The effect of misclassification on sample size for one and two-sample tests with binary endpoints

Area of Science:

• Biostatistics
• Statistical Methods
• Epidemiology

Background:

• Analysis of binary data increasingly incorporates misclassification methods.
• Study designs often neglect potential misclassification due to a lack of sample size formulas and software.
• Ignoring misclassification during design can lead to significant power loss when addressed only during analysis.

Purpose of the Study:

• To emphasize the necessity of adjusting sample size for misclassification in the design phase of studies analyzing binary data.
• To provide a practical sample size calculation procedure for studies with binary endpoints, accounting for misclassification.
• To illustrate the impact of misclassification on required sample sizes for one-sample and two-sample tests.

Main Methods:

• Development of sample size formulas for one-sample and two-sample tests for binary endpoints, incorporating misclassification.
• Implementation of the sample size procedure as an R function.
• Calculation of sample sizes based on presumed binomial parameters, desired power, sensitivity (Se), and specificity (Sp).

Main Results:

• Misclassification significantly impacts the required sample size in both one-sample and two-sample testing scenarios.
• The developed R function provides a tool for researchers to calculate appropriate sample sizes.
• Comparison of sample sizes with and without misclassification highlights the potential for power loss.

Conclusions:

• Integrating misclassification correction into the study design phase, through appropriate sample size adjustment, is crucial.
• The provided methodology and R function can help researchers avoid power loss and design more robust studies.
• Accurate estimation of sensitivity and specificity is vital for effective sample size calculation in the presence of misclassification.