Table 1 Methodologies for collecting real-world data [4] and examples of their application to multiple sclerosis (MS) studies

RCTs extensions

• Supplement trial data

• Extend RCT into real world

• Short study duration

• Limited patient population

• No information on rare events

• ENDORSE: EQ-5D and SF-36 in patients treated with DMF [10]

Registries

• Population-based collection of information

• Long-term data natural history and disease management

• Regional comparisons

• Non-randomized design

• Incomparable patient groupings

• Discontinuous visit schedule

• Varying practice patterns

• Inter-regional extrapolation not always appropriate

• Lyons MS database: disability progression [11]

• SMSreg: epidemiology and treatment outcomes [16, 19]

• MSBase: treatment outcomes (69 countries) [13–15, 17]

Prospective observational studies

• Pre-defined outcome measures in clinical practice

• Robust dataset powered to answer specific questions

• Hawthorne effect (patients behave differently because they know they are being observed)

• PANGAEA, TOP: fingolimod and natalizumab clinical trial [9, 20]

PASSs

• Voluntary or imposed by regulatory authorities for approval

• Ongoing monitoring of the benefit–risk profile

• No obligation for regulatory submission of protocols and study reports for voluntary PASSs

• PANGAEA: German voluntary PASS [20]

Administrative data

• Data required for reimbursement

• Quick, low-cost analyses

• Large patient populations provide information on rare events

• Privacy concerns limit access to data

• Incomplete or inaccurate data

• Costs and charges are not differentiated

• Pharmetrics Plus™ and Medco databases: relapses, treatment compliance, resource use and inpatient stays [6–8]

Health surveys

• Descriptive data

• Provide broadly generalizable data

• Not product-specific

• Subjective

• Relies on participant recollection

• NARCOMS survey: symptoms, comorbidities and health-related quality of life [18]

EMRs

• Real-time data collection

• Low cost

• Detailed information over long periods

• High-end statistical analysis tools required

• EMRs: diagnosis, disease progression, symptoms and treatment [12]