Exploratory vs. confirmatory studies in the field of diagnostic studies
Exploratory diagnostic study
Objective: To estimate diagnostic accuracy with confidence interval
Advantage of the design: lower number of cases.
Disadvantage: no proof of delivery.
If models are to be developed, e.g., combinations of several markers are to be linked to form a mathematical marker, the first study for this is always an exploratory study. The result must be validated in an independent study.
Confirmatory diagnostic study
Objective: To demonstrate that diagnostic accuracy is achieved.
An example of the evidence required by the authorities is the iFOBT test for colon cancer screening. Here, the German authorities (GBA) require a sensitivity of 25% and a specificity of 90%.
Example: Proof that its sensitivity is better than 60% and its specificity is better than 90%. Often, one of the two hypotheses is also formulated as a non-inferiority test: Proof that its sensitivity is better than 60% and its specificity does not fall significantly below 90%.
This study tests two linked null hypotheses, one for sensitivity and one for specificity. Both null hypotheses must be rejected to achieve the study objective. This necessitates a higher power than usual when calculating the sample size. If the overall power is set at 80%, the power for each individual test must be set at approximately 90% (0.9 x 0.9 ~ 0.8).
Advantage: Evidence can be provided.
Disadvantages of these designs: higher number of cases
