Performance of Synthetic Extracellular Volume Fraction in Different Cardiac Phenotypes From a Prospective Cohort of Patients Referred for Cardiac Magnetic Resonance

Background: A synthetic myocardial extracellular volume fraction (sECV) can be obtained without blood hematocrit (Hct) by using the linear relationship between Hct and the longitudinal relaxation time of blood. Concerns have been raised about the widespread clinical application of this approach.

Purpose: To assess the relationship between measured ECV (m-ECV) and sECV, using both a published model and a locally derived one.

Study type: Single-center, prospective.

Field strength/sequence: A 1.5 T/modified Look Locker (MOLLI) sequence.

Subjects: Fifty-two healthy subjects and 113 patients (76 with and 37 without a hypertrophic cardiac phenotype).

Assessment: Three ECV values were obtained for each patient: 1) measured ECV (m-ECV), using Hct from a venous blood sample; 2) Fent-synthetic ECV (F-sECV), using the equation proposed by Fent et al; and 3) Local-synthetic ECV (L-sECV), using the equation obtained from a local derivation cohort comprising 83 subjects.

Statistical tests: Shapiro-Wilk test, analysis of variance, Kruskal Wallis test, Pearson correlation, Bland-Altman analysis, univariate and multivariable regression analysis.

Results: In the validation cohort (N = 82), Bland-Altmann analysis revealed an excellent agreement between m-ECV and L-sECV with a statistically insignificant bias (-0.1%, limits of agreement: -2.8% and 2.6%; P = 0.528), while in the overall population (N = 165), the mean bias between m-ECV and F-sECV was small but significant (1.2%, limits of agreement: -1.5% and 3.9%, P < 0.05). F-sECV bias was significantly higher for measured Hct (m-Hct) values <0.372 (2.3% vs. 1.0%, P < 0.05). Among the phenotype subgroups, amyloidotic patients showed a higher bias compared to others, both with F-sECV and L-sECV (2.3% vs. 1.1%, P < 0.05 and 1.1% vs. 0.2%, P < 0.05, respectively).

Data conclusion: Although synthetic ECV performs well in an external cohort, the use of a local formula further improves the accuracy of ECV estimate over a broad spectrum of cardiac phenotypes. Local sECV performs better for a wider range of Hct values, compared to the published model. Amyloidosis is the only group associated with a lower accuracy.

Level of evidence: 5 TECHNICAL EFFICACY: Stage 2.