Abstract 15467: Variations in T1 Mapping and LV Functional Parameters in a Multi-Center Study: The Multi-Ethnic Study of Atherosclerosis (MESA)

Abstract

Introduction: T1 mapping techniques may be useful as an index of myocardial fibrosis, but have not previously been assessed in a multi-center study. We sought to assess site variation/reproducibility of MRI T1 mapping and LV functional measurements in the Multi-Ethnic Study of Atherosclerosis (MESA), a large multi-center study using MRI to evaluate cardiovascular disease.

Methods: A total of 2832 participants, age 54-95 y/o from six field centers (FC), underwent cardiac MRI during Exam 5 (2010-12). Images were transferred to the MRI reading center using ACR Medical Imaging Resource System (MIRC). Image quality scorecards were applied to the studies. LV, LVEF, T1 mapping values pre and at 12 minutes following Gadolinium administration and partition coefficient (λ= αR1myo / αR11bloodT1) were calculated. Of this data, 15% was used to assess variability across three readers by calculating the concordance correlation coefficients (CCC). Uni and multivariable regression analysis were used to evaluate variability of functional and T1 MRI values across different FC accounting for basic participants and image parameters.

Results: Each FC performed16-19% of the total number of studies. There was no difference in proportions of diagnostic and non-diagnostic studies across FC (p=0.626). In the inter-reader agreement pairwise analysis, the CCC ranged from 0.61-0.78 for LVEF and from 0.94-0.96 for LV mass. T1 at 12 min had statistically significant variation among FCs, even in the multivariate analysis adjusting for basic demographics and image acquisition parameters; whereas the partition coefficient demonstrated significant variation only in one site. LVmass and LVEF also showed slight but significant variation in some FCs (table).

Conclusion: Standardization of MRI processes led to great inter-reader reproducibility. There was still variability introduced by differences among sites that were not explained by basic population and imaging variables.