# Device-based day-to-day and observer variability to quantify dilation capacity in the retinal microcirculation

## TL;DR

This study evaluates the consistency of a method to measure retinal blood vessel dilation using different devices and observers.

## Contribution

The study introduces a detailed analysis of variability in retinal microcirculation measurements across devices and observers.

## Key findings

- Day-to-day variability was moderate to good for DVA 2.0 and low to moderate for DVA 3.0.
- Intraobserver and interobserver variability were very good for both devices.
- Correcting for baseline arteriolar diameter improved variability measurements.

## Abstract

Dynamic retinal analysis (DVA) is a validated method to quantify microvascular endothelial function. This study aimed to analyze day-to-day variability, intra- and interobserver variability and differences between two device generations.

DVA was performed on two separate days and on two devices each, the DVA 2.0 and the DVA 3.0. One reader analyzed 20 signals of maximum arteriolar (aFID) and venular flicker-light induced dilation (vFID) twice to investigate intraobserver variability. A second reader independently analyzed 20 aFID and vFID signals to quantify interobserver variability. The interclass correlation coefficient (ICC) and the 95% confidence interval were used to quantify reliability.

The analysis of 26 participants (mean age 43 ± 14 years) showed moderate to good day-to-day variability for aFID (ICC 0.81 (0.57, 0.92), p = 0.037) and vFID (0.91 (0.80, 0.96), p < 0.001) of DVA 2.0 and low to moderate day-to-day variability for aFID (0.79 (0.49, 0.91), p = 0.076) and vFID (0.87 (0.61, 0.95), p = 0.022) of DVA 3.0. The analyses showed very good intraobserver (aFID and vFID: 0.999 (0.998, 1), p < 0.001) and interobserver variability (aFID: 0.997 (0.993, 0.999), p < 0.001; vFID: 0.998 (0.971, 0.995), p < 0.001). The measurements with devices DVA 2.0 and DVA 3.0 showed a moderate interdevice variability for aFID (0.76 (0.57, 0.89), p = 0.042) and vFID (0.87 (0.74, 0.93), p < 0.001). The ICC of aFID improved for day-to-day variability and interdevice variability after correcting for the baseline diameter.

Consideration of arteriolar baseline diameter variations may further improve day-to-day and interdevice variability. This work underpins the necessity for standardized methods to support clinical implementation of the method and the need to consider arteriolar baseline diameters in future research and clinical applications.