Transcript Systematic reviews of genetic association studies

Systematic reviews of genetic
association studies
Robert Walton
Fiona Fong
15 March 2013
Outline of session
• Reasons for doing a systematic review
• Differences in methods between genetic
systematic review and conventional
• Assessment of bias
• Meta analysis
• A practical example of a genetic systematic
review in progress – Fiona Fong
Why do a genetic systematic review?
• Identify genes previously studied and positive or
negative associations with different outcomes
• Standardise statistical analysis
• Make sub group analyses
• Plan future work
• Make grant applications
• Publish!
Genetic systematic reviews are
generally well cited in the literature
The genetic basis for smoking behavior: a systematic review and meta-analysis
Marcus R Munafò, Taane G Clark, Elaine C Johnstone, Michael FG Murphy, Robert T Walton
Cited by 213
Human Genome Epidemiology
Network
• Provides online resources – links to suitable
papers
• Guidelines for performing and writing genetic
systematic reviews
• Center for disease control - Atlanta
What's so different about a
genetic systematic review?
Genetic systematic reviews are very similar to
systematic reviews of observational studies
• Very important to work out the question fully and
precisely
• Abstract reviewing paper selection and data extraction
are the same
• Meta analysis is very similar need to consider the
genetic question carefully too
• Interpretation of the results may need to take into
account an understanding of how genes work
Specific genetic factors to consider
when performing a review
• Linkage disequilibrium
• Hardy Weinberg equilibrium
• Different models of gene action
Assessment of bias
• Selection bias
– Extreme vs unselected cases
– Use of prevalent cases
– Using a phenotypic test
– Biased selection of controls
– Differential participation and dropout
Assessment of bias
• Information bias
–
–
–
–
–
–
Misclassification of genotype
Were the laboratory staff blind?
Using a phenotypic test
Biased selection of controls
Differential participation and dropout
Genotyping error
Assessment of bias
• Confounding
– Population stratification
• Family studies TDT
• Genomic controls
• But how much of a problem is it really?
– Other
Meta analysis of genetic studies
• Useful not just for summary estimate but to investigate
heterogeneity
• Meta regression
• Odds ratios, differences in means and standardised mean
differences
• Choice of genetic model
• Sensitivity analysis – Hardy Weinberg deviation
• Use of individual patient data
A practical example of a genetic
systematic review in progress
An example
• Our topic: Genetic factors and pre-eclampsia
• Register with PROSPERO
• Our new topic: Genetics factors and
complications of pre-eclampsia
Design
•
•
•
•
•
Protocol
Comprehensive search
Data extraction
Validity of studies
Meta-analysis
Inclusion criteria
Case control/cohort studies
Complications of pre-eclampsia
Medline
Maternal
Embase genotype(s) tested
Can
extract data
into 2x2 table
Cochrane
2 independent
reviewers
rd reviewercriteria
Exclusion
of references
3Handsearching
if discrepancy
Genome
wide association
from
reviews
/ includedstudies
No gold
standard!
studies
Study design – Newcastle
HuGENavigator
Ottawa Scale
Genetically ‘sound’ – STREGA
(STrengthening the REporting of
Genetic Association Studies)
Additional elements – Data extraction
Traditional meta-analysis
Intervention
Genetic meta-analysis
Control
TT
Observe
Pre-eclampsia
CC
Observe
No pre-eclampsia
Intervention
TC
Pre-eclampsia
Control
No pre-eclampsia
CCCC
+ TC
Outcome 1
Outcome 1
Outcome 2
Outcome 2
Dominant
Recessive
TTTT
+ TC
Which genetic model?
• 3 groups+
– Dominant (CC + TC vs TT)
– Recessive (CC vs TT + TC)
– Co-dominant (CC vs TT, CC vs TC, TT vs TC)
• Choose a model based on previous evidence
• Look at control group genotype frequencies to
determine minor allele (ie aa)
Additional elements – STREGA
STrengthening the REporting of Genetic Association studies
To enhance transparency of reporting
– Methods variables
• Population stratification (eg ethnicity)
• Nomenclature system
• Genotyping errors
– Data sources ie DNA processing
– Hardy Weinberg Equilibrium
Additional elements - HWE
Hardy Weinberg
Equilibrium
A concept of population
genetics
p2 + 2pq + q2 =1
p2 = genotype AA
2pq = genotype Aa
q2 = genotype aa
Our methodological quality
assessment table
Processing the results
What does this lead to?
• Successful systematic reviews of genetic studies
can collate evidence across all studied genetic
variants for a phenotype to form genetic
association evidence databases.
– Alzheimer disease (Alzgene database)
– Parkinson disease (PDGene database)
– Schizophrenia database (SzGene database)
The systematic review process
Formulate
research
question
Design
search
strategy
Nomenclature
Further selection of
primary studies
using inclusion
criteria
Extract data
Genetic model
(dominant?)
Retrieve papers
Quality
appraisal
Search
bibliographic
databases
HUGE
Identify possible
papers from
titles/abstracts
Synthesis
STREGA
Formulate research /
policy conclusions
Useful resources
• HuGENet handbook
– http://www.medicine.uottawa.ca/public-healthgenomics/web/assets/documents/HuGE_Review_Handbook_V1_0.pdf
• STREGA
– http://link.springer.com/article/10.1007%2Fs00439-008-0592-7
• PROSPERO
– http://www.crd.york.ac.uk/Prospero/
• Hardy Weinberg Equilibrium calculator
– http://www.tufts.edu/~mcourt01/lab_protocols.htm