280 A Snapshot into the Demonstration of Reproducibility of Cardiac Imaging Biomarkers

Abstract

The frequency of reported reproducibility statistics and the proportion of population used for reproducibility assessment is poorly examined within Cardiovascular Imaging (CVI) literature. We aimed to analyse the demonstration of reproducibility of various cardiac imaging modalities reported in original articles amongst 3 CVI journals. All articles published in JACC CVI, Circulation CVI and EHJ CVI in 2018 were reviewed. Non-original research or irrelevant articles were excluded. Included articles were then evaluated for demonstration of reproducibility, tools used to assess reproducibility, and the proportion of total population used in reproducibility analysis. More than half of the included articles (136/258, 52.7%) did not report reproducibility. Of those, 21.3% (26/122) referred to previous reproducibility data and 79.7% (96/122) demonstrated reproducibility. Differences in demonstration of reproducibility by modality are as follows: Echo 48% (58/122); CT 24% (9/38); CMR 31% (17/55); Nuclear 32% (8/25); miscellaneous 22% (4/18) (p=0.02). The tools used to demonstrate reproducibility were as follows: Intraclass correlation (ICC) 75% (72/96); Coefficient of Variance 17% (16/96); Bland-Altman 27% (26/96); Correlation 18% (17/96). The proportion of population used to demonstrate reproducibility ranged from 0.8-100%, median 18.6% IQR (7.8%-56.3%). Less than half of the manuscripts published in high-impact imaging journals reported or demonstrated reproducibility with significant variety among modalities. ICC was the most common tool used to report reproducibility. There is significant heterogeneity in the proportion of population used to assess reproducibility. Standardised reporting format for reproducibility will reduce variability among CVI articles.