The ARTS Marker of Arteriolosclerosis: Instrumental and Clinical Validation

Abstract Background The ARTS biomarker is a fully automated software container that predicts the presence of arteriolosclerosis based on in‐vivo MRI data and demographic features. The present study describes findings from the instrumental and clinical validation of ARTS conducted by the MarkVCID consortium. Method Instrumental validation of ARTS involved assessment of inter‐rater reliability, test‐retest repeatability, and inter‐scanner reproducibility. Inter‐rater reliability was assessed by means of the intraclass correlation (ICC) between ARTS scores generated from all MarkVCID sites using data on 20 older adults. Test‐retest repeatability was assessed by means of the ICC between test and retest ARTS scores from 41 older adults imaged twice. Inter‐scanner reproducibility was assessed by means of the ICC between ARTS scores on 20 older adults imaged on 4 different scanners. Clinical validation of ARTS included two groups of non‐demented older adults: N = 156 participating in community cohort studies at Rush Alzheimer’s Disease Center; and N = 906 participating in the Atherosclerosis Risk in Communities (ARIC) study. The primary prespecified hypothesis was that change in Trails B score (or perceptual speed score for Rush data) two years after baseline MRI is associated with baseline ARTS score, controlling for years of education. The secondary prespecified hypothesis was that baseline Trails B (or perceptual speed) is associated with baseline ARTS score, controlling for years of education. Results ARTS exhibited excellent inter‐rater reliability (ICC = 0.9999, p<10‐200, CI:[0.99987, 0.99997]), test‐retest repeatability (ICC = 0.996, p<10‐42, CI:[0.993, 0.998]), and inter‐scanner reproducibility (ICC = 0.955, p = 0.00018, CI:[0.622,0.989]). In Rush data, a higher baseline ARTS score was associated with a faster two‐year decline in perceptual speed (beta = ‐0.32, p = 0.04, CI:[‐0.63, ‐0.01]) (Fig.1A) and a lower baseline level of perceptual speed (beta = ‐1.04, p<10‐4, CI:[‐1.48, ‐0.60]) (Fig.1B). In ARIC data, a higher baseline ARTS score was associated with a faster annual increase in Trails B (beta = 3.7, SE = 1.7, p = 0.03) (Fig.2A), and a higher baseline Trails B (beta = 45.5, SE = 5.8, p<10‐4) (Fig.2B). Conclusion ARTS exhibited outstanding technical performance and baseline ARTS score was associated with subsequent decline in cognitive abilities known to be affected by small vessel disease. The above evidence supports ARTS as a marker of the risk of vascular contributions to cognitive impairment and dementia (VCID).