Intraprocedural MRI-based dosimetry during transarterial radioembolization of liver tumours with holmium-166 microspheres (EMERITUS-1): a phase I trial towards adaptive, image-controlled treatment delivery

Over the past two decades, transarterial radioembolization (TARE) has become an established treatment method, predominantly for hepatocellular carcinoma [1, 2] and colorectal cancer liver metastases [3‐5]. Its clinical value and safety have also been explored in cholangiocarcinoma [6] and liver metastases of various other origins, such as breast cancer [7]. Treatment involves injection of microspheres labelled with the beta-emitting isotope yttrium-90 (⁹⁰Y) or holmium-166 (¹⁶⁶Ho) into the hepatic artery under angiography guidance. The microspheres are transported by the blood flow and lodge in the arterioles in the tumour and healthy tissue because of their size (mean of 30 µm), potentially resulting in high absorbed tumour doses as a result of preferential arterial vascularization of tumour tissue (as opposed to the healthy liver tissue, which is primarily vascularized through the portal venous system) [8]. The treatment parameters such as injected activity and injection positions are partly based on a treatment simulation, and the resulting absorbed dose distribution is only assessed after treatment. As a result, not all tumours necessarily receive an adequate dose and, indeed, many patients have a suboptimal response to TARE [9].

It has become increasingly clear that dosimetry is a key aspect in TARE. A successful TARE treatment is characterized by a high tumour dose and a relative low dose in the healthy tissue. The importance of dosimetry is clearly illustrated in an ancillary study of the SARAH trial, in which it was established that patients of whom all tumours had received a mean dose of at least 100 Gy had a prolonged survival compared to the group that had received less than 100 Gy (median of 14.1 months vs. 6.1 months) [10]. Additionally, in the prospective DOSISPHERE-01 study, it was established that striving for a mean tumour dose ≥ 205 Gy led to an increased response rate of 71% vs. 36% using conventional dose prescription [11]. There is, however, currently no feedback on the actual dose distribution during treatment with TARE, as dosimetry is conventionally only performed after finishing the procedure by means of PET or SPECT imaging.

The ¹⁶⁶Ho microspheres commercially available for TARE facilitate multimodal imaging after treatment, as they can be visualized and quantified with both SPECT and MRI. The latter intrinsically results in high-resolution dose maps that can be acquired within a relatively short time frame [12, 13]. If a patient would be treated with TARE while positioned in the MRI scanner, MRI quantification of ¹⁶⁶Ho microspheres could be applied for acquisition of intraprocedural dosimetry, which could be used for an image-guided approach to TARE. In the present study, we investigate the feasibility and safety of performing ¹⁶⁶Ho TARE within an MRI scanner and explore the potential of intraprocedural MRI-based dosimetry for making TARE an image-guided procedure.

This phase I trial demonstrated the feasibility and safety of performing TARE with the patient positioned in the MRI scanner, allowing for intraprocedural dosimetry. Almost all catheter positions were stable for transport towards the MRI scanner, and MRI-based dosimetry could be performed in 18/19 tumours. No adverse events were attributed to treatment in the MRI. Moreover, we are the first to our best knowledge to acquire dosimetry data during actual TARE treatment delivery and to observe the changing dose distributions as more microspheres are injected.

The majority of the increase in mean absorbed tumour dose (14/18 tumours, 78%) was linearly correlated with the injected activity, which could indicate that most tumours were not saturated with microspheres at the end of treatment. In four out of 18 tumours (22%), we found an uptake pattern that resulted in a lower mean dose than expected based on the linear uptake pattern, albeit only possible to evaluate in the last fraction. This could suggest that these tumours were close to saturation with microspheres, and the maximum achievable mean tumour dose was approached. However, as this was always only the case in the last 30% of the injected activity, this finding may be confirmed in follow-up studies with more injection time points or an increased injected activity. This linear uptake pattern in the majority of the tumours suggests that a small amount of microspheres (such as the commercially available glass ⁹⁰Y microspheres) would result in a similar T/N ratio as a large amount of microspheres (such as the commercially available resin ⁹⁰Y microspheres or the ¹⁶⁶Ho microspheres). It also further validates the use of ¹⁶⁶Ho scout dose [16] as a predictor of the final dose distribution, which essentially consists of a small amount of ¹⁶⁶Ho microspheres. Most importantly, it suggests that, in most cases, the maximum achievable mean tumour dose has not been reached, a maximum which can only be achieved by microsphere saturation of the tumour microvasculature. This leaves a room for personalized treatment optimization, as an increased tumour dose has been correlated with an improved response rate [9]. Using an image-guided approach, more microspheres could be injected until the point of complete tumour saturation, as long as the healthy liver dose remains below the toxicity threshold.

Treatment personalization is one of the key advances to be made to improve treatment efficacy in the upcoming years, not only in TARE [11], but also in other radionuclide therapies such as [¹⁷⁷Lu]Lu-PSMA [18] or [¹⁷⁷Lu]Lu-octreotate radioligand therapy [19]. Further personalization of TARE has been recommended by two different expert groups [20, 21], mainly through improving activity prescription based on pre-treatment angiography and treatment simulation findings. Personalizing any form of radionuclide therapy involves adjusting treatment parameters to achieve a sufficiently high tumour dose, while maintaining a low healthy tissue dose. TARE is, however, vastly different from any other radionuclide therapy, as the microspheres have no biological half-life, which reduces the treatment complexity to transferring enough microspheres to the desired target volume, while avoiding deposition in healthy tissue. The microsphere distribution is a dynamic process based on arterial blood flow, catheter positioning and injection technique, as there is no receptor targeting. This is a complex process to predict beforehand, and dose distributions can also be dependent on the injected number of microspheres (i.e. specific activity) and flow redistributions as a result of decreasing flow in the tumour microvasculature. It would therefore be beneficial to have direct feedback on the dose distribution during the procedure. This concept of near real-time intraprocedural dosimetry facilitates a novel approach to TARE personalization, as it does not solely rely on absorbed dose predictions during a planning procedure, but it turns TARE into an adaptive treatment modality that can be adjusted based on intraprocedural findings. Image-guided ¹⁶⁶Ho TARE is one of the first examples that would fully embrace the theranostic characteristic of the device, in which every former distribution of the radioactive microspheres can be used to predict (diagnose) the distribution of the next therapeutic fraction.

As no definitive dose thresholds have been established that need to be achieved in order to expect a response after TARE [9], an image-guided approach should aim to achieve an as-high-as-possible tumour dose, as long as the healthy liver dose remains low. In healthy pigs, mean absorbed liver doses up to over 100 Gy were relatively well tolerated [22]. A relationship between healthy liver absorbed dose and toxicity has been established in patients with colorectal cancer metastases treated with ¹⁶⁶Ho TARE, even though no dose-limiting toxicity was observed [5]. The highest found healthy liver mean absorbed dose in that study was 55 Gy, and therefore, it has been recommended as the current healthy liver dose threshold. Similar safety data for patients with cirrhotic livers is expected in the near future from secondary analyses of the HEPAR Primary study [23, 24]. An additional advantage of intraprocedural dosimetry is that the physician can assess not only the mean tumour dose in near real time, but also the coverage of tumours with microspheres through the heterogeneity of the dose distribution. If (part of) a tumour is not reached, the catheter position could be altered empirically (currently with the use of X-ray guidance and cone-beam CT) in order to potentially improve the homogeneity of the intratumoural dose distribution (examples of which have been presented in this study).

Theoretically, imaging modalities other than MRI could also be used for an image-guided approach to TARE. The ¹⁶⁶Ho microspheres can also be visualized with SPECT, and CT-based quantification has been explored in a preclinical setting [25, 26], in patients with head-and-neck cancer after direct intratumoural injection [27], and has been explored in a single case after TARE [28]. Additionally, different types of ¹⁶⁶Ho microspheres with a higher density have been developed, which would increase the CT quantification capabilities [29, 30]. ⁹⁰Y-based microspheres used for TARE can be visualized by SPECT [31] and PET [32], and in a recent publication, the CT-based quantification of a novel radiopaque ⁹⁰Y microsphere was investigated in a preclinical model [33]. In general, nuclear medicine imaging modalities have much longer acquisition times than CT or MRI, making them less favourable for image guidance during TARE. MRI-based quantification has an advantage over CT with respect to soft-tissue contrast and is, currently, the most clinically established method [12], but on the other hand, a CBCT-guided approach would have the advantage of being more easily accessible in the hospital and would also facilitate the angiography procedure currently required to position the catheter for injection of microspheres (at the cost of extra radiation dose). All in all, both an MRI-based approach and a CT-based approach currently have their limitations, and research into the further development of either approach is required for full implementation in clinical practice.

The main limitation of our study is the requirement of a rather specific operation room design, and hospitals in which a hybrid OR is positioned directly adjacent to an MRI scanner are scarce. Additionally, our sample size was low, but deemed sufficient to demonstrate the feasibility and potential of an image-guided approach to TARE. Lastly, our current study design was based on using multiple vials of microspheres per injection position. This leads to a loss of activity in the system, as described in an earlier work [34]. An optimal image-guided procedure would use a system depending on only a single vial of activity, to minimize this loss.

Future studies are required to investigate the added value of this image-guided approach, both in terms of additional increase in achievable tumour dose and in patient response. This demands a novel phase I trial of the ¹⁶⁶Ho microspheres, in which the safety and efficacy of determining injected activity based on intraprocedural dosimetry of healthy liver tissue and tumours has to be established. If an added value of an image-guided approach can be demonstrated and the accessibility of the image-guided approach can be improved, this may enable the ultimate treatment personalization in TARE.

This study demonstrated the feasibility and safety of a first in-human application of TARE within a clinical MRI system. Intraprocedural MRI-based dosimetry enabled dynamic insight in the microsphere distribution during TARE and illustrated different uptake patterns between tumours. This proof of concept yields unique possibilities to better understand microsphere distribution in vivo and to potentially optimize treatment efficacy through treatment personalization in an image-guided setting.