Mendelian randomization

1
Mendelian randomization

• Hein Stigum
• http//folk.uio.no/heins/

Nov-09
H.S.
1
2
Cause versus Association

• Observe association, not necessarily causal
• Confounding
• Reverse causation
• Selection effect
• Time trends

Nov-09
H.S.
2
3
Definitions

• Association
• Observing E predicts D
• Causation
• Manipulating E predicts D

Nov-09
H.S.
3
4
Observational studies

• Successes
• Smoking ? lung cancer
• High blood pressure ? cardiovascular disease
• Failures
• Hormone replacement and cardiovascular disease
• Vitamin E and cancer

5
Randomized controlled trials

• Exposures
• Smoking, alcohol, air pollution, overweight
• Diseases
• cardiovascular disease, cancer, diabetes
• Limitations
• Unethical
• Long lead time
• Representative?

6
Problem

• Association ? causation
• Intervention on association may be useless
• Randomization not always feasible
• Need causal information from observational studies

Nov-09
H.S.
6
7
Mendelian randomization
Nov-09
H.S.
7
8
Mendels 2. law

• Law
• Alleles at two different loci are inherited
  independently
• Exception
• Loci close on the chromosome will be
  correlatedLinkage Disequilibrium (LD)

9
Observation versus trial

• Observational study
• all measured confounders adjusted for
• Randomized trial
• Strong effect, compliance
• Does not exist
• Does not exist
• R?D if and only if E causes D
• Medelian randomization
• Gene/ exposure association strong, or large N
• Should not exist, Mendels 2. law
• Must not exist, depends on the function of the
  gene
• G?D if and only if E causes D

U
2
E
D
R
1
3
U
2
E
D
G
1
3
Nov-09
H.S.
9
10
Ex Alcohol and blood pressure

• Observational study
• Alcohol use increases blood pressure
• Many lifestyle confounders
• Gene ALDH2, 2 alleles
• 2,2 type suffer nausea, headache after alcohol
• ? low alcohol regardless of lifestyle (U)
• Medelian randomization
• Gene ALDH2 is highly associated with alcohol
• Mendels 2. law, no ass. to obs. confounders
• OK, gene function is known
• Result 2,2 type BP -7.4 mmHg

U
2
A
BP
G
1
3
Nov-09
H.S.
10
11
Estimation

• If linear and single gene
• More genes 2 Stage Least Squares
• Regress E on G
• Find predicted values
• Regress D on predicted values
• Uses only the variation in D caused by G
• Stata, user written ivreg2

12
C-Reactive Proteine and insuline resistance
U Smoking Alcohol Physical activity socioeconomy

• Criteria
• Strong
• Not present, test theory
• Not present, weak theory

2
CRP level
Insulin resistance
CRP haplotype
1
3
Conclusion CRP does not cause insulin
resistance
13
Limitations, Violations of core conditions

• Gene/exposure association
• Suitable gene numbers increasing
• Reliable effect from many data sets
• Social pressure Ex lactate intolerance

U
E
D
G
1
14
Violations of core conditions

• 2., 3. G independent of U and D
• Pleiotropy
• Linkage disequilibrium
• OK, G a marker for G2
• Population stratification
• Stratify analysis

15
Sample size

• If GE effect is known (without error)
• Specify the ED effect you want to find
• Use equation and do sample size calculation on
  the GD-effect

16
Problem?

• Statement Multifactorial diseases can not be
  studied by Medelian randomization
• True or False?

• False! Study OK

17
Summing up

• Observational studies suffer from confounding and
  reverse causation
• Randomized trials are not always feasible
• Mendelian randomization is similar to RCT
• New randomizing genes are likely to be found
• Gene-exposure association weaker that in RCTs,
  need large N