Transcript Genetic Susceptibility Risk Models in Clinical Decision Making

Genetic Susceptibility Risk Models
in Clinical Decision Making
Susan M. Domchek, MD
Abramson Cancer Center
University of Pennsylvania
Oncologist
PCP
Self-referral
Cancer Risk Evaluation Program
Models
>10%
Genetic
Testing
+
<10%
Gail and
Claus
-
BSO
Screening studies
Prevention studies
PM
>25%
Screening studies
Chemoprevention
BRCA prediction models
• Logistic regression models (Couch,
Shattuck-Eidens, Frank)
• Bayesian formulations (BRCAPRO)
• Empiric tables (Frank 2002)
• Prevalence tables
• Unique attributes to each model
• Consideration of testing for women with
a probability of 10%
Limitations of family history
• Adoption
• Small family size, especially women
– Prevalence tables can be very helpful
• Early deaths
• Accuracy of cancer information
– Stomach cancer in women
– Obtain medical records whenever possible
Limitations of all models
•
•
•
•
Race/ethnicity data
How to handle DCIS
LCIS
“Other” cancers – pancreatic cancer,
melanoma, early prostate
1
2
d. 65
d. 58
3
4
BR
d. 49
5
d. 68
Variability in models
6
76
38
7
Pharynx Ca 63
d. 65
8
9
Prostate Ca 63
d. 65
Lung Ca 47
d. 47
6
7
40
37
47
1
49
2
BR Ca 42
d. 43
3
4
LCIS 49
50
53
5
BR Ca 50
52
Myriad Tables: 21.2% (47% in 2 <50)
Couch: 7.7% for family
BRCAPro: Dependent on proband – 55% vs 1.6%
What is the goal of prior probability
models?
• Identify candidates for testing for
BRCA1/BRCA2
– Do we care more about sensitivity or
specificity?
– Clinically: sensitivity
– Economically: specificity
• Stratify risk of hereditary syndromes
– In tested negative families should we do
counseling based on PP models?
“False” negative: what to counsel?
6
7
BR_CA 28
OV_CA d. 55
d. 48
15
62
BR_CA 52
14
59
Panc_CA 55
d. 60
63
3
2
Skin_CA 78
d. 55
64
d. 10
65
d. 7
60
87
Liver_CA
Panc_CA d. 45
d. 47
1
BR_CA 48
8
9
10
Eye_Melan 38
11
Which syndrome?
10
11
OV_CA 42
d. unk
What ovarian
cancer risk?
7
Pros_CA 80
d. 56
5
Colon_CA 61
d. 63
d. 83
12
6
13
d. 42
4
d. 64
3
2
Lung_CA 76 OV_CA 69
d. 81
d. 70
1
8
9
Can pathologic features help?
• BRCA1 mutation related breast cancers
– 90% are estrogen receptor negative
– High grade, aneuploid, “pushing margins”
– 3% are HER2/neu positive
• BRCA2 mutation related breast cancers
– More like sporadic tumors
– Approximately 50% are ER positve
– Only 3% HER2/neu positive
Probability of BRCA1 mutation by
age, ER status and grade
ER positive tumors
Age
All (%) Grade 1(%) Grade 2 (%)
Grade 3 (%)
<30
8
1.1
1.6
2.7
30-34
5
0.8
1.2
2.0
35-39
2
0.2
0.3
0.5
40-44
1.5
0.1
0.2
0.3
45-49
1
0.1
0.1
0.2
50-59
0.3
0.03
0.04
0.07
Lahkini et al, JCO 2002
Probability of BRCA1 mutation by
age, ER status and grade
ER positive tumors
Age
All
(%)
Grade
1(%)
Grade 2
(%)
Grade 3
(%)
<30
8
14.4
21.0
35.0
30-34
5
10.9
15.9
26.5
35-39
2
2.7
4.0
6.6
40-44
1.5
1.5
2.2
3.7
45-49
1
1.0
1.5
2.5
50-59
0.3
0.4
0.6
0.9
Claus and Gail in Hereditary Families
1
2
OV Ca 59
d. unk
d. 61
1
2
3
4
62
BR Ca 50
67
60
2
3
4
33
40
35
OV Ca 52
d. 54
1
BR Ca 35
36
Issues in clinical decision
making
• “Hereditary” patterns that test negative
• How to define them?
• Is breast cancer risk assessment
accurate?
• What is their ovarian cancer risk?
• Risk assessment in VUS?