Blinded sample size re-estimation in a comparative diagnostic accuracy study.

Area of Science:

• Biostatistics
• Medical Diagnostics
• Clinical Trials

Background:

• Sample size calculation in diagnostic accuracy studies is complex due to co-primary endpoints (sensitivity, specificity).
• Initial calculations rely on assumptions about disease prevalence and discordant test results, which can impact study power.
• Uncertainty in these nuisance parameters can lead to inefficient sample size estimations.

Purpose of the Study:

• To develop an optimal sample size calculation for unpaired and paired diagnostic studies with co-primary endpoints.
• To introduce a blinded adaptive design for sample size re-estimation, adjusting for nuisance parameters.
• To compare the adaptive design against fixed designs and existing methods.

Main Methods:

• Developed an optimal sample size calculation method for co-primary endpoints in diagnostic accuracy studies.
• Implemented a blinded adaptive design for sample size re-estimation in both unpaired and paired study designs.
• Conducted a simulation study to compare the adaptive design with fixed designs and evaluated a paired design example.

Main Results:

• The blinded adaptive design effectively controls type I error rates.
• The adaptive design accurately re-estimates nuisance parameters without significant bias and achieves target power.
• The proposed adaptive methods result in a smaller sample size compared to existing approaches.

Conclusions:

• Recommends applying optimal sample size calculation and blinded adaptive designs in confirmatory diagnostic accuracy studies.
• These methods address inefficiencies in traditional sample size calculations.
• Supports achieving study objectives by optimizing sample size and power.