals
High Blood pressure
Severe mental health:
- schizophrenia, depression
- > 5000 individuals
Migraine
15 000 individuals
Old age dementia,
~ 5000 individuals
Cancer
> 10 000 cases
Life style and socio-economic data
- education, economic state, smoking
Cardiovascular events
stroke, CHD
25 000 individuals
Life course events
Prescription medication dat
18 000 statin users
20 000 estrogen substitution th
Cause of death data
Cardiovascular risk factor data
100 000 individuals
National Biobank
201 858 individuals
National Biobanks Finland
130 000 individuals from population cohorts
70 000 individuals from disease collections
Social security number

13. www.fimm.fi
40 year health event follow-up through health registries

14. SISU-project
Sequencing Initiative Suomi
National
Biobank
Disease genes
discovery
Extensive health,
phenotype,
metabolomic data
Population
cohorts
Disease specific
collections
Genome wide genotype
data
70,000
Genome/exome sequences
~ 16,000
more than double in a year
200 000 individuals
4% of the population
Variation
reference/Imputation
The 200K

15. www.fimm.fi13/09/2015 15
Apply to an individual patient
SISu provides a reference
database of genomic variation

16. www.sisuproject.fi

17. www.fimm.fi
LPA loss-of-function variant protecting
from CHD

18. www.fimm.fi
Rare high-impact vs. common low-impact variants
18
Single rare variant
dominating the risk
(>5×risk, >0.5sd)
Multiple genetic and other risk
factors contributing (<2×risk,
<0.2sd)
Growing evidence
for full range of
variants
contributing to the
risks of common
diseases

19. www.fimm.fi
Low frequency and functional
candidates contribute to lipid variances
Ida SurakkaHDL-C LDL-C TC TG
Trait variance explained in FRCoreExome9702 cohort
Varianceexplained(r2)
0.000.050.100.150.20
Adding low frequency lead SNPs
Common SNPs
FL SNPs and new
functional candidate SNPs
0.128
0.195
0.188
0.093
0.119
0.163 0.162
0.082
0.021
0.067
0.056
0.027
Surakka 2015 Nat Genet

20. The evidence:
• Familial hypercholesterolemia (FH) although frequent is underdiagnosed
Some of the most centralized health care
systems worldwide NL, NO ,IS, UK
Finland:
10,000 – 20,000 undiagnosed
FH cases
Nordenstgaard et al Eur Heart J 2013

21. Risk factor distributions overlapping
Framingham risk score at baseline:
age, sex, total cholesterol, HDL, BMI, systolic blood pressure,
blood pressure treatment,
current smoking status, diabetes mellitus, family history of CHD
Incident CHD cases
Ripatti Lancet 2010

22. www.fimm.fi
CHD risk variants
22
CARDIoGRAMplusC4D Consortium, Nat Genet 2013

23. Predicting heart disease risk with
genetic risk scores
Incident cases
Non-cases
Refining the risk
estimates using
genetic dataRipatti Lancet 2010
Tikkanen ATVB 2013

24. Stage 2: Screening
based on genetic risk
score
Stage 1: Screening for
Framingham risk factors
100,000
screened
>20% Risk
N=16169
- 3745 cases
- 12423 non-
cases
Assume
treated
10-20% Risk
N=18966
- 2481 cases
- 16485 non-
cases
Assume no
treatment,
test GRS
>20% Risk
N=2196
- 694 cases
- 1502 non-cases
10-20% Risk
N=13545
- 1630 cases
- 11915 non-
cases
<10% Risk
N=3225
- 157 cases
- 3069 non-cases
<10% Risk
N=64865
- 2458 cases
- 62408 non-
cases
Assume no
treatment nor
further testing
Assuming:
- 20% risk reduction with statin treatment
- Treatment compliance and efficacy similar in Stage 1 and Stage 2 high risk groups
 694*0.2 = 139 cases prevented in 14 years / 100,000 individuals (119 in 10 years)
 Compared to 17 for lp(a) (Di Angelantonio, JAMA 2012) and 30 for CRP (Kaptoge, NEJM
2012)
Statin treatment
added
Tikkanen et al ATVB 2013
12%
reclassified,
28% of cases

25. www.fimm.fi
Examples
› Using GRS on top of
traditional risk factors
25
58-year-old female
AGE
58 Baseline examination
Total cholesterol 5.0
HDL cholesterol 1.4
Systolic blood pressure 169, treated
Non-smoker
No diabetes
No family history of CVD
CHD risk 13.7%
High genetic risk score
CHD risk 21%
70 S422 Fracture of upper end of humerus
73 N179 Acute kidney failure, unspecified
I2141 Non-ST elevation (NSTEMI) myocardial
infarction
I509 Heart failure, unspecified

26. www.fimm.fi
Using genome-wide marker data for
prediction
Abraham, under review
Training with Cardiogram+C4D
Testing with Finrisk
Replicating in independent
cohorts
C index
41,000 SNPs
28 SNPs

27. www.fimm.fi
KardioKompassi: communicating risk to citizens
27

28. www.fimm.fi
- Consent
- Assessment of
cardiovascular
disease risk
- Ordering DNA-test
Blood samplingKardiokompassi
web-service
DNA-test
Participant X
Pilot:
Communicating
the risk
Traditional
risk factors
200 healthy volunteers
FIMM, SITRA,
Finnish Red Cross
Elisabeth Widén

29. www.fimm.fi
OPINIONS ABOUT THE KARDIOKOMPASSI
APPLICATION
Disagree
(%)
No opinion
(%)
Agree
(%)
I learned useful information regarding my health males 16 4 80
females 12.7 12.7 74.5
My personal disease risk information was reassuring males 12 28 60
females 26.5 25.5 48.1
My personal disease risk information motivated me to take
better care of my health males 12 24 64
females 11.8 25.5 62.6
The information I received was worrying males 64 32 4
females 65.7 16.7 17.7
The information I received was interesting males 0 4 96
females 3.9 5.9 90.2
My personal genetic risk information was confusing males 52 32 16
females 60.8 24.5 14.7
I was indifferent to the information provided on my personal
genetic risk males 84 12 4
females 83.3 9.8 6.9
The information on my genetic risk, in particular, motivated
me to take better care of my health males 12 24 64
females 16.7 18.6 64.7

31. Evidence for strong statin response in
high genetic risk group
Mega et al, Lancet 2015
High genetic risk
Risk
reduction

33. www.fimm.fi
Acknowledgements
› Magnetic consortium:
 Analysis team: Johannes Kettunen,
Tom Haller, Rene Pool, Ayse
Demirkan, Rajesh Rawal, Tune Pers,
Tonu Esko, Mari Niemi, Taru
Tukiainen, Harm-Jan Westra
 Cohorts: KORA, Finrisk, NFBC, YFS,
HBCS, FTC, Estonian biobank, ERF,
Dutch Twin Registry
› ENGAGE lipids scan:
 Analysis team: Ida Surakka, Momoko
Horikoshi, Reedik Mägi, Antti-Pekka
Sarin, Anubha Mahajan, Vasiliki
Lagou, Letizia Marullo, Teresa
Ferreira, Benjamin Miraglio, Sanna
Timonen, Johannes Kettunen, Matti
Pirinen, Juha Karjalainen, Gudmar
Thorleifsson, Sara Hägg, Jouke-Jan
Hottenga, Aaron Isaacs, Claes
Ladenvall, Marian Beekman, Tõnu
Esko, Janina S Ried, Christopher P
Nelson, Christina Willenborg, Harm-
Jan Westra
 22 cohorts
dd.mm.yyyy 33

34. www.fimm.fi
2
Prediction: 13 SNP study
Emmi Tikkanen
Kaisa Silander
Leena Peltonen
Marju Orho-Melander
Amitabh Sharma
Olle Melander
Aki S. Havulinna
Markus Perola
Antti Jula
Veikko Salomaa
Candace Guiducci
Elena Gonzalez
Sekar Kathiresan
HUCH
Juha Sinisalo
Markku S Nieminen
Haartman Institute
Marja-Liisa Lokki
Ripatti Lancet 2010

35. www.fimm.fi
Prediction: 28 SNP study
Emmi Tikkanen
Aarno Palotie
Aki Havulinna
Veikko Salomaa
Tikkanen ATVB 2013

36. www.fimm.fi
Kardiokompassi & GRS trial teams
› FIMM:
Elisabeth Widén
Kimmo Pitkänen
Emmi Tikkanen
Timo Miettinen
Olli Kallioniemi
› Carea:
Pasi Pöllänen
› Duodecim:
Pekka Mustonen
› Mehiläinen:
Kristiina Hotakainen
› Finnish Red Cross
Blood Service:
Jukka Partanen
36

37. www.fimm.fi
Ida Surakka: Lipid GWAS
Emmi Tikkanen: Risk
Prediction