Statistical Considerations in the Design and Analysis of Precision Study for in vivo Diagnostics Devices
*Haiwen Shi, FDA/CDRH 

Keywords: Medical Devices, Diagnostics, Precision Study, Study Design, Random Effect ANOVA models, Variance Component Analysis

Agreement and precision analysis are two essential statistical methods used in diagnostics devices submission to FDA to establish substantial equivalence between subject device and predicate. The precision study is particularly important to evaluate the imprecision of repeated quantitative measurements of devices. The variability among measurements should be in clinical acceptable range. In this talk, I will discuss some statistical considerations in design and analysis precision study. My talk will focus on in vivo diagnostics devices such as Optical Coherence Tomography (OCT) devices. OCT is a medical imaging device primarily used in ophthalmic clinics. Its quantitative measurements including thickness of various retinal structures provide very useful information to help diagnose and monitor retinal diseases such as glaucoma. Therefore, the precision of the measurements are crucial to provide reliable information. Using OCT devices as an example, I will go through a couple of common designs for precision study in the submission to FDA, nested or crossed design. The difference between them is whether different or the same subjects are scanned under each device + operator configuration. The pros and cons of nested and crossed design will be discussed. In addition, I will talk about some common problems we met in the submissions to FDA for precision study.