Comparison of ferumoxytol and gadolinium chelate-enhanced MRI for assessment of sarcomas in children and adolescents

Abstract

Objectives

We compared the value of ferumoxytol (FMX)- and gadolinium (Gd)-enhanced MRI for assessment of sarcomas in paediatric/adolescent patients and hypothesized that tumour size and morphological features can be equally well assessed with both protocols.

Methods

We conducted a retrospective study of paediatric/adolescent patients with newly diagnosed bone or soft tissue sarcomas and both pre-treatment FMX- and Gd-MRI scans with maximal 4 weeks apart. Both protocols included T1- and T2-weighted sequences. One reader assessed tumour volumes, signal-to-noise ratios (SNR) of the primary tumour and adjacent tissues and contrast-to-noise ratios (CNR) with subsequent comparison between FMX and Gd-MRI scans. Additionally, four readers scored 15 diagnostic parameters using a Likert scale. Then, the results were pooled across readers and compared between FMX- and Gd-MRI scans. Statistical methods included multivariate analyses with different models.

Results

Twenty-two patients met inclusion criteria (16 males, 6 females; mean age 15.3±5.0). Tumour volume was not significantly different on T1-LAVA (p=0.721), T1-SE (p=0.290) and T2-FSE (p=0.609). Compared to Gd-MRI, FMX-MRI demonstrated significantly lower tumour SNR on T1-LAVA (p<0.001), equal tumour SNR on T1-SE (p=0.104) and T2-FSE (p=0.305), significantly higher tumour-to-marrow CNR (p<0.001) on T2-FSE as well as tumour-to-liver (p=0.021) and tumour-to-vessel (p=0.003) CNR on T1-LAVA. Contrast-enhanced peritumoural and marrow oedema were significantly more frequent on Gd-MRI (p<0.001/p=0.002). Tumour thrombi and neurovascular bundle involvement were assessed with a significantly higher confidence on FMX-MRI (both p<0.001).

Conclusions

FMX-MRI provides equal assessment of the extent of bone and soft tissue sarcomas compared to Gd-MRI with improved evaluation of neurovascular involvement and tumour thrombi making it a valid alternative for tumour staging in paediatric/adolescent sarcoma patients.

Introduction

Gadolinium-based contrast agents (GBCAs) are the standard contrast-agent for magnetic resonance imaging (MRI) since their introduction in 1988 [1]. However, GBCAs can be associated with side effects, such as nephrogenic systemic fibrosis (NSF) in patients with renal failure [2, 3] and gadolinium (Gd) deposition in the brain after repetitive administration [4]. Although recent studies showed that macrocyclic GBCAs release less Gd in the brain parenchyma than linear Gd-chelates [5–8], an U.S. food and drug administration (FDA) panel recommended adding a warning to all Gd-agents about retention in certain organs including the brain in September 2017. Three months later, the FDA required all GBCA manufacturers to conduct further studies on the safety of these agents [9]. These interventions support the increasing notion about unknown long-term risks of GBCAs on the human body with unclear clinical significance. Especially in children and adolescents with oncologic diseases who often require multiple MR scans to assess treatment response or continued remission, the repetitive application of Gd-chelates could induce long-term complications that are yet unknown.

Thus, there is a need for alternatives to Gd-chelates with comparable or better properties for disease assessment. An immediately clinically available alternative is the FDA-approved iron supplement ferumoxytol (Feraheme®; AMAG Pharmaceuticals). Ferumoxytol (FMX) is composed of ultrasmall superparamagnetic iron oxide nanoparticles, which can be detected on MRI [10–12] and are being used “off label” to enhance MR scans in patients [13–18]. After intravenous injection, ferumoxytol nanoparticles provide long-lasting positive (bright) T1-contrast enhancement, which can be used for MR angiography [19, 20], assessment of tumour perfusion [21] and whole-body imaging [22]. Subsequently, the nanoparticles are phagocytosed by macrophages in the liver, spleen and bone marrow which leads to a hypointense (dark) signal on T2w-images [23, 24]. Ferumoxytol is not excreted by the kidneys and elimination takes place entirely through macrophage phagocytosis with subsequent metabolization in the reticuloendothelial system. Therefore, ferumoxytol is not associated with any direct or indirect renal toxicity [10, 12, 25–27]. Additionally, ferumoxytol does not cross the blood brain barrier and is therefore not associated with any known risk of deposition in the brain thus far. It is also important to mention that ferumoxytol does not induce an iron overload at the regular dose of 5 mg/kg but that it is contraindicated in patients with a pre-existing iron overload due to e.g. hemosiderosis or hemochromatosis.

Our group has pioneered the “off-label” use of ferumoxytol as a contrast agent for MRI to examine inflammation [9], tumours [24] and stem cell tracking [10, 11, 25]. MR imaging is essential for the diagnosis of sarcomas because it sets up the foundation for surgical approaches and patient-specific treatment protocols [30]. To our knowledge, the value of ferumoxytol for the assessment of sarcomas has not been studied.

Therefore, the purpose of our study was to compare the value of ferumoxytol- and Gd-chelate-enhanced MRI for the assessment of sarcomas in paediatric and adolescent patients. We hypothesized that tumour size and morphological features can be equally assessed on FMX- and Gd-enhanced MRI scans.

Materials and Methods

Study population

The institutional review board at our institution approved this single centre retrospective study and informed consent was waived. Between June 2015 and December 2018, we included patients aged of 6–30 years with a histopathologically confirmed diagnosis of a bone or soft tissue sarcoma who had completed a FMX-enhanced (FMX-MRI) and Gd-enhanced (Gd-MRI) MRI scan, including a T1-liver acquisition with volume acceleration (T1-LAVA) and T2-fast spin echo (T2-FSE) sequence, before treatment initiation. Exclusion criteria were a scan interval between FMX- and Gd-MRI of more than 4 weeks, technically considered inadequate scans (e.g. motion artefacts, incomplete study) or missing contrast-enhancement. Most of our patients also received a T1-spin echo (T1-SE) sequence. However, a missing T1-SE sequence as part of either FMX-MRI or Gd-MRI or both protocols did not lead to patient exclusion, since the main goal of our study was to compare Fe- and Gd-enhanced sequences.

Magnetic Resonance Imaging

FMX-MRI scans were carried out on a 3T Signa PET/MRI scanner (GE Healthcare) and included axial fat-saturated T1-LAVA, T1-SE and fat-saturated T2-FSE sequences at 2–24 hours after a slow infusion of ferumoxytol (Feraheme®; AMAG Pharmaceuticals) at a dose of 5 mg Fe/kg (284.1 ± 108.1 mg; range: 115.0–466.0 mg). The non fat-saturated T1-SE sequence was not obtained in 8 of the 22 patients. These studies were obtained as part of a prospective clinical trial () with investigational new drug (IND) authorization from the FDA for using ferumoxytol nanoparticles “off label” for MR imaging of paediatric patients (IND111,154). Ferumoxytol was administered slowly 1:3 diluted in saline over 15 minutes, according to FDA protocols [31]. The FMX-protocol did not include pre-contrast scans as they were considered impractical because of prolonged scan times for the young patients.

Gd-MRI scans were carried out on a 3T Discovery MR750 or a 3T Signa HDxT (GE Healthcare) scanner. Pulse sequences comprised axial unenhanced T1-SE and fat-saturated T2-FSE sequences as well as axial fat-saturated T1-LAVA sequences directly after intravenous injection of 0.1 mmol Gd/kg gadobutrol (Gadavist®, Bayer HealthCare Pharmaceuticals). The non fat-saturated T1-SE sequence was not obtained in 3 of the 22 patients.

The most relevant characteristics of both used contrast agents (gadobutrol vs. ferumoxytol) are presented in Table 1. Detailed sequence parameters for both protocols are presented in Table 2.

Table 1:

Characteristics of gadobutrol and ferumoxytol

	Gadobutrol (Gadavist®)	Ferumoxytol (Feraheme®)
Clinically approved application	MR Imaging	Anaemia treatment
Size / Molecular weight	604.72 g/mol [1,2]	USPIO: 20–30 nm [3]
R1 relaxivity in plasma at 37°C at 3 Tesla	5.0 (L/(mmol x sec)) [4]	10.1 (L/(mmol x sec)) [5]
R2 relaxivity in plasma at 37°C at 3 Tesla	7.1 (L/(mmol x sec)) [4]	73.2 (L/(mmol x sec)) [5]
Dose	0.1 mmol Gd/kg (0.1 mL/kg)	1–5 mg Fe/kg dilution: 1:3 in sterile saline
Mode of i.v.-administration	i.v.-bolus injection	Slow i.v.-infusion over 15 minutes
Elimination	Renal excretion	Macrophage phagocytosis and metabolization in the reticuloendothelial system
Blood half-life in humans	1.5–2.5 hours (depending on patient’s age and renal clearance rate) [6]	10–14 hours [7]
Possible re-administration for follow up exam	After 24 hours	After 4–6 weeks
Signal change on T1-weighted sequence	Increased signal	Increased signal (signal decreased at very sequence high concentrations)
Signal change on T2-weighted sequence	Usually no change	Decreased signal
Signal change on T2*-weighted sequence	Decreased signal if given as a bolus	Decreased signal
Contraindications	- Previous anaphylaxis - Renal insufficiency	- Previous anaphylaxis - Hemosiderosis/hemochromatosis
Potential side effects	- Allergic-like reaction/ allergic reaction - Nephrogenic systemic fibrosis (NSF) - Gd-deposition in the brain	- Allergic-like reaction/ allergic reaction - Iron deposition in the reticuloendothelial system
Risk of allergic reactions	- 0.32 % of subjects [8,9]	- 0.2 % of subjects [10,11]

USPIO - Ultrasmall superparamagnetic iron oxide nanoparticle.

References:

^1.Frenzel T, Lengsfeld P, Schirmer H, Hütter J, Weinmann HJ (2008) Stability of Gadolinium-based Magnetic Resonance Imaging Contrast Agents in Human Serum at 37°C. Invest Radiol 43:817–828

^2.Tombach B, Bremer C, Reimer P, et al (2000) Pharmacokinetics of 1M gadobutrol in patients with chronic renal failure. Invest Radiol 35:35–40

^3.Simon G, von Vopelius-Feldt J, Fu Y, et al (2006) Ultrasmall Supraparamagnetic Iron Oxide-Enhanced Magnetic Resonance Imaging of Antigen-Induced Arthritis: A Comparative Study Between SHU 555 C, Ferumoxtran-10, and Ferumoxytol. Invest Radiol 41:45–51

^4.Rohrer M, Bauer H, Mintorovitch J, Requardt M, Weinmann HJ (2005) Comparison of magnetic properties of MRI contrast media solutions at different magnetic field strengths. Invest Radiol 40:715–724

^5.Knobloch G, Colgan T, Wiens CN, et al (2018) Relaxivity of Ferumoxytol at 1.5 T and 3.0 T. Invest Radiol 53:257–263

^6.Aime S, Caravan P (2009) Biodistribution of gadolinium-based contrast agents, including gadolinium deposition. J Magn Reson Imaging 30:1259–1267

^7.Daldrup-Link HE (2017) Ten Things You Might Not Know about Iron Oxide Nanoparticles. Radiology 284:616–629

^8.Power S, Talbot N, Kucharczyk W, Mandell DM (2016) Allergic-like Reactions to the MR Imaging Contrast Agent Gadobutrol: A Prospective Study of 32 991 Consecutive Injections. Radiology 281:72–77

^9.Clement O, Dewachter P, Mouton-Faivre C, et al (2018) Immediate Hypersensitivity to Contrast Agents: The French 5-year CIRTACI Study. EClinicalMedicine 1:51–61

^10.Toth GB, Varallyay CG, Horvath A, et al (2017) Current and Potential Imaging Applications of Ferumoxytol for Magnetic Resonance Imaging. Kidney Int 92:47–66

^11.Lu M, Cohen MH, Rieves D, Pazdur R (2010) FDA report: Ferumoxytol for intravenous iron therapy in adult patients with chronic kidney disease. Am J Hematol 85:315–319

Table 2:

Scan parameters

	FMX MRI			Gd-MRI
Sequence name	T1-LAVA	T1-SE	T2-FSE	T1-LAVA	T1-SE (unenhanced)	T2-FSE (unenhanced)
Sequence type	Spoiled 3D gradient echo	Spin echo	Fast spin echo	Spoiled 3D gradient echo	Spin echo	Fast spin echo
Image plane	Axial	Axial	Axial	Axial	Axial	Axial
Fat suppression	Yes	No	Yes	Yes	No	Yes
TE [ms]	1.7–1.9	8.3–12.6	46.8–114.0	1.7–3.8	5.3 −19.9	41.5–104.0
TR [ms]	4.1–8.7	485.0–795.0	2,500–12,861	3.1–8.4	474.0–889.0	3,000–12,000
FA [°]	12–15	111–142	110–142	12–15	90–140	90–180
Matrix size [pixel]	320 × 224–256	320–512 × 256–320	320–384 × 224–348	304–416 × 224–320	256–512 × 192–320	256–384 × 192–256
Slice thickness [mm]	0.5–3.0	2.5–5.0	3.0–6.0	1.0–4.4	3.0–6.0	3.0–7.0
Spacing [mm]	1.0–1.5	3.0–7.5	3.3–8.0	1.0–3.4	3.0–7.5	3.5–8.0

Ranges of values are presented. FMX-MRI – Ferumoxytol-enhanced MRI scan-protocol; Gd-MRI – Gadolinium-enhanced MRI scan-protocol (here T1-SE and T2-FSE were acquired before gadolinium injection); T1 - T1-weighted; T2 - T2-weighted; LAVA - liver acquisition with volume acceleration; SE – spin echo; FSE - fast spin echo; TE - echo time; TR - repetition time; FA - flip angle.

Quantitative Analyses

One radiologist (FS) with 6 years of experience in oncologic imaging performed the quantitative analyses. He measured the size of the primary tumour in the largest anterior-posterior, transverse and longitudinal dimension, as defined in Children’s Oncology Group (COG) protocols, on all three sequences of the Gd- and FMX-MRI scans. Subsequently, the tumour volume was calculated for each sequence in both groups using either the formula for a prolate ellipsoid (AP x transverse x longitudinal x 0.5: Ewing sarcoma, rhabdomyosarcoma, undifferentiated sarcoma, DSRCT, MPNST) or the product of all three dimensions (AP x transverse x longitudinal: osteosarcoma, GIST) based on the evaluation of target lesions for each tumour’s response criteria, suggested by the COG.

Additionally, he measured the signal-to-noise ratios ($\text{SNR}=\frac{meansignaltissue}{standarddeviationofbackgroundnoise}$) as previously described by Dietrich et al. [32] for the primary tumour and adjacent anatomical structures (reference tissues: muscle and bone marrow) on all sequences. In addition, SNRs were measured for two additional reference tissues on T1-LAVA sequences (vein in close vicinity to the tumour and liver - if acquired on both Gd- and FMX-MRI).

Based on these measurements, the absolute contrast-to-noise ratios (CNR= abs(SNR_a - SNR_b)) as previously described by Li et al. [33] were calculated as a quantitative measure of the contrast between the primary tumour and adjacent anatomical structures. Tumour-to-muscle CNRs and tumour-to-bone marrow CNRs were calculated for all sequences while tumour-to-vein CNRs and tumour-to-liver CNRs were calculated on the T1-LAVA sequence.

Regarding Gd-MRI, 17 studies were performed 1–42 days before biopsy of the primary tumour, 1 study was performed on the same day right before biopsy and 4 imaging studies were performed 4–35 days after biopsy. Regarding FMX-MRI, 6 studies were performed 5–62 days before biopsy of the primary, 2 studies were performed on the same day right before biopsy and 14 imaging studies were performed 1–43 days after biopsy. To make sure that measurements were not influenced by the biopsy area with potentially associated tissue changes (haematoma, oedema with associated contrast-enhancement), the biopsy area was excluded from the quantitative analyses.

Qualitative Analyses

Four radiologists (FS, AJT, CF, HDL) with 6 to 20 years of experience in oncologic imaging analysed images of all three sequences of both Gd- and FMX-MRI independently. The readers were blinded to histopathology, disease history, and other imaging findings. Sequences from Gd- and FMX-MRI of each patient were evaluated at least two weeks apart from each other to minimize recall bias.

A Likert scale from 1–5 (1=definitely absent, 2=probably absent, 3=uncertain, 4=probably present, 5=definitely present) was used to assess 15 different tumour parameters in 6 categories that are relevant for evaluation of sarcomas: Tumour extension (tumour presence in diaphysis/epiphysis/metaphysis, extraosseous tumour tissue, joint invasion); Tumour composition (tumour necrosis); Cortical involvement (cortical destruction, pathological fracture); Vascular involvement (encasement of neurovascular bundle, tumour thrombus); Tumour oedema (contrast-enhanced peritumoural oedema, contrast-enhanced bone marrow oedema); Tumour metastases (skip lesions, bone metastases, lymph node metastases). The results were pooled across all four readers and then analysed according their agreement on tumour assessment (affected – yes/no/uncertain) and confidence (uncertain/probably/definitely).

Statistics

All quantitative measurements were compared between corresponding sequences of FMX-MRI and Gd-MRI scans, using a generalized linear model while considering clustering within patients. For tumour volumes as well as SNR and CNR measurements, mean values, 95% confidence intervals (95%CI) and mean differences between FMX-MRI and Gd-MRI data with 95%CIs were calculated. When the estimated 95%CIs of the mean difference fell within the ±15% equivalence interval of the mean value, FMX- and Gd-MRI were considered equivalent and not significantly different. When the estimated 95%CIs crossed or fell outside the ±15% equivalence interval without crossing zero FMX- and Gd-MRI were considered significantly different for that measure.

All qualitative parameters were compared between FMX-MRI and Gd-MRI for all sequences combined using a mixed effect model while considering clustering within patients and readers. Agreement between FMX-MRI and Gd-MRI in terms of assessment of the parameters’ presence as well as the associated confidence level was assessed. After dichotomization of the data, the proportions of the parameters’ presence (“yes” (1) and “no” (0)) for FMX-MRI and Gd-MRI and their differences were estimated to rate assessment. Similarly, the proportions of the associated confidence (“definitely” (1) and “probably” (0)) for FMX-MRI and Gd-MRI and their differences were estimated to rate confidence. Significant differences were assumed for p≤0.05.

Apart from the main qualitative analyses using a mixed effect model, we also performed separate inter-reader agreement analyses for the 4 readers for Gd-MRI and FMX-MRI for each sequence (T1-LAVA, T1-SE, T2-FSE) and for each of the 15 parameters based on the original reading (Likert-scale from 1–5) as well as after dichotomization (“yes” (1) and “no” (0)) of the data to rate the parameters’ presence (assessment) and the associated confidence. Inter-reader agreement analyses were calculated by using Krippendorff’s alpha (k alpha) as the agreement index.

Results

Study population

We identified 22 patients that matched our inclusion and exclusion criteria (see Figure 1, which shows a study flowchart). Our cohort was comprised of 6 female and 16 male patients with bone sarcomas (osteosarcoma (n=12), Ewing sarcoma (n=2)) and soft tissue sarcomas (rhabdomyosarcoma (n=3), undifferentiated sarcoma (n=2), malignant peripheral nerve sheath tumour (MPNST), desmoplastic small round cell tumour (DSRCT), gastrointestinal stroma tumour (GIST) (each n=1)). The mean age was 15.3 ± 5.0 years (range: 7.3–25.3 years). Detailed demographic data are listed in Supplemental table 1.

Figure 1: Study flowchart:

This study flowchart illustrates patient in- and exclusion as well as the subsequent quantitative and qualitative analyses and used statistical methods.

Quantitative measurements

Tumour size and volume

Tumours could be equally well delineated on FMX- and Gd-enhanced scans without visually apparent differences in tumour size (see Figure 2). Mean differences in tumour volumes between FMX-MRI and Gd-MRI were 3.3±43.7 mm³ (T1-LAVA), 9.9±22.9 mm³ (T1-SE) and −11.5±103.0 mm³ (T2-FSE) with 95%CIs of −21.6–14.9 mm³, −8.5–28.5 mm³ and −55.5–32.5 mm³, respectively. All 95%CIs were within the equivalence interval for volume measurements and therefore tumour size measurements were not significantly different between FMX- and Gd-enhanced scans for any sequence (p=0.721, p=0.290 and p=0.609, respectively; see Table 3).

Figure 2: Tumour size comparison between Gd- and FMX-MRI:

Osteosarcoma of the right distal femur of an 18-year-old male on (a) coronal unenhanced T1-SE image or (b) ferumoxytol-enhanced T1-SE image illustrating equivalent tumour size. The white lines represent transverse and longitudinal dimensions in cm. (c) Mean tumour volume (anterior-posterior x transverse x longitudinal dimension) and 95% confidence intervals (95%CI) on T1-LAVA, T1-SE and T2-FSE sequences, using the Gd-MRI protocol or FMX-MRI protocol. Mean data were not significantly different between the two protocols (multivariate analysis with generalized linear model). Significant differences were assumed for p≤0.05.

Gd-MRI = gadolinium-enhanced MRI scan-protocol; FMX-MRI = ferumoxytol-enhanced MRI scan-protocol.

Table 3:

Results quantitative analysis – Tumour volume

Sequences	Mean FMX-MRI	FMX-MRI 95%CI		Mean Gd-MRI	Gd-MRI 95%CI		Mean Diff.	Difference 95%CI		±15% Equivalence Interval	p-value
		lower	upper		lower	upper		lower	upper
T1-LAVA	567.1	328.3	979.7	570.5	328.7	990.1	−3.3	−21.6	14.9	85.6	0.721
T1-SE	363.6	220.6	599.6	353.7	220.0	568.6	9.9	−8.5	28.5	54.6	0.290
T2-FSE	565.8	324.9	985.3	577.3	321.8	1035.8	−11.5	−55.5	32.5	86.6	0.609

Values in mm³. FMX-MRI = Ferumoxytol-enhanced MRI scan-protocol; Gd-MRI = Gadolinium-enhanced MRI scan-protocol; 95%CI = 95% confidence interval; when 95%CI of the differences between FMX- and Gd-MRI fell completely in bound of a ±15% equivalence interval, the observed differences in volume were defined to be practically equivalent to zero. Multivariate analysis with generalized linear model. Significant differences were assumed for p≤0.05.

Note: Mean and standard deviation are smaller for T1-SE compared to T1-LAVA and T2-FSE because T1-SE sequence was rarely missing in either Gd- and/or FMX-MRI.

Signal-to-noise ratio (SNR) of primary tumour and adjacent compartments

For results of the SNR-analyses see Supplemental results, which describe the results that demonstrate the signal-to-noise ratio (SNR) of the primary tumour and adjacent compartments.

Contrast between primary tumour and adjacent compartments

All primary tumours showed strong hyperintense Gd-enhancement on T1-weighted LAVA images, while tumour T1-enhancement after ferumoxytol administration was variable, with most tumours showing minor or no ferumoxytol enhancement on T1-LAVA images (see Figure 3). Accordingly, the mean tumour SNR was significantly higher on Gd- compared to FMX-enhanced T1-LAVA images (146.8 vs 43.5; p<0.001, see Supplemental table 2, which shows the results of the quantitative analysis - SNR) and the mean tumour-to-muscle CNR was significantly higher on Gd- compared to FMX-enhanced T1-LAVA images (78.2 vs 38.9; p=0.01, see Table 4). However, mean tumour SNR and tumour-to-muscle CNR data were not significantly different between FMX-enhanced and unenhanced T1-SE images or FMX-enhanced and unenhanced T2-FSE images (see Table 4 and Supplemental figure 1 and 2, which show the results of the quantitative analysis – CNR and SNR). Thus, while overall tumour T1-enhancement was lower on FMX-MRI, tumour delineation from surrounding muscle was equally well possible on FMX-enhanced T2-weighted scans.

Figure 3: Rhabdomyosarcoma of the right gluteus musculature (arrowheads) and adjacent subcutaneous soft tissue in an 11-year-old male, evaluated on Gd- and FMX-MRI:

(a) Axial gadolinium-enhanced T1-LAVA image shows strong hyperintense contrast enhancement of the tumour. (b) Ferumoxytol-enhanced T1-LAVA image shows only minor hyperintense tumour enhancement. (c) Unenhanced T2-FSE image and (d) ferumoxytol-enhanced T2-FSE image show the tumour equally well. (e) Mean tumour SNR and 95% confidence intervals (95%CI) on T1-LAVA, T1-SE and T2-FSE sequences, using the Gd-MRI protocol or FMX-MRI protocol. Mean tumour SNR is significantly higher on Gd- compared to FMX-MRI on T1-LAVA sequences while there is no significant difference on T1-SE and T2-FSE sequences (multivariate analysis with generalized linear model). Significant differences were assumed for p≤0.05.

Gd-MRI = gadolinium-enhanced MRI scan-protocol; FMX-MRI = ferumoxytol-enhanced MRI scan-protocol.

Table 4:

Results quantitative analysis – Contrast-to-noise ratio (CNR)

Contrast between tumour and	Mean CNR on FMX-MRI	FMX-MRI 95%CI		Mean CNR on Gd-MRI	Gd-MRI 95%CI		Mean CNR Diff.	Difference 95%CI		±15% Equivalence Interval	p-value
		lower	upper		lower	upper		lower	upper
T1-LAVA
• muscle	38.9	27.3	55.5	78.2	58.2	105.0	−39.3	−69.3	−9.2	-	0.010
• marrow	27.6	19.5	38.9	117.8	86.2	161.0	−90.2	−126.4	−54.0	-	<0.001
• liver	86.2	51.4	144.3	23.3	12.2	44.4	62.9	9.6	116.2	-	0.021
• vein	168.6	110.6	257.0	66.4	40.0	110.2	102.2	35.3	169.2	-	0.003
T1-SE
• muscle	45.2	27.7	73.6	43.2	28.8	64.7	1.9	−10.3	14.3	6.8	0.755
• marrow	73.4	56.4	95.5	44.6	30.9	64.2	28.8	11.9	45.8	-	0.001
T2-FSE
• muscle	41.8	28.1	62.2	40.0	26.7	59.9	1.8	−9.3	12.9	6.3	0.749
• marrow	67.9	44.3	104.1	41.2	28.8	59.1	26.7	5.4	48.0	-	0.014

Contrast-to-noise ratios (CNRs) between primary tumour and adjacent compartments. FMX-MRI = Ferumoxytol-enhanced MRI scan-protocol; Gd-MRI = Gadolinium-enhanced MRI scan-protocol; 95%CI = 95% confidence interval; ±15% Equivalence interval is presented for insignificant results; Multivariate analysis with generalized linear model, significant results are highlighted in bold. Significant differences were assumed for p≤0.05.

However, FMX-MRI improved delineation of tumour deposits in the bone marrow (see Figure 4), liver (see Figure 5) and vessels (see Figure 6). Tumour-to-marrow CNR data were significantly higher for FMX-enhanced T2-FSE compared to unenhanced T2-FSE scans (p=0.014), tumour-to-liver CNR data were significantly higher for FMX- compared to Gd-enhanced T1-LAVA scans (p=0.021); and tumour-to-vessel CNR data were significantly higher for FMX- compared to Gd-enhanced T1-LAVA scans (p=0.003, see Table 4 and Supplemental figure 1, which shows the results of the quantitative analysis – CNR).

Figure 4: Contrast between osteosarcoma of the right distal femur and adjacent bone marrow in an 18-year-old male:

(a) It is difficult to differentiate tumour and peri-lesional oedema (arrow) on the axial unenhanced T2-FSE image. (b) The ferumoxytol-enhanced T2-FSE image shows hypointense enhancement of the normal bone marrow (arrow) and unchanged signal of the tumour, thereby increasing the tumour-to-marrow contrast. (c) Contrast-to-noise (CNR) between tumour and bone marrow, measured as mean value and 95% confidence intervals (95%CI). CNR between tumour and marrow is significantly higher on FMX- compared to Gd-MRI on T2-FSE (multivariate analysis with generalized linear model). Significant differences were assumed for p≤0.05.

Gd-MRI = gadolinium-enhanced MRI scan-protocol; FMX-MRI = ferumoxytol-enhanced MRI scan-protocol.

Figure 5: Contrast between peritoneal metastases and liver (arrowheads) in a 25-year-old male with desmoplastic small round cell tumour:

(a) It is difficult to differentiate tumour and liver oedema on the axial Gd-enhanced T1-LAVA image. (b) The ferumoxytol-enhanced T1-LAVA image shows hypointense enhancement of the tumour compared to the liver, thereby increasing the tumour-to-liver contrast. (c) Unenhanced T2-FSE image shows hyperintense tumour and intermediate T2-signal of the liver. (d) Ferumoxytol-enhanced T2-FSE image shows hypointense enhancement of the liver, thereby increasing the tumour-to-liver contrast. (e) Contrast-to-noise ratio (CNR) between tumour and liver, measured as mean value and 95% confidence intervals (95%CI). CNR between tumour and liver is significantly higher on FMX-compared to Gd-MRI (multivariate analysis with generalized linear model). Significant differences were assumed for p≤0.05.

Gd-MRI = gadolinium-enhanced MRI scan-protocol; FMX-MRI = ferumoxytol-enhanced MRI scan-protocol.

Figure 6: Detection of a tumour thrombus (arrow) on Gd- and FMX-enhanced MRI in a 14-year-old male with an osteosarcoma of the right proximal tibia:

(a) It is difficult to detect a tumour thrombus on the axial Gd-enhanced T1-LAVA image. (b) The ferumoxytol-enhanced T1-LAVA image shows hyperintense vessel enhancement with improved delineation of the tumour thrombus. (c) Unenhanced T2-FSE image shows hyperintense tumour thrombus. (d) Ferumoxytol-enhanced T2-FSE image shows improved delineation of the thrombus due to hypointense enhancement of the vessel lumen. (e) Contrast-to-noise ratio (CNR) between tumour and vein, measured as mean value and 95% confidence intervals (95%CI). CNR between tumour and vein is significantly higher on FMX- compared to Gd-MRI (multivariate analysis with generalized linear model). Significant differences were assumed for p≤0.05.

Gd-MRI = gadolinium-enhanced MRI scan-protocol; FMX-MRI = ferumoxytol-enhanced MRI scan-protocol.

Qualitative measurements

Qualitative assessments of 15 diagnostic criteria analysed with a mixed effects model revealed either perfect or high agreement between FMX- and Gd-MRI for 13 criteria, including tumour presence in diaphysis/epiphysis/metaphysis, extraosseous tumour tissue, joint invasion, tumour necrosis, cortical destruction, pathological fracture, encasement of neurovascular bundle, tumour thrombus, skip lesions, bone metastases, and lymph node metastases (see Table 5). However, readers noted significantly more contrast-enhancement of peritumoural oedema on Gd- than FMX-MRI (0.61 vs 0.06; p=<0.001) and significantly more contrast-enhancement of marrow oedema with Gd- than FMX-MRI (0.36 vs 0.10; p=0.002; see Table 5). Histopathology results showed that out of 14 patients with bone sarcomas the diaphysis was involved in 10 patients, the metaphysis in 9 patients and the epiphysis in 8 patients while 12 patients had extraosseous tumour tissue. The neurovascular bundle was involved in 6 patients, a tumour thrombus was present in 3 patients, a pathologic fracture was present in 4 patients, the joint was invaded in 9 patients, cortical destruction was present in 15 patients, a tumour necrosis was present in 16 patients and lymph node metastases were present in 6 patients. A contrast-enhancement of the bone marrow oedema was recognized in 8 patients in Gd-MRI and 4 patients in Fe-MRI, a contrast-enhancement of the peritumoural oedema was registered in 13 patients in Gd-MRI and 2 patients in Fe-MRI.

Table 5:

Results qualitative analysis – Assessment and Confidence

	Assessment				Confidence
Parameter	Mean Diff.	Difference 95%CI		p-value	Mean Diff.	Difference 95%CI		p-value
		lower	upper			lower	upper
Tumour presence in
• Diaphysis	0	-	-	-	0	-	-	-
• Epiphysis	−0.03	−0.01	0.05	0.489	0.01	−0.11	0.13	0.876
• Metaphysis	0	-	-	-	0	-	-	-
Extraosseous tumour tissue	0	-	-	-	0	-	-	-
Joint invasion	0	-	-	-	0	-	-	-
Tumour necrosis	0.05	−0.03	0.09	0.371	0.06	−0.03	0.13	0.238
Cortical destruction	0	-	-	-	0	-	-	-
Pathological fracture	0	-	-	-	0	-	-	-
Encasement of neurovasc. bundle	0	-	-	-	0.24	0.14	0.31	<0.001
Tumour thrombus	0	-	-	-	0.42	0.34	0.48	<0.001
CE peritumoral oedema	−0.55	−0.75	−0.35	<0.001	0.13	−0.06	0.32	0.190
CE marrow oedema	−0.26	−0.41	−0.10	0.002	0.15	−0.01	0.32	0.064
Skip lesions	0	-	-	-	0.04	−0.03	0.11	0.244
Bone metastases	0	-	-	-	0	-	-	-
Lymph node metastases	0	-	-	-	0	-	-	-

Results for assessment and corresponding confidence evaluation for all 15 tumour parameters. Parameters with a mean difference of “0” showed perfect agreement between readers which indicates that these parameters can be equally analysed on both FMX- and Gd-MRI.

Assessment: Agreement between readers in terms of assessment of a specific parameter’s presence with mean differences of readers’ ratings, 95% confidence interval and according p-value. A positive mean difference shows that this specific parameter was rated as more present on FMX-MRI compared to Gd-MRI and vice versa.

Confidence: Agreement between readers regarding confidence of assessment with mean differences of readers’ ratings, 95% confidence interval and according p-value. A positive mean difference shows that this specific parameter was rated with more confidence on FMX-MRI compared to Gd-MRI.

FMX-MRI = Ferumoxytol-enhanced MRI scan-protocol; Gd-MRI = Gadolinium-enhanced MRI scan-protocol; CE = contrast-enhanced; 95%CI = 95% confidence interval; Multivariate analysis using a mixed effect model while considering clustering within patients and readers

Significant results are highlighted in bold. Significant differences were assumed for p≤0.05.

In addition, the reported confidence for assessment of tumour presence in diaphysis/epiphysis/metaphysis, extraosseous tumour tissue, joint invasion, tumour necrosis, cortical destruction, pathological fracture, contrast-enhanced peritumoural oedema, contrast-enhanced bone marrow oedema, skip lesions, bone metastases, and lymph node metastases showed either perfect or high agreement between FMX- and Gd-MRI (see Table 5). However, the reported reader confidence for evaluation of presence or absence of a tumour thrombus was significantly higher on FMX-MRI compared to Gd-MRI (0.62 vs 0.20; p<0.001) and the confidence for evaluation of involvement of the neurovascular bundle was significantly higher on FMX-MRI compared to Gd-MRI (0.76 vs 0.52; p<0.001; see Table 5).

The results of the separate inter-reader agreement analyses (k alpha + 95%CI) are presented in Supplemental table 3.

Discussion

Overall, we found an equivalent performance of ferumoxytol- and gadolinium-enhanced MRI scans for assessment of bone and soft tissue sarcomas in paediatric and adolescent patients. The tumour T1-enhancement was significantly more hyperintense on Gd-MRI compared to FMX-MRI scans. This caused decreased tumour-to-muscle contrast on T1-weighted FMX-scans, which however, could be compensated by unimpaired tumour-to-muscle contrast on T2-FSE scans. FMX-MRI scans provided improved tumour-to-liver contrast and improved tumour-to-vessel contrast compared to Gd-MRI scans, which facilitated the detection of vascular involvement and tumour thrombi.

To the best of our knowledge, our study is the first to compare ferumoxytol- and gadolinium-enhanced MRI for assessment of sarcomas. Several other authors compared iron- and gadolinium-based contrast agents for detection and characterization of other tumour types, such as brain tumours with corresponding leakiness of aberrant blood vessels [34–36] and focal liver lesions [37–40]. In accordance with our results, Dósa et al. [35] reported equal performance of FMX- and Gd-MRI for detection of intracranial tumours. Gahramanov et al. [34] found different enhancement patterns of malignant brain tumours with FMX- and Gd-enhanced MRI: Ferumoxytol-enhanced MRI demonstrated enhancement of growing tumours but not pseudoprogression, while Gd-MRI showed enhancement of both pathologies. Similarly, Hamilton et al. [36] found FMX-enhancement of dural metastases and not meningiomas, while Gd-MRI showed enhancement of both. The underlying reason for the observed differential nanoparticle enhancement is likely leaky micro-vessels in malignant tumours and not benign conditions. In the liver, previous investigators reported, in accordance with our findings, that iron oxide nanoparticles detected more liver metastases than Gd-MRI [37–40].

Aghighi et al. [41] investigated the value of FMX-MRI for the detection of tumour-associated macrophages (TAM) in paediatric lymphomas and bone sarcomas and found that ferumoxytol-tumour enhancement correlated with the degree of TAM infiltration which might allow to stratify patients with TAM-rich tumours to immune-targeted therapies and to monitor tumour response to these therapies. This study did not include comparisons with Gd-MRI, however.

The assessment of tumour size on MRI is clinically important because it determines the planning for surgical margins [42]. Previous studies reported a high correlation (Spearman correlation: 0.96) of unenhanced T1-weighted sequences (sequence type unknown) with tumour size on histopathology [42, 43]. Gadolinium enhancement did not show any benefit for tumour size determination [42]. Similarly, our data show that tumour size measurement was not impaired by the more limited (moderate) enhancement on T1-SE images on ferumoxytol-enhanced sequences. Ferumoxytol is phagocytosed by bone marrow cells and thereby improves the tumour-to-bone marrow contrast compared to unenhanced T2-weighted scans [44–46].

Aghighi et al. [47] described that ferumoxytol nanoparticles can leak into early tumour necrosis and cause a combined T1- and T2-signal effect on MR images. In our study, we did not find this to be a diagnostic problem. Confidence for the diagnosis of tumour necrosis was moderate on both Gd- and FMX-MRI. Our results are in concordance with a recent study by Crombé et al. [48] who described moderate interobserver agreement for the diagnosis of necrosis in soft-tissue sarcomas on MRI.

The diagnosis of a pathologic fracture and cortical destruction is associated with poor overall survival in patients with bone sarcomas [49]. Our results showed no significant difference between FMX- and Gd-MRI in the evaluation of these criteria.

Ferumoxytol enabled comprehensive imaging of the arterial and venous system in children with little vulnerability to issues of poor contrast, “missed contrast bolus” or renal elimination due to its dextran derivative coat which prolongs its intravascular half-life and decreases leakage into the extravascular space [50]. In patients with sarcoma, neurovascular involvement and the detection of a tumour thrombus influence the operative approach and overall prognosis [51]. Our data showed significantly improved vessel and tumour thrombus delineation on FMX-MRI. In the rare event that a patient might have contraindications for both Gd-chelates and ferumoxytol, non-contrast enhanced MRI techniques such as balanced steady state free precession (bSSFP) might be an alternative for tumour thrombi in local veins. Although bSSFP has shown high sensitivity for the detection of tumour thrombi of renal tumours [52], future studies have to show if the same applies to bone tumours.

In comparison to T1w-images, T2w- or STIR-sequences were previously described to overestimate bone tumour limits because of confounding T2-signal effects of peritumoural inflammatory tissue [42, 43]. We found significantly less ferumoxytol uptake in peritumoural inflammatory tissue compared to Gd-MRI, therefore improving delineation of tumour margins in the bone.

We acknowledge a few limitations to our study. First, our patient population is limited by a relatively small sample size of 22 patients. Nonetheless, sarcomas are rare, and our cohort is unique for having both pre-treatment FMX- and Gd-enhanced MRI scans in the same patient. Second, FMX- and Gd-MRIs took place on different MRI scanners. However, these scanners used near identical pulse sequence parameters.

In conclusion, ferumoxytol-enhanced MRI is a valid alternative to gadolinium-enhanced MRI for tumour staging in paediatric and adolescent sarcoma patients. FMX-MRI can enable whole body staging and local tumour staging in one session and in addition, avoids risk of gadolinium deposition in the brain and NSF.

Key points.

• In contrast to gadolinium, ferumoxytol does not bear the risk of a nephrogenic systemic fibrosis in patients with an impaired kidney function or brain deposition after repetitive administration. However, it is important to note that ferumoxytol has a higher risk of serious adverse events compared to gadolinium chelates. Iron oxide nanoparticle-induced complement activated pseudoallergies can be significantly reduced by a slow infusion of diluted ferumoxytol nanoparticles over 15 minutes as recommended by the FDA. Also, before ferumoxytol is applied, a pre-existing iron overload should be ruled out.

• Ferumoxytol-enhanced MRI allows an equal assessment of size and other diagnostic parameters compared to gadolinium-enhanced MRI and thus is a valid alternative MRI contrast agent for tumour staging in paediatric and adolescent sarcoma patients.

• Ferumoxytol-enhanced MRI allows an improved detection of tumour tissue within or in close vicinity to bone marrow, liver parenchyma or vessels compared to gadolinium-enhanced MRI.

Abbreviations

bSSFP

Balanced steady state free precession

CNR

Contrast-to-noise ratio

COG

Children’s oncology group

DSRCT

Desmoplastic small round cell tumour

FDA

U.S. food and drug administration

FMX

Ferumoxytol

FMX-MRI

Ferumoxytol-enhanced MRI scan-protocol

GBCAs

Gadolinium-based contrast agents

Gd

Gadolinium

Gd-MRI

Gadolinium-enhanced MRI scan-protocol

GIST

Gastrointestinal stroma tumour

IND

Investigational new drug

MPNST

Malignant peripheral nerve sheath tumour

MRI

Magnetic resonance imaging

NSF

Nephrogenic systemic fibrosis

SNR

Signal-to-noise ratio

T1-LAVA

T1-liver acquisition with volume acceleration

T1-SE

T1-weighted spin echo sequence

T2-FSE

T2-weighted fast spin echo sequence

TAM

Tumour-associated macrophages