Loss of base-to-apex circumferential strain gradient assessed by cardiovascular magnetic resonance in Fabry disease: relationship to T1 mapping, late gadolinium enhancement and hypertrophy

From June 2016 to May 2018, consecutive FD patients undergoing 3 T CMR at our institution were prospectively enrolled. The study was approved by the institutional research ethics board and written informed consent was obtained from all subjects. Inclusion criteria for patients included age ≥ 18 years and confirmed diagnosis of FD. Exclusion criteria included prior myocardial infarction or known coronary artery disease and contraindications to CMR. A reference healthy control group was also imaged, defined as volunteers without past medical history of cardiovascular disease. The final study population consisted of thirty-eight gene-positive FD patients (mean age 45.0 ± 14.5 years, 37% male) and eight healthy controls (mean age 40.1 ± 13.7 years, 63% male).

All CMR studies were performed using a 3 T scanner (Magnetom Skyra Fit, Siemens Healthineers, Erlangen, Germany or Biograph mMR, Siemens Healthineers). Multi-planar retrospectively gated cine balanced steady-state free precession (bSSFP) images were obtained including multiple long-axis planes and a stack of short-axis slices with coverage from cardiac base to apex (8 mm slice thickness and 2 mm inter-slice gap). Multi-planar LGE imaging was performed approximately 15 min after administration of 0.15 mmol/kg body weight of gadobutrol (Gadovist, Bayer Healthcare, Berlin, Germany) employing a phase sensitive inversion recovery gradient-recalled echo sequence (8 mm slice thickness and 2 mm inter-slice gap) at matching slice positions. Native and 12–15 min post-contrast mid-ventricular short-axis T1 maps were acquired using a bSSFP readout modified Look-Locker inversion recovery (MOLLI) technique with inline motion correction and map creation (native (5(3)3) and post-contrast (4(1)3(1)2) inversion groupings, 8 mm slice thickness) [16].

CMR analysis was performed by a fellowship-trained radiologist (SM, with 2 years of cardiovascular imaging experience) blinded to all identifying information using commercially available software (Circle cmr42 release 5.6.3; Circle Cardiovascular Imaging, Calgary, Canada). LV endocardial and epicardial borders were contoured on the stack of short-axis bSSFP images to assess for left ventricular (LV) volumes, LV ejection fraction (LVEF), and mass, and on short axis LGE images to assess for the extent of LGE. LVH was defined as LV mass indexed to body surface area (BSA) > 85 g/m² in males and > 81 g/m² in females [17]. All LGE images were visually assessed for the presence of LGE. Quantitative LGE analysis was performed using a signal intensity threshold of 4 standard deviations above visually normal remote myocardium, expressed as a percentage of total myocardial mass [18]. For feature tracking strain analysis, endocardial and epicardial borders were contoured at end-diastole on short-axis (at basal, mid-ventricle and apical levels) and long-axis (2-, 3- and 4-chamber) bSSFP images. Peak myocardial global longitudinal strain (GLS) was calculated by averaging all peak longitudinal myocardial values. Peak global CS (GCS) was obtained by averaging peak myocardial strain values across basal, mid and apical levels (Fig. 1). Base-to-apex LS and CS gradients were calculated as the peak gradient difference between basal and apical LS and CS, respectively [12]. T1 values were evaluated by drawing a single region of interest in the mid interventricular septum excluding any areas with LGE [6, 19]. Myocardial extracellular volume (ECV) was calculated with input of native and post-contrast myocardial and blood pool T1 values, and hematocrit obtained within 24 h of CMR by non-invasive testing [20, 21].

To assess intra-observer agreement for strain analysis, a random subset of 15 studies were re-analyzed by the same reader following a minimum 2-week interval after the first analysis, blinded to the results of the initial assessment and all identifying data. To assess inter-observer agreement, the same subset of studies was analyzed by a second experienced fellowship-trained reader (JD, with 2 years of cardiovascular imaging experience), blinded to all identifying information including the results of the initial assessment.

Native T1 values were significantly lower in FD patients compared to controls (1170.2 ± 37.5 vs. 1239.0 ± 18.0 ms, p < 0.001), even when the comparison was restricted to the LVH- LGE- subgroup (1172.2 ± 42.4 ms, p < 0.001), FD LVH+ subgroup (1160.1 ± 41.6 ms, p < 0.001), and FD LGE+ subgroup (1163.2 ± 29.9 ms, p < 0.001) (Fig. 2a). There was no significant difference in ECV between FD patients and controls (25.1 ± 2.9% vs. 25.9 ± 3.2%, p = 0.48), even when the comparison was restricted to the FD LVH- LGE- subgroup (25.3 ± 2.8%, p = 0.63), FD LVH+ subgroup (23.2 ± 2.5%, p = 0.09), and FD LGE + subgroup (24.9 ± 3.4%, p = 0.51) (Fig. 2b). Among FD patients, native T1 and ECV were both significantly lower in males compared to females (1154.6 ± 34.8 vs. 1179.3 ± 36.6 ms, p = 0.049 and 23.4 ± 2.1% vs. 26.0 ± 3.0%, p = 0.009).

There was no significant difference in GLS between FD patients and controls (− 15.3 ± 3.5% vs. -16.3 ± 1.5%, p = 0.45), even when the comparison was restricted to the FD LVH- LGE- subgroup (− 16.8 ± 2.8%, p = 0.59) (Fig. 2c). However, GLS was significantly lower in the FD LVH+ subgroup (− 11.4 ± 3.9%, p = 0.005) and FD LGE + subgroup (− 13.6 ± 3.3%, p = 0.045) when compared to controls. Among patients with FD, GLS correlated positively with LV mass indexed to BSA (r = 0.64, p = 0.02) and quantitative LGE (r = 0.508, p = 0.002), but did not correlate significantly with native T1 (r = 0.112, p = 0.51) or ECV (r = − 0.206, p = 0.23), Table 3. Among FD patients, there was no significant difference in GLS between males and females (− 14.0 ± 4.1% vs. -16.0 ± 2.9%, p = 0.09).

There was no significant difference in GCS between FD patients and controls (− 19.4 ± 3.0% vs. -19.5 ± 2.9%, p = 0.84), even when the comparison was restricted to the FD LVH- LGE- subgroup (− 20.2 ± 1.8% p = 0.52) and the FD LGE+ subgroup (− 18.4 ± 3.8%, p = 0.68) (Fig. 2d). The difference in GCS between the FD LVH+ subgroup (− 15.9 ± 3.9%) and controls approached statistical significance, p = 0.059. Among patients with FD, GCS correlated positively with LV mass indexed to body surface area (BSA) (r = 0.426, p = 0.01), but did not correlate significantly with quantitative LGE (r = 0.298, p = 0.08), native T1 (r = − 0.085, p = 0.62), or ECV (r = − 0.250, p = 0.14). Among FD patients, GCS was significantly lower in males compared to females (− 17.8 ± 3.4% vs. -20.4 ± 2.3%, p = 0.009).

There was no significant difference in base-to-apex LS gradient between FD patients and controls (7.5 ± 3.8% vs. 9.3 ± 3.5%, p = 0.24), even when the comparison was restricted to the FD LVH- LGE- subgroup (6.9 ± 4.1%, p = 0.18), FD LVH+ subgroup (7.7 ± 3.4%, p = 0.44), and FD LGE+ subgroup (8.3 ± 3.4%, p = 0.51) (Fig. 2e). Among patients with FD, base-to-apex LS gradient did not correlate significantly with LV mass indexed to BSA (r = − 0.101, p = 0.60), quantitative LGE (r = − 0.097, p = 0.61), native T1 (r = − 0.006, p = 0.98) or ECV (r = − 0.071, p = 0.71). Among FD patients, there was no significant difference in base-to-apex LS gradient between males and females (7.5 ± 3.7% vs. 7.6 ± 3.9%, p = 0.98).

Base-to-apex CS gradient was significantly lower in FD patients compared to controls (2.1 ± 3.7% vs. 6.5 ± 2.2%, p = 0.002), even when the comparison was restricted to the LVH- LGE- subgroup (2.7 ± 3.0%, p = 0.003), FD LVH+ subgroup (0.1 ± 3.2%, p < 0.001), and FD LGE+ subgroup (1.7 ± 4.4%, p = 0.008) (Fig. 2f). Among patients with FD, base-to-apex CS gradient did not correlate significantly with LV mass indexed to BSA (r = − 0.255, p = 0.13), quantitative LGE (r = − 0.089, p = 0.61), native T1 (r = − 0.243, p = 0.15) or ECV (r = − 0.265, p = 0.12). Among FD patients, there was no significant difference in base-to-apex CS gradient between males and females (2.2 ± 3.5% vs. 2.0 ± 3.9%, p = 0.93).

Base-to-apex CS gradient (odds ratio [OR] 0.42, 95% confidence interval [CI]: 0.20 to 0.86, p = 0.019) and native T1 (OR 0.96, 95% CI: 0.92 to 0.99, p = 0.010) were significant in differentiating FD LVH- LGE- patients from healthy controls in univariable models, Table 4. There was no evidence that sex was a modifying variable of the relationship between base-to-apex CS gradient (interaction term, p = 0.41) or native T1 (interaction term, p = 0.99) and diagnosis of FD. Base-to-apex CS gradient remained significant in a bivariable logistic regression model after adjusting for native T1 (OR 0.24, 95% CI: 0.06 to 0.99, p = 0.049). In this bivariable model, native T1 approached but did not reach statistical significance (OR 0.94, 95% CI: 0.88 to 1.00, p = 0.056). In a nested logistic regression model with native T1, model fit was significantly improved by the addition of base-to-apex CS gradient (χ²(df = 1) = 11.04, p < 0.001).

Intra- and inter-observer agreement were moderate to good for myocardial strain parameters: GLS (ICC 0.849 and 0.774, respectively), GCS (ICC 0.831 and 0.833, respectively), and base-to-apex CS gradient (ICC 0.737 and 0.613, respectively).