Testing non-inferiority (and equivalence) between two diagnostic procedures in paired-sample ordinal data

Area of Science:

• Biostatistics
• Medical Diagnostics
• Health Technology Assessment

Background:

• New diagnostic procedures must demonstrate non-inferiority to existing standards before adoption.
• Assessing diagnostic accuracy is crucial for cost-effective and convenient healthcare innovations.
• Ordinal test responses present unique challenges in evaluating diagnostic accuracy.

Purpose of the Study:

• To develop and validate statistical methods for assessing non-inferiority of new diagnostic procedures with ordinal outcomes.
• To provide practical tools for researchers and clinicians comparing diagnostic tests.
• To extend existing methods to handle ordinal data in diagnostic accuracy studies.

Main Methods:

• Defined two novel statistical definitions of non-inferiority for ordinal diagnostic tests.
• Developed two large-sample theory-based test procedures for non-inferiority.
• Employed Monte Carlo simulations to assess the finite sample performance of the proposed tests.
• Illustrated the application with a breast cancer screening example comparing digitized and plain films.

Main Results:

• The proposed asymptotic test procedures perform reasonably well, even with limited sample sizes.
• The methods provide a robust framework for evaluating diagnostic accuracy equivalence.
• The breast cancer screening example demonstrates the practical utility of the developed tests.

Conclusions:

• The study offers effective statistical tools for non-inferiority testing in diagnostic accuracy research with ordinal data.
• These methods facilitate the adoption of improved, more convenient, and less expensive diagnostic procedures.
• The approach is readily extendable to assess two-sided equivalence of diagnostic tests.