Abstract Background Stress myocardial blood flow (MBF) estimated from dynamic myocardial perfusion computed tomography (DynCTP-MBF) and cardiac magnetic resonance (CMR-MBF) due to their accessibility have emerged as promising modalities that could allow for the widespread use of non-invasive quantitative perfusion analysis in clinical practice. However, the underlying method of obtaining these measurements differs significantly between modalities and vendors. Consequently, the agreement between DynCTP-MBF and CMR-MBF is not well understood. Purpose To assess the agreement between DynCTP-MBF and CMR-MBF in pharmacological stress conditions. Methods We retrospectively enrolled consecutive patients who underwent both DynCTP-MBF and cardiac magnetic resonance (CMR) scans with pharmacological stress perfusion sequences within 23 days (IQR: 15–140). Stress MBF values were quantified through blinded analysis dedicated software tools for both CT and CMR. We investigated correlation and agreement of stress MBF on per-segment basis using Spearman’s method and Bland-Altman plots, respectively. We additionally examined proportional bias by regressing the difference between CMR-MBF and DynCTP-MBF on the average of the two measurements. We assessed inter-class correlation (ICC) using single fixed raters method for per-segment analyses and using average fixed raters method for per-segment and per-patient analyses. For concordance analysis, reduced CMR-MBF and DynCTP were defined as 1.94 ml/g/min and 1.35, respectively, based on published validation data. Results The cohort comprised 34 patients with a median age of 72 years (IQR: 64–76), 79% of whom were male, and 32% had a history of chronic coronary syndrome. Per-segment DynCTP-MBF and CMR-MBF were 1.06 (0.87–1.3) and 1.88 (1.54–2.29) mL/min/g, respectively. The mean difference (CMR minus DynCTP) were 0.85 (Bland–Altman limits of agreement +2.07, −0.37) and was more pronounced in patients with diabetes or hypertension but not significantly affected by nitroglycerin premedication (Figure 1). The regression analysis revealed a highly significant positive slope (β = 1.02, p 0.001) indicating DynCTP-MBF underestimation increases at higher flow levels. Spearman’s correlation was r = 0.41 (p 0.001), indicating a moderate positive but non-linear association (Figure 2). Intraclass correlation coefficients were 0.26, 0.48 and 0.48 for per segment, per-territory and per-patient analysis, respectively. The concordance index was 0.61, 0.67 and 0.7 for per segment, per-territory and per-patient analysis. For reference, the concordance of final physician ruling if the study was ‘positive’ for ischemia was 0.79. Conclusions Quantitative stress myocardial blood flow measured by DynCTP and CMR imaging demonstrate moderate concordance and non-linear correlation. Despite moderate correlation, the presence of proportional bias suggests these methods cannot be used interchangeably without calibration.Figure 1 Figure 2
Pieszko et al. (Thu,) studied this question.