Data quality/usability and population -based biobanks

Data quality/usability and Data quality/usability and populationpopulation-based biobanks-based biobanks

Paul Burton

Dept of Health Sciences Dept of Genetics

University of Leicester

StructureStructure of talk of talk Why does data quality/usability matter? UK Biobank as an illustration Statistical power of nested case-control studies Expected event rates in UK Biobank Biobank harmonisation Conclusions

Why does data Why does data quality/usability quality/usability

matter?matter?

Epidemiological analysisEpidemiological analysisat its simplestat its simplest

Odds ratio (OR) = (120*240)/(200*100) = 1.44 [1.04 – 2.0] May also adjust for a confounder

• e.g. high saturated fat intake [y/n] What is the impact of error in an outcome or an

explanatory variable or in a confounder?

Outcome CAD Not CAD

Smoker 120 200 Exposure Non-smoker 100 240

Systematic errorSystematic error

Some disease free smokers deny smoking Odds ratio (OR) = (120*250)/(190*100) = 1.58

Outcome CAD Not CAD

Smoker 120 200 190 Exposure Non-smoker 100 240 250

Random errorRandom error

At random, 10% of subjects state their exposure incorrectly

Odds ratio (OR) = (118*236)/(204*102) = 1.34

Outcome CAD Not CAD

Smoker 120 118 200 204 Exposure Non-smoker 100 102 240 236

The impact of errorsThe impact of errors Systematic errors in outcome or explanatory

variables systematic bias in either direction• True OR = 2 estimated OR = e.g. 1.5 or 2.7

Random errors in binary outcomes or any explanatory variables shrinkage bias• True OR = 2 estimated OR = e.g. 1.5

Random errors in confounding variables systematic bias in either direction• True OR = 2 estimated OR = e.g. 1.5 or 2.7

Errors in biobanksErrors in biobanks

Random errors• Loss of power is primary problem• Biobank sample sizes very large, so why is there a

problem?

Errors in biobanksErrors in biobanks

Random errors• But: why are biobank sample sizes so large?

NB Biobanks very large not nested case-control studies

• Need to detect small relative risks (e.g. OR=1.3)• Power generally limited (see later)• Small error effects catastrophic

Apparent causal effects easily created or destroyed

Errors in biobanksErrors in biobanks Systematic errors

• Small real effects a major issue again• Must understand data collection protocols, and

must attempt to optimise those protocols• UK Biobank• P3G Observatory

What is UK Biobank?What is UK Biobank?

A prospective cohort study 500,000 adults across UK Middle aged (40-69 years) A population-based biobank

• Not disease or exposure based• Recruitment via electronic GP lists

“Broad spectrum” not “fully representative” Individuals not families MRC, Wellcome Trust, DH, Scottish Executive

• £61M

Basic design featuresBasic design features

Longitudinal health tracking Nested case-control studies Long time-horizon

Owned by the Nation Central Administration – Manchester

• PI: Prof Rory Collins - Oxford 6 collaborating groups (RCCs) of university

scientists

Basic design featuresBasic design features

Statistical powerStatistical powerand sample sizeand sample size

Focus on power of nestedFocus on power of nestedcase-control analysescase-control analyses

Likely to be very common analyses Power limiting

Issues that are often ignored in Issues that are often ignored in standard power calculationsstandard power calculations

Multiple testing/low prior probability of association* Interactions* Unobserved frailty Misclassification*

• Genotype• Environmental determinant• Case-control status

Subgroup analyses* Population substructure

Power calculationsPower calculations Work with least powerful setting

• Binary disease, binary genotype, binary environmental exposure

Logistic regression analysis; interactions = departure from a multiplicative model

Complexity (arbitrary but reasonable)Frailty

variance Symmetrical genotyping error rate

Symmetrical environmental

error rate

Disease to non-diseased

error rate

Non-diseased to diseased error rate

Critical P-value

Power

100 fold 1% 20% 80% 0.2% 10-4 80%

Summarise power using “Minimum Summarise power using “Minimum Detectable Odds Ratios” (MDORs) Detectable Odds Ratios” (MDORs) calculated by ‘iterative simulation’calculated by ‘iterative simulation’

Estimate minimum ORs detectable with 80% power at stated level of statistical significance under specified scenario

Genetic main effectsGenetic main effectsNumber of cases

1,000 2,500 5,000 10,000 20,000 0.05 2.64 1.99 1.67 1.48 1.32 0.1 2.08 1.65 1.46 1.31 1.22 0.2 1.79 1.47 1.32 1.24 1.16

0.33 1.68 1.39 1.28 1.19 1.14

Prevalence of

genotypic exposure

0.5 1.65 1.39 1.27 1.19 1.13

Whole genome scanWhole genome scan Genetic main effect, p<10-7

Number of cases

2,500 5,000 10,000 20,000 0.05 2.30 1.86 1.65 1.41 0.1 1.87 1.60 1.42 1.27 0.2 1.63 1.44 1.30 1.21

0.33 1.54 1.39 1.24 1.18

Prevalence of genotypic exposure

0.5 1.52 1.35 1.24 1.16

Gene:environment interactionGene:environment interaction 20,000 cases

Genotype Prevalence

0.1 0.2 0.33 0.5 0.1 2.47 1.94 1.82 1.72 0.2 1.97 1.67 1.58 1.54

0.33 1.79 1.58 1.47 1.45

Prevalence of Environmental

Exposure

0.5 1.79 1.61 1.46 1.44

Summary – rule of thumbSummary – rule of thumb 80% power for genotype frequency = 0.1, (allele

frequency 0.05 under dominant model) • Genetic main effect 1.5, p=10-4 5,000 cases• Genetic main effect 1.3, p=10-4 10,000 cases• Genetic main effect 1.2, p=10-4 20,000 cases• Genetic main effect 1.4, p=10-7 10,000 cases• Genetic main effect 1.3, p=10-7 20,000 cases• G:E interaction with environmental exposure prevalance = 0.2 2.0, p=10-4 20,000 cases

Effect of realistic data errorsEffect of realistic data errorsScenario Genetic MDOR Environmental MDOR No error 1.20 1.20

Explanatory error only 1.25 1.48 Outcome error only 1.31 1.31 Both types of error 1.40 1.72

Expected event ratesExpected event ratesin UK Biobankin UK Biobank

Taking account ofTaking account of Age range at recruitment 40-69 years Recruitment over 5 years All cause mortality Disease incidence (“healthy cohort effect”) Migration overseas Comprehensive withdrawal (max 1/500 p.a.)

Time to achieve 1,000 cases





Bladder cancer 11 yrs 19 yrs 31 yrs - - Breast cancer (F) 4 yrs 6 yrs 10 yrs 17 yrs 40 yrs Colorectal cancer 5 yrs 9 yrs 14 yrs 22 yrs 42 yrs Prostate cancer (M) 6 yrs 9 yrs 14 yrs 22 yrs 41 yrs Lung cancer 7 yrs 12 yrs 19 yrs 34 yrs - Non-Hodgkins lymphoma 11 yrs 22 yrs - - - Ovarian cancer (F) 12 yrs 26 yrs - - - Stomach cancer 16 yrs 29 yrs - - - Stroke 5 yrs 8 yrs 12 yrs 18 yrs 28 yrs MI and coronary death 2 yrs 4 yrs 5 yrs 8 yrs 13 yrs Diabetes mellitus 2 yrs 3 yrs 4 yrs 6 yrs 10 yrs COPD 4 yrs 6 yrs 8 yrs 13 yrs 23 yrs Hip fracture 7 yrs 11 yrs 15 yrs 21 yrs 31 yrs Rheumatoid arthritis 7 yrs 14 yrs 27 yrs - - Alzheimer’s disease 7 yrs 10 yrs 13 yrs 18 yrs 23 yrs Parkinson’s disease 6 yrs 10 yrs 15 yrs 23 yrs 37 yrs

No need to contact subjectsNo need to contact subjects

Smaller sample sizesSmaller sample sizes 500K 250K 100K [1,] Bladder cancer 7500 3800 1500 [2,] Breast cancer:females only 21800 10900 4400 [3,] Colorectal cancer 21800 10900 4400 [4,] Prostate cancer:males only 22000 11000 4400 [5,] Lung cancer 12600 6300 2500 [6,] Non-Hodgkins lymphoma 4900 2400 1000 [7,] Ovarian cancer:females only 3800 1900 800 [8,] Stomach cancer 3900 1900 800 [9,] Stroke 34700 17400 6900 [10,] MI and coronary death 59200 29600 11800 [11,] Diabetes Mellitus 89300 44600 17900 [12,] COPD 35500 17700 7100 [13,] Hip fracture 37900 18900 7600 [14,] Rheumatoid.Arthritis 6800 3400 1400 [15,] Alzheimers Disease 73900 36900 14800 [16,] Parkinsons Disease 24900 12500 5000

Interim conclusionsInterim conclusions Having taken account of realistic bioclinical complexity, UK

Biobank is just large enough to be of great value as a stand-alone research infrastructure

Data quality, in particular errors in outcome or explanatory variables, or in confounders is crucial

Its value will be greatly augmented if it proves possible to set up a coherent and scientifically harmonized international network of Biobanks and large cohort studies

Harmonising Harmonising biobanks biobanks

internationallyinternationally

Why harmonise?Why harmonise?

Basic aim is to enable and promote data pooling, in a manner that recognises and takes appropriate account of systematic differences between studies.

Why harmonise?Why harmonise? Investigate less common (but not rare)

conditions• UKBB: Ca stomach 2,500 cases in 29 years

6 UKBB equivalents: 10,000 cases in 20 years Investigate smaller ORs

• GME 1.5 1.2 requires 5,000 20,000 4 UKBB equivalents

Analysis based on subsets – homogeneous classes of phenotype, or e.g. by sex

Why harmonise?Why harmonise? Earlier analyses

• UKBB: Alzheimers disease, 10,000 cases in 18 yrs 5 UKBB equivalents 9 years

Events at younger ages Broad range of environmental exposures Aim for 4-6 UKBB equivalents

• 2M – 3M recruits

Harmonisation initiativesHarmonisation initiatives Public Population Program in Genomics (P3G)

• Canada + Europe Tom Hudson, Bartha Knoppers, Leena Peltonen, Isabel Fortier …..

Population Biobanks• FP6 Co-ordination Action (PHOEBE – Promoting Harmonisation Of

Epidemiological Biobanks in Europe) Camilla Stoltenberg, Paul Burton, Leena Peltonen, George Davey Smith …..

Harmonisation in the P3G Harmonisation in the P3G ObservatoryObservatory

(from Isabel Fortier)(from Isabel Fortier) Description Comparison Harmonisation

Data quality crucial at every stage

Final conclusionsFinal conclusions

Power of individual biobanks is limited Minimisation of measurement error is crucial Harmonisation is crucial if we are to optimise

the value of biobanks internationally Harmonisation depends on a full understanding

of all aspects of data quality

Extra slidesExtra slides

Rarer genotypesRarer genotypes

Genetic main effects

Number of cases

Genotype prevalence

Environmental exposure

prevalence

Symmetrical genotyping error rate

Symmetrical environmental

error rate

Disease to non-diseased

error rate

Non-diseased to diseased error rate

Critical P-value

Minimum detectable*

OR for genetic effect

10000 0.01 0.2 1% 20% 80%1 0.2%1 10-4 2.37 20000 0.01 0.2 1% 20% 80%1 0.2%1 10-4 2.09 10000 0.002 0.2 1% 20% 80%1 0.2%1 10-4 [[8.2]] 20000 0.002 0.2 1% 20% 80%1 0.2%1 10-4 [[5.1]]

Gene:environment interactionGene:environment interaction 10,000 cases

Genotype Prevalence

0.1 0.2 0.33 0.5 0.1 3.03 2.36 2.11 2.05 0.2 2.32 1.95 1.87 1.78 0.33 2.15 1.80 1.68 1.64

Prevalence of Environmental

Exposure

0.5 2.16 1.86 1.70 1.70

Disease Gene Polymorphism Approximate frequency of the disease associated allele

Approximate odds ratio for disease associated allele

Thrombophilia F5 Leiden Arg506Gln

0.03 4

Crohn’s disease CARD15 3 SNPs 0.06(composite) 4.6

Alzheimer’s disease

APOE 2/3/4 0.15 3.3

Osteoporotic fractures

COL1A1 Sp1 restriction site

0.19 1.3

Type 2 diabetes KCNJ11 Glu23Lys 0.36 1.23

Type 1 diabetes CTLA4 Thr17Ala 0.36 1.27

Graves’ Disease CTLA4 Thr17Ala 0.36 1.6

Type 1 diabetes INS 5’ VNTR 0.67 1.2

Bladder Cancer GSTM1 Null (gene deletion)

0.70 1.28

Type 2 diabetes PPARG Pro12Ala 0.85 1.23

Hattersley AT, McCarthy MI. A question of standards: what makes a good genetic association study?Lancet 2005; in press.

Summarise power using MDORs Summarise power using MDORs calculated by ‘iterative simulation’calculated by ‘iterative simulation’ Want minimum ORs detectable with 80% power at stated

level of statistical significance• 1. Guess starting values for ORs• 2. Simulate population under specified scenario• 3. Sample required number of cases and controls• 4. Analyse resultant case-control study in standard way• 5. Repeat 2,3,4 1,000 times• 6. Use empirical statistical power results from the 1,000

analyses to update ORs to new values expected to generate a power of 80%

• Repeat 2-6 till all ORs have 80% power

Taking account ofTaking account of Age range at recruitment 40-69 years Recruitment over 5 years All cause mortality Disease incidence (“healthy cohort effect”) Migration overseas Comprehensive withdrawal (max 1/500 p.a.) Partial withdrawal (c.f. 1958 Birth Cohort)






Bladder cancer 12 yrs 21 yrs 39 yrs - - Breast cancer (F) 4 yrs 7 yrs 11 yrs 19 yrs - Colorectal cancer 6 yrs 10 yrs 15 yrs 25 yrs - Prostate cancer (M) 6 yrs 10 yrs 15 yrs 24 yrs - Lung cancer 7 yrs 13 yrs 22 yrs - - Non-Hodgkins lymphoma 12 yrs 26 yrs - - - Ovarian cancer (F) 13 yrs 33 yrs - - - Stomach cancer 17 yrs 36 yrs - - - Stroke 5 yrs 8 yrs 12 yrs 19 yrs 33 yrs MI and coronary death 2 yrs 4 yrs 5 yrs 8 yrs 14 yrs Diabetes mellitus 2 yrs 3 yrs 5 yrs 7 yrs 11 yrs COPD 4 yrs 6 yrs 9 yrs 14 yrs 27 yrs Hip fracture 7 yrs 12 yrs 16 yrs 24 yrs 36 yrs Rheumatoid arthritis 7 yrs 15 yrs 36 yrs - - Alzheimer’s disease 7 yrs 11 yrs 14 yrs 19 yrs 26 yrs Parkinson’s disease 7 yrs 11 yrs 17 yrs 26 yrs -

Necessary to contact subjectsNecessary to contact subjects

Study Sample size

Recruitment Age at recruitment

Web site

Cohort studies

EPIC Europe >500,000 1993-1997 45-74 http://www.iarc.fr/epic/ centers/iarc.html

ProtecT Study 120,000 1999-2006 50-69 http://www.epi.bris.ac.uk/ protect/index.htm

Kadoorie Study China

500,000 Ongoing 35-74 http://www.ctsu.ox.ac.uk

Prospective Study Mexico

200,000 Ongoing >40 http://www.ctsu.ox.ac.uk/projects/mexicoblood.shtml

UK Biobank 500,000 2006-2010 40-69 http://www.ukbiobank.ac.uk/

Birth Cohorts

Mother and Child Cohort Study (Norway)

100,000 babiesA

2001-2005 At birth http://www.fhi.no/

Danish National Birth Cohort

100,000 babiesB

1997-2002 At birth http://www.serum.dk/ sw9314.asp

Twin Cohorts

GenomEUtwin >600,000 twin pairs

Various Various http://www.genomeutwin.org/

Total populations

Decode Genetics >100,000 Iceland Various http://www.decode.com/ Estonian Genome Project

>100,000C Estonia Various http://www.geenivaramu.ee

Western Australian Genome Project

~2,000,000D Australia Various http://www.genepi.com.au/wagp

Data quality/usability and population -based biobanks

Documents