1. Automated -80°C archive: rapid access to samples

2. Rationale for enhanced phenotyping in large subsets of study participants More precise assessment of risk factors (e.g. internet-based diet questionnaire; mailed accelerometers for activity) Assessment of unmeasured risk factors (e.g. ambulatory monitoring of BP variability; mailed work/home environmental monitors) Information from imaging modalities (e.g. brain/heart/body MRI; bone/joint DEXA) Increased power for studying a wider range of associations with disease

3. Proposed imaging assessment visit in 100K participants during 2012-2017 Minutes MRI (3T) of brain 30 MRI (1.5T) of heart and body 30 3D carotid ultrasound 15 DEXA of bone and joint 15 12 lead ECG 15 Submit application in 1Q 2012 for pilot in 4-8,000 participants prior to main phase of imaging assessments (2012-17)

4. UK Biobank: Centralised follow-up of health outcomes Death and cancer registries In-patient and out-patient hospital episodes (including psychiatric) and related procedure registries Primary care records of health conditions, prescriptions, diagnostic tests and other investigations Other health-related: disease registries; dispensing records; imaging; screening; dental records Direct to participants: self-reported medical conditions; treatments actually being taken; degree of functional impairment; cognitive and psychological scores

5. Adjudication of health outcomes in order to confirm or refute (validate) and to sub-classify (phenotype) Enhancement of power to detect associations between risk factors and disease outcomes (false positive diagnoses have main adverse impact) Increased specificity of disease classification allows the detection of specific associations (i.e. risk factor may only be linked to disease sub-type) “Future-proofing” of the outcome data so that more detailed phenotyping is possible in future (i.e. retain data/samples to allow refined sub-typing)

6. Avoid false positives: Misclassification of disease status reduces power in observational studies

7. Adjudication of health outcomes in order to confirm or refute (validate) and to sub-classify (phenotype) Enhancement of power to detect associations between risk factors and disease outcomes (false positive diagnoses have main adverse impact) Increased specificity of disease classification allows the detection of specific associations (i.e. risk factor may only be linked to disease sub-type) “Future-proofing” of the outcome data so that more detailed phenotyping is possible in future (i.e. retain data/samples to allow refined sub-typing)

8. Increased specificity of disease classification Validating non-specific clinical diagnoses (e.g. confirmation of Parkinson’s disease requires information about symptoms and exclusion of secondary causes) Dissecting heterogeneous clinical syndromes (e.g. distinction of latent autoimmune diabetes in adult [LADA] from other type 2 diabetes requires autoantibody test) Distinguishing pathological disease subtypes (e.g. classification of aetiological subtypes of stroke requires CT/MRI brain scans, echocardiogram & carotid artery images) Distinguishing histological disease subtypes (e.g. cancer classification based on microscopic appearance, biochemical criteria, molecular signatures, receptor assays)

9. Different associations of lacunar and non-lacunar subtypes of ischaemic stroke with various risk factors

10. Increased specificity of disease classification Validating non-specific clinical diagnoses (e.g. confirmation of Parkinson’s disease requires information about symptoms and exclusion of secondary causes) Dissecting heterogeneous clinical syndromes (e.g. distinction of latent autoimmune diabetes in adult [LADA] from other type 2 diabetes requires autoantibody test) Distinguishing pathological disease subtypes (e.g. classification of aetiological subtypes of stroke requires CT/MRI brain scans, echocardiogram & carotid artery images) Distinguishing histological disease subtypes (e.g. cancer classification based on microscopic appearance, biochemical criteria, molecular signatures, receptor assays)

11. Reproductive factors and breast cancer risk by hormone receptor status of the tumour

12. Association of genetic variant with different histological subtypes of epithelial ovarian cancer

13. “Future-proofing” of health outcome data Rapid collection of information or samples reduces likelihood of their subsequent loss or degradation Pro-active approach to adjudication of disease outcomes reduces delays for use by researchers Centralised data/sample collection and adjudication should yield more reproducible disease diagnoses Internationally-accepted diagnostic criteria should enhance ability to combined data from studies Storage of imaging data and tissue samples would allow further analyses and classification in future

14. Scalability: semi-automatic approach to phenotyping required with large numbers of health outcomes UK Biobank: Expected numbers of incident health outcomes Condition201220172022 Diabetes10,00025,00040,000 MI/CHD death7,00017,00028,000 Stroke2,0005,0009,000 COPD3,0008,00014,000 Breast cancer2,5006,00010,000 Colorectal cancer1,5003,5007,000 Prostate cancer1,5003,5007,000 Lung cancer5002,0004,000 Hip fracture1,0002,5006,000 Alzheimer’s5003,0009,000

15. Staged approach to identification and adjudication of health outcomes Identification of probable incident cases Confirmation of cases Sub-classification of cases All Biobank participants Suspected cases only Confirmed cases only

16. Approach to adjudication: different characteristics of data required at different stages ApproachCharacteristicsPossible examples Ascertainment of suspected cases Cost-effective Feasible Geographically generalisable Scalable Cause-specific mortality Cancer incidence Hospital discharge records GP records Web self-report questionnaire Confirmation of “caseness” As above but somewhat higher cost/lower feasibility Existing morbidity registers (eg MINAP) Cross-referencing e-records Targeted blood sampling with cheap assays Classification of confirmed cases More involved and costlyTargeted blood sampling with costly assays Tumour collection/assays Specialised databases (eg imaging) Review of clinical records

17. Strategy for access to the UK Biobank Resource for health-related research in the public interest The Resource is expected to be used chiefly for a series of case-control studies of different outcomes within the cohort (based on anonymised data sets and/or samples) Timetable will be developed to indicate when sufficient cases of each condition are scheduled to have occurred Researchers can then develop proposals against this indicative timetable, which UK Biobank would use to coordinate calls for efficient use of the resource Review of disease-specific proposals by UK Biobank and its Access Sub-Committee (with the process kept under review by the Ethics & Governance Council) No preferential or exclusive access to the resource (access procedures: www.ukbiobank.ac.uk)

18. Advantage of “nested” case-control approach to sample and data analysis Assays need to be conducted only on relatively small subset (e.g. few thousand or tens of thousands) –Substantially reduces costs of assays –Minimises depletion of limited sample –Facilitates good quality control Assays need to be conducted only after sufficient numbers of relevant disease cases have occurred –Specific hypotheses will be clearer after delay –Wider range of assays possible at lower cost

19. UK Biobank: what makes it so special? Prospective: Assess the full effects of a particular exposure (e.g. smoking) on all health outcomes (e.g. cancer, vascular disease, lung disease) Detailed: Wide range of questions, measures and samples allows good assessment of exposures, and adjudication allows good disease phenotyping Big: Sufficiently large number of people develop a condition to allow reliable assessment of causes, and of interactions between different exposures Reliable identification of novel lifestyle, environmental and genetic causes of many different conditions will lead to new methods for preventing and treating disease