Hypothesis testing in noninferiority and equivalence MRMC ROC studies

Area of Science:

• Medical Imaging Analysis
• Statistical Methodology
• Diagnostic Accuracy Assessment

Background:

• Conventional multireader multicase receiver operating characteristic (MRMC ROC) methodologies are primarily designed for superiority testing of diagnostic accuracies.
• In practice, demonstrating noninferiority or equivalence of imaging modalities is often more relevant than proving superiority.
• Existing MRMC ROC frameworks require adjustments to accommodate noninferiority and equivalence hypothesis testing.

Purpose of the Study:

• To investigate and present appropriate adjustments to conventional MRMC ROC hypothesis testing for noninferiority and equivalence studies.
• To provide a framework for designing and analyzing studies that assess if one imaging modality is statistically noninferior or equivalent to another.
• To offer guidance on sample size calculations for such studies.

Main Methods:

• Three methodological adjustments to the Obuchowski-Rockette (OR)/Dorfman-Berbaum-Metz (DBM) MRMC ROC method are proposed.
• These include defining null/alternative hypotheses, data analysis methods, and study sizing procedures for noninferiority/equivalence.
• A clinical example illustrates the application of these methods for noninferiority MRMC ROC studies.

Main Results:

• Detailed procedures for analysis and sample size computation for noninferiority MRMC ROC studies are provided.
• It is demonstrated that an equivalence hypothesis test is equivalent to conducting two simultaneous noninferiority tests.
• The interplay between effect size, noninferiority margin, and sample size is explored.

Conclusions:

• Conventional MRMC ROC methodology can be effectively adjusted for noninferiority and equivalence hypothesis testing.
• These adjusted methods are essential for accurate comparisons of diagnostic performance between imaging modalities.
• Confidence intervals for differences in diagnostic accuracies should accompany hypothesis testing for comprehensive interpretation.