Related Experiment Videos
Sample size determination for diagnostic accuracy studies involving binormal ROC curve indices
1Department of Biostatistics and Epidemiology, Cleveland Clinic Foundation, OH 44195-5196, USA.
Statistics in Medicine
|July 15, 1997
Summary
This study introduces a unified sample size computation method for Receiver Operating Characteristic (ROC) curves and their indices. The approach uses Taylor series expansions for accurate variance and covariance estimation in diagnostic test accuracy assessments.
Area of Science:
- Biostatistics
- Medical Imaging Analysis
- Diagnostic Test Evaluation
Background:
- Receiver Operating Characteristic (ROC) curves are essential for assessing diagnostic test accuracy.
- The area under the ROC curve (AUC) is a common but criticized accuracy measure due to equal weighting of false positive rates.
- Alternative ROC indices, like partial AUC and sensitivity at a fixed false positive rate (FPR), offer more nuanced evaluations.
Purpose of the Study:
- To present a unified method for calculating sample sizes for binormal ROC curves and their associated indices.
- To address the limitations of the full AUC by incorporating alternative ROC measures.
- To provide a statistically robust framework for designing studies that evaluate diagnostic test accuracy.
Main Methods:
- Utilizing Taylor series expansions to derive approximate large-sample variance and covariance estimates for binormal ROC curve parameters.
- Developing a unified approach applicable to various ROC indices, including partial AUC and sensitivity at a fixed FPR.
- Applying the method to real-world examples from diagnostic radiology to demonstrate its utility.
Main Results:
- The proposed method provides a unified framework for sample size computation for binormal ROC curves and their indices.
- The use of Taylor series expansions yields reliable estimates for variance and covariance of ROC parameters.
- Demonstrated applicability and effectiveness through illustrative examples in diagnostic radiology.
Conclusions:
- The presented unified approach offers a valuable tool for researchers and clinicians in determining appropriate sample sizes for studies evaluating diagnostic tests.
- This method enhances the design of studies by accounting for specific ROC indices beyond the traditional full AUC.
- The findings contribute to more precise and efficient study designs in diagnostic accuracy research, particularly in fields like radiology.