| Retrospective |
Large databases that include patients over a broad range of disease severity and patient demographics Existence of healthy normal controls in a known-groups proof-of-concept analysis Data collection already done Ability to use other analyses to inform on the desired objectives Can sometimes use patient data to match the desired intended use patient profile Cost may be less than the cost of a prospectively designed study. May be only method of acquiring sufficient data on rare disease populations
|
Databases may be private with restricted access and restrictions on use Subjects are not typically randomized and control of biases not guaranteed or may not even be possible Healthy controls not always available requiring a prospective collection that would match the database Influenced by other analyses Patient population may not be equivalent for the intended use of the mp-QIB May not be able to conduct an external cross-validation requiring prospectively acquired data with different patient population
|
| Prospective |
Can specify the subject population(s) Greater control of bias if the sample size is sufficiently large Greater trust in the results Piggyback on current therapeutic intervention study can provide data for known-groups validity on measuring disease progression versus known or standard of care intervention (SOC)
|
Does not require external organizational approval with restricted use. Expensive and requires all of the costs with study start-up and recruiting that can be Slow recruitment when there is no therapeutic benefit can stop a study
|
| Combination Retrospective-then-Prospective85
|
Uses available data for mp-QIB development and prospectively acquired data for validation Retrospective data analysis can provide information on the optimal patient inclusion/exclusion criteria in the prospective study Prospective numbers and data costs minimized when used as validation compared to use as both development, testing and validation
|
|
