Nonparametric Sample Size Estimation for Sensitivity and Specificity with Multiple Observations per Subject.

Area of Science:

• Biostatistics
• Medical Diagnostics
• Statistical Methods

Background:

• Diagnostic tests often involve multiple observations per subject, leading to clustered binary data.
• Existing sample size calculation methods may not adequately account for this clustered nature and varying observations.
• Accurate sample size estimation is crucial for reliable diagnostic test evaluation.

Purpose of the Study:

• To propose a novel sample size calculation approach for estimating sensitivity and specificity of diagnostic tests with multiple observations per subject.
• To develop a sample size formula that accounts for clustered binary data and varying number of observations.
• To evaluate the performance of the proposed formula compared to a standard parametric approach.

Main Methods:

• Derivation of a sample size formula for sensitivity and specificity using the sign test for clustered binary data.
• Accounting for multiple observations per subject and potential variations in their number.
• Conducting simulation studies to assess finite sample performance and compare with a parametric method.

Main Results:

• The proposed sample size formula demonstrated empirical powers closer to the nominal level than the parametric method.
• Simulation studies indicated that increased variability in the number of observations per subject and higher intracluster correlation necessitate a larger sample size.
• The new formula provides a more accurate estimation for diagnostic tests with clustered observations.

Conclusions:

• The proposed nonparametric sample size formula is effective for diagnostic tests with multiple observations per subject.
• The method offers improved accuracy over traditional parametric approaches, especially when observation counts vary.
• Researchers should consider the variability of observations and intracluster correlation when designing studies for diagnostic test evaluation.