Assessment Of The Reproducibility Of Semi-Automated Flow Mediated Dilation Measurements Performed On Separate Days

Objective: Assessment of endothelial function has the potential to diagnose cardiovascular disease earlier than traditional methods. The gold standard non-invasive measure of endothelial function is brachial artery flow-mediated dilatation (FMD). However, this technique is limited by significant measurement variability and requiring extensive operator training. Semi-autonomous FMD has the potential to overcome these problems. This study assessed the reproducibility of separate-day semi-autonomous FMD measurements by investigators who received basic training on the correct use of the device. Design and method: Forty three healthy participants were recruited for serial FMD readings using the UNEX EF 38G device. The cohort was largely Caucasian and composed of younger normotensive adults. Two readings were taken 20 minutes apart with a further reading taken on a separate day. The device produced automatic FMD outputs which were manually analysed using edge-detection software. The reproducibility of FMD measurements performed on separate days was compared for automated and manual readings. Results: 25% of scans were not diagnostic (final sample size = 32). Automated results correlated poorly (r = 0.106) whereas manual results demonstrated good correlation (r = 0.622). Poor reproducibility and high measurement variance were shown for automated analyses (intraclass correlation coefficient (ICCC) = 0.103, coefficient of variation (CV) = 45.71%). In contrast, manually analysed scans demonstrated good reproducibility and low measurement variance (ICCC = 0.679, CV = 12.43%). Same day measurements had improved reproducibility for both automatic (ICCC = 0.334) and manual (ICCC = 0.815) results when compared to separate day readings. Conclusions: Manually analysed FMD scans had good reproducibility, indicating that semi-autonomous FMD requires manual analysis before independent automatic interrogation. The lower reproducibility of separate-day measurements is most probably due to the inherent biological variability of endothelial function. The high non-diagnostic scan rate most likely reflects the investigator's basic and limited training. Thus, we strongly advocate for more extensive training before performing assessments in a research context.