Does McNemar's test compare the sensitivities and specificities of two diagnostic tests?

Area of Science:

• Biostatistics
• Medical Diagnostics
• Statistical Inference

Background:

• McNemar's test is frequently applied to compare paired binary data, particularly for evaluating diagnostic test accuracy.
• Current practice often involves separate analysis of diseased and non-diseased groups to compare sensitivities and specificities, respectively.

Purpose of the Study:

• To provide a theoretical framework for the separate analysis of diagnostic test performance.
• To investigate the validity of McNemar's test under both conditional independence and dependence assumptions.
• To assess the impact of conditional dependence on statistical power.

Main Methods:

• Derivation of McNemar's test statistic under the null hypothesis, considering both conditional independence and dependence.
• Power analyses were conducted to evaluate the influence of varying degrees of conditional dependence.

Main Results:

• The study establishes a theoretical basis for comparing diagnostic test sensitivities and specificities separately.
• McNemar's test validity is examined irrespective of the conditional independence assumption.
• Conditional dependence significantly affects the power of McNemar's test.

Conclusions:

• The findings support the practice of analyzing sensitivities and specificities in separate diseased and non-diseased groups for diagnostic test evaluation.
• Understanding the impact of conditional dependence is crucial for accurate interpretation of McNemar's test results.
• This research offers a more robust framework for applying McNemar's test in medical diagnostic accuracy studies.