Impact of ferumoxytol vs gadolinium on 4D flow cardiovascular magnetic resonance measurements in small children with congenital heart disease

This was a single-center retrospective, Institutional Review Board-approved study of 44 sequential CMR studies performed from October 2019 to October 2021. All CMR studies were clinically indicated and protocolled to answer salient clinical questions at the time of acquisition, per lab standard, on a 1.5 T CMR scanner (Aera, Siemens Healthineers, Erlangen, Germany). All included studies had both clinically relevant 2D-PC flows performed as well as a single 4D Flow CMR retrospectively gated acquisition. Acquisition parameters are found in Table 1 for both 2D-PC and 4D Flow acquisitions for 1.3 mm isotropic (up to 7 kg) and 1.8 mm isotropic (up to 25 kg) voxels, per lab standard. Contrast agent selection was at the discretion of the overseeing imager at the time of acquisition in accordance with institutional policy, which includes reviewing an information sheet in discussion with the patient and family members regarding safety and risks of contrast. In general, very ill infants, patients who required additional non-CMR studies during one sedation, and those with any history of drug allergies did not receive ferumoxytol during this study period. Clinically appropriate velocity encoding (VENC) for 2D-PC and 4D Flow CMR varied and were chosen by the overseeing imager at the time of acquisition, and based on velocities from the most recent echocardiogram while accounting for physiologic changes anticipated with sedation, if appropriate. The overseeing imager at the time of acquisition collected 2D-PC data for all clinically relevant vessels, thus the type of vessels where flow was measured varied patient-to-patient and by diagnosis. The 4D Flow measurements were then performed for each corresponding 2D-PC acquisition. Gadolinium-enhanced studies were performed with 0.1 mmol/kg of Dotarem (Guerbet, Villepinte, France) bolus for purposes of magnetic resonance angiogram (MRA) acquisition, with 4D Flow acquisitions performed immediately following administration and 2D-PC completed within 18 ± 6 min on average. Ferumoxytol-enhanced studies were performed with 3 mg/kg of ferumoxytol (max 510 mg/dose) infused over 15 min just prior to entry to CMR suite. 2D-PC and 4D Flow CMR were completed within an average of 38 ± 8 min from the start of the scan. No studies were excluded for reasons of quality or artifacts.

Measurements of both 4D Flow and 2D-PC datasets were obtained in the great vessels (aorta, main pulmonary artery, or proximal branch pulmonary arteries), systemic veins (right or left superior vena cava and inferior vena cava), and pulmonary veins (right and left upper and lower pulmonary veins), Fig. 1. Following selection of the cohort, the patients were randomized prior to analysis, without identification of type of contrast administered for each individual study. 4D Flow measurements were performed after background-phase correction on de-identified datasets by a single investigator trained in PC measurement (Arterys Inc., San Francisco, California, USA). All 2D-PC flow measurements were performed by the non-invasive imaging cardiologist at the time of image acquisition and were reported in the medical record. A subset of 40% of studies were made available for a second, blinded reader to perform 4D Flow CMR measurements for interobserver reliability analysis. Intraobserver reliability measurements were performed at the location of the ascending aorta vs. sinotubular junction, distal right pulmonary artery vs. proximal right pulmonary artery, and high superior vena cava vs. low superior vena cava.

The CNR will vary with exogenous contrast administration, slice thickness and spatial resolution, which can affect accuracy of flow measurements regardless of contrast type selected. Therefore, the CNR was calculated, per 4D Flow acquisition, to assess the differences in signal intensity of the blood pool among the cohort in a similar manner to Mukai et al. [15]. Measurements were performed at end systole at the level of the aortic root in the transverse plane where corresponding 4D Flow measurements were obtained. The maximum aorta signal was obtained in the aortic root and the mean chest signal was obtained from sampling a portion of the distal lung fields away from central vasculature in the thoracic cavity. The standard deviation of noise in air was calculated from the mean of two samples outside the thoracic cavity, one anterior and midline to the chest wall and one anterior and rightward. CNR was calculated by dividing the difference between the maximum aorta signal and the mean chest signal by the standard deviation of noise in air, Fig. 2. Measurements were performed blinded to type of contrast agent administered and spatial resolution.

Subgroup analysis was performed comparing two ventricle CHD patients to a subset of 10 single ventricle patients comprised of five patients with ferumoxytol and five patients with gadolinium-enhanced images. Additionally, vessels with low flows (< 1.5 L/min) were compared between the two ventricle and single ventricle cohorts. From the total cohort, the accuracy of ferumoxytol was compared to gadolinium when measuring arterial flows (aorta and pulmonary arteries) vs. venous flows (pulmonary veins and vena cavae).

All statistical analyses were performed with Prism 8 (Graphpad, San Diego, California, USA) with the exception of interobserver analysis, which was performed with MedCalc 12.2.1.0 (MedCalce Software, Ostend, Belgium). Demographic data (including age, body surface area (BSA), and vessel diameter) and CNR for the two contrast agents were compared using an unpaired t test. Ferumoxytol and gadolinium-enhanced 4D Flow CMR measurements were compared as percent difference from 2D-PC flow measurements using an unpaired t test. Simple linear regression was used to assess 4D Flow measurements in comparison to 2D-PC flow measurements from both ferumoxytol and gadolinium-enhanced images. For all tests, a P value < 0.05 was considered statistically significant. As part of the interobserver analysis, the intraclass correlation coefficient (ICC) reflected the average level of agreement between 4D flow measurements obtained by the primary investigator and a secondary reader. Similarly, for intraobserver analysis, the ICC reflected the average level of agreement between 4D flow measurements obtained in close proximity from the same vessel by the primary investigator.

Measurements were obtained from 21 ferumoxytol-enhanced and 23 gadolinium-enhanced CMR studies. The patients included in this study were those with predominantly complex CHD including single ventricle anatomy, conotruncal anomalies, anomalous systemic or pulmonary venous drainage, and septal/atrioventricular valve anomalies. A subgroup analysis of 10 CMR studies focused on variability in measurements specifically in the single ventricle population.

The ferumoxytol and gadolinium administration groups were not statistically different with respect to average age (2.1 ± 1.6 vs. 2.3 ± 1.9 years), BSA (0.50 ± 0.2 m² vs. 0.52 ± 0.2 m²), proportion of single ventricle patients (24% vs. 22%), and average vessel diameter (great vessels (16.3 ± 5 vs. 15.7 ± 5.9 mm), branch pulmonary arteries (9 ± 3.1 vs. 9.5 ± 2.7 mm), pulmonary veins (7.2 ± 1.6 vs. 7.6 ± 2 mm), and systemic veins (12.7 ± 3.9 vs. 10.7 ± 3.8 mm), Table 2. There was only one patient with significant semilunar valve disease, classified as moderate to severe truncal valve insufficiency in the gadolinium cohort. There was no statistical difference in the proportion of patients in either cohort that received moderate sedation vs. general anesthesia (p = 0.217). All 4D Flow CMR acquisitions were performed free-breathing with respiratory navigation, regardless of type of anesthesia administered. No contrast reactions were noted following administration of either ferumoxytol or gadolinium. Lastly, given the higher expected CNR, more patients given ferumoxytol were imaged with a higher resolution compared to gadolinium-based contrast.

Ferumoxytol-enhanced 4D Flow CMR measurements demonstrated less percent difference from 2D-PC flow measurements when compared to gadolinium-enhanced studies (mean difference = 7.9 vs. 13, p = 0.01). This finding was even more pronounced in the single ventricle cohort with improved accuracy of flow measurements obtained from ferumoxytol-enhanced images compared to gadolinium (mean difference = 6.1 vs. 17.6, p = 0.001), Fig. 3A. In the subgroup analysis of vessels in single ventricle patients with low flows, defined as less than 1.5 L/min, ferumoxytol-enhanced 4D Flow CMR measurements had stronger correlation to 2D-PC flow than gadolinium-enhanced measurements, Fig. 3B.

In the subgroup analysis of arterial (aorta and pulmonary artery) vs. venous (pulmonary veins and vena cava), ferumoxytol-enhanced 4D Flow CMR measurements demonstrated less percent difference from 2D-PC flow measurements when compared to gadolinium-enhanced studies, Fig. 4A. This finding was more pronounced in the venous flow measurements (mean difference = 6.1 vs. 16.6, p = 0.001). Additional analysis was performed to assess the difference between the various types of vessels analyzed, Fig. 4B. Ferumoxytol-enhanced 4D Flow CMR measurements demonstrated less percent difference from 2D-PC flow measurements, which was statistically significant in the branch pulmonary arteries, systemic veins, and pulmonary veins (p = 0.008, p = 0.028, and p = 0.030 respectively).

Pulmonary venous flow measurements were isolated and analysis of 4D Flow CMR measurements acquired with low (150–200 cm/s) vs. high (250–375 cm/s) VENC was performed, Fig. 5. Ferumoxytol-enhanced 4D flow measurements demonstrated less percent difference from 2D-PC flow measurements in the low VENC cohort (mean difference = 6.8 vs. 20.3, p = 0.003), but there was no statistical difference in the high VENC cohort. In summary, all ferumoxytol-enhanced 4D flow measurements across all vessels had < 10% difference from 2D-PC measurements, compared to gadolinium which had > 10% difference.

There was a significant difference between the CNR in ferumoxytol studies with a spatial resolution of 1.3 mm (92 ± 34, n = 14) vs. 1.8 mm (185 ± 81, n = 6; p = 0.002). There was no significant difference in the CNR between gadolinium-enhanced studies performed at a spatial resolution of 1.3 mm vs. 1.8 mm. The CNR in ferumoxytol studies with a spatial resolution of 1.3 mm (92 ± 34, n = 14) was still higher than the CNR in gadolinium studies with a spatial resolution of 1.8 mm (69 ± 21, n = 20; p = 0.036), Fig. 6.

The ICC values reflecting the average level of agreement between 4D Flow CMR measurements obtained by the primary investigator and a secondary observer were over 95%, demonstrating excellent agreement between observers. Specifically, the ICC for ferumoxytol-enhanced 4D Flow CMR measurements was 0.997 (0.994–0.999) and similarly 0.985 (0.970–0.992) for gadolinium-enhanced 4D Flow CMR. Bland–Altman analysis in Fig. 7A demonstrates excellent agreement between observers for both contrast agents. The ICC values reflecting the average level of agreement between 4D Flow CMR measurements obtained by the primary investigator were over 95%, demonstrating excellent agreement. Specifically, the ICC for ferumoxytol-enhanced 4D Flow CMR measurements was 0.999 (0.998–0.999) and similarly 0.998 (0.995–0.999) for gadolinium-enhanced measurements. Bland–Altman analysis in Fig. 7B demonstrates excellent agreement between repeat measurements sampled in selected vessels by the primary investigator.

The study is limited in that it was performed in a single center in a retrospective manner. Because the choice to administer gadolinium vs. ferumoxytol was clinical in nature, and not systematically varied, this may have introduced bias into the cohorts receiving each contrast agent. However, there were no significant differences in any clinical variable between the two cohorts, so likely any bias was minimized. The sample size was small, although large enough to demonstrate significant reproducibility of 4D Flow CMR measurements in comparison to 2D-PC flow. Implications of the small sample size may lead to type II error and failure to reject the null hypothesis. The inequal distribution of gadolinium-enhanced studies performed at a spatial resolution of 1.3 mm (n = 3) vs. 1.8 mm (n = 20) may have contributed to the insignificant difference in the CNR between these two groups. Additionally, the 4D Flow CMR measurements were performed on a single platform with customized eddy-current phase correction to minimize vendor variability. Further investigation to determine variability between different vendor platforms and different background correction methods are needed. Additionally, the 2D-PC flow measurements were performed in a clinical setting by various trained non-invasive imaging cardiologists where the spatial resolution, velocity encoding, and scanning time were determined independently based on patient-specific clinical indications. In contrast, the 4D Flow CMR measurements were performed by an independent operator in a research setting. There were undoubtedly more patients in the gadolinium cohort with a spatial resolution of 1.8 mm, which is a limitation of the retrospective nature of this study. However, our results demonstrate the improved ability to augment CNR with ferumoxytol, even at a spatial resolution of 1.3 mm in comparison to gadolinium enhanced studies at 1.8 mm.