Importance of data collection for inherited disease research

1
Importance of data collection for inherited
disease research

• John L. Hopper
• Centre for Molecular, Environmental, Genetic, and
  Analytic (MEGA) Epidemiology
• The University of Melbourne

2
Question of critical clinical importance

• What are the disease risks associated with
  particular genetic variants?
• Answering this is no simple matter
• Importance of adjusting for ascertainment
• Need to use valid statistical methods
• Clarity about who inference is being made
• Imprecision and bias of estimates

3
Designs

• Population-based sampling
• Case-control
• Case-control-family
• Family
• Community sampling
• Cohort
• Twin studies
• Opportunistic sampling
• Multiple-case families
• Tumour banks

4
Basics of Statistical Inference
RANDOM
POPULATION
SAMPLE
PARAMETER ESTIMATE
UNKNOWN PARAMETER
INFERENCE
5
DESIGN,DESIGNDESIGN
6

• ascertainmentASCERTAINMENTascertainmentASCERTAINM
  ENTascertainmentASCERTAINMENTASCERTAINMENTascertai
  nmentASCERTAINMENTascertainmentASCERTAINMENTascert
  ainmentASCERTAINMENTascertainment

7
CONCLUSION

• SAMPLING ISSUES
• ARE
• IMPORTANT

8
Australian Breast Cancer Family Study

• Population-based sampling
• - 1,600 Cases (Vic NSW Cancer Registries)
• - 1,000 Controls (Electoral Rolls)
• Cases - unselected for family history
• - over-sampled for early onset (50 lt age 40)
• Relatives (1o 2o) of cases and controls
• Epidemiological data (by questionnaire)
• Genetic and molecular (blood, tissue)
• Part of NIH-funded Breast Cancer Family Registry
• John, Hopper, et al., Br Cancer Res 2004

9
BRCA1 1bp del 1876 (C)
81
88
75
70
?Breast(42)
63
74
47
65
61
57
73
53
69
42
60
64
50
61
58
Breast(48)
Prostate(65)
?Breast(53)
?Larynx(60)
?Lung(50)
_
40
41
37
34
38
Breast(38)

3159
10
BRCA2 2bp del 6503 (TT)
64
57
75
6
6
76
78
58
0
12
60
67
59
Breast(62)
Breast(55)
Breast(38)
Bowel(58)
()

?Bowel(67)
38
54
52
50
41
44
41
Breast(38)
_
_
_




12
12
10
8
3172
11

• That was the outlier
• It does not represent a
• typical BRCA1/2 family!

12
hereditary
familial
sporadic
13
lt1
1 in 2
gt2 affected
5
1 in 8
2 affected
Familial
35
1 in 16
1 affected
1 in 20
0 affected
Sporadic
60
Hereditary
14
Penetrance

• Average age-specific cumulative risk in defined
  sets of carriers
• Risk associated with mutation(s)
• Mutations rare so need family design
• Estimate is of average risk over the (types of)
  mutations in the population
• Have to take into account how mutation-segregating
  families were ascertained
• Clinic-based families have about 3 times less
  information per carrier family than
  population-based families

15
BRCA1
Multiple-case families family cancer clinics
Population-based study Antoniou et al (2003)
16
BRCA2
Multiple-case families family cancer clinics
Population-based study Antoniou et al (2003)
17
Colorectal Cancer Risk for hMLH1 and hMSH2
Mutation Carriers
100
INCORRECT ESTIMATES from MISANALYSIS of data
from multiple-case families from family cancer
clinics
80
60
Cumulative risk of colorectal cancer ()
40
Population-based study Jenkins et al (2005)
20
0
20
30
40
50
60
70
Age (years)
18
Conclusions

• Population-based studies often best resource to
  minimise problems
• Must be analysed and interpreted properly
• Epidemiology provides solid theoretical basis and
  tools
• Genetics arent trained in epidemiology
• Multi-disciplinary approach needed
• Large informative studies needed

19
Solutions?

• Expert committee to gather data and estimate
  risks
• e.g. Antoniou et al. (Cambridge) for breast
  cancer
• Jenkins et al. (Melbourne) for colorectal cancer,
  using Colon CFR and other resources

20
Model/Predict Risk for Carriers

• Hazard ratio increased risk cf. population
• Age of individual at risk
• Age at onset of proband
• Mode of family ascertainment/ FH
• Type other characteristics of the mutation
• Grantham scores
• Functional assays
• Gives the clinician the best predictor of risk
• (not Deleterious? yes, no, dont know!)

21
Australasian Colorectal Cancer Family Study

• Population-based sampling
• - 800 Cases (Vic Cancer Registry)
• - 400 Controls (Electoral Rolls spouses)
• Cases - unselected for family history
• - over-sampled for early onset (50 lt age 45)
• Relatives (1o 2o) of cases and controls
• Epidemiological data (by questionnaire)
• Genetic and molecular (blood, tissue)
• 500 multiple-case families from Australia NZ
• Part of NIH-funded Colon Cancer Family Registry

22
SEATTLE
LOS ANGELES
HAWAII
ADVISORY COMMITTEE
MAYO
STEERING COMMITTEE
ONTARIO
AUSTRALASIA
WORKING GROUPS
RESEARCH GROUPS
EXTERNAL RESEARCH GROUPS
INFORMATICS SUPPORT CENTER
CENTRAL BIOSPECIMENS REPOSITORY
NCI PROGRAM
23
(No Transcript)
24

• Great opportunity for Australia to show
  leadership