Optimal sampling ratios in comparative diagnostic trials

TL;DR

This paper develops an adaptive two-stage method to determine optimal sampling ratios in diagnostic trials, improving power and reducing sample size without relying on pilot data or parametric assumptions.

Contribution

It introduces a novel two-stage procedure for estimating optimal sampling ratios without pilot data or distributional assumptions, enhancing trial efficiency.

Findings

Method increases statistical power in diagnostic trials.

Significant reduction in required sample size.

Validated through theoretical analysis and simulations.

Abstract

In this paper we focus on comparative diagnostic trials which are frequently employed to compare two markers with continuous or ordinal results. We derive explicit expressions for the optimal sampling ratio based on a common variance structure shared by existing summary statistics of the receiver operating characteristic (ROC) curve. Estimating the optimal ratio requires either pilot data or parametric model assumptions; however, pilot data are often unavailable at the planning stage of diagnostic trials. In the absence of pilot data, some distributions have to be assumed for carrying out the calculation. An optimal ratio from an incorrect distributional assumption may lead to an underpowered study. We propose a two-stage procedure to adaptively estimate the optimal ratio in comparative diagnostic trials without pilot data or assuming parametric distributions. We illustrate the properties of the proposed method through theoretical proofs and extensive simulation studies. We use an example in cancer diagnostic studies to illustrate the application of our method. We find that our method increases the power, or reduces the required overall sample size dramatically.