Cardiac Amyloid Imaging with ¹⁸F-Florbetaben PET: A Pilot Study

Patient Population and Study Protocol

Ten patients with cardiac amyloidosis (5 AL and 5 ATTR) were prospectively recruited to undergo PET imaging and transthoracic echocardiography. All amyloid patients (both AL and ATTR) had echocardiographic features consistent with cardiac amyloidosis. All ATTR patients had endomyocardial biopsy–proven cardiac amyloid and wild-type ATTR, otherwise known as senile cardiac amyloidosis. Patients with the AL subtype had biopsy from other organs (kidney, bone marrow, prostate, rectum) and met the consensus criteria of the International Society of Amyloidosis for cardiac involvement (5). The subtype of amyloidosis was determined by tandem mass spectrometry in 6 patients and immunohistochemistry or immunofluorescence in 4 patients. Genetic testing for transthyretin gene mutations was negative in all ATTR cases.

As per guidelines, echocardiographic definition of cardiac amyloidosis included LV wall thickness greater than 12 mm, granular sparkling of the myocardium, right ventricular (RV) free wall thickness greater than 5 mm, increased atrioventricular valve or interatrial septal thickness, pericardial effusion, and LV diastolic dysfunction (based on mitral inflow, pulmonary venous flow, and tissue Doppler imaging of septal and lateral mitral annulus) (5).

Because LV hypertrophy due to hypertensive heart disease is a frequent differential diagnosis for cardiac amyloidosis, 4 hypertensive controls with echocardiographic evidence of concentric LV hypertrophy were also prospectively recruited.

None of the patients had echocardiographic or cardiac MRI features that would suggest hypertrophic cardiomyopathy or variants.

Laboratory examinations included glomerular filtration rates calculated by the “Modification of Diet in Renal Disease” formula (6), cardiac troponin-I, and B-type natriuretic peptide.

Exclusion criteria for all amyloidosis and hypertensive control patients included the presence of moderate or severe valvular heart disease, congenital heart disease, severe claustrophobia, or pregnancy or an inability to lie flat for 80 min. Additional exclusion criteria for the hypertensive controls included previously known congestive cardiac failure, myocardial infarction, and an LV ejection fraction of less than 50%.

The study was approved by the institutional ethics committee, and all patients provided written informed consent.

Cardiac PET Data Acquisition

Cardiac PET (Biograph mCT, 128-slice; Siemens) was performed in list mode for 80 min. As the distribution of ¹⁸F-florbetaben in the heart has not been previously reported, a long acquisition time was required to enable selection of the optimal imaging protocol for future studies. Low-dose CT (tube current, 40 mA⋅s; tube voltage, 120 kV) was performed through the heart during tidal respiration for attenuation correction and to aid in anatomic localization of radiotracer activity. The PET acquisition commenced at the same time as injection of a single intravenous 4 MBq/kg (0.1 mCi/kg) bolus of ¹⁸F-florbetaben, without any observed hemodynamic effects. The average activity received by the patients was 259 MBq (7 mCi), with an SD of 41 MBq (1.1 mCi). Dynamic images were reconstructed from the list-mode data using 80 frames of 1 min each, with a matrix of 200 × 200, using ordered-subset expectation maximization iterative technique (3 iterations and 21 subsets). Static images were reconstructed with a matrix of 400 × 400 and a slice thickness of 3 mm. The estimated whole-body effective radiation dose to each patient was 5.8 mSv.

Cardiac PET Analyses

Qualitative assessment of myocardial ¹⁸F-florbetaben uptake was performed visually using static images reconstructed from PET data acquired between 15 and 75 min after the injection of ¹⁸F-florbetaben. To further assess contrast between myocardial ¹⁸F-florbetaben uptake and the background blood pool over time (the target-to-background ratio), the SUV_mean in the LV myocardium and blood pool was plotted every minute and displayed graphically. Therefore, a higher target-to-background ratio is represented by a greater separation between the myocardial SUV curve and the blood-pool SUV curve over time. LV myocardial SUV was calculated by placing a volume of interest around the heart on summed images between 5 and 10 min after radiotracer injection, when there was a maximal difference between myocardial and ventricular luminal blood-pool SUV (Fig. 1). Using an automated isocontour function in commercially available software (syngo.via; Siemens Healthcare), we transposed the obtained myocardial contour to the remainder of the frames for dynamic analysis (Supplemental Fig. 1; supplemental materials are available at http://jnm.snmjournals.org). We used the default isocontour threshold of 40% SUV_max because it corresponded best with the actual boundaries of the myocardium, as is generally the case with PET tumor imaging as recommended by guidelines (7). Blood-pool SUV was measured using a 1-cm-diameter region of interest in the mid descending thoracic aorta (extended over 2 cm in the z-axis). This region of interest can also be generated automatically using the PET analysis software, providing the benefits of reproducibility, ease of performance, and accuracy of localization of blood-pool activity.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1.

Example of software-automated volume-of-interest isocontouring performed on 5- to 10-min postflorbetaben images of patient with cardiac amyloidosis.

To quantify myocardial accumulation of ¹⁸F-florbetaben, both ¹⁸F-florbetaben retention index (RI) and percentage myocardial ¹⁸F-florbetaben retention were calculated. ¹⁸F-florbetaben RI was calculated as the mean myocardial ¹⁸F-florbetaben SUV between 15 and 20 min after injection divided by the integral of the arterial time–activity curve between 0 and 17.5 min after injection, the midpoint of 15–20 min (8). Calculation of percentage myocardial ¹⁸F-florbetaben retention was based on the change in SUV_mean on summed-framed images of the LV myocardium between 0 and 5 min and 15 and 20 min after radiotracer injection using the following formula:where SUV_0–5 is mean myocardial SUV between 0 and 5 min and SUV_15–20 is mean myocardial SUV between 15 and 20 min.

Echocardiography

With the patients at rest, a transthoracic echocardiogram was obtained using a commercially available ultrasound system (Vivid E9, 4V probe; GE Healthcare) and digitally stored for offline analysis (EchoPAC, version 113; GE Healthcare). A complete 2D, color, pulsed, continuous-wave Doppler echocardiogram was obtained using standard techniques (9,10).

LV mass index was calculated (7) and corrected for body surface area (11). Relative wall thickness was calculated (10). LV end-diastolic volume index and end-systolic volume index were calculated using the biplane Simpson method of disks and corrected for body surface area. LV ejection fraction was derived. Stroke volume was calculated from LV outflow tract cross-sectional area and the velocity–time integral and indexed to body surface area. Cardiac output was derived from the calculated stroke volume and heart rate.

RV free wall thickness was measured at the level of the tricuspid annulus in the 2D subcostal view (12). Changes in RV end-diastolic area, end-systolic area, and fractional area were measured (12).

Mitral inflow early (E wave) and late (A wave) diastolic velocities and pulmonary venous peak systolic and diastolic velocities were recorded using pulsed-wave Doppler echocardiography in the apical 4-chamber view.

Pulsed-wave tissue Doppler velocities were recorded at the septal and lateral mitral annulus in the apical 4-chamber view; mitral annular systolic (s′) and early diastolic (e′) velocities were obtained (13). LV filling pressure (E/e′) was calculated as the ratio of transmitral E wave to average mitral annular e′ velocity. Tricuspid annular s′ and e′ velocities were also recorded.

2D speckle-tracking analyses were performed on standard gray-scale images in the apical 2-, 3-, and 4-chamber views. LV global longitudinal strain was calculated from the 3 individual apical global longitudinal strain curves, whereas RV free wall longitudinal strain was obtained from the apical 4-chamber view (13).

Analysis of all PET results was masked to the echocardiographic results. As the PET analyses were automated, there were no intra- or interobserver measurement variabilities for myocardial SUV, background blood-pool SUV, or myocardial ¹⁸F-florbetaben RI. Analysis of all echocardiographic results was also masked to the PET results. Previous work has reported intra- and interobserver measurement variabilities of 1.2% ± 0.5% and 0.9% ± 1.0%, respectively, for longitudinal strain, expressed as mean absolute difference (14). Using a similar methodology, the intra- and interobserver variabilities for measurement of global longitudinal strain in the present study were 0.8% ± 0.3% and 1.0% ± 0.5%, respectively, expressed as mean absolute difference ± SD.

Statistical Analysis

Continuous variables are presented as mean ± SD, and categoric variables as frequencies and percentages. When comparing categoric variables, we used the exact χ² test using the Monte Carlo method with 99% confidence interval and 10,000 samples to calculate the P value. The Kruskal–Wallis test was used to compare 3 groups of continuous variables, and the Mann–Whitney U test was used to compare 2 groups of continuous variables. For significant results, post hoc multiple comparisons were performed using Bonferroni adjustments. Pearson correlation was used to determine the association between 2 continuous variables. To determine the correlation between myocardial ¹⁸F-florbetaben RI and biventricular strain, myocardial ¹⁸F-florbetaben RI was first inverse-transformed. Multiple linear regression analyses were then used to determine the independent association between the inverse transformed myocardial ¹⁸F-florbetaben RI and global LV/RV free wall longitudinal strain. To prevent overfitting of the data, no more than two of the most significant univariable predictors were sequentially entered into the multiple linear regression models. In addition, a tolerance of less than 0.5 (equating to model R > 0.7) was set to avoid multicolinearity. To determine the relative contribution of each predictor in the multivariable analyses, standardized β-values are presented. A 2-tailed P value of less than 0.05 was considered significant. All statistical analyses were performed using SPSS Statistics (version 22.0; IBM) for Windows (Microsoft).