The partial testing design: a less costly way to test equivalence for sensitivity and specificity

Area of Science:

• Radiology
• Medical Imaging
• Biostatistics

Background:

• Digital mammography is increasingly used, but its equivalence to analogue mammography requires rigorous testing.
• Breast cancer screening involves balancing sensitivity and specificity, with sample size calculations crucial for study power.
• Asymptomatic women undergoing screening present challenges due to the rarity of breast cancer, impacting sample size requirements.

Purpose of the Study:

• To propose a novel, cost-effective study design for comparing the sensitivity and specificity of digital versus analogue mammography.
• To achieve equivalent statistical power to traditional paired designs with reduced costs.
• To address the challenges of sample size calculations in rare disease screening.

Main Methods:

• A new study design is proposed, leveraging less expensive analogue mammograms.
• Digital mammograms are selectively withheld from subjects testing negative on analogue mammograms to reduce costs.
• A follow-up analogue mammogram, guided by a natural history model, is used to ascertain disease status for non-biopsied subjects.

Main Results:

• The proposed design allows for achieving the same statistical power as a fully paired design.
• Cost savings are realized by optimizing the use of more expensive digital mammograms.
• The design effectively utilizes information from subjects with varying disease likelihoods.

Conclusions:

• The proposed design offers a less costly method for testing the equivalence of digital and analogue mammography.
• This approach enhances the feasibility of large-scale mammography comparison studies.
• The design is distinct from double sampling, focusing on comparing two imperfect diagnostic tests.