Narrative literature analysis

Gadolinium reduction strategies for neuroimaging

Most studies published on Gd reduction strategies are primarily focused on static contrast- enhanced (CE) T1-weighted (T1w) imaging. Several studies have compared the use of half- dose imaging (e.g., 0.05 mmol/kg body weight) to full dose (e.g., 0.1 mmol/kg body weight) imaging in neuro MRI protocols. Initial studies compared the diagnostic certainty of detecting brain metastasis for different doses of GBCA. These have shown that for spin-echo MR imaging high dose GBCA was an efficient way to improve the detection of brain metastases, in particular of small metastases (Åkeson, 1995). However, other studies in the same period showed that half dose imaging with magnetization transfer did not lead to significant differences in contrast enhancement for extra-axial tumours (e.g. meningiomas) upon visual inspection when compared with standard dose imaging (Haba, 2001; Han, 1998).

At 3.0 T half-dose imaging using gadopentetate has shown to yield in significantly higher contrast-to-noise ratio (1.3-fold higher) compared to full-dose imaging at 1.5 T (Krautmacher, 2005). However, it should be not that the older studies addressing half dose imaging such as the 1995 study by Åkeson used linear GBCA’s which are currently no longer on available on the European Market because of suspended marketing authorizations due to the potential risk of gadolinium retention in the human body (Åkeson, 1995). This also applies to the study by Khoury Chalouhi et al. which performed an intraindividual and interindividual comparison between 0.075 mmol/kg and 0.1 mmol/kg of gadoterate meglumine for cranial MRI because of the higher relaxivity of this GBCA agent (Khoury Chalouhi, 2014). Also two recent studies evaluated the use of half dose imaging of the high- relaxivity linear GBCA gadobenate dimeglumine decmonstrated that half dose compared to full dose imaging was non-inferior with regard to visual lesion delineation, internal morphology, and contrast enhancement at 1.5 T and 3.0 T (DeLano, 2021), and small or ring- enhancing metastases can be better visualized on half-dose gadobenate dimeglumine delayed CE T2 FLAIR for than on half dose CE-T1w brain MRI scans (Jin, 2021). It should be noted that the findings of studies on linear GBCA’s cannot be extrapolated to the macrocyclic GBCAs that are currently used in neuroimaging because of the restricted use within the EU.

With regard to macrocyclic agents, the literature on reduced contrast dose is still limited and mainly based on the recent findings of the LEADER-75 trial. The LEADER-75 trial is an international prospective multicentre open-label crossover study that evaluated the use of three-quarter dose high-relaxivity gadobutrol (0.075 mmol/kg) compared with standard- dose gadoterate (0.1 mmol/kg) in adults with known or suspected CNS pathology undergoing CE brain imaging at 1.5T and 3T (Liu, 2021). Efficacy analysis in 141 patients found that improvement of reduced-dose gadobutrol over unenhanced images was noninferior to improvement of standard-dose gadoterate over unenhanced images. The authors used a 20% noninferiority margin for three primary efficacy measures using mean

readings (p ≤ 0.025). The total number of lesions detected by mean reading was 301 for reduced-dose gadobutrol versus 291 for standard-dose gadoterate. The sensitivity (58.7%), specificity (91.8%), and accuracy (70.2%) for malignancy from majority reading were identical for reduced-dose gadobutrol and standard-dose gadoterate. No differences in mean reader confidence (3.3 ± 0.6 for both reduced-dose gadobutrol and standard-dose gadoterate) and reader preference were found (95% CI, -0.10 to 0.11). Albeit that the LEADER-75 trial is the first study to demonstrate convincing evidence for reduced dose imaging for macrocyclic GBCA in neuroimaging with out compromising reader confidence and reader preference, there are several knowledge gaps that remain. These include the potential influence of field strength, the potential differences in CE in CNS pathologies that were underrepresented in the LEADER-75 sample (e.g. 41% of the sample consisted of meningiomas and 24% of metastases), and influence of sequence design and acquisition.

In addition to static CE-T1W for brain imaging, reduced dose imaging has also been investigated for brain perfusion using dynamic susceptibility contrast (DSC). In DSC half dose imaging has shown to lead to a more accurate arterial input function (Filice, 2017) and CBV maps of comparable diagnostic quality as the corresponding images acquired with a full dose imaging (Crisi, 2017). However, in acute stroke half-dose DSC imaging was found to result in poor image quality in 40.7% of the cases receiving half-dose GBCA (0.1 ml/kg gadobenate dimeglumine body weight) vs. 6.3% of patients who received full GBCA dose (0.2 ml/kg gadobenate dimeglumine body weight), and may thus adversely affect stroke patient triage for thrombectomy (Heit, 2021). For DSC MRI in neuroimaging the field dependency remains to be investigated, and reduced dose imaging for this application remains controversial.

More recently, several recent studies have evaluated the clinical performance of deep learning (DL)–based methods for brain MRI reducing contrast dose up to 10-fold (Gong, 2021). However the missing enhancement in small lesions indicates the need for further improvements in DL based algorithms or dosage design (Ammari, 2022; Luo, 2021). To date, DL strategies to minimize the dosage of GBCAs in brain MRI are still in its infancy and additional studies on the (potential) loss of diagnostic information are warranted.

Beyond reducing GBCA dose, the omission of the need of GBCA-based sequences in MRI scan protocols is also widely studied, in particular for the follow-up of extra-axial brain masses. The majority of the studies have focused on vestibular schwannomas, including evaluation of the diagnostic accuracy of non-CE gradient-echo constructive interference in steady state (CISS) and coronal T2w imaging in the setting of screening (Abele, 2014). A recent meta-analysis evaluated non-CE imaging for diagnosis and monitoring of vestibular schwannomas (Kim, 2019). In this meta-analysis six studies evaluated measurement difference, five articles focused on diagnostic accuracy and eight studied adverse effects. The studies showed that a non-CE MRI scan protocol with T2w imaging is highly accurate and highly reliable for diagnosing and monitoring vestibular schwannoma in comparison with CE-T1w imaging. In addition to vestibular schwannomas, also for meningiomas it has been shown that dimensions measured on pre-contrast T2 have similar results compared to measurements on CE-T1w imaging (Rahatli, 2019). One study leaving out CE-sequences in MRI protocols in children (Marsault, 2019). This study investigated the use of non-CE MRI for the follow-up of optic pathway gliomas in children, suggesting that tumour volume variation may be sufficient to assess optic pathway glioma progression (Marsault, 2019).

In addition to extra-axial masses, another disease group in which comparative studies on non-CE versus CE MRI protocols have been evaluated is multiple sclerosis. For multiple sclerosis three studies have evaluated the use of MRI protocols using non-contrast sequences for the detection of new brain lesions on follow-up imaging (Eichinger, 2021). These studies indicate that considering the very low incidence rate of new enhancing lesions in patients with non-progressive disease on follow-up, the routine administration of contrast in follow-up MRI scans is of limited value and does not change the diagnosis interval of disease progression. Finally, one study was identified that compared the use of non-CE MRI protocols to protocols using GBCAs for stroke. This study evaluated non-CE MR venography compared to conventional CE-T1w imaging and 3D gradient echo CE-T1w imaging, demonstrating that unenhanced MR venography had slightly lower sensitivity, specificity and accuracy for detecting cortical venous and/or dural sinus thrombosis (Sari, 2015) compared to the contrast-enhanced protocols.

Gadolinium reduction strategies for cardiovascular imaging

Thoracic aorta

Sequence designs such state-state free pression (SSFP) that enable combining non-breath- hold acquisitions with cardiac gating and respiratory triggering are widely used for vascular imaging with high resolution and high contrast between blood in the aorta and coronaries compared to surrounding tissue (Amano, 2008; Deshpande, 2001; François, 2008; Krishnam, 2008). International joint-society guidelines in the field of cardiology, cardiothoracic surgery and imaging have stated that non-CE MRA and CE-MRA are both acceptable imaging studies to measure the aorta in patients with thoracic aorta disease and adults with congenital heart disease (Baumgartner, 2010; Hiratzka, 2016). Direct comparison between non-CE MRA to CE-MRA for the assessment of the dimensions of the thoracic aorta has been performed in several studies demonstrating that diagnostic image quality can be achieved without the need for Gadolinium (Bannas, 2013; Groth, 2012; Pennig, 2021; Veldhoen, 2017; Von Tengg-Kobligk, 2009).

Supra-aortic vasculature

Several studies have compared various non-CE MRA techniques for blood flow-related luminography, such as gradient-echo based time-of-flight (ToF), with CE-MRA for evaluating stenosis of the supra-aortic arteries. These studies show that non-CE MRA techniques are promising alternatives for stenosis grading, in particular for distinguishing surgically treatable internal carotid artery stenosis, without significantly compromising image quality or diagnostic accuracy (Babiarz, 2009; Lim, 2008; Liu, 2019; Peters, 2019; Zhang, 2020).

With regard to half dose imaging, half-dose (0.05 mmol/kg body weight) CE MRA and full- dose (0.1 mmol/kg body weight) CE-MRA have been evaluated with regard to SNR, CNR at both 1.5 T and 3T, demonstrating that dose-reduction of cervical CE-MRA is feasible at 3T without compromising angiographic quality with regard to stenosis evaluation (Dehkharghani, 2015). Low dose time-resolved CE-MRA has been evaluated compared to non-contrast ToF MRA and high-resolution CE-MRA, showing that time-resolved MRA has a good image quality and accurate stenosis grading compared to high-resolution CE-MRA and might be more useful than ToF-MRA (Lee YJ, 2015).

Two studies evaluated ultralow-dose Gd (2-3 mL) time-resolved MRA versus standard dose (0.1 mmol/kg) CE-MRA for the evaluation of supra-aortic arterial stenosis at 3T. These showed that image quality and diagnostic agreement for stenotic disease in ultralow dose time-resolved MRA scans using 2-3 mL were not inferior to standard dose CE-MRA (Bak, 2017; Lohan, 2009). However, Gd doses below 2 mL were considered limited in spatial resolution leading to a tendency of overestimating stenosis grade. Also for Gd doses as low as 0.047 mmol/kg MRA of the supra-aortic arteries can be performed at 3T, without compromising image quality, acquisition speed, or spatial resolution (Tomasian, 2008).

However, it should be noted that the imaging quality at local centres will depend on the local MRI physics expertise to implement non-CE MRA or ultra-low CE-MRA techniques.

Abdominal vasculature, peripheral arteries, and vascular malformations

For the abdominal aorta and pelvic vasculature only one study was identified that compared low dose (Takahashi, 2004) to standard dose CE MRA. Three studies have been published that evaluated hepatic vasculature using non-contrast MRA compared to CE sequences (Kumar, 2021; Luk, 2017; Puippe, 2012), indicating that CE-MRI is not superior in depicting hepatic anatomy.

Several comparative studies have been published that evaluated non-contrast MRA for the assessment of renal artery stenosis, in particular balances steady-state free precession MR angiography (b-SSFP MRA) has shown promise in diagnosing renal artery stenosis (Aydin, 2017; Braidy, 2012; Glockner, 2010; Khoo, 2011; Lal, 2021; Maki, 2007). Also the three- dimensional Fast Imaging Employing Steady-State Acquisition (3D-FIESTA) sequence has been compared to CE-MRA and digital subtraction angiography (Gaudiano, 2014), suggesting that also 3D FIESTA sequence could be a useful tool in evaluating RAS. Further studies are needed to evaluate whether non-contrast MRA can truly replace CE-MRA to determine the presence of significant renal artery stenosis. Five studies were identified that evaluated non- contrast MRA for the evaluation of peripheral arterial occlusive disease. Although some studies were promising with regard to potential of non-contrast MRA techniques (Hodnett, 2011; Knobloch, 2021; Thierfelder, 2014), also concerns were expressed with regard to the rate of non-diagnostic vessel segments being considerably higher for non-contrast MRA than for CE-MRA (Diop, 2013; Schubert, 2016).

A specific indication of vascular imaging is the evaluation of vascular malformations. There is limited information in this disease group. One study evaluated a low dose CE protocol for diagnostic accuracy for treatment planning and follow-up but did not compare to standard dose MRI (Anzidei, 2011). For coil-embolized intra-cranial aneurysms it has been suggested that non-contrast ToF MRA can be used as a diagnostic alternative to CE ToF MRA (Behme, 2016).

Heart

With regards to cardiac imaging, various studies have been published that evaluated the possibilities of non-contrast imaging for various applications. This is of particular relevance in cardiac imaging considering that high-risk populations with chronic kidney disease often are referred for cardiac imaging due to concomitant cardiovascular disease as part of cardiorenal syndrome. Few studies have evaluated the possibilities of detecting myocardial fibrosis using non-Gd protocols (Graham-Brown, 2018) or with lowered Gd dose using higher-relaxivity contrast media such as gadobenate dimeglumine (Cheong, 2015; Galea, 2014). With regards to these studies, it can be concluded that the need of GBCA is of great relevance for the detection of myocardial disease as the distribution of the Gd chelate to the increased extracellular volume in the equilibrium phase is the pathophysiological marker of delayed enhancement imaging, which as this moment cannot be reliable replaced by existing non-contrast sequences.

Although low dose GBCA protocols can visualize myocardial fibrosis, standard dose protocols did result in overall better image quality and should be routinely preferred (Galea, 2014). One study evaluated non-contrast coronary MRA for the detection of significant coronary artery disease combined with subsequent Gd adenosine stress perfusion imaging of the heart (Heer, 2013), indicating that additional stress perfusion imaging with Gd substantially improved the diagnostic accuracy of detecting significant coronary artery disease. A specific group in which the application of reduced dose (Faggioni, 2012; Montalt-Tordera, 2021) and non-CE protocols (Chang, 2013; Elzayat, 2018; Isaak, 2021) have been evaluated are patients with congenital heart disease. For visualization of anatomy of the great vessels in congenital heart disease, non-contrast MRA protocols can be used as alternative to contrast-enhanced (CE) MRA protocols. One study described the potential of DL for the improvement of contrast in low-dose MRA studies in patients with congenital heart disease (Montalt- Tordera, 2021).

Gadolinium reduction strategies for musculoskeletal imaging

Four studies were identified that evaluated half dose imaging for musculoskeletal indications involving the assessment of synovitis or tenosynovitis (Schueller-Weidekamm, 2013), bone and soft-tissue disease in children (Colafati, 2018), bone and soft tissue tumours (Costelloe, 2011) and the evaluation of cartilage. These studies indicate that half dose GBCA protocols may be used while maintaining image quality (Rehnitz, 2020), however the limited number of studies indicate the need for additional research on this topic.

Studies evaluating the added value of GBCAs to musculoskeletal imaging protocols were mainly focused on detecting synovitis in patients with osteoarthritis and in spinal imaging. Albeit non-CE sequences can visualize synovitis, these are limited with an underestimation for detecting synovitis in patients with osteoarthritis (Crema, 2013; de Vries, 2021; Eshed, 2015) and inflammatory arthritis (Hemke, 2013). A recent meta-analysis aimed at determining the correlation between knee synovitis assessed on non-CE and CE MRI with histology in patients with knee osteoarthritis found that CE MRI scores correlated best with inflammatory infiltrates of synovial tissue, while paucity of current evidence warrants further studies on non-contrast for detecting knee synovitis (Shakoor, 2020).

With regards to spinal imaging, only three studies have been published that evaluated the added diagnostic value of Gd to spinal imaging protocols. One study investigated the differentiation of epidural fibrosis from disc herniation (Passavanti, 2020), one the characterization of vertebral marrow infiltrative lesions (Zidan, 2014), and one debated the added value of post-Gd images in contrast to non-enhanced scans for diagnosis of spondylitis and its complications (Prasetyo, 2020).

Gadolinium reduction strategies for body imaging

Prostate

Several studies in the past few years have been performed that have investigated the performance of non-CE MRI of the prostate versus CE prostate MRI protocols. These studies have focused on the sensitivity and specificity of detecting prostate cancer using non- contrast imaging protocols (T2w + DWI [diffusion weighted imaging] sequences) compared to CE-imaging protocols (T2w + DWI + DCE [dynamic contrast enhanced imaging]). For the non-contrast imaging protocols the ranges for sensitivity and specificity of detecting clinically significant prostate cancer were respectively 63%-95% and 71-88%, compared to sensitivity and specificity ranging between 73-95% and 45-85% for protocols including Gd (Alabousi, 2019; Bass, 2021; Cuocolo, 2021; Knaapila, 2020; Kuhl, 2017; Liang, 2020; Niu, 2018; Park, 2021; Tamada, 2021). An overview of the main studies investigating the performance of non-contrast MRI of the prostate vs contrast MRI was recently summarized by Pecoraro (2021). For intra-procedural prostate imaging for identification of ablation zone extent, non-contrast T2*w-MRI in one study has shown to be comparable to CE T1w-MRI suggesting that this might be a method for repeated intra-procedural monitoring of the thermal ablation zone without the need for Gd (Sun, 2021). With regard to lowered Gd dose strategies for prostate cancer, only one small study in 17 patients was identified that evaluated whether administration of low doses of Gd for DCE MRI can be as effective as a standard dose in distinguishing prostate cancer from benign tissue (He, 2018).

Liver

Several studies have been performed that compared half-dose imaging with standard-dose imaging in liver MRI. Most studies focused on gadobenate dimeglumine which has high T1 relaxivity and can be used for both dynamic and delayed liver MRI. A blinded intra-individual study evaluating standard and low dose liver MRI with gadobenate, found that albeit the standard dose yields greater relative enhancement, there is overall little improvement in subjective image quality (Kamali, 2020). Evaluation of enhancement patterns and characterization showed that half-dose and standard-dose liver MRI with gadopentetate dimeglumine found that 62 out of 64 lesions (97%) were identically characterized based on similar contrast enhancement compared to standard-dose gadodiamide (De Corato, 1999).

One study comparing half-dose gadobenate dimeglumine to standard dose gadopentetate dimeglumine showed that the half-dose imaging resulted in similar diagnostic information on dynamic imaging as well as the possibility of delayed imaging in the hepatobiliary phase (Schneider, 2003). Quarter-dose (0.025 mmol/kg) with gadobenate dimeglumine was compared retrospectively for image quality with half-dose imaging for abdominal MRI in patients at risk for nephrogenic systemic fibrosis, showing that the overall enhancement quality of the quarter dose was rated as good in all phases of enhancement, but was significantly lower than that for half-dose imaging (De Campos, 2011).

A recent meta-analysis study for surveillance MRI of hepatocellular carcinoma (HCC) using shortened MRI protocols (also referred to as abbreviated MRI) assessed the pooled sensitivity and specificity of contrast-enhanced hepatobiliary phase (HBP) abbreviated MRI (T2w, DWI, CE-T1w in HBP) and non-contrast abbreviated MRI (T2w, DWI, T1w dual-gradient echo imaging) (Kim, 2021). In this study there was a good overall diagnostic performance for detecting both any-stage HCC and early-stage HC, and the contrast-enhanced HBP abbreviated MRI showed a significantly higher sensitivity for detecting HCC than the non- contrast abbreviated MRI (87% vs. 82%) but had a significantly lower specificity (93% vs. 98%) (p = 0.03). The main limitation of the non-contrast abbreviated MRI is the relatively low lesion-to-liver contrast.

Also for liver metastases detection, non-contrast MRI protocols have been studied indicating that in particular DWI is an important sequence that improved mean specificity, positive predictive, negative predictive, and accuracy values for lesions either as small or greater than 1 cm (Colagrande, 2016). A comparative study in 175 patients with histologically confirmed 401 liver metastases and 73 benign liver lesions found no significant differences in sensitivity (range = 95.2-99.6%), specificity (range = 77.3-100%), positive predictive value (range = 92.9-100%) or negative predictive value (range = 87.5-95.7%) between the non- contrast MRI and the full MRI protocol with contrast (Hwang, 2019). These studies indicate that non-contrast liver MRI that includes DWI may serve as an alternative to contrast- enhanced MRI for detecting and characterizing liver metastases in patients with relatively high risk of liver metastases.

Finally, one study in patients with suspected possible choledocholithiasis evaluated the comparative performance of non-contrast MRI with half-Fourier acquisition single-shot turbo spin echo (HASTE) versus contrast-enhanced MRI/3D-magnetic resonance cholangiopancreatography (MRCP) (Kang, 2017). In this study the abbreviated non-contrast MRI with HASTE and full contrast-enhanced MRI/3D-MRCP resulted in high accuracy for choledocholithiasis (91.1-94.3% vs. 91.9-92.7%) and no differences in sensitivity or specificity were found, indicating that in patients with suspected choledocholithiasis, performance of non-contrast abdominal MRI with HASTE is similar to contrast-enhanced MRI with 3D-MRCP, offering potential for decreased scanning time and improved patient tolerability (Kang, 2017).

Other body imaging applications

Only few studies have been published that evaluate non-contrast MRI protocols for other body applications such as renal (Mawi, 2021), pancreatic (Lee, 2019), gastro-intestinal (Cattapan, 2019; Goshima, 2009; Kim SJ, 2019), and adnexal (Sahin, 2021) imaging. Albeit these studies are promising about the possibility of leaving out Gd-based sequences in MRI protocols without compromising diagnostic confidence, more studies are needed before specific recommendations on non-contrast MRI strategies for body imaging can be made.